From f096cc7dc1cdae1698eb7a340cd8c7f5ea0b1166 Mon Sep 17 00:00:00 2001 From: Nicolai Hähnle Date: Sun, 11 Oct 2009 12:40:07 +0200 Subject: r300g: Fix fragment program constants upload on R300 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Signed-off-by: Nicolai Hähnle --- src/gallium/drivers/r300/r300_emit.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 77ce431cdc3..570b4c5ef7f 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -212,7 +212,7 @@ void r300_emit_fragment_program_code(struct r300_context* r300, } if (constants->Count) { - OUT_CS_ONE_REG(R300_PFS_PARAM_0_X, constants->Count * 4); + OUT_CS_REG_SEQ(R300_PFS_PARAM_0_X, constants->Count * 4); for(i = 0; i < constants->Count; ++i) { const float * data = get_shader_constant(r300, &constants->Constants[i], externals); OUT_CS(pack_float24(data[0])); -- cgit v1.2.3 From 23c0c820e2767324546d450d2a7aa7bf1f70c36f Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Wed, 14 Oct 2009 11:42:05 +1000 Subject: r300g: fix case where texture unit 0 is disabled but unit 1 is enabled. to reproduce, start texrect, disable 0 texture in menu. Signed-off-by: Dave Airlie --- src/gallium/drivers/r300/r300_emit.c | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 570b4c5ef7f..99deb50400f 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -683,7 +683,8 @@ validate: /* ...textures... */ for (i = 0; i < r300->texture_count; i++) { tex = r300->textures[i]; - assert(tex && tex->buffer && "texture is marked, but NULL!"); + if (!tex) + continue; if (!r300->winsys->add_buffer(r300->winsys, tex->buffer, RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0)) { r300->context.flush(&r300->context, 0, NULL); @@ -770,12 +771,13 @@ validate: if (r300->dirty_state & (R300_ANY_NEW_SAMPLERS | R300_ANY_NEW_TEXTURES)) { for (i = 0; i < MIN2(r300->sampler_count, r300->texture_count); i++) { - if (r300->dirty_state & - ((R300_NEW_SAMPLER << i) | (R300_NEW_TEXTURE << i))) { - r300_emit_texture(r300, - r300->sampler_states[i], - r300->textures[i], - i); + if (r300->dirty_state & + ((R300_NEW_SAMPLER << i) | (R300_NEW_TEXTURE << i))) { + if (r300->textures[i]) + r300_emit_texture(r300, + r300->sampler_states[i], + r300->textures[i], + i); r300->dirty_state &= ~((R300_NEW_SAMPLER << i) | (R300_NEW_TEXTURE << i)); dirty_tex++; -- cgit v1.2.3 From 210481ae16e966865dcf9f1fd5f5dfabf4dc28bc Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Wed, 14 Oct 2009 15:13:25 +1000 Subject: r300g: attempt to make bo space check sane. This attempts to make r300g do proper bo space checking as opposed to whatever it was doing now. Signed-off-by: Dave Airlie --- src/gallium/drivers/r300/r300_context.c | 9 +++++++++ src/gallium/drivers/r300/r300_emit.c | 3 +++ src/gallium/drivers/r300/r300_winsys.h | 6 ++++++ src/gallium/winsys/drm/radeon/core/radeon_r300.c | 23 ++++++++++++++++++++--- 4 files changed, 38 insertions(+), 3 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c index 9cc455135db..e6bc80e48f8 100644 --- a/src/gallium/drivers/r300/r300_context.c +++ b/src/gallium/drivers/r300/r300_context.c @@ -136,6 +136,13 @@ r300_is_buffer_referenced( struct pipe_context *pipe, return PIPE_REFERENCED_FOR_READ | PIPE_REFERENCED_FOR_WRITE; } +static void r300_flush_cb(void *data) +{ + struct r300_context* const cs_context_copy = data; + + cs_context_copy->context.flush(&cs_context_copy->context, 0, NULL); +} + struct pipe_context* r300_create_context(struct pipe_screen* screen, struct r300_winsys* r300_winsys) { @@ -190,6 +197,8 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen, r300_init_state_functions(r300); r300_emit_invariant_state(r300); + + r300->winsys->set_flush_cb(r300->winsys, r300_flush_cb, r300); r300->dirty_state = R300_NEW_KITCHEN_SINK; r300->dirty_hw++; diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 99deb50400f..64748ad8f81 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -658,6 +658,9 @@ void r300_emit_dirty_state(struct r300_context* r300) r300_update_derived_state(r300); + /* Clean out BOs. */ + r300->winsys->reset_bos(r300->winsys); + /* XXX check size */ validate: /* Color buffers... */ diff --git a/src/gallium/drivers/r300/r300_winsys.h b/src/gallium/drivers/r300/r300_winsys.h index f18ad75a47d..540f8eca92e 100644 --- a/src/gallium/drivers/r300/r300_winsys.h +++ b/src/gallium/drivers/r300/r300_winsys.h @@ -92,6 +92,12 @@ struct r300_winsys { /* Flush the CS. */ void (*flush_cs)(struct r300_winsys* winsys); + + /* winsys flush - callback from winsys when flush required */ + void (*set_flush_cb)(struct r300_winsys *winsys, + void (*flush_cb)(void *), void *data); + + void (*reset_bos)(struct r300_winsys *winsys); }; struct pipe_context* r300_create_context(struct pipe_screen* screen, diff --git a/src/gallium/winsys/drm/radeon/core/radeon_r300.c b/src/gallium/winsys/drm/radeon/core/radeon_r300.c index d2d84f1a8f0..3587892e003 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_r300.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.c @@ -22,6 +22,17 @@ #include "radeon_r300.h" +static void radeon_r300_set_flush_cb(struct r300_winsys *winsys, + void (*flush_cb)(void *), + void *data) +{ + struct radeon_winsys_priv* priv = + (struct radeon_winsys_priv*)winsys->radeon_winsys; + + radeon_cs_space_set_flush(priv->cs, flush_cb, + data); +} + static boolean radeon_r300_add_buffer(struct r300_winsys* winsys, struct pipe_buffer* pbuffer, uint32_t rd, @@ -95,6 +106,13 @@ static void radeon_r300_write_cs_reloc(struct r300_winsys* winsys, } } +static void radeon_r300_reset_bos(struct r300_winsys *winsys) +{ + struct radeon_winsys_priv* priv = + (struct radeon_winsys_priv*)winsys->radeon_winsys; + radeon_cs_space_reset_bos(priv->cs); +} + static void radeon_r300_end_cs(struct r300_winsys* winsys, const char* file, const char* function, @@ -119,9 +137,6 @@ static void radeon_r300_flush_cs(struct r300_winsys* winsys) radeon_cs_print(priv->cs, stderr); } - /* Clean out BOs. */ - radeon_cs_space_reset_bos(priv->cs); - /* Reset CS. * Someday, when we care about performance, we should really find a way * to rotate between two or three CS objects so that the GPU can be @@ -203,6 +218,8 @@ radeon_create_r300_winsys(int fd, struct radeon_winsys* old_winsys) winsys->write_cs_reloc = radeon_r300_write_cs_reloc; winsys->end_cs = radeon_r300_end_cs; winsys->flush_cs = radeon_r300_flush_cs; + winsys->reset_bos = radeon_r300_reset_bos; + winsys->set_flush_cb = radeon_r300_set_flush_cb; memcpy(winsys, old_winsys, sizeof(struct radeon_winsys)); -- cgit v1.2.3 From c1bee7bdea470b6b5dcebef9aacc8fe4feca687c Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Wed, 14 Oct 2009 16:53:12 +1000 Subject: r300g: fixup arb occulsion query support. 1: add rv530 support - num z pipes cap - add proper start/finish query options for rv530 2: convert to use linked list properly. 3: add flushing required check. 4: initial Z top disabling support. TODO: make it actually work on my rv530. --- src/gallium/drivers/r300/r300_chipset.h | 2 + src/gallium/drivers/r300/r300_context.c | 10 ++-- src/gallium/drivers/r300/r300_context.h | 4 +- src/gallium/drivers/r300/r300_emit.c | 69 ++++++++++++++++++++---- src/gallium/drivers/r300/r300_flush.c | 10 +++- src/gallium/drivers/r300/r300_query.c | 42 ++++++++------- src/gallium/drivers/r300/r300_reg.h | 15 ++++-- src/gallium/drivers/r300/r300_screen.c | 1 + src/gallium/drivers/r300/r300_state.c | 11 ++-- src/gallium/drivers/r300/r300_winsys.h | 3 ++ src/gallium/winsys/drm/radeon/core/radeon_r300.c | 10 ++++ 11 files changed, 134 insertions(+), 43 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_chipset.h b/src/gallium/drivers/r300/r300_chipset.h index 322d4a57e41..f015a4243d3 100644 --- a/src/gallium/drivers/r300/r300_chipset.h +++ b/src/gallium/drivers/r300/r300_chipset.h @@ -36,6 +36,8 @@ struct r300_capabilities { int num_vert_fpus; /* The number of fragment pipes */ int num_frag_pipes; + /* The number of z pipes */ + int num_z_pipes; /* Whether or not TCL is physically present */ boolean has_tcl; /* Whether or not this is an RV515 or newer; R500s have many differences diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c index e6bc80e48f8..7a9c098e30f 100644 --- a/src/gallium/drivers/r300/r300_context.c +++ b/src/gallium/drivers/r300/r300_context.c @@ -99,11 +99,9 @@ static void r300_destroy_context(struct pipe_context* context) { context->screen->buffer_destroy(r300->oqbo); /* If there are any queries pending or not destroyed, remove them now. */ - if (r300->query_list) { - foreach_s(query, temp, r300->query_list) { - remove_from_list(query); - FREE(query); - } + foreach_s(query, temp, &r300->query_list) { + remove_from_list(query); + FREE(query); } FREE(r300->blend_color_state); @@ -201,6 +199,6 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen, r300->winsys->set_flush_cb(r300->winsys, r300_flush_cb, r300); r300->dirty_state = R300_NEW_KITCHEN_SINK; r300->dirty_hw++; - + make_empty_list(&r300->query_list); return &r300->context; } diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index 086633f732d..9b0094b63ca 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -172,6 +172,8 @@ struct r300_query { unsigned int count; /* The offset of this query into the query buffer, in bytes. */ unsigned offset; + /* if we've flushed the query */ + boolean flushed; /* Linked list members. */ struct r300_query* prev; struct r300_query* next; @@ -237,7 +239,7 @@ struct r300_context { /* Occlusion query buffer. */ struct pipe_buffer* oqbo; /* Query list. */ - struct r300_query* query_list; + struct r300_query query_list; /* Various CSO state objects. */ /* Blend state. */ diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 64748ad8f81..3d28249c164 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -323,28 +323,30 @@ void r300_emit_fb_state(struct r300_context* r300, void r300_emit_query_begin(struct r300_context* r300, struct r300_query* query) { + struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps; CS_LOCALS(r300); /* XXX This will almost certainly not return good results * for overlapping queries. */ - BEGIN_CS(2); + BEGIN_CS(4); + if (caps->family == CHIP_FAMILY_RV530) { + OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL); + } else { + OUT_CS_REG(R300_SU_REG_DEST, R300_RASTER_PIPE_SELECT_ALL); + } OUT_CS_REG(R300_ZB_ZPASS_DATA, 0); END_CS; } -void r300_emit_query_end(struct r300_context* r300, - struct r300_query* query) + +static void r300_emit_query_finish(struct r300_context *r300, + struct r300_query *query) { struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps; CS_LOCALS(r300); - if (!r300->winsys->add_buffer(r300->winsys, r300->oqbo, - 0, RADEON_GEM_DOMAIN_GTT)) { - debug_printf("r300: There wasn't room for the OQ buffer!?" - " Oh noes!\n"); - } - assert(caps->num_frag_pipes); + BEGIN_CS(6 * caps->num_frag_pipes + 2); /* I'm not so sure I like this switch, but it's hard to be elegant * when there's so many special cases... @@ -394,6 +396,55 @@ void r300_emit_query_end(struct r300_context* r300, } +static void rv530_emit_query_single(struct r300_context *r300, + struct r300_query *query) +{ + CS_LOCALS(r300); + + BEGIN_CS(8); + OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_0); + OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1); + OUT_CS_RELOC(r300->oqbo, query->offset, 0, RADEON_GEM_DOMAIN_GTT, 0); + OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL); + END_CS; +} + +static void rv530_emit_query_double(struct r300_context *r300, + struct r300_query *query) +{ + CS_LOCALS(r300); + + BEGIN_CS(14); + OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_0); + OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1); + OUT_CS_RELOC(r300->oqbo, query->offset, 0, RADEON_GEM_DOMAIN_GTT, 0); + OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_1); + OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1); + OUT_CS_RELOC(r300->oqbo, query->offset + sizeof(uint32_t), 0, RADEON_GEM_DOMAIN_GTT, 0); + OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL); + END_CS; +} + +void r300_emit_query_end(struct r300_context* r300, + struct r300_query* query) +{ + struct r300_capabilities *caps = r300_screen(r300->context.screen)->caps; + + if (!r300->winsys->add_buffer(r300->winsys, r300->oqbo, + 0, RADEON_GEM_DOMAIN_GTT)) { + debug_printf("r300: There wasn't room for the OQ buffer!?" + " Oh noes!\n"); + } + + if (caps->family == CHIP_FAMILY_RV530) { + if (caps->num_z_pipes == 2) + rv530_emit_query_double(r300, query); + else + rv530_emit_query_single(r300, query); + } else + r300_emit_query_finish(r300, query); +} + void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs) { CS_LOCALS(r300); diff --git a/src/gallium/drivers/r300/r300_flush.c b/src/gallium/drivers/r300/r300_flush.c index 0dff1c6f4fb..a8ab0d72120 100644 --- a/src/gallium/drivers/r300/r300_flush.c +++ b/src/gallium/drivers/r300/r300_flush.c @@ -26,9 +26,10 @@ static void r300_flush(struct pipe_context* pipe, unsigned flags, struct pipe_fence_handle** fence) { - struct r300_context* r300 = r300_context(pipe); - CS_LOCALS(r300); + struct r300_context *r300 = r300_context(pipe); + struct r300_query *query; + CS_LOCALS(r300); /* We probably need to flush Draw, but we may have been called from * within Draw. This feels kludgy, but it might be the best thing. */ if (!r300->draw->flushing) { @@ -41,8 +42,13 @@ static void r300_flush(struct pipe_context* pipe, r300->dirty_state = R300_NEW_KITCHEN_SINK; r300->dirty_hw = 0; } + /* reset flushed query */ + foreach(query, &r300->query_list) { + query->flushed = TRUE; + } } + void r300_init_flush_functions(struct r300_context* r300) { r300->context.flush = r300_flush; diff --git a/src/gallium/drivers/r300/r300_query.c b/src/gallium/drivers/r300/r300_query.c index 2880d34877f..b01313648bf 100644 --- a/src/gallium/drivers/r300/r300_query.c +++ b/src/gallium/drivers/r300/r300_query.c @@ -24,13 +24,13 @@ #include "r300_emit.h" -static struct pipe_query* r300_create_query(struct pipe_context* pipe, +static struct pipe_query *r300_create_query(struct pipe_context *pipe, unsigned query_type) { - struct r300_context* r300 = r300_context(pipe); - struct r300_screen* r300screen = r300_screen(r300->context.screen); - unsigned query_size = r300screen->caps->num_frag_pipes * 4; - struct r300_query* q, * qptr; + struct r300_context *r300 = r300_context(pipe); + struct r300_screen *r300screen = r300_screen(r300->context.screen); + unsigned query_size; + struct r300_query *q, *qptr; q = CALLOC_STRUCT(r300_query); @@ -39,13 +39,16 @@ static struct pipe_query* r300_create_query(struct pipe_context* pipe, q->active = FALSE; - if (!r300->query_list) { - r300->query_list = q; - } else if (!is_empty_list(r300->query_list)) { - qptr = last_elem(r300->query_list); + if (r300screen->caps->family == CHIP_FAMILY_RV530) + query_size = r300screen->caps->num_z_pipes * sizeof(uint32_t); + else + query_size = r300screen->caps->num_frag_pipes * sizeof(uint32_t); + + if (!is_empty_list(&r300->query_list)) { + qptr = last_elem(&r300->query_list); q->offset = qptr->offset + query_size; - insert_at_tail(r300->query_list, q); } + insert_at_tail(&r300->query_list, q); /* XXX */ if (q->offset >= 4096) { @@ -74,9 +77,10 @@ static void r300_begin_query(struct pipe_context* pipe, map = pipe->screen->buffer_map(pipe->screen, r300->oqbo, PIPE_BUFFER_USAGE_CPU_WRITE); map += q->offset / 4; - *map = ~0; + *map = ~0U; pipe->screen->buffer_unmap(pipe->screen, r300->oqbo); + q->flushed = FALSE; r300_emit_dirty_state(r300); r300_emit_query_begin(r300, q); } @@ -98,28 +102,30 @@ static boolean r300_get_query_result(struct pipe_context* pipe, { struct r300_context* r300 = r300_context(pipe); struct r300_screen* r300screen = r300_screen(r300->context.screen); - struct r300_query* q = (struct r300_query*)query; + struct r300_query *q = (struct r300_query*)query; unsigned flags = PIPE_BUFFER_USAGE_CPU_READ; uint32_t* map; - uint32_t temp; + uint32_t temp = 0; unsigned i; - if (wait) { + if (q->flushed == FALSE) pipe->flush(pipe, 0, NULL); - } else { + if (!wait) { flags |= PIPE_BUFFER_USAGE_DONTBLOCK; } map = pipe->screen->buffer_map(pipe->screen, r300->oqbo, flags); + if (!map) + return FALSE; map += q->offset / 4; for (i = 0; i < r300screen->caps->num_frag_pipes; i++) { - if (*map == ~0) { + if (*map == ~0U) { /* Looks like our results aren't ready yet. */ if (wait) { debug_printf("r300: Despite waiting, OQ results haven't" " come in yet.\n"); } - temp = ~0; + temp = ~0U; break; } temp += *map; @@ -127,7 +133,7 @@ static boolean r300_get_query_result(struct pipe_context* pipe, } pipe->screen->buffer_unmap(pipe->screen, r300->oqbo); - if (temp == ~0) { + if (temp == ~0U) { /* Our results haven't been written yet... */ return FALSE; } diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h index 3abff5db622..ae94bb9b9fc 100644 --- a/src/gallium/drivers/r300/r300_reg.h +++ b/src/gallium/drivers/r300/r300_reg.h @@ -1172,6 +1172,13 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. /* SU Depth Offset value */ #define R300_SU_DEPTH_OFFSET 0x42c4 +#define R300_SU_REG_DEST 0x42c8 +# define R300_RASTER_PIPE_SELECT_0 (1 << 0) +# define R300_RASTER_PIPE_SELECT_1 (1 << 1) +# define R300_RASTER_PIPE_SELECT_2 (1 << 2) +# define R300_RASTER_PIPE_SELECT_3 (1 << 3) +# define R300_RASTER_PIPE_SELECT_ALL 0xf + /* BEGIN: Rasterization / Interpolators - many guesses */ @@ -2095,6 +2102,10 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. #define R500_FG_ALPHA_VALUE 0x4be0 # define R500_FG_ALPHA_VALUE_MASK 0x0000ffff +#define RV530_FG_ZBREG_DEST 0x4be8 +# define RV530_FG_ZBREG_DEST_PIPE_SELECT_0 (1 << 0) +# define RV530_FG_ZBREG_DEST_PIPE_SELECT_1 (1 << 1) +# define RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL (3 << 0) /* gap */ /* Fragment program parameters in 7.16 floating point */ @@ -3313,10 +3324,6 @@ enum { #define R200_3D_DRAW_IMMD_2 0xC0003500 -/* XXX Oh look, stuff not brought over from docs yet */ - -#define R300_SU_REG_DEST 0x42C8 - #endif /* _R300_REG_H */ /* *INDENT-ON* */ diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c index 7d154576e04..5381651c77b 100644 --- a/src/gallium/drivers/r300/r300_screen.c +++ b/src/gallium/drivers/r300/r300_screen.c @@ -395,6 +395,7 @@ struct pipe_screen* r300_create_screen(struct r300_winsys* r300_winsys) caps->pci_id = r300_winsys->pci_id; caps->num_frag_pipes = r300_winsys->gb_pipes; + caps->num_z_pipes = r300_winsys->z_pipes; r300_parse_chipset(caps); diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index 3cef285dee8..d8533ac168e 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -190,6 +190,7 @@ static void* r300_create_dsa_state(struct pipe_context* pipe, const struct pipe_depth_stencil_alpha_state* state) { + struct r300_context* r300 = r300_context(pipe); struct r300_dsa_state* dsa = CALLOC_STRUCT(r300_dsa_state); /* Depth test setup. */ @@ -247,11 +248,15 @@ static void* R300_FG_ALPHA_FUNC_ENABLE; dsa->alpha_reference = CLAMP(state->alpha.ref_value * 1023.0f, 0, 1023); - } else { - /* XXX need to fix this to be dynamically set - dsa->z_buffer_top = R300_ZTOP_ENABLE; */ } + dsa->z_buffer_top = R300_ZTOP_ENABLE; + /* XXX TODO: add frag prog rules for ztop disable */ + if (state->alpha.enabled && state->alpha.func != PIPE_FUNC_ALWAYS) + dsa->z_buffer_top = R300_ZTOP_DISABLE; + if (!is_empty_list(&r300->query_list)) + dsa->z_buffer_top = R300_ZTOP_DISABLE; + return (void*)dsa; } diff --git a/src/gallium/drivers/r300/r300_winsys.h b/src/gallium/drivers/r300/r300_winsys.h index 540f8eca92e..864a6146b22 100644 --- a/src/gallium/drivers/r300/r300_winsys.h +++ b/src/gallium/drivers/r300/r300_winsys.h @@ -48,6 +48,9 @@ struct r300_winsys { /* GB pipe count */ uint32_t gb_pipes; + /* Z pipe count (rv530 only) */ + uint32_t z_pipes; + /* GART size. */ uint32_t gart_size; diff --git a/src/gallium/winsys/drm/radeon/core/radeon_r300.c b/src/gallium/winsys/drm/radeon/core/radeon_r300.c index 3587892e003..7ea5d1fb4e7 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_r300.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.c @@ -164,6 +164,16 @@ static void do_ioctls(struct r300_winsys* winsys, int fd) } winsys->gb_pipes = target; + /* get Z pipes */ + info.request = RADEON_INFO_NUM_Z_PIPES; + retval = drmCommandWriteRead(fd, DRM_RADEON_INFO, &info, sizeof(info)); + if (retval) { + fprintf(stderr, "%s: Failed to get GB pipe count, " + "error number %d\n", __FUNCTION__, retval); + exit(1); + } + winsys->z_pipes = target; + /* Then, get PCI ID */ info.request = RADEON_INFO_DEVICE_ID; retval = drmCommandWriteRead(fd, DRM_RADEON_INFO, &info, sizeof(info)); -- cgit v1.2.3 From 47791697ab6eb6965f0ba8ba3f20373b3753ca2a Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Wed, 14 Oct 2009 17:14:43 +1000 Subject: r300g: convert query to a state for emitting. This means we don't emit in the begin query but when we have to flush. Similiar to classic. TODO: make query object actually work. --- src/gallium/drivers/r300/r300_context.h | 2 ++ src/gallium/drivers/r300/r300_emit.c | 15 ++++++++++++--- src/gallium/drivers/r300/r300_query.c | 8 ++++++-- 3 files changed, 20 insertions(+), 5 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index 9b0094b63ca..3a01869ba10 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -141,6 +141,7 @@ struct r300_viewport_state { #define R300_NEW_VERTEX_FORMAT 0x04000000 #define R300_NEW_VERTEX_SHADER 0x08000000 #define R300_NEW_VIEWPORT 0x10000000 +#define R300_NEW_QUERY 0x20000000 #define R300_NEW_KITCHEN_SINK 0x1fffffff /* The next several objects are not pure Radeon state; they inherit from @@ -239,6 +240,7 @@ struct r300_context { /* Occlusion query buffer. */ struct pipe_buffer* oqbo; /* Query list. */ + struct r300_query *query_current; struct r300_query query_list; /* Various CSO state objects. */ diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 3d28249c164..babbe0dd74a 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -320,12 +320,16 @@ void r300_emit_fb_state(struct r300_context* r300, END_CS; } -void r300_emit_query_begin(struct r300_context* r300, - struct r300_query* query) +void r300_emit_query_start(struct r300_context *r300) + { - struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps; + struct r300_capabilities *caps = r300_screen(r300->context.screen)->caps; + struct r300_query *query = r300->query_current; CS_LOCALS(r300); + if (!query) + return; + /* XXX This will almost certainly not return good results * for overlapping queries. */ BEGIN_CS(4); @@ -772,6 +776,11 @@ validate: goto validate; } + if (r300->dirty_state & R300_NEW_QUERY) { + r300_emit_query_start(r300); + r300->dirty_state &= ~R300_NEW_QUERY; + } + if (r300->dirty_state & R300_NEW_BLEND) { r300_emit_blend_state(r300, r300->blend_state); r300->dirty_state &= ~R300_NEW_BLEND; diff --git a/src/gallium/drivers/r300/r300_query.c b/src/gallium/drivers/r300/r300_query.c index b01313648bf..fb4340ff3d1 100644 --- a/src/gallium/drivers/r300/r300_query.c +++ b/src/gallium/drivers/r300/r300_query.c @@ -74,6 +74,8 @@ static void r300_begin_query(struct pipe_context* pipe, struct r300_context* r300 = r300_context(pipe); struct r300_query* q = (struct r300_query*)query; + assert(r300->query_current == NULL); + map = pipe->screen->buffer_map(pipe->screen, r300->oqbo, PIPE_BUFFER_USAGE_CPU_WRITE); map += q->offset / 4; @@ -81,8 +83,8 @@ static void r300_begin_query(struct pipe_context* pipe, pipe->screen->buffer_unmap(pipe->screen, r300->oqbo); q->flushed = FALSE; - r300_emit_dirty_state(r300); - r300_emit_query_begin(r300, q); + r300->query_current = q; + r300->dirty_state |= R300_NEW_QUERY; } static void r300_end_query(struct pipe_context* pipe, @@ -93,6 +95,8 @@ static void r300_end_query(struct pipe_context* pipe, r300_emit_dirty_state(r300); r300_emit_query_end(r300, q); + + r300->query_current = NULL; } static boolean r300_get_query_result(struct pipe_context* pipe, -- cgit v1.2.3 From 51d1cf55da6f8b8a215814589a189b6e5e537fe5 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Wed, 14 Oct 2009 17:44:19 +1000 Subject: r300g: port over last parts of oq support. Add support for begin/end in each CS so we don't get any other processes rendering in between. TODO: blame other parts of driver for this not working like Z. --- src/gallium/drivers/r300/r300_context.h | 2 ++ src/gallium/drivers/r300/r300_emit.c | 11 +++++++++-- src/gallium/drivers/r300/r300_emit.h | 3 +-- src/gallium/drivers/r300/r300_flush.c | 2 ++ src/gallium/drivers/r300/r300_query.c | 7 ++----- 5 files changed, 16 insertions(+), 9 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index 3a01869ba10..7826ed14525 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -175,6 +175,8 @@ struct r300_query { unsigned offset; /* if we've flushed the query */ boolean flushed; + /* if begin has been emitted */ + boolean begin_emitted; /* Linked list members. */ struct r300_query* prev; struct r300_query* next; diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index babbe0dd74a..6e616cd5b2c 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -340,6 +340,7 @@ void r300_emit_query_start(struct r300_context *r300) } OUT_CS_REG(R300_ZB_ZPASS_DATA, 0); END_CS; + query->begin_emitted = TRUE; } @@ -429,10 +430,16 @@ static void rv530_emit_query_double(struct r300_context *r300, END_CS; } -void r300_emit_query_end(struct r300_context* r300, - struct r300_query* query) +void r300_emit_query_end(struct r300_context* r300) { struct r300_capabilities *caps = r300_screen(r300->context.screen)->caps; + struct r300_query *query = r300->query_current; + + if (!query) + return; + + if (query->begin_emitted == FALSE) + return; if (!r300->winsys->add_buffer(r300->winsys, r300->oqbo, 0, RADEON_GEM_DOMAIN_GTT)) { diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h index c4002b8e5d0..b62aa9fec50 100644 --- a/src/gallium/drivers/r300/r300_emit.h +++ b/src/gallium/drivers/r300/r300_emit.h @@ -58,8 +58,7 @@ void r300_emit_fb_state(struct r300_context* r300, void r300_emit_query_begin(struct r300_context* r300, struct r300_query* query); -void r300_emit_query_end(struct r300_context* r300, - struct r300_query* query); +void r300_emit_query_end(struct r300_context* r300); void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs); diff --git a/src/gallium/drivers/r300/r300_flush.c b/src/gallium/drivers/r300/r300_flush.c index a8ab0d72120..241ea71d6b1 100644 --- a/src/gallium/drivers/r300/r300_flush.c +++ b/src/gallium/drivers/r300/r300_flush.c @@ -36,6 +36,8 @@ static void r300_flush(struct pipe_context* pipe, draw_flush(r300->draw); } + r300_emit_query_end(r300); + if (r300->dirty_hw) { FLUSH_CS; r300_emit_invariant_state(r300); diff --git a/src/gallium/drivers/r300/r300_query.c b/src/gallium/drivers/r300/r300_query.c index fb4340ff3d1..2b0fbfb7d22 100644 --- a/src/gallium/drivers/r300/r300_query.c +++ b/src/gallium/drivers/r300/r300_query.c @@ -88,14 +88,11 @@ static void r300_begin_query(struct pipe_context* pipe, } static void r300_end_query(struct pipe_context* pipe, - struct pipe_query* query) + struct pipe_query* query) { struct r300_context* r300 = r300_context(pipe); - struct r300_query* q = (struct r300_query*)query; - - r300_emit_dirty_state(r300); - r300_emit_query_end(r300, q); + r300_emit_query_end(r300); r300->query_current = NULL; } -- cgit v1.2.3 From 88b697fb0aaaab8479716763510f56b1053ddb37 Mon Sep 17 00:00:00 2001 From: Dave Airlie Date: Wed, 14 Oct 2009 18:24:34 +1000 Subject: r300g: remove buffer add that should be unnecessary. This should be handled in the emit fine --- src/gallium/drivers/r300/r300_emit.c | 6 ------ 1 file changed, 6 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 6e616cd5b2c..feffadd0ee8 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -441,12 +441,6 @@ void r300_emit_query_end(struct r300_context* r300) if (query->begin_emitted == FALSE) return; - if (!r300->winsys->add_buffer(r300->winsys, r300->oqbo, - 0, RADEON_GEM_DOMAIN_GTT)) { - debug_printf("r300: There wasn't room for the OQ buffer!?" - " Oh noes!\n"); - } - if (caps->family == CHIP_FAMILY_RV530) { if (caps->num_z_pipes == 2) rv530_emit_query_double(r300, query); -- cgit v1.2.3 From a82fc97c643c4309a10cfefb108c4c0f11a2e55a Mon Sep 17 00:00:00 2001 From: Corbin Simpson Date: Wed, 14 Oct 2009 20:06:38 -0700 Subject: r300g: Move ZTOP to its own state atom. It may seem pointless, but this avoids a fair amount of predicted CSO pain. --- src/gallium/drivers/r300/r300_context.h | 8 +++++++- src/gallium/drivers/r300/r300_emit.c | 2 +- src/gallium/drivers/r300/r300_state_derived.c | 8 ++++---- src/gallium/drivers/r300/r300_surface.h | 1 - 4 files changed, 12 insertions(+), 7 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index d2e88755037..2acce0fd4a4 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -62,7 +62,6 @@ struct r300_dsa_state { uint32_t z_buffer_control; /* R300_ZB_CNTL: 0x4f00 */ uint32_t z_stencil_control; /* R300_ZB_ZSTENCILCNTL: 0x4f04 */ uint32_t stencil_ref_mask; /* R300_ZB_STENCILREFMASK: 0x4f08 */ - uint32_t z_buffer_top; /* R300_ZB_ZTOP: 0x4f14 */ uint32_t stencil_ref_bf; /* R500_ZB_STENCILREFMASK_BF: 0x4fd4 */ }; @@ -124,6 +123,10 @@ struct r300_viewport_state { uint32_t vte_control; /* R300_VAP_VTE_CNTL: 0x20b0 */ }; +struct r300_ztop_state { + uint32_t z_buffer_top; /* R300_ZB_ZTOP: 0x4f14 */ +}; + #define R300_NEW_BLEND 0x00000001 #define R300_NEW_BLEND_COLOR 0x00000002 #define R300_NEW_CLIP 0x00000004 @@ -281,6 +284,9 @@ struct r300_context { struct r300_vertex_shader* vs; /* Viewport state. */ struct r300_viewport_state* viewport_state; + /* ZTOP state. */ + struct r300_ztop_state ztop_state; + /* Bitmask of dirty state objects. */ uint32_t dirty_state; /* Flag indicating whether or not the HW is dirty. */ diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index feffadd0ee8..2c3bba952d8 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -106,7 +106,7 @@ void r300_emit_dsa_state(struct r300_context* r300, OUT_CS(dsa->z_buffer_control); OUT_CS(dsa->z_stencil_control); OUT_CS(dsa->stencil_ref_mask); - OUT_CS_REG(R300_ZB_ZTOP, dsa->z_buffer_top); + OUT_CS_REG(R300_ZB_ZTOP, r300->ztop_state.z_buffer_top); if (r300screen->caps->is_r500) { /* OUT_CS_REG(R500_ZB_STENCILREFMASK_BF, dsa->stencil_ref_bf); */ } diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c index 5d323a26b14..f0861a9cf1d 100644 --- a/src/gallium/drivers/r300/r300_state_derived.c +++ b/src/gallium/drivers/r300/r300_state_derived.c @@ -446,7 +446,7 @@ static void r300_update_rs_block(struct r300_context* r300) static void r300_update_ztop(struct r300_context* r300) { - r300->dsa_state->z_buffer_top = R300_ZTOP_ENABLE; + r300->ztop_state.z_buffer_top = R300_ZTOP_ENABLE; /* This is important enough that I felt it warranted a comment. * @@ -467,11 +467,11 @@ static void r300_update_ztop(struct r300_context* r300) * ~C. */ if (r300->dsa_state->alpha_function) { - r300->dsa_state->z_buffer_top = R300_ZTOP_DISABLE; + r300->ztop_state.z_buffer_top = R300_ZTOP_DISABLE; } else if (r300_fragment_shader_writes_depth(r300->fs)) { - r300->dsa_state->z_buffer_top = R300_ZTOP_DISABLE; + r300->ztop_state.z_buffer_top = R300_ZTOP_DISABLE; } else if (r300->query_current) { - r300->dsa_state->z_buffer_top = R300_ZTOP_DISABLE; + r300->ztop_state.z_buffer_top = R300_ZTOP_DISABLE; } } diff --git a/src/gallium/drivers/r300/r300_surface.h b/src/gallium/drivers/r300/r300_surface.h index f9e98b2ec9c..d5998e6e6da 100644 --- a/src/gallium/drivers/r300/r300_surface.h +++ b/src/gallium/drivers/r300/r300_surface.h @@ -54,7 +54,6 @@ static struct r300_dsa_state dsa_clear_state = { .z_buffer_control = 0x0, .z_stencil_control = 0x0, .stencil_ref_mask = R300_STENCILWRITEMASK_MASK, - .z_buffer_top = R300_ZTOP_ENABLE, .stencil_ref_bf = 0x0, }; -- cgit v1.2.3 From e4f21be13ac14edd89e865138a9e36b266425b39 Mon Sep 17 00:00:00 2001 From: Corbin Simpson Date: Thu, 15 Oct 2009 16:58:26 -0700 Subject: r300g: Set logical ID for each emitted texture/sampler. multitexarray works on my r300, but texrect doesn't. --- src/gallium/drivers/r300/r300_emit.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 2c3bba952d8..f3adc0968e8 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -531,7 +531,8 @@ void r300_emit_texture(struct r300_context* r300, CS_LOCALS(r300); BEGIN_CS(16); - OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), sampler->filter0); + OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), sampler->filter0 | + (offset << 28)); OUT_CS_REG(R300_TX_FILTER1_0 + (offset * 4), sampler->filter1); OUT_CS_REG(R300_TX_BORDER_COLOR_0 + (offset * 4), sampler->border_color); -- cgit v1.2.3 From fc8a156cfc539b9c04dc3527e4fc61cb4b0b688e Mon Sep 17 00:00:00 2001 From: Corbin Simpson Date: Fri, 16 Oct 2009 08:39:59 -0700 Subject: r300g: Use a hash table to look up vertex info. Need to move rs_block to this, too. Also, I'm getting massive amounts of flicker for some reason; I bet we've gotta re-re-examine PSC and friends. :C --- src/gallium/drivers/r300/r300_context.c | 18 ++++++- src/gallium/drivers/r300/r300_context.h | 10 +++- src/gallium/drivers/r300/r300_emit.c | 26 +++++----- src/gallium/drivers/r300/r300_render.c | 2 +- src/gallium/drivers/r300/r300_state_derived.c | 72 ++++++++++++++++++++------- src/gallium/drivers/r300/r300_state_derived.h | 4 ++ 6 files changed, 97 insertions(+), 35 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c index b243f88bb5b..a1156d2de65 100644 --- a/src/gallium/drivers/r300/r300_context.c +++ b/src/gallium/drivers/r300/r300_context.c @@ -89,10 +89,23 @@ static boolean r300_draw_arrays(struct pipe_context* pipe, unsigned mode, return r300_draw_elements(pipe, NULL, 0, mode, start, count); } -static void r300_destroy_context(struct pipe_context* context) { +static enum pipe_error r300_clear_hash_table(void* key, void* value, + void* data) +{ + FREE(key); + FREE(value); + return PIPE_OK; +} + +static void r300_destroy_context(struct pipe_context* context) +{ struct r300_context* r300 = r300_context(context); struct r300_query* query, * temp; + u_hash_table_foreach(r300->shader_hash_table, r300_clear_hash_table, + NULL); + u_hash_table_destroy(r300->shader_hash_table); + draw_destroy(r300->draw); /* Free the OQ BO. */ @@ -167,6 +180,9 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen, r300->context.is_texture_referenced = r300_is_texture_referenced; r300->context.is_buffer_referenced = r300_is_buffer_referenced; + r300->shader_hash_table = u_hash_table_create(r300_shader_key_hash, + r300_shader_key_compare); + r300->blend_color_state = CALLOC_STRUCT(r300_blend_color_state); r300->rs_block = CALLOC_STRUCT(r300_rs_block); r300->scissor_state = CALLOC_STRUCT(r300_scissor_state); diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index 2acce0fd4a4..2a62c67fc73 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -30,12 +30,14 @@ #include "tgsi/tgsi_scan.h" +#include "util/u_hash_table.h" #include "util/u_memory.h" #include "util/u_simple_list.h" #include "r300_clear.h" #include "r300_query.h" #include "r300_screen.h" +#include "r300_state_derived.h" #include "r300_winsys.h" struct r300_fragment_shader; @@ -248,6 +250,12 @@ struct r300_context { struct r300_query *query_current; struct r300_query query_list; + /* Shader hash table. Used to store vertex formatting information, which + * depends on the combination of both currently loaded shaders. */ + struct u_hash_table* shader_hash_table; + /* Vertex formatting information. */ + struct r300_vertex_format* vertex_info; + /* Various CSO state objects. */ /* Blend state. */ struct r300_blend_state* blend_state; @@ -278,8 +286,6 @@ struct r300_context { /* Vertex buffers for Gallium. */ struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS]; int vertex_buffer_count; - /* Vertex information. */ - struct r300_vertex_format vertex_info; /* Vertex shader. */ struct r300_vertex_shader* vs; /* Viewport state. */ diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index f3adc0968e8..e6092cda9bf 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -551,7 +551,7 @@ void r300_emit_vertex_buffer(struct r300_context* r300) DBG(r300, DBG_DRAW, "r300: Preparing vertex buffer %p for render, " "vertex size %d\n", r300->vbo, - r300->vertex_info.vinfo.size); + r300->vertex_info->vinfo.size); /* Set the pointer to our vertex buffer. The emitted values are this: * PACKET3 [3D_LOAD_VBPNTR] * COUNT [1] @@ -562,8 +562,8 @@ void r300_emit_vertex_buffer(struct r300_context* r300) BEGIN_CS(7); OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, 3); OUT_CS(1); - OUT_CS(r300->vertex_info.vinfo.size | - (r300->vertex_info.vinfo.size << 8)); + OUT_CS(r300->vertex_info->vinfo.size | + (r300->vertex_info->vinfo.size << 8)); OUT_CS(r300->vbo_offset); OUT_CS_RELOC(r300->vbo, 0, RADEON_GEM_DOMAIN_GTT, 0, 0); END_CS; @@ -575,30 +575,30 @@ void r300_emit_vertex_format_state(struct r300_context* r300) CS_LOCALS(r300); BEGIN_CS(26); - OUT_CS_REG(R300_VAP_VTX_SIZE, r300->vertex_info.vinfo.size); + OUT_CS_REG(R300_VAP_VTX_SIZE, r300->vertex_info->vinfo.size); OUT_CS_REG_SEQ(R300_VAP_VTX_STATE_CNTL, 2); - OUT_CS(r300->vertex_info.vinfo.hwfmt[0]); - OUT_CS(r300->vertex_info.vinfo.hwfmt[1]); + OUT_CS(r300->vertex_info->vinfo.hwfmt[0]); + OUT_CS(r300->vertex_info->vinfo.hwfmt[1]); OUT_CS_REG_SEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2); - OUT_CS(r300->vertex_info.vinfo.hwfmt[2]); - OUT_CS(r300->vertex_info.vinfo.hwfmt[3]); + OUT_CS(r300->vertex_info->vinfo.hwfmt[2]); + OUT_CS(r300->vertex_info->vinfo.hwfmt[3]); /* for (i = 0; i < 4; i++) { * debug_printf("hwfmt%d: 0x%08x\n", i, - * r300->vertex_info.vinfo.hwfmt[i]); + * r300->vertex_info->vinfo.hwfmt[i]); * } */ OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_0, 8); for (i = 0; i < 8; i++) { - OUT_CS(r300->vertex_info.vap_prog_stream_cntl[i]); + OUT_CS(r300->vertex_info->vap_prog_stream_cntl[i]); /* debug_printf("prog_stream_cntl%d: 0x%08x\n", i, - * r300->vertex_info.vap_prog_stream_cntl[i]); */ + * r300->vertex_info->vap_prog_stream_cntl[i]); */ } OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_EXT_0, 8); for (i = 0; i < 8; i++) { - OUT_CS(r300->vertex_info.vap_prog_stream_cntl_ext[i]); + OUT_CS(r300->vertex_info->vap_prog_stream_cntl_ext[i]); /* debug_printf("prog_stream_cntl_ext%d: 0x%08x\n", i, - * r300->vertex_info.vap_prog_stream_cntl_ext[i]); */ + * r300->vertex_info->vap_prog_stream_cntl_ext[i]); */ } END_CS; } diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c index b56f7a3d1e6..4e778e1e573 100644 --- a/src/gallium/drivers/r300/r300_render.c +++ b/src/gallium/drivers/r300/r300_render.c @@ -67,7 +67,7 @@ r300_render_get_vertex_info(struct vbuf_render* render) r300_update_derived_state(r300); - return &r300->vertex_info.vinfo; + return &r300->vertex_info->vinfo; } static boolean r300_render_allocate_vertices(struct vbuf_render* render, diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c index f0861a9cf1d..53027777d68 100644 --- a/src/gallium/drivers/r300/r300_state_derived.c +++ b/src/gallium/drivers/r300/r300_state_derived.c @@ -29,6 +29,27 @@ /* r300_state_derived: Various bits of state which are dependent upon * currently bound CSO data. */ +struct r300_shader_key { + struct r300_vertex_shader* vs; + struct r300_fragment_shader* fs; +}; + +unsigned r300_shader_key_hash(void* key) { + struct r300_shader_key* shader_key = (struct r300_shader_key*)key; + unsigned vs = (unsigned)shader_key->vs; + unsigned fs = (unsigned)shader_key->fs; + + return (vs << 16) | (fs & 0xffff); +} + +int r300_shader_key_compare(void* key1, void* key2) { + struct r300_shader_key* shader_key1 = (struct r300_shader_key*)key1; + struct r300_shader_key* shader_key2 = (struct r300_shader_key*)key2; + + return (shader_key1->vs == shader_key2->vs) && + (shader_key1->fs == shader_key2->fs); +} + /* Set up the vs_tab and routes. */ static void r300_vs_tab_routes(struct r300_context* r300, struct r300_vertex_format* vformat) @@ -247,23 +268,41 @@ static void r300_vertex_psc(struct r300_context* r300, /* Update the vertex format. */ static void r300_update_vertex_format(struct r300_context* r300) { - struct r300_vertex_format vformat; + struct r300_shader_key* key; + struct r300_vertex_format* vformat; + void* value; int i; - memset(&vformat, 0, sizeof(struct r300_vertex_format)); - for (i = 0; i < 16; i++) { - vformat.vs_tab[i] = -1; - vformat.fs_tab[i] = -1; - } + key = CALLOC_STRUCT(r300_shader_key); + key->vs = r300->vs; + key->fs = r300->fs; - r300_vs_tab_routes(r300, &vformat); + value = u_hash_table_get(r300->shader_hash_table, (void*)key); + if (value) { + debug_printf("r300: Hash table hit! vs: %p fs: %p\n", key->vs, + key->fs); + vformat = (struct r300_vertex_format*)value; + } else { + debug_printf("r300: Hash table miss... vs: %p fs: %p\n", key->vs, + key->fs); + vformat = CALLOC_STRUCT(r300_vertex_format); + + for (i = 0; i < 16; i++) { + vformat->vs_tab[i] = -1; + vformat->fs_tab[i] = -1; + } + + r300_vs_tab_routes(r300, vformat); + r300_vertex_psc(r300, vformat); - r300_vertex_psc(r300, &vformat); + if (u_hash_table_set(r300->shader_hash_table, (void*)key, + (void*)vformat) != PIPE_OK) { + debug_printf("r300: Hash table insertion error!\n"); + } + } - if (memcmp(&r300->vertex_info, &vformat, - sizeof(struct r300_vertex_format))) { - memcpy(&r300->vertex_info, &vformat, - sizeof(struct r300_vertex_format)); + if (r300->vertex_info != vformat) { + r300->vertex_info = vformat; r300->dirty_state |= R300_NEW_VERTEX_FORMAT; } } @@ -271,7 +310,7 @@ static void r300_update_vertex_format(struct r300_context* r300) /* Set up the mappings from GB to US, for RS block. */ static void r300_update_fs_tab(struct r300_context* r300) { - struct r300_vertex_format* vformat = &r300->vertex_info; + struct r300_vertex_format* vformat = r300->vertex_info; struct tgsi_shader_info* info = &r300->fs->info; int i, cols = 0, texs = 0, cols_emitted = 0; int* tab = vformat->fs_tab; @@ -337,7 +376,7 @@ static void r300_update_rs_block(struct r300_context* r300) { struct r300_rs_block* rs = r300->rs_block; struct tgsi_shader_info* info = &r300->fs->info; - int* tab = r300->vertex_info.fs_tab; + int* tab = r300->vertex_info->fs_tab; int col_count = 0, fp_offset = 0, i, tex_count = 0; int rs_tex_comp = 0; memset(rs, 0, sizeof(struct r300_rs_block)); @@ -477,10 +516,7 @@ static void r300_update_ztop(struct r300_context* r300) void r300_update_derived_state(struct r300_context* r300) { - if (r300->dirty_state & - (R300_NEW_FRAGMENT_SHADER | R300_NEW_VERTEX_SHADER)) { - r300_update_vertex_format(r300); - } + r300_update_vertex_format(r300); if (r300->dirty_state & R300_NEW_VERTEX_FORMAT) { r300_update_fs_tab(r300); diff --git a/src/gallium/drivers/r300/r300_state_derived.h b/src/gallium/drivers/r300/r300_state_derived.h index 71a4a47b003..05ad535e2de 100644 --- a/src/gallium/drivers/r300/r300_state_derived.h +++ b/src/gallium/drivers/r300/r300_state_derived.h @@ -25,6 +25,10 @@ struct r300_context; +unsigned r300_shader_key_hash(void* key); + +int r300_shader_key_compare(void* key1, void* key2); + void r300_update_derived_state(struct r300_context* r300); #endif /* R300_STATE_DERIVED_H */ -- cgit v1.2.3 From 16a06fea73b1e6e8857f7568762bfc56dcfe2940 Mon Sep 17 00:00:00 2001 From: Corbin Simpson Date: Sun, 18 Oct 2009 15:54:39 -0700 Subject: r300g: Fix up a bunch of warnings. --- src/gallium/drivers/r300/r300_emit.c | 24 +++++++++++------------- src/gallium/drivers/r300/r300_emit.h | 1 + src/gallium/drivers/r300/r300_flush.c | 2 ++ src/gallium/drivers/r300/r300_screen.c | 2 +- src/gallium/drivers/r300/r300_state.c | 4 ++-- src/gallium/drivers/r300/r300_state_derived.c | 10 ++++------ 6 files changed, 21 insertions(+), 22 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index e6092cda9bf..df2046bd0c3 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -320,8 +320,7 @@ void r300_emit_fb_state(struct r300_context* r300, END_CS; } -void r300_emit_query_start(struct r300_context *r300) - +static void r300_emit_query_start(struct r300_context *r300) { struct r300_capabilities *caps = r300_screen(r300->context.screen)->caps; struct r300_query *query = r300->query_current; @@ -334,9 +333,9 @@ void r300_emit_query_start(struct r300_context *r300) * for overlapping queries. */ BEGIN_CS(4); if (caps->family == CHIP_FAMILY_RV530) { - OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL); + OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL); } else { - OUT_CS_REG(R300_SU_REG_DEST, R300_RASTER_PIPE_SELECT_ALL); + OUT_CS_REG(R300_SU_REG_DEST, R300_RASTER_PIPE_SELECT_ALL); } OUT_CS_REG(R300_ZB_ZPASS_DATA, 0); END_CS; @@ -345,7 +344,7 @@ void r300_emit_query_start(struct r300_context *r300) static void r300_emit_query_finish(struct r300_context *r300, - struct r300_query *query) + struct r300_query *query) { struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps; CS_LOCALS(r300); @@ -388,7 +387,7 @@ static void r300_emit_query_finish(struct r300_context *r300, OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1); OUT_CS_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 0), 0, RADEON_GEM_DOMAIN_GTT, 0); - break; + break; default: debug_printf("r300: Implementation error: Chipset reports %d" " pixel pipes!\n", caps->num_frag_pipes); @@ -398,11 +397,10 @@ static void r300_emit_query_finish(struct r300_context *r300, /* And, finally, reset it to normal... */ OUT_CS_REG(R300_SU_REG_DEST, 0xF); END_CS; - } static void rv530_emit_query_single(struct r300_context *r300, - struct r300_query *query) + struct r300_query *query) { CS_LOCALS(r300); @@ -415,7 +413,7 @@ static void rv530_emit_query_single(struct r300_context *r300, } static void rv530_emit_query_double(struct r300_context *r300, - struct r300_query *query) + struct r300_query *query) { CS_LOCALS(r300); @@ -442,10 +440,10 @@ void r300_emit_query_end(struct r300_context* r300) return; if (caps->family == CHIP_FAMILY_RV530) { - if (caps->num_z_pipes == 2) - rv530_emit_query_double(r300, query); - else - rv530_emit_query_single(r300, query); + if (caps->num_z_pipes == 2) + rv530_emit_query_double(r300, query); + else + rv530_emit_query_single(r300, query); } else r300_emit_query_finish(r300, query); } diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h index b62aa9fec50..7e469ea0c77 100644 --- a/src/gallium/drivers/r300/r300_emit.h +++ b/src/gallium/drivers/r300/r300_emit.h @@ -58,6 +58,7 @@ void r300_emit_fb_state(struct r300_context* r300, void r300_emit_query_begin(struct r300_context* r300, struct r300_query* query); + void r300_emit_query_end(struct r300_context* r300); void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs); diff --git a/src/gallium/drivers/r300/r300_flush.c b/src/gallium/drivers/r300/r300_flush.c index 241ea71d6b1..d60652a0219 100644 --- a/src/gallium/drivers/r300/r300_flush.c +++ b/src/gallium/drivers/r300/r300_flush.c @@ -20,7 +20,9 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. */ +#include "r300_emit.h" #include "r300_flush.h" +#include "r300_state_invariant.h" static void r300_flush(struct pipe_context* pipe, unsigned flags, diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c index 44770d1acaa..cc499d400a4 100644 --- a/src/gallium/drivers/r300/r300_screen.c +++ b/src/gallium/drivers/r300/r300_screen.c @@ -311,7 +311,7 @@ r300_get_tex_transfer(struct pipe_screen *screen, { struct r300_texture *tex = (struct r300_texture *)texture; struct r300_transfer *trans; - unsigned offset; /* in bytes */ + unsigned offset = 0; /* in bytes */ /* XXX Add support for these things */ if (texture->target == PIPE_TEXTURE_CUBE) { diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index 8359850966b..0a982a9d5db 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -329,7 +329,7 @@ static void r300_delete_fs_state(struct pipe_context* pipe, void* shader) { struct r300_fragment_shader* fs = (struct r300_fragment_shader*)shader; rc_constants_destroy(&fs->code.constants); - FREE(fs->state.tokens); + FREE((void*)fs->state.tokens); FREE(shader); } @@ -697,7 +697,7 @@ static void r300_delete_vs_state(struct pipe_context* pipe, void* shader) rc_constants_destroy(&vs->code.constants); draw_delete_vertex_shader(r300->draw, vs->draw); - FREE(vs->state.tokens); + FREE((void*)vs->state.tokens); FREE(shader); } else { draw_delete_vertex_shader(r300->draw, diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c index 2c624766bb9..1468b9d36ed 100644 --- a/src/gallium/drivers/r300/r300_state_derived.c +++ b/src/gallium/drivers/r300/r300_state_derived.c @@ -265,7 +265,9 @@ static void r300_vertex_psc(struct r300_context* r300, } /* Set the last vector in the PSC. */ - i--; + if (i) { + i -= 1; + } vformat->vap_prog_stream_cntl[i >> 1] |= (R300_LAST_VEC << (i & 1 ? 16 : 0)); } @@ -336,17 +338,14 @@ static void r300_update_fs_tab(struct r300_context* r300, * the rasterization of vertices into fragments. This is also the part of the * chipset that locks up if any part of it is even slightly wrong. */ static void r300_update_rs_block(struct r300_context* r300, - struct r300_vertex_format* vformat, struct r300_rs_block* rs) { struct tgsi_shader_info* info = &r300->fs->info; - int* tab = vformat->fs_tab; int col_count = 0, fp_offset = 0, i, tex_count = 0; int rs_tex_comp = 0; if (r300_screen(r300->context.screen)->caps->is_r500) { for (i = 0; i < info->num_inputs; i++) { - assert(tab[i] != -1); switch (info->input_semantic_name[i]) { case TGSI_SEMANTIC_COLOR: rs->ip[col_count] |= @@ -387,7 +386,6 @@ static void r300_update_rs_block(struct r300_context* r300, } } else { for (i = 0; i < info->num_inputs; i++) { - assert(tab[i] != -1); switch (info->input_semantic_name[i]) { case TGSI_SEMANTIC_COLOR: rs->ip[col_count] |= @@ -480,7 +478,7 @@ static void r300_update_derived_shader_state(struct r300_context* r300) r300_vertex_psc(r300, vformat); r300_update_fs_tab(r300, vformat); - r300_update_rs_block(r300, vformat, rs_block); + r300_update_rs_block(r300, rs_block); value->vformat = vformat; value->rs_block = rs_block; -- cgit v1.2.3 From 6a448a525baf81173f92ee8c3074b98baa54397b Mon Sep 17 00:00:00 2001 From: Corbin Simpson Date: Wed, 21 Oct 2009 06:31:36 -0700 Subject: r300g: Cleanup header includes. --- src/gallium/drivers/r300/r300_chipset.c | 1 + src/gallium/drivers/r300/r300_chipset.h | 6 +++--- src/gallium/drivers/r300/r300_clear.c | 3 +++ src/gallium/drivers/r300/r300_clear.h | 4 +--- src/gallium/drivers/r300/r300_context.c | 16 +++++++++++++++- src/gallium/drivers/r300/r300_context.h | 13 ------------- src/gallium/drivers/r300/r300_emit.c | 7 ++++++- src/gallium/drivers/r300/r300_emit.h | 7 ------- src/gallium/drivers/r300/r300_flush.c | 7 +++++++ src/gallium/drivers/r300/r300_flush.h | 7 ------- src/gallium/drivers/r300/r300_fs.c | 6 +++++- src/gallium/drivers/r300/r300_fs.h | 7 +++---- src/gallium/drivers/r300/r300_query.c | 8 +++++++- src/gallium/drivers/r300/r300_query.h | 4 ---- src/gallium/drivers/r300/r300_render.c | 5 ++++- src/gallium/drivers/r300/r300_screen.c | 7 +++++++ src/gallium/drivers/r300/r300_screen.h | 5 ----- src/gallium/drivers/r300/r300_state.c | 4 ++++ src/gallium/drivers/r300/r300_state_derived.c | 8 +++++++- src/gallium/drivers/r300/r300_state_inlines.h | 2 ++ src/gallium/drivers/r300/r300_state_invariant.c | 5 ++++- src/gallium/drivers/r300/r300_state_invariant.h | 6 +----- src/gallium/drivers/r300/r300_texture.c | 6 ++++++ src/gallium/drivers/r300/r300_texture.h | 3 --- 24 files changed, 86 insertions(+), 61 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_chipset.c b/src/gallium/drivers/r300/r300_chipset.c index 9b763e2cb52..51fdb82ff34 100644 --- a/src/gallium/drivers/r300/r300_chipset.c +++ b/src/gallium/drivers/r300/r300_chipset.c @@ -21,6 +21,7 @@ * USE OR OTHER DEALINGS IN THE SOFTWARE. */ #include "r300_chipset.h" + #include "util/u_debug.h" /* r300_chipset: A file all to itself for deducing the various properties of diff --git a/src/gallium/drivers/r300/r300_chipset.h b/src/gallium/drivers/r300/r300_chipset.h index f015a4243d3..0633a8b8a72 100644 --- a/src/gallium/drivers/r300/r300_chipset.h +++ b/src/gallium/drivers/r300/r300_chipset.h @@ -33,11 +33,11 @@ struct r300_capabilities { /* Chipset family */ int family; /* The number of vertex floating-point units */ - int num_vert_fpus; + unsigned num_vert_fpus; /* The number of fragment pipes */ - int num_frag_pipes; + unsigned num_frag_pipes; /* The number of z pipes */ - int num_z_pipes; + unsigned num_z_pipes; /* Whether or not TCL is physically present */ boolean has_tcl; /* Whether or not this is an RV515 or newer; R500s have many differences diff --git a/src/gallium/drivers/r300/r300_clear.c b/src/gallium/drivers/r300/r300_clear.c index 8b9cb819ae6..02d6d504fc0 100644 --- a/src/gallium/drivers/r300/r300_clear.c +++ b/src/gallium/drivers/r300/r300_clear.c @@ -21,6 +21,9 @@ * USE OR OTHER DEALINGS IN THE SOFTWARE. */ #include "r300_clear.h" +#include "r300_context.h" + +#include "util/u_clear.h" /* Clears currently bound buffers. */ void r300_clear(struct pipe_context* pipe, diff --git a/src/gallium/drivers/r300/r300_clear.h b/src/gallium/drivers/r300/r300_clear.h index cd5900565e8..b8fcdf273c7 100644 --- a/src/gallium/drivers/r300/r300_clear.h +++ b/src/gallium/drivers/r300/r300_clear.h @@ -23,9 +23,7 @@ #ifndef R300_CLEAR_H #define R300_CLEAR_H -#include "util/u_clear.h" - -#include "r300_context.h" +struct pipe_context; void r300_clear(struct pipe_context* pipe, unsigned buffers, diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c index 7b370b3e955..4ba11a026ea 100644 --- a/src/gallium/drivers/r300/r300_context.c +++ b/src/gallium/drivers/r300/r300_context.c @@ -20,10 +20,24 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. */ -#include "r300_context.h" +#include "draw/draw_context.h" + +#include "pipe/p_inlines.h" + +#include "tgsi/tgsi_scan.h" +#include "util/u_hash_table.h" +#include "util/u_memory.h" +#include "util/u_simple_list.h" + +#include "r300_clear.h" +#include "r300_context.h" #include "r300_flush.h" +#include "r300_query.h" +#include "r300_screen.h" +#include "r300_state_derived.h" #include "r300_state_invariant.h" +#include "r300_winsys.h" static boolean r300_draw_range_elements(struct pipe_context* pipe, struct pipe_buffer* indexBuffer, diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index 2d608d6afc7..61a57df35e9 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -23,23 +23,10 @@ #ifndef R300_CONTEXT_H #define R300_CONTEXT_H -#include "draw/draw_context.h" #include "draw/draw_vertex.h" #include "pipe/p_context.h" -#include "tgsi/tgsi_scan.h" - -#include "util/u_hash_table.h" -#include "util/u_memory.h" -#include "util/u_simple_list.h" - -#include "r300_clear.h" -#include "r300_query.h" -#include "r300_screen.h" -#include "r300_state_derived.h" -#include "r300_winsys.h" - struct r300_fragment_shader; struct r300_vertex_shader; diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index df2046bd0c3..258c38fefdc 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -22,10 +22,15 @@ /* r300_emit: Functions for emitting state. */ -#include "r300_emit.h" +#include "util/u_math.h" +#include "r300_context.h" +#include "r300_cs.h" +#include "r300_emit.h" #include "r300_fs.h" +#include "r300_screen.h" #include "r300_state_derived.h" +#include "r300_state_inlines.h" #include "r300_vs.h" void r300_emit_blend_state(struct r300_context* r300, diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h index 7e469ea0c77..02ac5bebbd9 100644 --- a/src/gallium/drivers/r300/r300_emit.h +++ b/src/gallium/drivers/r300/r300_emit.h @@ -23,13 +23,6 @@ #ifndef R300_EMIT_H #define R300_EMIT_H -#include "util/u_math.h" - -#include "r300_context.h" -#include "r300_cs.h" -#include "r300_screen.h" -#include "r300_state_inlines.h" - struct rX00_fragment_program_code; struct r300_vertex_program_code; diff --git a/src/gallium/drivers/r300/r300_flush.c b/src/gallium/drivers/r300/r300_flush.c index d60652a0219..c222ea09b17 100644 --- a/src/gallium/drivers/r300/r300_flush.c +++ b/src/gallium/drivers/r300/r300_flush.c @@ -20,6 +20,13 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. */ +#include "draw/draw_context.h" +#include "draw/draw_private.h" + +#include "util/u_simple_list.h" + +#include "r300_context.h" +#include "r300_cs.h" #include "r300_emit.h" #include "r300_flush.h" #include "r300_state_invariant.h" diff --git a/src/gallium/drivers/r300/r300_flush.h b/src/gallium/drivers/r300/r300_flush.h index 9a83d89daab..0e9e6106bb7 100644 --- a/src/gallium/drivers/r300/r300_flush.h +++ b/src/gallium/drivers/r300/r300_flush.h @@ -23,13 +23,6 @@ #ifndef R300_FLUSH_H #define R300_FLUSH_H -#include "draw/draw_private.h" - -#include "pipe/p_context.h" - -#include "r300_context.h" -#include "r300_cs.h" - void r300_init_flush_functions(struct r300_context* r300); #endif /* R300_FLUSH_H */ diff --git a/src/gallium/drivers/r300/r300_fs.c b/src/gallium/drivers/r300/r300_fs.c index 546ad545a53..2db185fd80e 100644 --- a/src/gallium/drivers/r300/r300_fs.c +++ b/src/gallium/drivers/r300/r300_fs.c @@ -21,10 +21,14 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. */ -#include "r300_fs.h" +#include "tgsi/tgsi_dump.h" +#include "r300_context.h" +#include "r300_screen.h" +#include "r300_fs.h" #include "r300_tgsi_to_rc.h" +#include "radeon_code.h" #include "radeon_compiler.h" static void find_output_registers(struct r300_fragment_program_compiler * compiler, diff --git a/src/gallium/drivers/r300/r300_fs.h b/src/gallium/drivers/r300/r300_fs.h index 04453274aa3..8dfb1397387 100644 --- a/src/gallium/drivers/r300/r300_fs.h +++ b/src/gallium/drivers/r300/r300_fs.h @@ -24,14 +24,13 @@ #ifndef R300_FS_H #define R300_FS_H -#include "tgsi/tgsi_dump.h" +#include "pipe/p_state.h" + +#include "tgsi/tgsi_scan.h" -#include "r300_context.h" #include "r3xx_fs.h" #include "r5xx_fs.h" -#include "radeon_code.h" - struct r300_fragment_shader { /* Parent class */ struct pipe_shader_state state; diff --git a/src/gallium/drivers/r300/r300_query.c b/src/gallium/drivers/r300/r300_query.c index 2b0fbfb7d22..007f11efae6 100644 --- a/src/gallium/drivers/r300/r300_query.c +++ b/src/gallium/drivers/r300/r300_query.c @@ -20,9 +20,15 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. */ -#include "r300_query.h" +#include "util/u_memory.h" +#include "util/u_simple_list.h" +#include "r300_context.h" +#include "r300_screen.h" +#include "r300_cs.h" #include "r300_emit.h" +#include "r300_query.h" +#include "r300_reg.h" static struct pipe_query *r300_create_query(struct pipe_context *pipe, unsigned query_type) diff --git a/src/gallium/drivers/r300/r300_query.h b/src/gallium/drivers/r300/r300_query.h index 4f50e8f8440..48876da3123 100644 --- a/src/gallium/drivers/r300/r300_query.h +++ b/src/gallium/drivers/r300/r300_query.h @@ -23,10 +23,6 @@ #ifndef R300_QUERY_H #define R300_QUERY_H -#include "r300_context.h" -#include "r300_cs.h" -#include "r300_reg.h" - struct r300_context; static INLINE struct r300_query* r300_query(struct pipe_query* q) diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c index 4e778e1e573..79a33b53cb1 100644 --- a/src/gallium/drivers/r300/r300_render.c +++ b/src/gallium/drivers/r300/r300_render.c @@ -20,8 +20,11 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. */ -#include "draw/draw_pipe.h" +#include "draw/draw_context.h" #include "draw/draw_vbuf.h" + +#include "pipe/p_inlines.h" + #include "util/u_memory.h" #include "r300_cs.h" diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c index cc499d400a4..a9058a25e5c 100644 --- a/src/gallium/drivers/r300/r300_screen.c +++ b/src/gallium/drivers/r300/r300_screen.c @@ -20,7 +20,14 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. */ +#include "pipe/p_inlines.h" +#include "util/u_memory.h" +#include "util/u_simple_screen.h" + +#include "r300_context.h" #include "r300_screen.h" +#include "r300_texture.h" +#include "r300_winsys.h" /* Return the identifier behind whom the brave coders responsible for this * amalgamation of code, sweat, and duct tape, routinely obscure their names. diff --git a/src/gallium/drivers/r300/r300_screen.h b/src/gallium/drivers/r300/r300_screen.h index 2a0e41fbc3b..41df31f670f 100644 --- a/src/gallium/drivers/r300/r300_screen.h +++ b/src/gallium/drivers/r300/r300_screen.h @@ -23,14 +23,9 @@ #ifndef R300_SCREEN_H #define R300_SCREEN_H -#include "pipe/p_inlines.h" #include "pipe/p_screen.h" -#include "util/u_memory.h" -#include "util/u_simple_screen.h" #include "r300_chipset.h" -#include "r300_texture.h" -#include "r300_winsys.h" struct r300_screen { /* Parent class */ diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index 0a982a9d5db..c4c9e5d7c25 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -20,8 +20,11 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. */ +#include "draw/draw_context.h" + #include "util/u_debug.h" #include "util/u_math.h" +#include "util/u_memory.h" #include "util/u_pack_color.h" #include "tgsi/tgsi_parse.h" @@ -31,6 +34,7 @@ #include "r300_context.h" #include "r300_reg.h" +#include "r300_screen.h" #include "r300_state_inlines.h" #include "r300_fs.h" #include "r300_vs.h" diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c index 6e7679c39fe..e749bd9cb94 100644 --- a/src/gallium/drivers/r300/r300_state_derived.c +++ b/src/gallium/drivers/r300/r300_state_derived.c @@ -20,9 +20,15 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. */ -#include "r300_state_derived.h" +#include "draw/draw_context.h" + +#include "util/u_math.h" +#include "util/u_memory.h" +#include "r300_context.h" #include "r300_fs.h" +#include "r300_screen.h" +#include "r300_state_derived.h" #include "r300_state_inlines.h" #include "r300_vs.h" diff --git a/src/gallium/drivers/r300/r300_state_inlines.h b/src/gallium/drivers/r300/r300_state_inlines.h index d7b57e1b227..9135c55ec44 100644 --- a/src/gallium/drivers/r300/r300_state_inlines.h +++ b/src/gallium/drivers/r300/r300_state_inlines.h @@ -24,6 +24,8 @@ #ifndef R300_STATE_INLINES_H #define R300_STATE_INLINES_H +#include "draw/draw_vertex.h" + #include "pipe/p_format.h" #include "r300_reg.h" diff --git a/src/gallium/drivers/r300/r300_state_invariant.c b/src/gallium/drivers/r300/r300_state_invariant.c index 3865730d635..4865f160581 100644 --- a/src/gallium/drivers/r300/r300_state_invariant.c +++ b/src/gallium/drivers/r300/r300_state_invariant.c @@ -21,9 +21,12 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. */ +#include "r300_context.h" +#include "r300_cs.h" +#include "r300_reg.h" +#include "r300_screen.h" #include "r300_state_invariant.h" - struct pipe_viewport_state r300_viewport_identity = { .scale = {1.0, 1.0, 1.0, 1.0}, .translate = {0.0, 0.0, 0.0, 0.0}, diff --git a/src/gallium/drivers/r300/r300_state_invariant.h b/src/gallium/drivers/r300/r300_state_invariant.h index 5bea6779fe5..05cff0d6dfe 100644 --- a/src/gallium/drivers/r300/r300_state_invariant.h +++ b/src/gallium/drivers/r300/r300_state_invariant.h @@ -23,11 +23,7 @@ #ifndef R300_STATE_INVARIANT_H #define R300_STATE_INVARIANT_H -#include "r300_chipset.h" -#include "r300_context.h" -#include "r300_cs.h" -#include "r300_reg.h" -#include "r300_state_inlines.h" +struct r300_context; void r300_emit_invariant_state(struct r300_context* r300); diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c index 7ea4c33fa9a..339fbb6242f 100644 --- a/src/gallium/drivers/r300/r300_texture.c +++ b/src/gallium/drivers/r300/r300_texture.c @@ -20,6 +20,12 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. */ +#include "pipe/p_screen.h" + +#include "util/u_math.h" +#include "util/u_memory.h" + +#include "r300_context.h" #include "r300_texture.h" static void r300_setup_texture_state(struct r300_texture* tex) diff --git a/src/gallium/drivers/r300/r300_texture.h b/src/gallium/drivers/r300/r300_texture.h index 992dad77ab4..2e58bda7168 100644 --- a/src/gallium/drivers/r300/r300_texture.h +++ b/src/gallium/drivers/r300/r300_texture.h @@ -23,11 +23,8 @@ #ifndef R300_TEXTURE_H #define R300_TEXTURE_H -#include "pipe/p_screen.h" #include "pipe/p_video_state.h" -#include "util/u_math.h" -#include "r300_context.h" #include "r300_reg.h" struct r300_texture; -- cgit v1.2.3 From 0e44884aada4e4bd6384245d9ae065da5aca7f3d Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Wed, 28 Oct 2009 02:19:52 +0100 Subject: r300g: fix blending and do some optimizations Signed-off-by: Corbin Simpson --- src/gallium/drivers/r300/r300_emit.c | 1 + src/gallium/drivers/r300/r300_state.c | 101 ++++++++++++++++++---------------- 2 files changed, 54 insertions(+), 48 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 258c38fefdc..096165a2926 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -31,6 +31,7 @@ #include "r300_screen.h" #include "r300_state_derived.h" #include "r300_state_inlines.h" +#include "r300_texture.h" #include "r300_vs.h" void r300_emit_blend_state(struct r300_context* r300, diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index a3e1bc621a5..5d28837ef7e 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -49,46 +49,47 @@ static void* r300_create_blend_state(struct pipe_context* pipe, { struct r300_blend_state* blend = CALLOC_STRUCT(r300_blend_state); + if (state->blend_enable) { - unsigned eqRGB = state->rgb_func; - unsigned srcRGB = state->rgb_src_factor; - unsigned dstRGB = state->rgb_dst_factor; - - unsigned eqA = state->alpha_func; - unsigned srcA = state->alpha_src_factor; - unsigned dstA = state->alpha_dst_factor; - - if (srcA != srcRGB || - dstA != dstRGB || - eqA != eqRGB) { - blend->alpha_blend_control = - r300_translate_blend_function(eqA) | - (r300_translate_blend_factor(srcA) << - R300_SRC_BLEND_SHIFT) | - (r300_translate_blend_factor(dstA) << - R300_DST_BLEND_SHIFT); - blend->blend_control |= R300_ALPHA_BLEND_ENABLE | - R300_SEPARATE_ALPHA_ENABLE; - } else { - blend->alpha_blend_control = R300_COMB_FCN_ADD_CLAMP | - (R300_BLEND_GL_ONE << R300_SRC_BLEND_SHIFT) | - (R300_BLEND_GL_ZERO << R300_DST_BLEND_SHIFT); - } - } - if (state->blend_enable) { - /* XXX for now, always do separate alpha... - * is it faster to do it with one reg? */ - blend->blend_control |= R300_READ_ENABLE | - r300_translate_blend_function(state->rgb_func) | - (r300_translate_blend_factor(state->rgb_src_factor) << - R300_SRC_BLEND_SHIFT) | - (r300_translate_blend_factor(state->rgb_dst_factor) << - R300_DST_BLEND_SHIFT); - } else { - blend->blend_control = - R300_COMB_FCN_ADD_CLAMP | - (R300_BLEND_GL_ONE << R300_SRC_BLEND_SHIFT) | - (R300_BLEND_GL_ZERO << R300_DST_BLEND_SHIFT); + unsigned eqRGB = state->rgb_func; + unsigned srcRGB = state->rgb_src_factor; + unsigned dstRGB = state->rgb_dst_factor; + + unsigned eqA = state->alpha_func; + unsigned srcA = state->alpha_src_factor; + unsigned dstA = state->alpha_dst_factor; + + /* despite the name, ALPHA_BLEND_ENABLE has nothing to do with alpha, + * this is just the crappy D3D naming */ + blend->blend_control = R300_ALPHA_BLEND_ENABLE | + r300_translate_blend_function(eqRGB) | + ( r300_translate_blend_factor(srcRGB) << R300_SRC_BLEND_SHIFT) | + ( r300_translate_blend_factor(dstRGB) << R300_DST_BLEND_SHIFT); + + /* optimization: some operations do not require the destination color */ + if (eqRGB == PIPE_BLEND_MIN || eqA == PIPE_BLEND_MIN || + eqRGB == PIPE_BLEND_MAX || eqA == PIPE_BLEND_MAX || + dstRGB != PIPE_BLENDFACTOR_ZERO || + dstA != PIPE_BLENDFACTOR_ZERO || + srcRGB == PIPE_BLENDFACTOR_DST_COLOR || + srcRGB == PIPE_BLENDFACTOR_DST_ALPHA || + srcRGB == PIPE_BLENDFACTOR_INV_DST_COLOR || + srcRGB == PIPE_BLENDFACTOR_INV_DST_ALPHA || + srcA == PIPE_BLENDFACTOR_DST_ALPHA || + srcA == PIPE_BLENDFACTOR_INV_DST_ALPHA) + blend->blend_control |= R300_READ_ENABLE; + + /* XXX implement the optimization with DISCARD_SRC_PIXELS*/ + /* XXX implement the optimization with SRC_ALPHA_?_NO_READ */ + + /* separate alpha */ + if (srcA != srcRGB || dstA != dstRGB || eqA != eqRGB) { + blend->blend_control |= R300_SEPARATE_ALPHA_ENABLE; + blend->alpha_blend_control = + r300_translate_blend_function(eqA) | + (r300_translate_blend_factor(srcA) << R300_SRC_BLEND_SHIFT) | + (r300_translate_blend_factor(dstA) << R300_DST_BLEND_SHIFT); + } } /* PIPE_LOGICOP_* don't need to be translated, fortunately. */ @@ -122,25 +123,29 @@ static void r300_delete_blend_state(struct pipe_context* pipe, FREE(state); } +/* Convert float to 10bit integer */ +static unsigned float_to_fixed10(float f) +{ + return CLAMP((unsigned)(f * 1023.9f), 0, 1023); +} + /* Set blend color. * Setup both R300 and R500 registers, figure out later which one to write. */ static void r300_set_blend_color(struct pipe_context* pipe, const struct pipe_blend_color* color) { struct r300_context* r300 = r300_context(pipe); - ubyte ur, ug, ub, ua; - - ur = float_to_ubyte(color->color[0]); - ug = float_to_ubyte(color->color[1]); - ub = float_to_ubyte(color->color[2]); - ua = float_to_ubyte(color->color[3]); util_pack_color(color->color, PIPE_FORMAT_A8R8G8B8_UNORM, &r300->blend_color_state->blend_color); - /* XXX this is wrong */ - r300->blend_color_state->blend_color_red_alpha = ur | (ua << 16); - r300->blend_color_state->blend_color_green_blue = ub | (ug << 16); + /* XXX if FP16 blending is enabled, we should use the FP16 format */ + r300->blend_color_state->blend_color_red_alpha = + float_to_fixed10(color->color[0]) | + (float_to_fixed10(color->color[3]) << 16); + r300->blend_color_state->blend_color_green_blue = + float_to_fixed10(color->color[2]) | + (float_to_fixed10(color->color[1]) << 16); r300->dirty_state |= R300_NEW_BLEND_COLOR; } -- cgit v1.2.3 From a1d726aae8fcacfa1eb1d76ce9c46adaafeaf4a4 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Wed, 28 Oct 2009 02:21:49 +0100 Subject: r300g: fix the WRAP_T mode when using 1D textures Signed-off-by: Corbin Simpson --- src/gallium/drivers/r300/r300_emit.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 096165a2926..8bfa2932c98 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -532,10 +532,17 @@ void r300_emit_texture(struct r300_context* r300, struct r300_texture* tex, unsigned offset) { + uint32_t filter0 = sampler->filter0; CS_LOCALS(r300); + /* to emulate 1D textures through 2D ones correctly */ + if (tex->tex.height[0] == 1) { + filter0 &= ~R300_TX_WRAP_T_MASK; + filter0 |= R300_TX_WRAP_T(R300_TX_CLAMP_TO_EDGE); + } + BEGIN_CS(16); - OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), sampler->filter0 | + OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), filter0 | (offset << 28)); OUT_CS_REG(R300_TX_FILTER1_0 + (offset * 4), sampler->filter1); OUT_CS_REG(R300_TX_BORDER_COLOR_0 + (offset * 4), sampler->border_color); -- cgit v1.2.3 From bcfde429139476c2d04baddaf671651cfc860145 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Wed, 28 Oct 2009 02:43:51 +0100 Subject: r300g: fix emitting the stencil-ref and alpha-ref values Signed-off-by: Corbin Simpson DSA really needs its head examined someday. ~ C. --- src/gallium/drivers/r300/r300_emit.c | 16 ++++++++++------ src/gallium/drivers/r300/r300_reg.h | 2 ++ src/gallium/drivers/r300/r300_state.c | 24 +++++++++++++++++++----- 3 files changed, 31 insertions(+), 11 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 8bfa2932c98..2a8e4a9f410 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -102,19 +102,23 @@ void r300_emit_dsa_state(struct r300_context* r300, struct r300_screen* r300screen = r300_screen(r300->context.screen); CS_LOCALS(r300); - BEGIN_CS(r300screen->caps->is_r500 ? 8 : 8); + BEGIN_CS(r300screen->caps->is_r500 ? 10 : 8); OUT_CS_REG(R300_FG_ALPHA_FUNC, dsa->alpha_function); - /* XXX figure out the r300 counterpart for this */ - if (r300screen->caps->is_r500) { - /* OUT_CS_REG(R500_FG_ALPHA_VALUE, dsa->alpha_reference); */ - } + + /* not needed since we use the 8bit alpha ref */ + /*if (r300screen->caps->is_r500) { + OUT_CS_REG(R500_FG_ALPHA_VALUE, dsa->alpha_reference); + }*/ + OUT_CS_REG_SEQ(R300_ZB_CNTL, 3); OUT_CS(dsa->z_buffer_control); OUT_CS(dsa->z_stencil_control); OUT_CS(dsa->stencil_ref_mask); OUT_CS_REG(R300_ZB_ZTOP, r300->ztop_state.z_buffer_top); + + /* XXX it seems r3xx doesn't support STENCILREFMASK_BF */ if (r300screen->caps->is_r500) { - /* OUT_CS_REG(R500_ZB_STENCILREFMASK_BF, dsa->stencil_ref_bf); */ + OUT_CS_REG(R500_ZB_STENCILREFMASK_BF, dsa->stencil_ref_bf); } END_CS; } diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h index e920b2a5e77..babc3c709e0 100644 --- a/src/gallium/drivers/r300/r300_reg.h +++ b/src/gallium/drivers/r300/r300_reg.h @@ -2416,6 +2416,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_Z_WRITE_ENABLE (1 << 2) # define R300_Z_SIGNED_COMPARE (1 << 3) # define R300_STENCIL_FRONT_BACK (1 << 4) +# define R500_STENCIL_ZSIGNED_MAGNITUDE (1 << 5) +# define R500_STENCIL_REFMASK_FRONT_BACK (1 << 6) #define R300_ZB_ZSTENCILCNTL 0x4f04 /* functions */ diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index 5d28837ef7e..5db8c69dec9 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -198,6 +198,8 @@ static void* r300_create_dsa_state(struct pipe_context* pipe, const struct pipe_depth_stencil_alpha_state* state) { + struct r300_capabilities *caps = + r300_screen(r300_context(pipe)->context.screen)->caps; struct r300_dsa_state* dsa = CALLOC_STRUCT(r300_dsa_state); /* Depth test setup. */ @@ -242,9 +244,16 @@ static void* (r300_translate_stencil_op(state->stencil[1].zfail_op) << R300_S_BACK_ZFAIL_OP_SHIFT); - dsa->stencil_ref_bf = (state->stencil[1].ref_value) | - (state->stencil[1].valuemask << R300_STENCILMASK_SHIFT) | - (state->stencil[1].writemask << R300_STENCILWRITEMASK_SHIFT); + /* XXX it seems r3xx doesn't support STENCILREFMASK_BF */ + if (caps->is_r500) + { + dsa->z_buffer_control |= R500_STENCIL_REFMASK_FRONT_BACK; + dsa->stencil_ref_bf = (state->stencil[1].ref_value) | + (state->stencil[1].valuemask << + R300_STENCILMASK_SHIFT) | + (state->stencil[1].writemask << + R300_STENCILWRITEMASK_SHIFT); + } } } @@ -253,8 +262,13 @@ static void* dsa->alpha_function = r300_translate_alpha_function(state->alpha.func) | R300_FG_ALPHA_FUNC_ENABLE; - dsa->alpha_reference = CLAMP(state->alpha.ref_value * 1023.0f, - 0, 1023); + + /* XXX figure out why emitting 10bit alpha ref causes CS to dump */ + /* always use 8bit alpha ref */ + dsa->alpha_function |= float_to_ubyte(state->alpha.ref_value); + + if (caps->is_r500) + dsa->alpha_function |= R500_FG_ALPHA_FUNC_8BIT; } return (void*)dsa; -- cgit v1.2.3 From ace78d90ded52d8fe4b3b077abf9a4db381dce16 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Thu, 29 Oct 2009 23:48:59 +0100 Subject: r300g: fix crash in r300_is_texture_referenced Also, a subtle fix in emitting a texture state --- src/gallium/drivers/r300/r300_context.c | 2 +- src/gallium/drivers/r300/r300_emit.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c index e45564b54e8..02f201b49ad 100644 --- a/src/gallium/drivers/r300/r300_context.c +++ b/src/gallium/drivers/r300/r300_context.c @@ -80,7 +80,7 @@ r300_is_texture_referenced(struct pipe_context *pipe, struct pipe_texture *texture, unsigned face, unsigned level) { - struct pipe_buffer* buf; + struct pipe_buffer* buf = 0; r300_get_texture_buffer(texture, &buf, NULL); diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 2a8e4a9f410..3b0b41e486f 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -540,7 +540,7 @@ void r300_emit_texture(struct r300_context* r300, CS_LOCALS(r300); /* to emulate 1D textures through 2D ones correctly */ - if (tex->tex.height[0] == 1) { + if (tex->tex.target == PIPE_TEXTURE_1D) { filter0 &= ~R300_TX_WRAP_T_MASK; filter0 |= R300_TX_WRAP_T(R300_TX_CLAMP_TO_EDGE); } -- cgit v1.2.3 From 63c9450ae776ff4207422442dd8c3d9d13a05e7a Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Fri, 30 Oct 2009 18:19:25 +0100 Subject: r300g: add precalculating of pixel pitch, add a new NPOT flag --- src/gallium/drivers/r300/r300_context.h | 8 ++++++++ src/gallium/drivers/r300/r300_emit.c | 16 ++++++++------- src/gallium/drivers/r300/r300_texture.c | 35 +++++++++++++++++++++++---------- 3 files changed, 42 insertions(+), 17 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index 4d73567bbea..cee0734d213 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -181,6 +181,9 @@ struct r300_texture { /* Offsets into the buffer. */ unsigned offset[PIPE_MAX_TEXTURE_LEVELS]; + /* A pitch for each mip-level */ + unsigned pitch[PIPE_MAX_TEXTURE_LEVELS]; + /* Size of one zslice or face based on the texture target */ unsigned layer_size[PIPE_MAX_TEXTURE_LEVELS]; @@ -197,6 +200,11 @@ struct r300_texture { /* Total size of this texture, in bytes. */ unsigned size; + /* Whether this texture has non-power-of-two dimensions. + * It can be either a regular texture or a rectangle one. + */ + boolean is_npot; + /* Pipe buffer backing this texture. */ struct pipe_buffer* buffer; diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 3b0b41e486f..be38fbc619e 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -285,21 +285,22 @@ void r300_emit_fb_state(struct r300_context* r300, struct pipe_framebuffer_state* fb) { struct r300_texture* tex; - unsigned pixpitch; + struct pipe_surface* surf; int i; CS_LOCALS(r300); BEGIN_CS((10 * fb->nr_cbufs) + (fb->zsbuf ? 10 : 0) + 4); for (i = 0; i < fb->nr_cbufs; i++) { - tex = (struct r300_texture*)fb->cbufs[i]->texture; + surf = fb->cbufs[i]; + tex = (struct r300_texture*)surf->texture; assert(tex && tex->buffer && "cbuf is marked, but NULL!"); - pixpitch = r300_texture_get_stride(tex, 0) / tex->tex.block.size; + /* XXX I still need to figure out how to set the mipmap level here */ OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0 + (4 * i), 1); OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); OUT_CS_REG_SEQ(R300_RB3D_COLORPITCH0 + (4 * i), 1); - OUT_CS_RELOC(tex->buffer, pixpitch | + OUT_CS_RELOC(tex->buffer, tex->pitch[surf->level] | r300_translate_colorformat(tex->tex.format), 0, RADEON_GEM_DOMAIN_VRAM, 0); @@ -308,9 +309,9 @@ void r300_emit_fb_state(struct r300_context* r300, } if (fb->zsbuf) { - tex = (struct r300_texture*)fb->zsbuf->texture; + surf = fb->zsbuf; + tex = (struct r300_texture*)surf->texture; assert(tex && tex->buffer && "zsbuf is marked, but NULL!"); - pixpitch = r300_texture_get_stride(tex, 0) / tex->tex.block.size; OUT_CS_REG_SEQ(R300_ZB_DEPTHOFFSET, 1); OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); @@ -318,7 +319,8 @@ void r300_emit_fb_state(struct r300_context* r300, OUT_CS_REG(R300_ZB_FORMAT, r300_translate_zsformat(tex->tex.format)); OUT_CS_REG_SEQ(R300_ZB_DEPTHPITCH, 1); - OUT_CS_RELOC(tex->buffer, pixpitch, 0, RADEON_GEM_DOMAIN_VRAM, 0); + OUT_CS_RELOC(tex->buffer, tex->pitch[surf->level], 0, + RADEON_GEM_DOMAIN_VRAM, 0); } OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c index 3e90fea6c8a..7199918a849 100644 --- a/src/gallium/drivers/r300/r300_texture.c +++ b/src/gallium/drivers/r300/r300_texture.c @@ -33,21 +33,15 @@ static void r300_setup_texture_state(struct r300_texture* tex, boolean is_r500) { struct r300_texture_state* state = &tex->state; struct pipe_texture *pt = &tex->tex; - unsigned stride; state->format0 = R300_TX_WIDTH((pt->width[0] - 1) & 0x7ff) | R300_TX_HEIGHT((pt->height[0] - 1) & 0x7ff); - if (!util_is_power_of_two(pt->width[0]) || - !util_is_power_of_two(pt->height[0])) { - + if (tex->is_npot) { /* rectangles love this */ state->format0 |= R300_TX_PITCH_EN; - - stride = r300_texture_get_stride(tex, 0) / pt->block.size; - state->format2 = (stride - 1) & 0x1fff; - } - else { + state->format2 = (tex->pitch[0] - 1) & 0x1fff; + } else { /* power of two textures (3D, mipmaps, and no pitch) */ state->format0 |= R300_TX_DEPTH(util_logbase2(pt->depth[0]) & 0xf) | R300_TX_NUM_LEVELS(pt->last_level & 0xf); @@ -143,6 +137,12 @@ static void r300_setup_miptree(struct r300_texture* tex) tex->size = tex->offset[i] + size; tex->layer_size[i] = layer_size; + if (tex->is_npot) { + tex->pitch[i] = stride / base->block.size; + } else { + tex->pitch[i] = base->width[i]; + } + debug_printf("r300: Texture miptree: Level %d " "(%dx%dx%d px, pitch %d bytes)\n", i, base->width[i], base->height[i], base->depth[i], @@ -150,6 +150,12 @@ static void r300_setup_miptree(struct r300_texture* tex) } } +static void r300_setup_flags(struct r300_texture* tex) +{ + tex->is_npot = !util_is_power_of_two(tex->tex.width[0]) || + !util_is_power_of_two(tex->tex.height[0]); +} + /* Create a new texture. */ static struct pipe_texture* r300_texture_create(struct pipe_screen* screen, @@ -165,8 +171,8 @@ static struct pipe_texture* pipe_reference_init(&tex->tex.reference, 1); tex->tex.screen = screen; + r300_setup_flags(tex); r300_setup_miptree(tex); - r300_setup_texture_state(tex, r300_screen(screen)->caps->is_r500); tex->buffer = screen->buffer_create(screen, 1024, @@ -234,6 +240,13 @@ static struct pipe_texture* { struct r300_texture* tex; + /* Support only 2D textures without mipmaps */ + if (base->target != PIPE_TEXTURE_2D || + base->depth[0] != 1 || + base->last_level != 0) { + return NULL; + } + tex = CALLOC_STRUCT(r300_texture); if (!tex) { return NULL; @@ -244,7 +257,9 @@ static struct pipe_texture* tex->tex.screen = screen; tex->stride_override = *stride; + tex->pitch[0] = *stride / base->block.size; + r300_setup_flags(tex); r300_setup_texture_state(tex, r300_screen(screen)->caps->is_r500); pipe_buffer_reference(&tex->buffer, buffer); -- cgit v1.2.3 From c9928ac3ee5dc0d10127388f9312779a6c59da7c Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Sat, 31 Oct 2009 07:23:00 +0100 Subject: r300g: correct the pitch calculation for smaller mipmaps --- src/gallium/drivers/r300/r300_emit.c | 2 +- src/gallium/drivers/r300/r300_texture.c | 7 +------ 2 files changed, 2 insertions(+), 7 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index be38fbc619e..22cf9cac2a3 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -305,7 +305,7 @@ void r300_emit_fb_state(struct r300_context* r300, RADEON_GEM_DOMAIN_VRAM, 0); OUT_CS_REG(R300_US_OUT_FMT_0 + (4 * i), - r300_translate_out_fmt(fb->cbufs[i]->format)); + r300_translate_out_fmt(surf->format)); } if (fb->zsbuf) { diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c index 7199918a849..aea25cf71dd 100644 --- a/src/gallium/drivers/r300/r300_texture.c +++ b/src/gallium/drivers/r300/r300_texture.c @@ -136,12 +136,7 @@ static void r300_setup_miptree(struct r300_texture* tex) tex->offset[i] = align(tex->size, 32); tex->size = tex->offset[i] + size; tex->layer_size[i] = layer_size; - - if (tex->is_npot) { - tex->pitch[i] = stride / base->block.size; - } else { - tex->pitch[i] = base->width[i]; - } + tex->pitch[i] = stride / base->block.size; debug_printf("r300: Texture miptree: Level %d " "(%dx%dx%d px, pitch %d bytes)\n", -- cgit v1.2.3 From 2db46af8758bf77a2748460f617d0ead5b08a454 Mon Sep 17 00:00:00 2001 From: Maciej Cencora Date: Wed, 21 Oct 2009 21:17:43 +0200 Subject: r300g: split constant buffer and shader emittion --- src/gallium/drivers/r300/r300_context.c | 2 - src/gallium/drivers/r300/r300_context.h | 16 ++-- src/gallium/drivers/r300/r300_emit.c | 157 ++++++++++++++++++++------------ src/gallium/drivers/r300/r300_emit.h | 21 +++-- src/gallium/drivers/r300/r300_state.c | 54 +++++------ 5 files changed, 152 insertions(+), 98 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c index 02f201b49ad..f974147ea43 100644 --- a/src/gallium/drivers/r300/r300_context.c +++ b/src/gallium/drivers/r300/r300_context.c @@ -22,8 +22,6 @@ #include "draw/draw_context.h" -#include "pipe/p_inlines.h" - #include "tgsi/tgsi_scan.h" #include "util/u_hash_table.h" diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index cee0734d213..b1738452de0 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -26,6 +26,7 @@ #include "draw/draw_vertex.h" #include "pipe/p_context.h" +#include "pipe/p_inlines.h" struct r300_fragment_shader; struct r300_vertex_shader; @@ -119,10 +120,10 @@ struct r300_ztop_state { #define R300_NEW_BLEND 0x00000001 #define R300_NEW_BLEND_COLOR 0x00000002 #define R300_NEW_CLIP 0x00000004 -#define R300_NEW_CONSTANTS 0x00000008 -#define R300_NEW_DSA 0x00000010 -#define R300_NEW_FRAMEBUFFERS 0x00000020 -#define R300_NEW_FRAGMENT_SHADER 0x00000040 +#define R300_NEW_DSA 0x00000008 +#define R300_NEW_FRAMEBUFFERS 0x00000010 +#define R300_NEW_FRAGMENT_SHADER 0x00000020 +#define R300_NEW_FRAGMENT_SHADER_CONSTANTS 0x00000040 #define R300_NEW_RASTERIZER 0x00000080 #define R300_NEW_RS_BLOCK 0x00000100 #define R300_NEW_SAMPLER 0x00000200 @@ -132,9 +133,10 @@ struct r300_ztop_state { #define R300_ANY_NEW_TEXTURES 0x03fc0000 #define R300_NEW_VERTEX_FORMAT 0x04000000 #define R300_NEW_VERTEX_SHADER 0x08000000 -#define R300_NEW_VIEWPORT 0x10000000 -#define R300_NEW_QUERY 0x20000000 -#define R300_NEW_KITCHEN_SINK 0x3fffffff +#define R300_NEW_VERTEX_SHADER_CONSTANTS 0x10000000 +#define R300_NEW_VIEWPORT 0x20000000 +#define R300_NEW_QUERY 0x40000000 +#define R300_NEW_KITCHEN_SINK 0x7fffffff /* The next several objects are not pure Radeon state; they inherit from * various Gallium classes. */ diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 22cf9cac2a3..de27f0939b0 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -178,18 +178,15 @@ static uint32_t pack_float24(float f) } void r300_emit_fragment_program_code(struct r300_context* r300, - struct rX00_fragment_program_code* generic_code, - struct r300_constant_buffer* externals) + struct rX00_fragment_program_code* generic_code) { struct r300_fragment_program_code * code = &generic_code->code.r300; - struct rc_constant_list * constants = &generic_code->constants; int i; CS_LOCALS(r300); BEGIN_CS(15 + code->alu.length * 4 + - (code->tex.length ? (1 + code->tex.length) : 0) + - (constants->Count ? (1 + constants->Count * 4) : 0)); + (code->tex.length ? (1 + code->tex.length) : 0)); OUT_CS_REG(R300_US_CONFIG, code->config); OUT_CS_REG(R300_US_PIXSIZE, code->pixsize); @@ -221,32 +218,41 @@ void r300_emit_fragment_program_code(struct r300_context* r300, OUT_CS(code->tex.inst[i]); } - if (constants->Count) { - OUT_CS_REG_SEQ(R300_PFS_PARAM_0_X, constants->Count * 4); - for(i = 0; i < constants->Count; ++i) { - const float * data = get_shader_constant(r300, &constants->Constants[i], externals); - OUT_CS(pack_float24(data[0])); - OUT_CS(pack_float24(data[1])); - OUT_CS(pack_float24(data[2])); - OUT_CS(pack_float24(data[3])); - } - } + END_CS; +} + +void r300_emit_fs_constant_buffer(struct r300_context* r300, + struct rc_constant_list* constants) +{ + int i; + CS_LOCALS(r300); + + if (constants->Count == 0) + return; + BEGIN_CS(constants->Count * 4 + 1); + OUT_CS_REG_SEQ(R300_PFS_PARAM_0_X, constants->Count * 4); + for(i = 0; i < constants->Count; ++i) { + const float * data = get_shader_constant(r300, + &constants->Constants[i], + &r300->shader_constants[PIPE_SHADER_FRAGMENT]); + OUT_CS(pack_float24(data[0])); + OUT_CS(pack_float24(data[1])); + OUT_CS(pack_float24(data[2])); + OUT_CS(pack_float24(data[3])); + } END_CS; } void r500_emit_fragment_program_code(struct r300_context* r300, - struct rX00_fragment_program_code* generic_code, - struct r300_constant_buffer* externals) + struct rX00_fragment_program_code* generic_code) { struct r500_fragment_program_code * code = &generic_code->code.r500; - struct rc_constant_list * constants = &generic_code->constants; int i; CS_LOCALS(r300); BEGIN_CS(13 + - ((code->inst_end + 1) * 6) + - (constants->Count ? (3 + (constants->Count * 4)) : 0)); + ((code->inst_end + 1) * 6)); OUT_CS_REG(R500_US_CONFIG, 0); OUT_CS_REG(R500_US_PIXSIZE, code->max_temp_idx); OUT_CS_REG(R500_US_CODE_RANGE, @@ -266,18 +272,30 @@ void r500_emit_fragment_program_code(struct r300_context* r300, OUT_CS(code->inst[i].inst5); } - if (constants->Count) { - OUT_CS_REG(R500_GA_US_VECTOR_INDEX, R500_GA_US_VECTOR_INDEX_TYPE_CONST); - OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, constants->Count * 4); - for (i = 0; i < constants->Count; i++) { - const float * data = get_shader_constant(r300, &constants->Constants[i], externals); - OUT_CS_32F(data[0]); - OUT_CS_32F(data[1]); - OUT_CS_32F(data[2]); - OUT_CS_32F(data[3]); - } - } + END_CS; +} + +void r500_emit_fs_constant_buffer(struct r300_context* r300, + struct rc_constant_list* constants) +{ + int i; + CS_LOCALS(r300); + + if (constants->Count == 0) + return; + BEGIN_CS(constants->Count * 4 + 2); + OUT_CS_REG(R500_GA_US_VECTOR_INDEX, R500_GA_US_VECTOR_INDEX_TYPE_CONST); + OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, constants->Count * 4); + for (i = 0; i < constants->Count; i++) { + const float * data = get_shader_constant(r300, + &constants->Constants[i], + &r300->shader_constants[PIPE_SHADER_FRAGMENT]); + OUT_CS_32F(data[0]); + OUT_CS_32F(data[1]); + OUT_CS_32F(data[2]); + OUT_CS_32F(data[3]); + } END_CS; } @@ -621,8 +639,7 @@ void r300_emit_vertex_format_state(struct r300_context* r300) } void r300_emit_vertex_program_code(struct r300_context* r300, - struct r300_vertex_program_code* code, - struct r300_constant_buffer* constants) + struct r300_vertex_program_code* code) { int i; struct r300_screen* r300screen = r300_screen(r300->context.screen); @@ -635,12 +652,7 @@ void r300_emit_vertex_program_code(struct r300_context* r300, return; } - if (code->constants.Count) { - BEGIN_CS(14 + code->length + (code->constants.Count * 4)); - } else { - BEGIN_CS(11 + code->length); - } - + BEGIN_CS(11 + code->length); /* R300_VAP_PVS_CODE_CNTL_0 * R300_VAP_PVS_CONST_CNTL * R300_VAP_PVS_CODE_CNTL_1 @@ -658,20 +670,6 @@ void r300_emit_vertex_program_code(struct r300_context* r300, for (i = 0; i < code->length; i++) OUT_CS(code->body.d[i]); - if (code->constants.Count) { - OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG, - (r300screen->caps->is_r500 ? - R500_PVS_CONST_START : R300_PVS_CONST_START)); - OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, code->constants.Count * 4); - for (i = 0; i < code->constants.Count; i++) { - const float * data = get_shader_constant(r300, &code->constants.Constants[i], constants); - OUT_CS_32F(data[0]); - OUT_CS_32F(data[1]); - OUT_CS_32F(data[2]); - OUT_CS_32F(data[3]); - } - } - OUT_CS_REG(R300_VAP_CNTL, R300_PVS_NUM_SLOTS(10) | R300_PVS_NUM_CNTLRS(5) | R300_PVS_NUM_FPUS(r300screen->caps->num_vert_fpus) | @@ -683,7 +681,40 @@ void r300_emit_vertex_program_code(struct r300_context* r300, void r300_emit_vertex_shader(struct r300_context* r300, struct r300_vertex_shader* vs) { - r300_emit_vertex_program_code(r300, &vs->code, &r300->shader_constants[PIPE_SHADER_VERTEX]); + r300_emit_vertex_program_code(r300, &vs->code); +} + +void r300_emit_vs_constant_buffer(struct r300_context* r300, + struct rc_constant_list* constants) +{ + int i; + struct r300_screen* r300screen = r300_screen(r300->context.screen); + CS_LOCALS(r300); + + if (!r300screen->caps->has_tcl) { + debug_printf("r300: Implementation error: emit_vertex_shader called," + " but has_tcl is FALSE!\n"); + return; + } + + if (constants->Count == 0) + return; + + BEGIN_CS(constants->Count * 4 + 3); + OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG, + (r300screen->caps->is_r500 ? + R500_PVS_CONST_START : R300_PVS_CONST_START)); + OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, constants->Count * 4); + for (i = 0; i < constants->Count; i++) { + const float * data = get_shader_constant(r300, + &constants->Constants[i], + &r300->shader_constants[PIPE_SHADER_VERTEX]); + OUT_CS_32F(data[0]); + OUT_CS_32F(data[1]); + OUT_CS_32F(data[2]); + OUT_CS_32F(data[3]); + } + END_CS; } void r300_emit_viewport_state(struct r300_context* r300, @@ -822,13 +853,22 @@ validate: if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER) { if (r300screen->caps->is_r500) { - r500_emit_fragment_program_code(r300, &r300->fs->code, &r300->shader_constants[PIPE_SHADER_FRAGMENT]); + r500_emit_fragment_program_code(r300, &r300->fs->code); } else { - r300_emit_fragment_program_code(r300, &r300->fs->code, &r300->shader_constants[PIPE_SHADER_FRAGMENT]); + r300_emit_fragment_program_code(r300, &r300->fs->code); } r300->dirty_state &= ~R300_NEW_FRAGMENT_SHADER; } + if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER_CONSTANTS) { + if (r300screen->caps->is_r500) { + r500_emit_fs_constant_buffer(r300, &r300->fs->code.constants); + } else { + r300_emit_fs_constant_buffer(r300, &r300->fs->code.constants); + } + r300->dirty_state &= ~R300_NEW_FRAGMENT_SHADER_CONSTANTS; + } + if (r300->dirty_state & R300_NEW_FRAMEBUFFERS) { r300_emit_fb_state(r300, &r300->framebuffer_state); r300->dirty_state &= ~R300_NEW_FRAMEBUFFERS; @@ -887,6 +927,11 @@ validate: r300->dirty_state &= ~R300_NEW_VERTEX_SHADER; } + if (r300->dirty_state & R300_NEW_VERTEX_SHADER_CONSTANTS) { + r300_emit_vs_constant_buffer(r300, &r300->vs->code.constants); + r300->dirty_state &= ~R300_NEW_VERTEX_SHADER_CONSTANTS; + } + /* XXX assert(r300->dirty_state == 0); */ diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h index 02ac5bebbd9..6befca72ced 100644 --- a/src/gallium/drivers/r300/r300_emit.h +++ b/src/gallium/drivers/r300/r300_emit.h @@ -23,6 +23,9 @@ #ifndef R300_EMIT_H #define R300_EMIT_H +#include "r300_context.h" +#include "radeon_code.h" + struct rX00_fragment_program_code; struct r300_vertex_program_code; @@ -39,12 +42,16 @@ void r300_emit_dsa_state(struct r300_context* r300, struct r300_dsa_state* dsa); void r300_emit_fragment_program_code(struct r300_context* r300, - struct rX00_fragment_program_code* generic_code, - struct r300_constant_buffer* externals); + struct rX00_fragment_program_code* generic_code); + +void r300_emit_fs_constant_buffer(struct r300_context* r300, + struct rc_constant_list* constants); void r500_emit_fragment_program_code(struct r300_context* r300, - struct rX00_fragment_program_code* generic_code, - struct r300_constant_buffer* externals); + struct rX00_fragment_program_code* generic_code); + +void r500_emit_fs_constant_buffer(struct r300_context* r300, + struct rc_constant_list* constants); void r300_emit_fb_state(struct r300_context* r300, struct pipe_framebuffer_state* fb); @@ -72,8 +79,10 @@ void r300_emit_vertex_buffer(struct r300_context* r300); void r300_emit_vertex_format_state(struct r300_context* r300); void r300_emit_vertex_program_code(struct r300_context* r300, - struct r300_vertex_program_code* code, - struct r300_constant_buffer* constants); + struct r300_vertex_program_code* code); + +void r300_emit_vs_constant_buffer(struct r300_context* r300, + struct rc_constant_list* constants); void r300_emit_vertex_shader(struct r300_context* r300, struct r300_vertex_shader* vs); diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index 3ac627e9599..4cf01389d2b 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -166,31 +166,6 @@ static void r300_set_clip_state(struct pipe_context* pipe, } } -static void - r300_set_constant_buffer(struct pipe_context* pipe, - uint shader, uint index, - const struct pipe_constant_buffer* buffer) -{ - struct r300_context* r300 = r300_context(pipe); - - /* This entire chunk of code seems ever-so-slightly baked. - * It's as if I've got pipe_buffer* matryoshkas... */ - if (buffer && buffer->buffer && buffer->buffer->size) { - void* map = pipe->winsys->buffer_map(pipe->winsys, buffer->buffer, - PIPE_BUFFER_USAGE_CPU_READ); - memcpy(r300->shader_constants[shader].constants, - map, buffer->buffer->size); - pipe->winsys->buffer_unmap(pipe->winsys, buffer->buffer); - - r300->shader_constants[shader].count = - buffer->buffer->size / (sizeof(float) * 4); - } else { - r300->shader_constants[shader].count = 0; - } - - r300->dirty_state |= R300_NEW_CONSTANTS; -} - /* Create a new depth, stencil, and alpha state based on the CSO dsa state. * * This contains the depth buffer, stencil buffer, alpha test, and such. @@ -345,7 +320,7 @@ static void r300_bind_fs_state(struct pipe_context* pipe, void* shader) r300->fs = fs; - r300->dirty_state |= R300_NEW_FRAGMENT_SHADER; + r300->dirty_state |= R300_NEW_FRAGMENT_SHADER | R300_NEW_FRAGMENT_SHADER_CONSTANTS; } /* Delete fragment shader state. */ @@ -702,7 +677,7 @@ static void r300_bind_vs_state(struct pipe_context* pipe, void* shader) draw_bind_vertex_shader(r300->draw, vs->draw); r300->vs = vs; - r300->dirty_state |= R300_NEW_VERTEX_SHADER; + r300->dirty_state |= R300_NEW_VERTEX_SHADER | R300_NEW_VERTEX_SHADER_CONSTANTS; } else { draw_bind_vertex_shader(r300->draw, (struct draw_vertex_shader*)shader); @@ -726,6 +701,31 @@ static void r300_delete_vs_state(struct pipe_context* pipe, void* shader) } } +static void r300_set_constant_buffer(struct pipe_context *pipe, + uint shader, uint index, + const struct pipe_constant_buffer *buf) +{ + struct r300_context* r300 = r300_context(pipe); + void *mapped; + + if (buf == NULL || buf->buffer->size == 0 || + (mapped = pipe_buffer_map(pipe->screen, buf->buffer, PIPE_BUFFER_USAGE_CPU_READ)) == NULL) + { + r300->shader_constants[shader].count = 0; + return; + } + + assert((buf->buffer->size % 4 * sizeof(float)) == 0); + memcpy(r300->shader_constants[shader].constants, mapped, buf->buffer->size); + r300->shader_constants[shader].count = buf->buffer->size / (4 * sizeof(float)); + pipe_buffer_unmap(pipe->screen, buf->buffer); + + if (shader == PIPE_SHADER_VERTEX) + r300->dirty_state |= R300_NEW_VERTEX_SHADER_CONSTANTS; + else if (shader == PIPE_SHADER_FRAGMENT) + r300->dirty_state |= R300_NEW_FRAGMENT_SHADER_CONSTANTS; +} + void r300_init_state_functions(struct r300_context* r300) { r300->context.create_blend_state = r300_create_blend_state; -- cgit v1.2.3 From 3d73852121f13832f6bc87918798ff96589d0349 Mon Sep 17 00:00:00 2001 From: Maciej Cencora Date: Sun, 1 Nov 2009 18:50:52 +0100 Subject: r300g: fix geometry corruptions PVS flush is needed before changing the vertex shader or vertex shader constants. --- src/gallium/drivers/r300/r300_emit.c | 16 ++++++++++++++-- 1 file changed, 14 insertions(+), 2 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index de27f0939b0..5b03c1aa6c1 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -652,7 +652,7 @@ void r300_emit_vertex_program_code(struct r300_context* r300, return; } - BEGIN_CS(11 + code->length); + BEGIN_CS(9 + code->length); /* R300_VAP_PVS_CODE_CNTL_0 * R300_VAP_PVS_CONST_CNTL * R300_VAP_PVS_CODE_CNTL_1 @@ -674,7 +674,6 @@ void r300_emit_vertex_program_code(struct r300_context* r300, R300_PVS_NUM_CNTLRS(5) | R300_PVS_NUM_FPUS(r300screen->caps->num_vert_fpus) | R300_PVS_VF_MAX_VTX_NUM(12)); - OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0); END_CS; } @@ -749,6 +748,15 @@ void r300_flush_textures(struct r300_context* r300) END_CS; } +static void r300_flush_pvs(struct r300_context* r300) +{ + CS_LOCALS(r300); + + BEGIN_CS(2); + OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0); + END_CS; +} + /* Emit all dirty state. */ void r300_emit_dirty_state(struct r300_context* r300) { @@ -922,6 +930,10 @@ validate: r300->dirty_state &= ~R300_NEW_VERTEX_FORMAT; } + if (r300->dirty_state & (R300_NEW_VERTEX_SHADER | R300_NEW_VERTEX_SHADER_CONSTANTS)) { + r300_flush_pvs(r300); + } + if (r300->dirty_state & R300_NEW_VERTEX_SHADER) { r300_emit_vertex_shader(r300, r300->vs); r300->dirty_state &= ~R300_NEW_VERTEX_SHADER; -- cgit v1.2.3 From 1f630fa0167ed799556a764178772c096a3ddeba Mon Sep 17 00:00:00 2001 From: Corbin Simpson Date: Sun, 1 Nov 2009 11:54:52 -0800 Subject: r300g: Miscellania. Avoid draw segfaults, s/true/TRUE/, etc. Cleared out my git stash. --- src/gallium/drivers/r300/r300_context.h | 2 +- src/gallium/drivers/r300/r300_debug.c | 6 +++--- src/gallium/drivers/r300/r300_emit.c | 2 +- src/gallium/drivers/r300/r300_state.c | 28 +++++++++++++++++++++------- src/gallium/drivers/r300/r300_vs.c | 4 ++-- 5 files changed, 28 insertions(+), 14 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index b1738452de0..ae7015634cc 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -339,7 +339,7 @@ void r300_init_surface_functions(struct r300_context* r300); static INLINE boolean DBG_ON(struct r300_context * ctx, unsigned flags) { - return (ctx->debug & flags) ? true : false; + return (ctx->debug & flags) ? TRUE : FALSE; } static INLINE void DBG(struct r300_context * ctx, unsigned flags, const char * fmt, ...) diff --git a/src/gallium/drivers/r300/r300_debug.c b/src/gallium/drivers/r300/r300_debug.c index 421253ca722..2a6ed54ac9b 100644 --- a/src/gallium/drivers/r300/r300_debug.c +++ b/src/gallium/drivers/r300/r300_debug.c @@ -49,7 +49,7 @@ static struct debug_option debug_options[] = { void r300_init_debug(struct r300_context * ctx) { const char * options = debug_get_option("RADEON_DEBUG", 0); - boolean printhint = false; + boolean printhint = FALSE; size_t length; struct debug_option * opt; @@ -71,14 +71,14 @@ void r300_init_debug(struct r300_context * ctx) if (!opt->name) { debug_printf("Unknown debug option: %s\n", options); - printhint = true; + printhint = TRUE; } options += length; } if (!ctx->debug) - printhint = true; + printhint = TRUE; } if (printhint || ctx->debug & DBG_HELP) { diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 5b03c1aa6c1..79972dbb49e 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -800,7 +800,7 @@ validate: for (i = 0; i < r300->texture_count; i++) { tex = r300->textures[i]; if (!tex) - continue; + continue; if (!r300->winsys->add_buffer(r300->winsys, tex->buffer, RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0)) { r300->context.flush(&r300->context, 0, NULL); diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index 4cf01389d2b..af063d4b20a 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -281,7 +281,9 @@ static void { struct r300_context* r300 = r300_context(pipe); - draw_flush(r300->draw); + if (r300->draw) { + draw_flush(r300->draw); + } r300->framebuffer_state = *state; @@ -444,10 +446,13 @@ static void r300_bind_rs_state(struct pipe_context* pipe, void* state) struct r300_context* r300 = r300_context(pipe); struct r300_rs_state* rs = (struct r300_rs_state*)state; - draw_flush(r300->draw); - draw_set_rasterizer_state(r300->draw, &rs->rs); + if (r300->draw) { + draw_flush(r300->draw); + draw_set_rasterizer_state(r300->draw, &rs->rs); + } r300->rs_state = rs; + /* XXX Clean these up when we move to atom emits */ r300->dirty_state |= R300_NEW_RASTERIZER; r300->dirty_state |= R300_NEW_RS_BLOCK; r300->dirty_state |= R300_NEW_SCISSOR; @@ -623,8 +628,10 @@ static void r300_set_vertex_buffers(struct pipe_context* pipe, r300->vertex_buffer_count = count; - draw_flush(r300->draw); - draw_set_vertex_buffers(r300->draw, count, buffers); + if (r300->draw) { + draw_flush(r300->draw); + draw_set_vertex_buffers(r300->draw, count, buffers); + } } static void r300_set_vertex_elements(struct pipe_context* pipe, @@ -633,8 +640,15 @@ static void r300_set_vertex_elements(struct pipe_context* pipe, { struct r300_context* r300 = r300_context(pipe); - draw_flush(r300->draw); - draw_set_vertex_elements(r300->draw, count, elements); + memcpy(r300->vertex_elements, elements, + sizeof(struct pipe_vertex_element) * count); + + r300->vertex_element_count = count; + + if (r300->draw) { + draw_flush(r300->draw); + draw_set_vertex_elements(r300->draw, count, elements); + } } static void* r300_create_vs_state(struct pipe_context* pipe, diff --git a/src/gallium/drivers/r300/r300_vs.c b/src/gallium/drivers/r300/r300_vs.c index eca85879a7f..74ef416dc14 100644 --- a/src/gallium/drivers/r300/r300_vs.c +++ b/src/gallium/drivers/r300/r300_vs.c @@ -37,7 +37,7 @@ static void set_vertex_inputs_outputs(struct r300_vertex_program_compiler * c) struct tgsi_shader_info* info = &vs->info; struct tgsi_parse_context parser; struct tgsi_full_declaration * decl; - boolean pointsize = false; + boolean pointsize = FALSE; int out_colors = 0; int colors = 0; int out_generic = 0; @@ -52,7 +52,7 @@ static void set_vertex_inputs_outputs(struct r300_vertex_program_compiler * c) for (i = 0; i < info->num_outputs; i++) { switch (info->output_semantic_name[i]) { case TGSI_SEMANTIC_PSIZE: - pointsize = true; + pointsize = TRUE; break; case TGSI_SEMANTIC_COLOR: out_colors++; -- cgit v1.2.3 From ad96c0d851f6c3696fa6ae0c1f6ad56e849bc739 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Tue, 3 Nov 2009 16:48:48 +0100 Subject: r300g: add color channel masking Signed-off-by: Corbin Simpson --- src/gallium/drivers/r300/r300_context.h | 1 + src/gallium/drivers/r300/r300_emit.c | 6 +++--- src/gallium/drivers/r300/r300_state.c | 14 ++++++++++++++ src/gallium/drivers/r300/r300_state_invariant.c | 3 +-- 4 files changed, 19 insertions(+), 5 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index ae7015634cc..8c65c04d01a 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -34,6 +34,7 @@ struct r300_vertex_shader; struct r300_blend_state { uint32_t blend_control; /* R300_RB3D_CBLEND: 0x4e04 */ uint32_t alpha_blend_control; /* R300_RB3D_ABLEND: 0x4e08 */ + uint32_t color_channel_mask; /* R300_RB3D_COLOR_CHANNEL_MASK: 0x4e0c */ uint32_t rop; /* R300_RB3D_ROPCNTL: 0x4e18 */ uint32_t dither; /* R300_RB3D_DITHER_CTL: 0x4e50 */ }; diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 79972dbb49e..8fe9a688860 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -38,10 +38,11 @@ void r300_emit_blend_state(struct r300_context* r300, struct r300_blend_state* blend) { CS_LOCALS(r300); - BEGIN_CS(7); - OUT_CS_REG_SEQ(R300_RB3D_CBLEND, 2); + BEGIN_CS(8); + OUT_CS_REG_SEQ(R300_RB3D_CBLEND, 3); OUT_CS(blend->blend_control); OUT_CS(blend->alpha_blend_control); + OUT_CS(blend->color_channel_mask); OUT_CS_REG(R300_RB3D_ROPCNTL, blend->rop); OUT_CS_REG(R300_RB3D_DITHER_CTL, blend->dither); END_CS; @@ -313,7 +314,6 @@ void r300_emit_fb_state(struct r300_context* r300, tex = (struct r300_texture*)surf->texture; assert(tex && tex->buffer && "cbuf is marked, but NULL!"); - /* XXX I still need to figure out how to set the mipmap level here */ OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0 + (4 * i), 1); OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index af063d4b20a..242ec9f365c 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -100,6 +100,20 @@ static void* r300_create_blend_state(struct pipe_context* pipe, (state->logicop_func) << R300_RB3D_ROPCNTL_ROP_SHIFT; } + /* Color Channel Mask */ + if (state->colormask & PIPE_MASK_R) { + blend->color_channel_mask |= RB3D_COLOR_CHANNEL_MASK_RED_MASK0; + } + if (state->colormask & PIPE_MASK_G) { + blend->color_channel_mask |= RB3D_COLOR_CHANNEL_MASK_GREEN_MASK0; + } + if (state->colormask & PIPE_MASK_B) { + blend->color_channel_mask |= RB3D_COLOR_CHANNEL_MASK_BLUE_MASK0; + } + if (state->colormask & PIPE_MASK_A) { + blend->color_channel_mask |= RB3D_COLOR_CHANNEL_MASK_ALPHA_MASK0; + } + if (state->dither) { blend->dither = R300_RB3D_DITHER_CTL_DITHER_MODE_LUT | R300_RB3D_DITHER_CTL_ALPHA_DITHER_MODE_LUT; diff --git a/src/gallium/drivers/r300/r300_state_invariant.c b/src/gallium/drivers/r300/r300_state_invariant.c index 4865f160581..7e4d5c7c72d 100644 --- a/src/gallium/drivers/r300/r300_state_invariant.c +++ b/src/gallium/drivers/r300/r300_state_invariant.c @@ -84,7 +84,7 @@ void r300_emit_invariant_state(struct r300_context* r300) END_CS; /* XXX unsorted stuff from surface_fill */ - BEGIN_CS(64 + (caps->has_tcl ? 5 : 0) + (caps->is_r500 ? 4 : 0)); + BEGIN_CS(62 + (caps->has_tcl ? 5 : 0) + (caps->is_r500 ? 4 : 0)); /* Flush PVS. */ OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0); @@ -125,7 +125,6 @@ void r300_emit_invariant_state(struct r300_context* r300) OUT_CS_REG(R300_SC_HYPERZ, 0x0000001C); OUT_CS_REG(R300_SC_EDGERULE, 0x2DA49525); OUT_CS_REG(R300_RB3D_CCTL, 0x00000000); - OUT_CS_REG(RB3D_COLOR_CHANNEL_MASK, 0x0000000F); OUT_CS_REG(R300_RB3D_AARESOLVE_CTL, 0x00000000); if (caps->is_r500) { OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 0x00000000); -- cgit v1.2.3 From 07190888bdc41f53bf8ea30c9e2ee4a61b42d802 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Tue, 3 Nov 2009 16:50:09 +0100 Subject: r300g: set the correct offset in a colorbuffer surface Suggested by Joakim Sindholt. Also, put flushing of colorbuffers _before_ the framebuffer state setup, suggested by docs. Signed-off-by: Corbin Simpson --- src/gallium/drivers/r300/r300_emit.c | 17 +++++++++-------- 1 file changed, 9 insertions(+), 8 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 8fe9a688860..fc823ad31fe 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -309,13 +309,20 @@ void r300_emit_fb_state(struct r300_context* r300, CS_LOCALS(r300); BEGIN_CS((10 * fb->nr_cbufs) + (fb->zsbuf ? 10 : 0) + 4); + OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, + R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS | + R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D); + OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT, + R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE | + R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE); + for (i = 0; i < fb->nr_cbufs; i++) { surf = fb->cbufs[i]; tex = (struct r300_texture*)surf->texture; assert(tex && tex->buffer && "cbuf is marked, but NULL!"); OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0 + (4 * i), 1); - OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); + OUT_CS_RELOC(tex->buffer, surf->offset, 0, RADEON_GEM_DOMAIN_VRAM, 0); OUT_CS_REG_SEQ(R300_RB3D_COLORPITCH0 + (4 * i), 1); OUT_CS_RELOC(tex->buffer, tex->pitch[surf->level] | @@ -332,7 +339,7 @@ void r300_emit_fb_state(struct r300_context* r300, assert(tex && tex->buffer && "zsbuf is marked, but NULL!"); OUT_CS_REG_SEQ(R300_ZB_DEPTHOFFSET, 1); - OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0); + OUT_CS_RELOC(tex->buffer, surf->offset, 0, RADEON_GEM_DOMAIN_VRAM, 0); OUT_CS_REG(R300_ZB_FORMAT, r300_translate_zsformat(tex->tex.format)); @@ -341,12 +348,6 @@ void r300_emit_fb_state(struct r300_context* r300, RADEON_GEM_DOMAIN_VRAM, 0); } - OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, - R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS | - R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D); - OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT, - R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE | - R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE); END_CS; } -- cgit v1.2.3 From 4671005a4317fa37aea8786740470a40906fbfa7 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Tue, 3 Nov 2009 16:58:39 +0100 Subject: r300g: fix the size of CS when emitting the fragprog constant buffer Signed-off-by: Corbin Simpson --- src/gallium/drivers/r300/r300_emit.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index fc823ad31fe..6415c59c2d5 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -285,7 +285,7 @@ void r500_emit_fs_constant_buffer(struct r300_context* r300, if (constants->Count == 0) return; - BEGIN_CS(constants->Count * 4 + 2); + BEGIN_CS(constants->Count * 4 + 3); OUT_CS_REG(R500_GA_US_VECTOR_INDEX, R500_GA_US_VECTOR_INDEX_TYPE_CONST); OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, constants->Count * 4); for (i = 0; i < constants->Count; i++) { -- cgit v1.2.3 From c2e47191d72e16aaa1fae4f47bbed7639c2ff201 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Wed, 4 Nov 2009 10:56:44 +0100 Subject: r300g: add polygon mode Signed-off-by: Corbin Simpson --- src/gallium/drivers/r300/r300_context.h | 1 + src/gallium/drivers/r300/r300_emit.c | 3 ++- src/gallium/drivers/r300/r300_state.c | 27 +++++++++++++++++++ src/gallium/drivers/r300/r300_state_inlines.h | 36 +++++++++++++++++++++++++ src/gallium/drivers/r300/r300_state_invariant.c | 3 +-- 5 files changed, 67 insertions(+), 3 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index 8c65c04d01a..850e5a41c91 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -78,6 +78,7 @@ struct r300_rs_state { uint32_t line_stipple_config; /* R300_GA_LINE_STIPPLE_CONFIG: 0x4328 */ uint32_t line_stipple_value; /* R300_GA_LINE_STIPPLE_VALUE: 0x4260 */ uint32_t color_control; /* R300_GA_COLOR_CONTROL: 0x4278 */ + uint32_t polygon_mode; /* R300_GA_POLY_MODE: 0x4288 */ }; struct r300_rs_block { diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 6415c59c2d5..69ce5966e80 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -483,7 +483,7 @@ void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs) { CS_LOCALS(r300); - BEGIN_CS(20); + BEGIN_CS(22); OUT_CS_REG(R300_VAP_CNTL_STATUS, rs->vap_control_status); OUT_CS_REG(R300_GA_POINT_SIZE, rs->point_size); OUT_CS_REG_SEQ(R300_GA_POINT_MINMAX, 2); @@ -499,6 +499,7 @@ void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs) OUT_CS_REG(R300_GA_LINE_STIPPLE_CONFIG, rs->line_stipple_config); OUT_CS_REG(R300_GA_LINE_STIPPLE_VALUE, rs->line_stipple_value); OUT_CS_REG(R300_GA_COLOR_CONTROL, rs->color_control); + OUT_CS_REG(R300_GA_POLY_MODE, rs->polygon_mode); END_CS; } diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index 242ec9f365c..658a8cba132 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -398,25 +398,52 @@ static void* r300_create_rs_state(struct pipe_context* pipe, rs->line_control = pack_float_16_6x(state->line_width) | R300_GA_LINE_CNTL_END_TYPE_COMP; + /* XXX I think there is something wrong with the polygon mode, + * XXX re-test when r300g is in a better shape */ + + /* Enable polygon mode */ + if (state->fill_cw != PIPE_POLYGON_MODE_FILL || + state->fill_ccw != PIPE_POLYGON_MODE_FILL) { + rs->polygon_mode = R300_GA_POLY_MODE_DUAL; + } + /* Radeons don't think in "CW/CCW", they think in "front/back". */ if (state->front_winding == PIPE_WINDING_CW) { rs->cull_mode = R300_FRONT_FACE_CW; + /* Polygon offset */ if (state->offset_cw) { rs->polygon_offset_enable |= R300_FRONT_ENABLE; } if (state->offset_ccw) { rs->polygon_offset_enable |= R300_BACK_ENABLE; } + + /* Polygon mode */ + if (rs->polygon_mode) { + rs->polygon_mode |= + r300_translate_polygon_mode_front(state->fill_cw); + rs->polygon_mode |= + r300_translate_polygon_mode_back(state->fill_ccw); + } } else { rs->cull_mode = R300_FRONT_FACE_CCW; + /* Polygon offset */ if (state->offset_ccw) { rs->polygon_offset_enable |= R300_FRONT_ENABLE; } if (state->offset_cw) { rs->polygon_offset_enable |= R300_BACK_ENABLE; } + + /* Polygon mode */ + if (rs->polygon_mode) { + rs->polygon_mode |= + r300_translate_polygon_mode_front(state->fill_ccw); + rs->polygon_mode |= + r300_translate_polygon_mode_back(state->fill_cw); + } } if (state->front_winding & state->cull_mode) { rs->cull_mode |= R300_CULL_FRONT; diff --git a/src/gallium/drivers/r300/r300_state_inlines.h b/src/gallium/drivers/r300/r300_state_inlines.h index 176e59f281f..52b9650fc16 100644 --- a/src/gallium/drivers/r300/r300_state_inlines.h +++ b/src/gallium/drivers/r300/r300_state_inlines.h @@ -504,4 +504,40 @@ r300_translate_vertex_data_swizzle(enum pipe_format format) { (0xf << R300_WRITE_ENA_SHIFT)); } +static INLINE uint32_t +r300_translate_polygon_mode_front(unsigned mode) { + switch (mode) + { + case PIPE_POLYGON_MODE_FILL: + return R300_GA_POLY_MODE_FRONT_PTYPE_TRI; + case PIPE_POLYGON_MODE_LINE: + return R300_GA_POLY_MODE_FRONT_PTYPE_LINE; + case PIPE_POLYGON_MODE_POINT: + return R300_GA_POLY_MODE_FRONT_PTYPE_POINT; + + default: + debug_printf("r300: Bad polygon mode %i in %s\n", mode, + __FUNCTION__); + return R300_GA_POLY_MODE_FRONT_PTYPE_TRI; + } +} + +static INLINE uint32_t +r300_translate_polygon_mode_back(unsigned mode) { + switch (mode) + { + case PIPE_POLYGON_MODE_FILL: + return R300_GA_POLY_MODE_BACK_PTYPE_TRI; + case PIPE_POLYGON_MODE_LINE: + return R300_GA_POLY_MODE_BACK_PTYPE_LINE; + case PIPE_POLYGON_MODE_POINT: + return R300_GA_POLY_MODE_BACK_PTYPE_POINT; + + default: + debug_printf("r300: Bad polygon mode %i in %s\n", mode, + __FUNCTION__); + return R300_GA_POLY_MODE_BACK_PTYPE_TRI; + } +} + #endif /* R300_STATE_INLINES_H */ diff --git a/src/gallium/drivers/r300/r300_state_invariant.c b/src/gallium/drivers/r300/r300_state_invariant.c index 7e4d5c7c72d..c07e6ae676d 100644 --- a/src/gallium/drivers/r300/r300_state_invariant.c +++ b/src/gallium/drivers/r300/r300_state_invariant.c @@ -84,7 +84,7 @@ void r300_emit_invariant_state(struct r300_context* r300) END_CS; /* XXX unsorted stuff from surface_fill */ - BEGIN_CS(62 + (caps->has_tcl ? 5 : 0) + (caps->is_r500 ? 4 : 0)); + BEGIN_CS(60 + (caps->has_tcl ? 5 : 0) + (caps->is_r500 ? 4 : 0)); /* Flush PVS. */ OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0); @@ -114,7 +114,6 @@ void r300_emit_invariant_state(struct r300_context* r300) /* XXX this big chunk should be refactored into rs_state */ OUT_CS_REG(R300_GA_SOLID_RG, 0x00000000); OUT_CS_REG(R300_GA_SOLID_BA, 0x00000000); - OUT_CS_REG(R300_GA_POLY_MODE, 0x00000000); OUT_CS_REG(R300_GA_ROUND_MODE, 0x00000001); OUT_CS_REG(R300_GA_OFFSET, 0x00000000); OUT_CS_REG(R300_GA_FOG_SCALE, 0x3DBF1412); -- cgit v1.2.3 From 3445f476977ae403cef9ca15661fa0f96ff50eca Mon Sep 17 00:00:00 2001 From: Maciej Cencora Date: Sun, 25 Oct 2009 13:53:25 +0100 Subject: r300g: VBOs WIP --- src/gallium/drivers/r300/Makefile | 1 + src/gallium/drivers/r300/r300_context.c | 14 +- src/gallium/drivers/r300/r300_context.h | 27 +-- src/gallium/drivers/r300/r300_emit.c | 108 +++++++++++- src/gallium/drivers/r300/r300_emit.h | 12 ++ src/gallium/drivers/r300/r300_render.c | 135 +++++++-------- src/gallium/drivers/r300/r300_state.c | 13 +- src/gallium/drivers/r300/r300_state_derived.c | 14 +- src/gallium/drivers/r300/r300_vbo.c | 226 ++++++++++++++++++++++++++ src/gallium/drivers/r300/r300_vbo.h | 36 ++++ 10 files changed, 477 insertions(+), 109 deletions(-) create mode 100644 src/gallium/drivers/r300/r300_vbo.c create mode 100644 src/gallium/drivers/r300/r300_vbo.h (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/Makefile b/src/gallium/drivers/r300/Makefile index f73d80de882..d13bb7a36b1 100644 --- a/src/gallium/drivers/r300/Makefile +++ b/src/gallium/drivers/r300/Makefile @@ -17,6 +17,7 @@ C_SOURCES = \ r300_state.c \ r300_state_derived.c \ r300_state_invariant.c \ + r300_vbo.c \ r300_vs.c \ r300_texture.c \ r300_tgsi_to_rc.c diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c index f974147ea43..b520e5929e2 100644 --- a/src/gallium/drivers/r300/r300_context.c +++ b/src/gallium/drivers/r300/r300_context.c @@ -104,6 +104,7 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen, struct r300_winsys* r300_winsys) { struct r300_context* r300 = CALLOC_STRUCT(r300_context); + struct r300_screen* r300screen = r300_screen(screen); if (!r300) return NULL; @@ -119,9 +120,16 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen, r300->context.clear = r300_clear; - r300->context.draw_arrays = r300_draw_arrays; - r300->context.draw_elements = r300_draw_elements; - r300->context.draw_range_elements = r300_swtcl_draw_range_elements; + if (r300screen->caps->has_tcl) + { + r300->context.draw_arrays = r300_draw_arrays; + r300->context.draw_elements = r300_draw_elements; + r300->context.draw_range_elements = r300_draw_range_elements; + } + else + { + assert(0); + } r300->context.is_texture_referenced = r300_is_texture_referenced; r300->context.is_buffer_referenced = r300_is_buffer_referenced; diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index 850e5a41c91..a6748852d83 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -216,18 +216,19 @@ struct r300_texture { struct r300_texture_state state; }; -struct r300_vertex_format { +struct r300_vertex_info { /* Parent class */ struct vertex_info vinfo; - /* R300_VAP_PROG_STREAK_CNTL_[0-7] */ - uint32_t vap_prog_stream_cntl[8]; - /* R300_VAP_PROG_STREAK_CNTL_EXT_[0-7] */ - uint32_t vap_prog_stream_cntl_ext[8]; /* Map of vertex attributes into PVS memory for HW TCL, * or GA memory for SW TCL. */ int vs_tab[16]; /* Map of rasterizer attributes from GB through RS to US. */ int fs_tab[16]; + + /* R300_VAP_PROG_STREAK_CNTL_[0-7] */ + uint32_t vap_prog_stream_cntl[8]; + /* R300_VAP_PROG_STREAK_CNTL_EXT_[0-7] */ + uint32_t vap_prog_stream_cntl_ext[8]; }; extern struct pipe_viewport_state r300_viewport_identity; @@ -256,7 +257,7 @@ struct r300_context { * depends on the combination of both currently loaded shaders. */ struct util_hash_table* shader_hash_table; /* Vertex formatting information. */ - struct r300_vertex_format* vertex_info; + struct r300_vertex_info* vertex_info; /* Various CSO state objects. */ /* Blend state. */ @@ -285,12 +286,6 @@ struct r300_context { /* Texture states. */ struct r300_texture* textures[8]; int texture_count; - /* Vertex buffers for Gallium. */ - struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS]; - int vertex_buffer_count; - /* Vertex elements for Gallium. */ - struct pipe_vertex_element vertex_elements[PIPE_MAX_ATTRIBS]; - int vertex_element_count; /* Vertex shader. */ struct r300_vertex_shader* vs; /* Viewport state. */ @@ -298,6 +293,14 @@ struct r300_context { /* ZTOP state. */ struct r300_ztop_state ztop_state; + /* Vertex buffers for Gallium. */ + struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; + int vbuf_count; + /* Vertex elements for Gallium. */ + struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; + int aos_count; + unsigned hw_prim; + /* Bitmask of dirty state objects. */ uint32_t dirty_state; /* Flag indicating whether or not the HW is dirty. */ diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 69ce5966e80..92e6ec606ca 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -582,7 +582,48 @@ void r300_emit_texture(struct r300_context* r300, END_CS; } -void r300_emit_vertex_buffer(struct r300_context* r300) +void r300_emit_aos(struct r300_context* r300, unsigned offset) +{ + struct pipe_vertex_buffer *vbuf = r300->vertex_buffer; + struct pipe_vertex_element *velem = r300->vertex_element; + CS_LOCALS(r300); + int i; + unsigned packet_size = (r300->aos_count * 3 + 1) / 2; + BEGIN_CS(2 + packet_size + r300->aos_count * 2); + OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, packet_size); + OUT_CS(r300->aos_count); + for (i = 0; i < r300->aos_count - 1; i += 2) { + int buf_num1 = velem[i].vertex_buffer_index; + int buf_num2 = velem[i+1].vertex_buffer_index; + assert(vbuf[buf_num1].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0); + assert(vbuf[buf_num2].stride % 4 == 0 && pf_get_size(velem[i+1].src_format) % 4 == 0); + OUT_CS((pf_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num1].stride << 6) | + (pf_get_size(velem[i+1].src_format) << 14) | (vbuf[buf_num2].stride << 22)); + OUT_CS(vbuf[buf_num1].buffer_offset + velem[i].src_offset + + offset * vbuf[buf_num1].stride); + OUT_CS(vbuf[buf_num2].buffer_offset + velem[i+1].src_offset + + offset * vbuf[buf_num2].stride); + } + if (r300->aos_count & 1) { + int buf_num = velem[i].vertex_buffer_index; + assert(vbuf[buf_num].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0); + OUT_CS((pf_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num].stride << 6)); + OUT_CS(vbuf[buf_num].buffer_offset + velem[i].src_offset + + offset * vbuf[buf_num].stride); + } + + for (i = 0; i < r300->aos_count; i++) { + cs_winsys->write_cs_reloc(cs_winsys, + vbuf[velem[i].vertex_buffer_index].buffer, + RADEON_GEM_DOMAIN_GTT, + 0, + 0); + cs_count -= 2; + } + END_CS; +} +#if 0 +void r300_emit_draw_packet(struct r300_context* r300) { CS_LOCALS(r300); @@ -605,6 +646,65 @@ void r300_emit_vertex_buffer(struct r300_context* r300) OUT_CS_RELOC(r300->vbo, 0, RADEON_GEM_DOMAIN_GTT, 0, 0); END_CS; } +#endif +void r300_emit_draw_arrays(struct r300_context *r300, + unsigned count) +{ + CS_LOCALS(r300); + assert(count < 65536); + + BEGIN_CS(4); + OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, count); + OUT_CS_PKT3(R300_PACKET3_3D_DRAW_VBUF_2, 0); + OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (count << 16) | + r300->hw_prim); + END_CS; +} + +void r300_emit_draw_elements(struct r300_context *r300, + struct pipe_buffer* indexBuffer, + unsigned indexSize, + unsigned minIndex, + unsigned maxIndex, + unsigned start, + unsigned count) +{ + CS_LOCALS(r300); + assert(indexSize == 4 || indexSize == 2); + assert(count < 65536); + assert((start * indexSize) % 4 == 0); + + uint32_t size_dwords; + uint32_t skip_dwords = indexSize * start / sizeof(uint32_t); + assert(skip_dwords == 0); + + BEGIN_CS(10); + OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, maxIndex); + OUT_CS_PKT3(R300_PACKET3_3D_DRAW_INDX_2, 0); + if (indexSize == 4) { + size_dwords = count + start; + OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (count << 16) | + R300_VAP_VF_CNTL__INDEX_SIZE_32bit | r300->hw_prim); + } else { + size_dwords = (count + start + 1) / 2; + OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | + (count << 16) | r300->hw_prim); + } + + OUT_CS_PKT3(R300_PACKET3_INDX_BUFFER, 2); + OUT_CS(R300_INDX_BUFFER_ONE_REG_WR | (R300_VAP_PORT_IDX0 >> 2) | + (0 << R300_INDX_BUFFER_SKIP_SHIFT)); + OUT_CS(skip_dwords); + OUT_CS(size_dwords); + cs_winsys->write_cs_reloc(cs_winsys, + indexBuffer, + RADEON_GEM_DOMAIN_GTT, + 0, + 0); + cs_count -= 2; + + END_CS; +} void r300_emit_vertex_format_state(struct r300_context* r300) { @@ -771,8 +871,6 @@ void r300_emit_dirty_state(struct r300_context* r300) return; } - r300_update_derived_state(r300); - /* Clean out BOs. */ r300->winsys->reset_bos(r300->winsys); @@ -823,7 +921,7 @@ validate: goto validate; } } else { - debug_printf("No VBO while emitting dirty state!\n"); + // debug_printf("No VBO while emitting dirty state!\n"); } if (!r300->winsys->validate(r300->winsys)) { r300->context.flush(&r300->context, 0, NULL); @@ -951,7 +1049,7 @@ validate: */ /* Finally, emit the VBO. */ - r300_emit_vertex_buffer(r300); + //r300_emit_vertex_buffer(r300); r300->dirty_hw++; } diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h index 6befca72ced..b4fdfecde01 100644 --- a/src/gallium/drivers/r300/r300_emit.h +++ b/src/gallium/drivers/r300/r300_emit.h @@ -29,6 +29,8 @@ struct rX00_fragment_program_code; struct r300_vertex_program_code; +void r300_emit_aos(struct r300_context* r300, unsigned offset); + void r300_emit_blend_state(struct r300_context* r300, struct r300_blend_state* blend); @@ -38,6 +40,16 @@ void r300_emit_blend_color_state(struct r300_context* r300, void r300_emit_clip_state(struct r300_context* r300, struct pipe_clip_state* clip); +void r300_emit_draw_arrays(struct r300_context *r300, unsigned count); + +void r300_emit_draw_elements(struct r300_context *r300, + struct pipe_buffer* indexBuffer, + unsigned indexSize, + unsigned minIndex, + unsigned maxIndex, + unsigned start, + unsigned count); + void r300_emit_dsa_state(struct r300_context* r300, struct r300_dsa_state* dsa); diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c index 634c803f2a8..86aaf841dd6 100644 --- a/src/gallium/drivers/r300/r300_render.c +++ b/src/gallium/drivers/r300/r300_render.c @@ -31,6 +31,7 @@ #include "util/u_memory.h" #include "util/u_prim.h" +#include "r300_vbo.h" #include "r300_cs.h" #include "r300_context.h" #include "r300_emit.h" @@ -69,98 +70,64 @@ uint32_t r300_translate_primitive(unsigned prim) } } -/* This is the fast-path drawing & emission for HW TCL. */ -boolean r300_draw_range_elements(struct pipe_context* pipe, - struct pipe_buffer* indexBuffer, - unsigned indexSize, - unsigned minIndex, - unsigned maxIndex, - unsigned mode, - unsigned start, - unsigned count) +static boolean setup_vertex_buffers(struct r300_context *r300) { - struct r300_context* r300 = r300_context(pipe); - uint32_t prim = r300_translate_primitive(mode); - struct pipe_vertex_buffer* aos = r300->vertex_buffers; - unsigned aos_count = r300->vertex_buffer_count; - short* indices; - unsigned packet_size; - unsigned i; - bool invalid = FALSE; - - CS_LOCALS(r300); - - if (!u_trim_pipe_prim(mode, &count)) { - return FALSE; - } + unsigned vbuf_count = r300->aos_count; + struct pipe_vertex_buffer *vbuf= r300->vertex_buffer; + struct pipe_vertex_element *velem= r300->vertex_element; + bool invalid = false; validate: - for (i = 0; i < aos_count; i++) { - if (!r300->winsys->add_buffer(r300->winsys, aos[i].buffer, - RADEON_GEM_DOMAIN_GTT, 0)) { - pipe->flush(pipe, 0, NULL); + for (int i = 0; i < vbuf_count; i++) { + if (!r300->winsys->add_buffer(r300->winsys, vbuf[velem[i].vertex_buffer_index].buffer, + RADEON_GEM_DOMAIN_GTT, 0)) { + r300->context.flush(&r300->context, 0, NULL); goto validate; } } + if (!r300->winsys->validate(r300->winsys)) { - pipe->flush(pipe, 0, NULL); + r300->context.flush(&r300->context, 0, NULL); if (invalid) { /* Well, hell. */ debug_printf("r300: Stuck in validation loop, gonna quit now."); exit(1); } - invalid = TRUE; + invalid = true; goto validate; } - r300_emit_dirty_state(r300); + return invalid; +} - packet_size = (aos_count >> 1) * 3 + (aos_count & 1) * 2; - - BEGIN_CS(3 + packet_size + (aos_count * 2)); - OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, packet_size); - OUT_CS(aos_count); - for (i = 0; i < aos_count - 1; i += 2) { - OUT_CS(aos[i].stride | - (aos[i].stride << 8) | - (aos[i + 1].stride << 16) | - (aos[i + 1].stride << 24)); - OUT_CS(aos[i].buffer_offset + start * 4 * aos[i].stride); - OUT_CS(aos[i + 1].buffer_offset + start * 4 * aos[i + 1].stride); - } - if (aos_count & 1) { - OUT_CS(aos[i].stride | (aos[i].stride << 8)); - OUT_CS(aos[i].buffer_offset + start * 4 * aos[i].stride); - } - for (i = 0; i < aos_count; i++) { - OUT_CS_RELOC(aos[i].buffer, 0, RADEON_GEM_DOMAIN_GTT, 0, 0); - } - END_CS; +/* This is the fast-path drawing & emission for HW TCL. */ +boolean r300_draw_range_elements(struct pipe_context* pipe, + struct pipe_buffer* indexBuffer, + unsigned indexSize, + unsigned minIndex, + unsigned maxIndex, + unsigned mode, + unsigned start, + unsigned count) +{ + struct r300_context* r300 = r300_context(pipe); - if (indexBuffer) { - indices = (short*)pipe_buffer_map(pipe->screen, indexBuffer, - PIPE_BUFFER_USAGE_CPU_READ); + r300_update_derived_state(r300); - /* Set the starting point. */ - indices += start; + setup_vertex_buffers(r300); - BEGIN_CS(2 + (count+1)/2); - OUT_CS_PKT3(R300_PACKET3_3D_DRAW_INDX_2, (count + 1)/2); - OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (count << 16) | prim); - for (i = 0; i < count - 1; i += 2) { - OUT_CS(indices[i + 1] << 16 | indices[i]); - } - if (count % 2) { - OUT_CS(indices[count - 1]); - } - END_CS; - } else { - BEGIN_CS(2); - OUT_CS_PKT3(R300_PACKET3_3D_DRAW_VBUF_2, 0); - OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (count << 16) | - prim); - END_CS; - } + setup_vertex_attributes(r300); + + setup_index_buffer(r300, indexBuffer, indexSize); + + r300->hw_prim = r300_translate_primitive(mode); + + r300_emit_dirty_state(r300); + + r300_emit_aos(r300, 0); + + r300_emit_draw_elements(r300, indexBuffer, indexSize, minIndex, maxIndex, + start, count); return TRUE; } @@ -178,7 +145,23 @@ boolean r300_draw_elements(struct pipe_context* pipe, boolean r300_draw_arrays(struct pipe_context* pipe, unsigned mode, unsigned start, unsigned count) { - return pipe->draw_elements(pipe, NULL, 0, mode, start, count); + struct r300_context* r300 = r300_context(pipe); + + r300_update_derived_state(r300); + + setup_vertex_buffers(r300); + + setup_vertex_attributes(r300); + + r300->hw_prim = r300_translate_primitive(mode); + + r300_emit_dirty_state(r300); + + r300_emit_aos(r300, start); + + r300_emit_draw_arrays(r300, count); + + return TRUE; } /**************************************************************************** @@ -196,7 +179,9 @@ boolean r300_swtcl_draw_range_elements(struct pipe_context* pipe, unsigned start, unsigned count) { + assert(0); struct r300_context* r300 = r300_context(pipe); +#if 0 int i; if (!u_trim_pipe_prim(mode, &count)) { @@ -236,7 +221,7 @@ boolean r300_swtcl_draw_range_elements(struct pipe_context* pipe, draw_set_mapped_element_buffer_range(r300->draw, 0, start, start + count - 1, NULL); } - +#endif return TRUE; } diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index bed886fad04..e0b85ab768b 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -666,10 +666,9 @@ static void r300_set_vertex_buffers(struct pipe_context* pipe, { struct r300_context* r300 = r300_context(pipe); - memcpy(r300->vertex_buffers, buffers, + memcpy(r300->vertex_buffer, buffers, sizeof(struct pipe_vertex_buffer) * count); - - r300->vertex_buffer_count = count; + r300->vbuf_count = count; if (r300->draw) { draw_flush(r300->draw); @@ -683,10 +682,10 @@ static void r300_set_vertex_elements(struct pipe_context* pipe, { struct r300_context* r300 = r300_context(pipe); - memcpy(r300->vertex_elements, elements, - sizeof(struct pipe_vertex_element) * count); - - r300->vertex_element_count = count; + memcpy(r300->vertex_element, + elements, + sizeof(struct pipe_vertex_element) * count); + r300->aos_count = count; if (r300->draw) { draw_flush(r300->draw); diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c index 7d000e9e2da..14d7bb094c5 100644 --- a/src/gallium/drivers/r300/r300_state_derived.c +++ b/src/gallium/drivers/r300/r300_state_derived.c @@ -63,7 +63,7 @@ int r300_shader_key_compare(void* key1, void* key2) { /* Set up the vs_tab and routes. */ static void r300_vs_tab_routes(struct r300_context* r300, - struct r300_vertex_format* vformat) + struct r300_vertex_info* vformat) { struct r300_screen* r300screen = r300_screen(r300->context.screen); struct vertex_info* vinfo = &vformat->vinfo; @@ -219,7 +219,7 @@ static void r300_vs_tab_routes(struct r300_context* r300, /* Update the PSC tables. */ static void r300_vertex_psc(struct r300_context* r300, - struct r300_vertex_format* vformat) + struct r300_vertex_info* vformat) { struct r300_screen* r300screen = r300_screen(r300->context.screen); struct vertex_info* vinfo = &vformat->vinfo; @@ -282,7 +282,7 @@ static void r300_vertex_psc(struct r300_context* r300, /* Set up the mappings from GB to US, for RS block. */ static void r300_update_fs_tab(struct r300_context* r300, - struct r300_vertex_format* vformat) + struct r300_vertex_info* vformat) { struct tgsi_shader_info* info = &r300->fs->info; int i, cols = 0, texs = 0, cols_emitted = 0; @@ -455,13 +455,13 @@ static void r300_update_rs_block(struct r300_context* r300, /* Update the vertex format. */ static void r300_update_derived_shader_state(struct r300_context* r300) { - struct r300_shader_key* key; - struct r300_vertex_format* vformat; + struct r300_vertex_info* vformat; struct r300_rs_block* rs_block; - struct r300_shader_derived_value* value; int i; /* + struct r300_shader_key* key; + struct r300_shader_derived_value* value; key = CALLOC_STRUCT(r300_shader_key); key->vs = r300->vs; key->fs = r300->fs; @@ -486,7 +486,7 @@ static void r300_update_derived_shader_state(struct r300_context* r300) } */ /* XXX This will be refactored ASAP. */ - vformat = CALLOC_STRUCT(r300_vertex_format); + vformat = CALLOC_STRUCT(r300_vertex_info); rs_block = CALLOC_STRUCT(r300_rs_block); for (i = 0; i < 16; i++) { diff --git a/src/gallium/drivers/r300/r300_vbo.c b/src/gallium/drivers/r300/r300_vbo.c new file mode 100644 index 00000000000..e0326412865 --- /dev/null +++ b/src/gallium/drivers/r300/r300_vbo.c @@ -0,0 +1,226 @@ +/* + * Copyright 2009 Maciej Cencora + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#include "r300_vbo.h" + +#include "pipe/p_format.h" + +#include "r300_cs.h" +#include "r300_context.h" +#include "r300_reg.h" +#include "r300_winsys.h" + +static void translate_vertex_format(enum pipe_format format, + unsigned nr_comps, + unsigned component_size, + unsigned dst_loc, + uint32_t *hw_fmt1, + uint32_t *hw_fmt2) +{ + uint32_t fmt1 = 0; + + switch (pf_type(format)) + { + case PIPE_FORMAT_TYPE_FLOAT: + assert(component_size == 4); + fmt1 = R300_DATA_TYPE_FLOAT_1 + nr_comps - 1; + break; + case PIPE_FORMAT_TYPE_UNORM: + case PIPE_FORMAT_TYPE_SNORM: + case PIPE_FORMAT_TYPE_USCALED: + case PIPE_FORMAT_TYPE_SSCALED: + if (component_size == 1) + { + assert(nr_comps == 4); + fmt1 = R300_DATA_TYPE_BYTE; + } + else if (component_size == 2) + { + if (nr_comps == 2) + fmt1 = R300_DATA_TYPE_SHORT_2; + else if (nr_comps == 4) + fmt1 = R300_DATA_TYPE_SHORT_4; + else + assert(0); + } + else + { + assert(0); + } + + if (pf_type(format) == PIPE_FORMAT_TYPE_SNORM) + { + fmt1 |= R300_SIGNED; + } + else if (pf_type(format) == PIPE_FORMAT_TYPE_SSCALED) + { + fmt1 |= R300_SIGNED; + fmt1 |= R300_NORMALIZE; + } + else if (pf_type(format) == PIPE_FORMAT_TYPE_USCALED) + { + fmt1 |= R300_NORMALIZE; + } + break; + default: + assert(0); + break; + } + + *hw_fmt1 = fmt1 | (dst_loc << R300_DST_VEC_LOC_SHIFT); + *hw_fmt2 = (pf_swizzle_x(format) << R300_SWIZZLE_SELECT_X_SHIFT) | + (pf_swizzle_y(format) << R300_SWIZZLE_SELECT_Y_SHIFT) | + (pf_swizzle_z(format) << R300_SWIZZLE_SELECT_Z_SHIFT) | + (pf_swizzle_w(format) << R300_SWIZZLE_SELECT_W_SHIFT) | + (0xf << R300_WRITE_ENA_SHIFT); +} + +static INLINE void setup_vertex_attribute(struct r300_vertex_info *vinfo, + struct pipe_vertex_element *vert_elem, + unsigned attr_num) +{ + uint32_t hw_fmt1, hw_fmt2; + translate_vertex_format(vert_elem->src_format, + vert_elem->nr_components, + pf_size_x(vert_elem->src_format), + attr_num, + &hw_fmt1, + &hw_fmt2); + + if (attr_num % 2 == 0) + { + vinfo->vap_prog_stream_cntl[attr_num >> 1] = hw_fmt1; + vinfo->vap_prog_stream_cntl_ext[attr_num >> 1] = hw_fmt2; + } + else + { + vinfo->vap_prog_stream_cntl[attr_num >> 1] |= hw_fmt1 << 16; + vinfo->vap_prog_stream_cntl_ext[attr_num >> 1] |= hw_fmt2 << 16; + } +} + +static void finish_vertex_attribs_setup(struct r300_vertex_info *vinfo, + unsigned attribs_num) +{ + uint32_t last_vec_bit = (attribs_num % 2 == 0) ? (R300_LAST_VEC << 16) : R300_LAST_VEC; + + assert(attribs_num > 0 && attribs_num <= 16); + vinfo->vap_prog_stream_cntl[(attribs_num - 1) >> 1] |= last_vec_bit; +} + +void setup_vertex_attributes(struct r300_context *r300) +{ + for (int i=0; iaos_count; i++) + { + struct pipe_vertex_element *vert_elem = &r300->vertex_element[i]; + + setup_vertex_attribute(r300->vertex_info, vert_elem, i); + } + + finish_vertex_attribs_setup(r300->vertex_info, r300->aos_count); +} + +static void setup_vertex_array(struct r300_context *r300, struct pipe_vertex_element *element) +{ +} + +static void finish_vertex_arrays_setup(struct r300_context *r300) +{ +} + +static bool format_is_supported(enum pipe_format format, int nr_components) +{ + if (pf_layout(format) != PIPE_FORMAT_LAYOUT_RGBAZS) + return false; + + if ((pf_size_x(format) != pf_size_y(format)) || + (pf_size_x(format) != pf_size_z(format)) || + (pf_size_x(format) != pf_size_w(format))) + return false; + + /* Following should be supported as long as stride is 4 bytes aligned */ + if (pf_size_x(format) != 1 && nr_components != 4) + return false; + + if (pf_size_x(format) != 2 && !(nr_components == 2 || nr_components == 4)) + return false; + + if (pf_size_x(format) == 3 || pf_size_x(format) > 4) + return false; + + return true; +} + +static INLINE int get_buffer_offset(struct r300_context *r300, + unsigned int buf_nr, + unsigned int elem_offset) +{ + return r300->vertex_buffer[buf_nr].buffer_offset + elem_offset; +} + +/** + */ +static void setup_vertex_buffers(struct r300_context *r300) +{ + for (int i=0; iaos_count; i++) + { + struct pipe_vertex_element *vert_elem = &r300->vertex_element[i]; + if (!format_is_supported(vert_elem->src_format, vert_elem->nr_components)) + { + assert(0); + /* use translate module to convert the data */ + /* + struct pipe_buffer *buf; + const unsigned int max_index = r300->vertex_buffers[vert_elem->vertex_buffer_index].max_index; + buf = pipe_buffer_create(r300->context.screen, 4, usage, vert_elem->nr_components * max_index * sizeof(float)); + */ + } + + if (get_buffer_offset(r300, vert_elem->vertex_buffer_index, vert_elem->src_offset) % 4 != 0) + { + /* need to align buffer */ + assert(0); + } + setup_vertex_array(r300, vert_elem); + } + + finish_vertex_arrays_setup(r300); +} + +void setup_index_buffer(struct r300_context *r300, + struct pipe_buffer* indexBuffer, + unsigned indexSize) +{ + assert(indexSize = 2); + + if (!r300->winsys->add_buffer(r300->winsys, indexBuffer, RADEON_GEM_DOMAIN_GTT, 0)) + { + assert(0); + } + + if (!r300->winsys->validate(r300->winsys)) + { + assert(0); + } +} + diff --git a/src/gallium/drivers/r300/r300_vbo.h b/src/gallium/drivers/r300/r300_vbo.h new file mode 100644 index 00000000000..7afa75899cf --- /dev/null +++ b/src/gallium/drivers/r300/r300_vbo.h @@ -0,0 +1,36 @@ +/* + * Copyright 2009 Maciej Cencora + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * on the rights to use, copy, modify, merge, publish, distribute, sub + * license, and/or sell copies of the Software, and to permit persons to whom + * the Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice (including the next + * paragraph) shall be included in all copies or substantial portions of the + * Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL + * THE AUTHOR(S) AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM, + * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR + * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE + * USE OR OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef R300_VBO_H +#define R300_VBO_H + +struct r300_context; +struct pipe_buffer; + +void setup_vertex_attributes(struct r300_context *r300); + +void setup_index_buffer(struct r300_context *r300, + struct pipe_buffer* indexBuffer, + unsigned indexSize); + +#endif -- cgit v1.2.3 From 96b729f926fafeca6479eed0933bc4275fb7843b Mon Sep 17 00:00:00 2001 From: Corbin Simpson Date: Sat, 7 Nov 2009 09:47:01 -0800 Subject: r300g: Don't pass hw_prim around in the context. And some other fixes. --- src/gallium/drivers/r300/r300_context.h | 1 - src/gallium/drivers/r300/r300_emit.c | 58 ------------------------- src/gallium/drivers/r300/r300_emit.h | 10 ----- src/gallium/drivers/r300/r300_render.c | 76 +++++++++++++++++++++++++++++---- src/gallium/drivers/r300/r300_vbo.c | 10 +++-- 5 files changed, 74 insertions(+), 81 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index a6748852d83..8d14c53f492 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -299,7 +299,6 @@ struct r300_context { /* Vertex elements for Gallium. */ struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; int aos_count; - unsigned hw_prim; /* Bitmask of dirty state objects. */ uint32_t dirty_state; diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 92e6ec606ca..ec1d521800c 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -647,64 +647,6 @@ void r300_emit_draw_packet(struct r300_context* r300) END_CS; } #endif -void r300_emit_draw_arrays(struct r300_context *r300, - unsigned count) -{ - CS_LOCALS(r300); - assert(count < 65536); - - BEGIN_CS(4); - OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, count); - OUT_CS_PKT3(R300_PACKET3_3D_DRAW_VBUF_2, 0); - OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (count << 16) | - r300->hw_prim); - END_CS; -} - -void r300_emit_draw_elements(struct r300_context *r300, - struct pipe_buffer* indexBuffer, - unsigned indexSize, - unsigned minIndex, - unsigned maxIndex, - unsigned start, - unsigned count) -{ - CS_LOCALS(r300); - assert(indexSize == 4 || indexSize == 2); - assert(count < 65536); - assert((start * indexSize) % 4 == 0); - - uint32_t size_dwords; - uint32_t skip_dwords = indexSize * start / sizeof(uint32_t); - assert(skip_dwords == 0); - - BEGIN_CS(10); - OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, maxIndex); - OUT_CS_PKT3(R300_PACKET3_3D_DRAW_INDX_2, 0); - if (indexSize == 4) { - size_dwords = count + start; - OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (count << 16) | - R300_VAP_VF_CNTL__INDEX_SIZE_32bit | r300->hw_prim); - } else { - size_dwords = (count + start + 1) / 2; - OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | - (count << 16) | r300->hw_prim); - } - - OUT_CS_PKT3(R300_PACKET3_INDX_BUFFER, 2); - OUT_CS(R300_INDX_BUFFER_ONE_REG_WR | (R300_VAP_PORT_IDX0 >> 2) | - (0 << R300_INDX_BUFFER_SKIP_SHIFT)); - OUT_CS(skip_dwords); - OUT_CS(size_dwords); - cs_winsys->write_cs_reloc(cs_winsys, - indexBuffer, - RADEON_GEM_DOMAIN_GTT, - 0, - 0); - cs_count -= 2; - - END_CS; -} void r300_emit_vertex_format_state(struct r300_context* r300) { diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h index b4fdfecde01..7c83c5166de 100644 --- a/src/gallium/drivers/r300/r300_emit.h +++ b/src/gallium/drivers/r300/r300_emit.h @@ -40,16 +40,6 @@ void r300_emit_blend_color_state(struct r300_context* r300, void r300_emit_clip_state(struct r300_context* r300, struct pipe_clip_state* clip); -void r300_emit_draw_arrays(struct r300_context *r300, unsigned count); - -void r300_emit_draw_elements(struct r300_context *r300, - struct pipe_buffer* indexBuffer, - unsigned indexSize, - unsigned minIndex, - unsigned maxIndex, - unsigned start, - unsigned count); - void r300_emit_dsa_state(struct r300_context* r300, struct r300_dsa_state* dsa); diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c index cbda30227da..6f7c6453342 100644 --- a/src/gallium/drivers/r300/r300_render.c +++ b/src/gallium/drivers/r300/r300_render.c @@ -31,13 +31,13 @@ #include "util/u_memory.h" #include "util/u_prim.h" -#include "r300_vbo.h" #include "r300_cs.h" #include "r300_context.h" #include "r300_emit.h" #include "r300_reg.h" #include "r300_render.h" #include "r300_state_derived.h" +#include "r300_vbo.h" /* r300_render: Vertex and index buffer primitive emission. */ #define R300_MAX_VBO_SIZE (1024 * 1024) @@ -70,6 +70,70 @@ uint32_t r300_translate_primitive(unsigned prim) } } +static void r300_emit_draw_arrays(struct r300_context *r300, + unsigned mode, + unsigned count) +{ + CS_LOCALS(r300); + assert(count < 65536); + + BEGIN_CS(4); + OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, count); + OUT_CS_PKT3(R300_PACKET3_3D_DRAW_VBUF_2, 0); + OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (count << 16) | + r300_translate_primitive(mode)); + END_CS; +} + +static void r300_emit_draw_elements(struct r300_context *r300, + struct pipe_buffer* indexBuffer, + unsigned indexSize, + unsigned minIndex, + unsigned maxIndex, + unsigned mode, + unsigned start, + unsigned count) +{ + CS_LOCALS(r300); + assert(indexSize == 4 || indexSize == 2); + assert(count < 65536); + assert((start * indexSize) % 4 == 0); + + uint32_t size_dwords; + uint32_t skip_dwords = indexSize * start / sizeof(uint32_t); + assert(skip_dwords == 0); + + BEGIN_CS(10); + OUT_CS_REG(R300_VAP_VF_MAX_VTX_INDX, maxIndex); + OUT_CS_PKT3(R300_PACKET3_3D_DRAW_INDX_2, 0); + if (indexSize == 4) { + size_dwords = count + start; + OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (count << 16) | + R300_VAP_VF_CNTL__INDEX_SIZE_32bit | + r300_translate_primitive(mode)); + } else { + size_dwords = (count + start + 1) / 2; + OUT_CS(R300_VAP_VF_CNTL__PRIM_WALK_INDICES | (count << 16) | + r300_translate_primitive(mode)); + } + + OUT_CS_PKT3(R300_PACKET3_INDX_BUFFER, 2); + OUT_CS(R300_INDX_BUFFER_ONE_REG_WR | (R300_VAP_PORT_IDX0 >> 2) | + (0 << R300_INDX_BUFFER_SKIP_SHIFT)); + OUT_CS(skip_dwords); + OUT_CS(size_dwords); + /* XXX hax */ + cs_winsys->write_cs_reloc(cs_winsys, + indexBuffer, + RADEON_GEM_DOMAIN_GTT, + 0, + 0); + cs_count -= 2; + + END_CS; +} + + static boolean setup_vertex_buffers(struct r300_context *r300) { unsigned vbuf_count = r300->aos_count; @@ -123,14 +187,12 @@ boolean r300_draw_range_elements(struct pipe_context* pipe, setup_index_buffer(r300, indexBuffer, indexSize); - r300->hw_prim = r300_translate_primitive(mode); - r300_emit_dirty_state(r300); r300_emit_aos(r300, 0); r300_emit_draw_elements(r300, indexBuffer, indexSize, minIndex, maxIndex, - start, count); + mode, start, count); return TRUE; } @@ -159,13 +221,11 @@ boolean r300_draw_arrays(struct pipe_context* pipe, unsigned mode, setup_vertex_attributes(r300); - r300->hw_prim = r300_translate_primitive(mode); - r300_emit_dirty_state(r300); r300_emit_aos(r300, start); - r300_emit_draw_arrays(r300, count); + r300_emit_draw_arrays(r300, mode, count); return TRUE; } @@ -186,8 +246,8 @@ boolean r300_swtcl_draw_range_elements(struct pipe_context* pipe, unsigned count) { assert(0); - struct r300_context* r300 = r300_context(pipe); #if 0 + struct r300_context* r300 = r300_context(pipe); int i; if (!u_trim_pipe_prim(mode, &count)) { diff --git a/src/gallium/drivers/r300/r300_vbo.c b/src/gallium/drivers/r300/r300_vbo.c index 37b5c9224fc..ab6f5c59426 100644 --- a/src/gallium/drivers/r300/r300_vbo.c +++ b/src/gallium/drivers/r300/r300_vbo.c @@ -56,7 +56,8 @@ static INLINE void setup_vertex_attribute(struct r300_vertex_info *vinfo, static void finish_vertex_attribs_setup(struct r300_vertex_info *vinfo, unsigned attribs_num) { - uint32_t last_vec_bit = (attribs_num % 2 == 0) ? (R300_LAST_VEC << 16) : R300_LAST_VEC; + uint32_t last_vec_bit = (attribs_num % 2 == 0) ? + (R300_LAST_VEC << 16) : R300_LAST_VEC; assert(attribs_num > 0 && attribs_num <= 16); vinfo->vap_prog_stream_cntl[(attribs_num - 1) >> 1] |= last_vec_bit; @@ -64,10 +65,11 @@ static void finish_vertex_attribs_setup(struct r300_vertex_info *vinfo, void setup_vertex_attributes(struct r300_context *r300) { - for (int i=0; iaos_count; i++) - { - struct pipe_vertex_element *vert_elem = &r300->vertex_element[i]; + struct pipe_vertex_element *vert_elem; + int i; + for (i = 0; i < r300->aos_count; i++) { + vert_elem = &r300->vertex_element[i]; setup_vertex_attribute(r300->vertex_info, vert_elem, i); } -- cgit v1.2.3 From b6f93e2607f1bbc5b2f478f0a57d7786dd7d73a5 Mon Sep 17 00:00:00 2001 From: Corbin Simpson Date: Sun, 8 Nov 2009 11:32:32 -0800 Subject: r300g: Enable PSC/RS dump with new debugging flags. --- src/gallium/drivers/r300/r300_emit.c | 28 ++++++++++++++++------------ 1 file changed, 16 insertions(+), 12 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index ec1d521800c..b3d9db676a6 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -510,6 +510,8 @@ void r300_emit_rs_block_state(struct r300_context* r300, struct r300_screen* r300screen = r300_screen(r300->context.screen); CS_LOCALS(r300); + DBG(r300, DBG_DRAW, "r300: RS emit:\n"); + BEGIN_CS(21); if (r300screen->caps->is_r500) { OUT_CS_REG_SEQ(R500_RS_IP_0, 8); @@ -518,7 +520,7 @@ void r300_emit_rs_block_state(struct r300_context* r300, } for (i = 0; i < 8; i++) { OUT_CS(rs->ip[i]); - /* debug_printf("ip %d: 0x%08x\n", i, rs->ip[i]); */ + DBG(r300, DBG_DRAW, " : ip %d: 0x%08x\n", i, rs->ip[i]); } OUT_CS_REG_SEQ(R300_RS_COUNT, 2); @@ -532,11 +534,11 @@ void r300_emit_rs_block_state(struct r300_context* r300, } for (i = 0; i < 8; i++) { OUT_CS(rs->inst[i]); - /* debug_printf("inst %d: 0x%08x\n", i, rs->inst[i]); */ + DBG(r300, DBG_DRAW, " : inst %d: 0x%08x\n", i, rs->inst[i]); } - /* debug_printf("count: 0x%08x inst_count: 0x%08x\n", rs->count, - * rs->inst_count); */ + DBG(r300, DBG_DRAW, " : count: 0x%08x inst_count: 0x%08x\n", + rs->count, rs->inst_count); END_CS; } @@ -653,6 +655,8 @@ void r300_emit_vertex_format_state(struct r300_context* r300) int i; CS_LOCALS(r300); + DBG(r300, DBG_DRAW, "r300: VAP/PSC emit:\n"); + BEGIN_CS(26); OUT_CS_REG(R300_VAP_VTX_SIZE, r300->vertex_info->vinfo.size); @@ -662,22 +666,22 @@ void r300_emit_vertex_format_state(struct r300_context* r300) OUT_CS_REG_SEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2); OUT_CS(r300->vertex_info->vinfo.hwfmt[2]); OUT_CS(r300->vertex_info->vinfo.hwfmt[3]); - /* for (i = 0; i < 4; i++) { - * debug_printf("hwfmt%d: 0x%08x\n", i, - * r300->vertex_info->vinfo.hwfmt[i]); - * } */ + for (i = 0; i < 4; i++) { + DBG(r300, DBG_DRAW, " : hwfmt%d: 0x%08x\n", i, + r300->vertex_info->vinfo.hwfmt[i]); + } OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_0, 8); for (i = 0; i < 8; i++) { OUT_CS(r300->vertex_info->vap_prog_stream_cntl[i]); - /* debug_printf("prog_stream_cntl%d: 0x%08x\n", i, - * r300->vertex_info->vap_prog_stream_cntl[i]); */ + DBG(r300, DBG_DRAW, " : prog_stream_cntl%d: 0x%08x\n", i, + r300->vertex_info->vap_prog_stream_cntl[i]); } OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_EXT_0, 8); for (i = 0; i < 8; i++) { OUT_CS(r300->vertex_info->vap_prog_stream_cntl_ext[i]); - /* debug_printf("prog_stream_cntl_ext%d: 0x%08x\n", i, - * r300->vertex_info->vap_prog_stream_cntl_ext[i]); */ + DBG(r300, DBG_DRAW, " : prog_stream_cntl_ext%d: 0x%08x\n", i, + r300->vertex_info->vap_prog_stream_cntl_ext[i]); } END_CS; } -- cgit v1.2.3 From 11d9edf4c9c75d5a41fb0a1757441ad315330bea Mon Sep 17 00:00:00 2001 From: Corbin Simpson Date: Sun, 8 Nov 2009 11:45:57 -0800 Subject: r300g: Unify context names for counts. From the SW TCL fixups. --- src/gallium/drivers/r300/r300_context.c | 2 +- src/gallium/drivers/r300/r300_context.h | 4 ++-- src/gallium/drivers/r300/r300_emit.c | 16 ++++++++++------ src/gallium/drivers/r300/r300_render.c | 2 +- src/gallium/drivers/r300/r300_state.c | 4 ++-- src/gallium/drivers/r300/r300_vbo.c | 5 +++-- 6 files changed, 19 insertions(+), 14 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c index 43d7ff3ed36..ae23329b83f 100644 --- a/src/gallium/drivers/r300/r300_context.c +++ b/src/gallium/drivers/r300/r300_context.c @@ -158,6 +158,7 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen, /* Open up the OQ BO. */ r300->oqbo = screen->buffer_create(screen, 4096, PIPE_BUFFER_USAGE_VERTEX, 4096); + make_empty_list(&r300->query_list); r300_init_flush_functions(r300); @@ -172,6 +173,5 @@ struct pipe_context* r300_create_context(struct pipe_screen* screen, r300->winsys->set_flush_cb(r300->winsys, r300_flush_cb, r300); r300->dirty_state = R300_NEW_KITCHEN_SINK; r300->dirty_hw++; - make_empty_list(&r300->query_list); return &r300->context; } diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index 8d14c53f492..f954ba7f9aa 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -295,10 +295,10 @@ struct r300_context { /* Vertex buffers for Gallium. */ struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS]; - int vbuf_count; + int vertex_buffer_count; /* Vertex elements for Gallium. */ struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS]; - int aos_count; + int vertex_element_count; /* Bitmask of dirty state objects. */ uint32_t dirty_state; diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index b3d9db676a6..eeb97a2d370 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -584,17 +584,20 @@ void r300_emit_texture(struct r300_context* r300, END_CS; } +/* XXX I can't read this and that's not good */ void r300_emit_aos(struct r300_context* r300, unsigned offset) { struct pipe_vertex_buffer *vbuf = r300->vertex_buffer; struct pipe_vertex_element *velem = r300->vertex_element; CS_LOCALS(r300); int i; - unsigned packet_size = (r300->aos_count * 3 + 1) / 2; - BEGIN_CS(2 + packet_size + r300->aos_count * 2); + unsigned aos_count = r300->vertex_element_count; + + unsigned packet_size = (aos_count * 3 + 1) / 2; + BEGIN_CS(2 + packet_size + aos_count * 2); OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, packet_size); - OUT_CS(r300->aos_count); - for (i = 0; i < r300->aos_count - 1; i += 2) { + OUT_CS(aos_count); + for (i = 0; i < aos_count - 1; i += 2) { int buf_num1 = velem[i].vertex_buffer_index; int buf_num2 = velem[i+1].vertex_buffer_index; assert(vbuf[buf_num1].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0); @@ -606,7 +609,7 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset) OUT_CS(vbuf[buf_num2].buffer_offset + velem[i+1].src_offset + offset * vbuf[buf_num2].stride); } - if (r300->aos_count & 1) { + if (aos_count & 1) { int buf_num = velem[i].vertex_buffer_index; assert(vbuf[buf_num].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0); OUT_CS((pf_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num].stride << 6)); @@ -614,7 +617,8 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset) offset * vbuf[buf_num].stride); } - for (i = 0; i < r300->aos_count; i++) { + /* XXX bare CS reloc */ + for (i = 0; i < aos_count; i++) { cs_winsys->write_cs_reloc(cs_winsys, vbuf[velem[i].vertex_buffer_index].buffer, RADEON_GEM_DOMAIN_GTT, diff --git a/src/gallium/drivers/r300/r300_render.c b/src/gallium/drivers/r300/r300_render.c index fa057324f85..1ff3e64b449 100644 --- a/src/gallium/drivers/r300/r300_render.c +++ b/src/gallium/drivers/r300/r300_render.c @@ -140,7 +140,7 @@ static boolean r300_setup_vertex_buffers(struct r300_context *r300) struct pipe_vertex_element *velem = r300->vertex_element; validate: - for (int i = 0; i < r300->aos_count; i++) { + for (int i = 0; i < r300->vertex_element_count; i++) { if (!r300->winsys->add_buffer(r300->winsys, vbuf[velem[i].vertex_buffer_index].buffer, RADEON_GEM_DOMAIN_GTT, 0)) { diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index e0b85ab768b..d1eced61db1 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -668,7 +668,7 @@ static void r300_set_vertex_buffers(struct pipe_context* pipe, memcpy(r300->vertex_buffer, buffers, sizeof(struct pipe_vertex_buffer) * count); - r300->vbuf_count = count; + r300->vertex_buffer_count = count; if (r300->draw) { draw_flush(r300->draw); @@ -685,7 +685,7 @@ static void r300_set_vertex_elements(struct pipe_context* pipe, memcpy(r300->vertex_element, elements, sizeof(struct pipe_vertex_element) * count); - r300->aos_count = count; + r300->vertex_element_count = count; if (r300->draw) { draw_flush(r300->draw); diff --git a/src/gallium/drivers/r300/r300_vbo.c b/src/gallium/drivers/r300/r300_vbo.c index 5ad6b9c2157..a6a159667a3 100644 --- a/src/gallium/drivers/r300/r300_vbo.c +++ b/src/gallium/drivers/r300/r300_vbo.c @@ -71,12 +71,13 @@ void setup_vertex_attributes(struct r300_context *r300) struct pipe_vertex_element *vert_elem; int i; - for (i = 0; i < r300->aos_count; i++) { + for (i = 0; i < r300->vertex_element_count; i++) { vert_elem = &r300->vertex_element[i]; setup_vertex_attribute(r300->vertex_info, vert_elem, i); } - finish_vertex_attribs_setup(r300->vertex_info, r300->aos_count); + finish_vertex_attribs_setup(r300->vertex_info, + r300->vertex_element_count); } static INLINE int get_buffer_offset(struct r300_context *r300, -- cgit v1.2.3 From beea241374a91b8aab81db175b28e98c2b4835d9 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Thu, 19 Nov 2009 01:35:08 +0100 Subject: r300g: set better values in the R300_VAP_CNTL register --- src/gallium/drivers/r300/r300_emit.c | 33 ++++++++++++++++++++++++++++----- 1 file changed, 28 insertions(+), 5 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index eeb97a2d370..2a8d32242bc 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -690,12 +690,35 @@ void r300_emit_vertex_format_state(struct r300_context* r300) END_CS; } +/* XXX This should probably go to util ... */ +/* Return the number of bits set in the given number. */ +static unsigned bitcount(unsigned n) +{ + unsigned bits; + for (bits = 0; n > 0; n = n >> 1) { + bits += n & 1; + } + return bits; +} + +/* XXX ... and this one too. */ +#define MIN3(x, y, z) MIN2(MIN2(x, y), z) + void r300_emit_vertex_program_code(struct r300_context* r300, struct r300_vertex_program_code* code) { int i; struct r300_screen* r300screen = r300_screen(r300->context.screen); unsigned instruction_count = code->length / 4; + + int vtx_mem_size = r300screen->caps->is_r500 ? 128 : 72; + int input_count = MAX2(bitcount(code->InputsRead), 1); + int output_count = MAX2(bitcount(code->OutputsWritten), 1); + int temp_count = MAX2(code->num_temporaries, 1); + int pvs_num_slots = MIN3(vtx_mem_size / input_count, + vtx_mem_size / output_count, 10); + int pvs_num_controllers = MIN2(6, vtx_mem_size / temp_count); + CS_LOCALS(r300); if (!r300screen->caps->has_tcl) { @@ -708,8 +731,7 @@ void r300_emit_vertex_program_code(struct r300_context* r300, /* R300_VAP_PVS_CODE_CNTL_0 * R300_VAP_PVS_CONST_CNTL * R300_VAP_PVS_CODE_CNTL_1 - * See the r5xx docs for instructions on how to use these. - * XXX these could be optimized to select better values... */ + * See the r5xx docs for instructions on how to use these. */ OUT_CS_REG_SEQ(R300_VAP_PVS_CODE_CNTL_0, 3); OUT_CS(R300_PVS_FIRST_INST(0) | R300_PVS_XYZW_VALID_INST(instruction_count - 1) | @@ -722,10 +744,11 @@ void r300_emit_vertex_program_code(struct r300_context* r300, for (i = 0; i < code->length; i++) OUT_CS(code->body.d[i]); - OUT_CS_REG(R300_VAP_CNTL, R300_PVS_NUM_SLOTS(10) | - R300_PVS_NUM_CNTLRS(5) | + OUT_CS_REG(R300_VAP_CNTL, R300_PVS_NUM_SLOTS(pvs_num_slots) | + R300_PVS_NUM_CNTLRS(pvs_num_controllers) | R300_PVS_NUM_FPUS(r300screen->caps->num_vert_fpus) | - R300_PVS_VF_MAX_VTX_NUM(12)); + R300_PVS_VF_MAX_VTX_NUM(12) | + (r300screen->caps->is_r500 ? R500_TCL_STATE_OPTIMIZATION : 0)); END_CS; } -- cgit v1.2.3 From 6a3eb1f91b4ccd4ee7ac6b91505e0dfa476922d4 Mon Sep 17 00:00:00 2001 From: Corbin Simpson Date: Fri, 20 Nov 2009 14:10:45 -0800 Subject: r300g: Use MAX3 and MIN3. --- src/gallium/drivers/r300/r300_emit.c | 3 --- src/gallium/drivers/r300/r300_state_derived.c | 2 +- 2 files changed, 1 insertion(+), 4 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 2a8d32242bc..4cd5074379f 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -701,9 +701,6 @@ static unsigned bitcount(unsigned n) return bits; } -/* XXX ... and this one too. */ -#define MIN3(x, y, z) MIN2(MIN2(x, y), z) - void r300_emit_vertex_program_code(struct r300_context* r300, struct r300_vertex_program_code* code) { diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c index 7166694edf4..b4d0eeaf8c5 100644 --- a/src/gallium/drivers/r300/r300_state_derived.c +++ b/src/gallium/drivers/r300/r300_state_derived.c @@ -458,7 +458,7 @@ static void r300_update_rs_block(struct r300_context* r300, rs->count = (rs_tex_comp) | (col_count << R300_IC_COUNT_SHIFT) | R300_HIRES_EN; - rs->inst_count = MAX2(MAX2(col_count - 1, tex_count - 1), 0); + rs->inst_count = MAX3(col_count - 1, tex_count - 1, 0); } /* Update the vertex format. */ -- cgit v1.2.3 From 06ec216d191e160494dd0a922ab0395418a78402 Mon Sep 17 00:00:00 2001 From: Corbin Simpson Date: Fri, 20 Nov 2009 14:10:59 -0800 Subject: r300g: Clean up bitcounting function. I didn't see this in u_math; surely somebody else has this wheel reinvented elsewhere. --- src/gallium/drivers/r300/r300_emit.c | 15 ++++++++++----- 1 file changed, 10 insertions(+), 5 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 4cd5074379f..c50c989f01a 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -690,14 +690,19 @@ void r300_emit_vertex_format_state(struct r300_context* r300) END_CS; } -/* XXX This should probably go to util ... */ +/* XXX This should go to util ... */ /* Return the number of bits set in the given number. */ static unsigned bitcount(unsigned n) { - unsigned bits; - for (bits = 0; n > 0; n = n >> 1) { - bits += n & 1; + unsigned bits = 0; + + while (n) { + if (n & 1) { + bits++; + } + n >>= 1; } + return bits; } @@ -714,7 +719,7 @@ void r300_emit_vertex_program_code(struct r300_context* r300, int temp_count = MAX2(code->num_temporaries, 1); int pvs_num_slots = MIN3(vtx_mem_size / input_count, vtx_mem_size / output_count, 10); - int pvs_num_controllers = MIN2(6, vtx_mem_size / temp_count); + int pvs_num_controllers = MIN2(vtx_mem_size / temp_count, 6); CS_LOCALS(r300); -- cgit v1.2.3 From f4041b37e2d305cff0a97eb836250e9f8b1840a8 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Sat, 14 Nov 2009 22:14:42 +0100 Subject: r300g: fix rectangle textures on r3xx Adapted from Maciej Cencora's patch. --- src/gallium/drivers/r300/r300_emit.c | 22 ++++++++++++++++++++-- src/gallium/drivers/r300/r300_state.c | 8 ++++++++ 2 files changed, 28 insertions(+), 2 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index c50c989f01a..0bdf58202f0 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -129,7 +129,9 @@ static const float * get_shader_constant( struct rc_constant * constant, struct r300_constant_buffer * externals) { - static const float zero[4] = { 0.0, 0.0, 0.0, 0.0 }; + static float vec[4] = { 0.0, 0.0, 0.0, 0.0 }; + struct pipe_texture *tex; + switch(constant->Type) { case RC_CONSTANT_EXTERNAL: return externals->constants[constant->u.External]; @@ -137,10 +139,26 @@ static const float * get_shader_constant( case RC_CONSTANT_IMMEDIATE: return constant->u.Immediate; + case RC_CONSTANT_STATE: + switch (constant->u.State[0]) + { + /* R3xx-specific */ + case RC_STATE_R300_TEXRECT_FACTOR: + tex = &r300->textures[constant->u.State[1]]->tex; + vec[0] = 1.0 / tex->width[0]; + vec[1] = 1.0 / tex->height[0]; + vec[2] = vec[3] = 1; + break; + + default: + assert(0); + } + return vec; + default: debug_printf("r300: Implementation error: Unhandled constant type %i\n", constant->Type); - return zero; + return vec; } } diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index d1eced61db1..00f10ffd73c 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -571,6 +571,7 @@ static void r300_set_sampler_textures(struct pipe_context* pipe, struct pipe_texture** texture) { struct r300_context* r300 = r300_context(pipe); + boolean is_r500 = r300_screen(r300->context.screen)->caps->is_r500; int i; /* XXX magic num */ @@ -585,6 +586,13 @@ static void r300_set_sampler_textures(struct pipe_context* pipe, pipe_texture_reference((struct pipe_texture**)&r300->textures[i], texture[i]); r300->dirty_state |= (R300_NEW_TEXTURE << i); + + /* R300-specific - set the texrect factor in a fragment shader */ + if (!is_r500 && r300->textures[i]->is_npot) { + /* XXX It would be nice to re-emit just 1 constant, + * XXX not all of them */ + r300->dirty_state |= R300_NEW_FRAGMENT_SHADER_CONSTANTS; + } } } -- cgit v1.2.3 From 6a95996abb33a040f957ffedf3824afcc98a9e71 Mon Sep 17 00:00:00 2001 From: Corbin Simpson Date: Fri, 20 Nov 2009 14:55:22 -0800 Subject: r300g: Texrect factor cleanup. (0, 0, 0, 1) is a much saner default value, and texrect factors only need to be (1/s, 1/t, 0, 1). --- src/gallium/drivers/r300/r300_emit.c | 24 ++++++++++++++---------- 1 file changed, 14 insertions(+), 10 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 0bdf58202f0..37e75ba0616 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -129,7 +129,7 @@ static const float * get_shader_constant( struct rc_constant * constant, struct r300_constant_buffer * externals) { - static float vec[4] = { 0.0, 0.0, 0.0, 0.0 }; + static float vec[4] = { 0.0, 0.0, 0.0, 1.0 }; struct pipe_texture *tex; switch(constant->Type) { @@ -140,26 +140,30 @@ static const float * get_shader_constant( return constant->u.Immediate; case RC_CONSTANT_STATE: - switch (constant->u.State[0]) - { - /* R3xx-specific */ + switch (constant->u.State[0]) { + /* Factor for converting rectangle coords to + * normalized coords. Should only show up on non-r500. */ case RC_STATE_R300_TEXRECT_FACTOR: tex = &r300->textures[constant->u.State[1]]->tex; vec[0] = 1.0 / tex->width[0]; vec[1] = 1.0 / tex->height[0]; - vec[2] = vec[3] = 1; break; default: - assert(0); + debug_printf("r300: Implementation error: " + "Unknown RC_CONSTANT type %d\n", constant->u.State[0]); } - return vec; + break; default: - debug_printf("r300: Implementation error: Unhandled constant type %i\n", - constant->Type); - return vec; + debug_printf("r300: Implementation error: " + "Unhandled constant type %d\n", constant->Type); } + + /* This should either be (0, 0, 0, 1), which should be a relatively safe + * RGBA or STRQ value, or it could be one of the RC_CONSTANT_STATE + * state factors. */ + return vec; } /* Convert a normal single-precision float into the 7.16 format -- cgit v1.2.3 From 015e7e7724a64d3d9e02e57f6a8eb88a6441f596 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Fri, 20 Nov 2009 05:17:00 +0100 Subject: r300g: emit R300_TEX_ENABLE to indicate there are no textures bound Previously, this reg wasn't emitted at all if texture_count == 0. --- src/gallium/drivers/r300/r300_emit.c | 15 +++++++++++++-- src/gallium/drivers/r300/r300_emit.h | 2 ++ 2 files changed, 15 insertions(+), 2 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 37e75ba0616..6d702c00272 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -837,13 +837,22 @@ void r300_emit_viewport_state(struct r300_context* r300, END_CS; } +void r300_emit_texture_count(struct r300_context* r300) +{ + CS_LOCALS(r300); + + BEGIN_CS(2); + OUT_CS_REG(R300_TX_ENABLE, (1 << r300->texture_count) - 1); + END_CS; + +} + void r300_flush_textures(struct r300_context* r300) { CS_LOCALS(r300); - BEGIN_CS(4); + BEGIN_CS(2); OUT_CS_REG(R300_TX_INVALTAGS, 0); - OUT_CS_REG(R300_TX_ENABLE, (1 << r300->texture_count) - 1); END_CS; } @@ -997,6 +1006,8 @@ validate: /* Samplers and textures are tracked separately but emitted together. */ if (r300->dirty_state & (R300_ANY_NEW_SAMPLERS | R300_ANY_NEW_TEXTURES)) { + r300_emit_texture_count(r300); + for (i = 0; i < MIN2(r300->sampler_count, r300->texture_count); i++) { if (r300->dirty_state & ((R300_NEW_SAMPLER << i) | (R300_NEW_TEXTURE << i))) { diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h index 7c83c5166de..3797d3d332a 100644 --- a/src/gallium/drivers/r300/r300_emit.h +++ b/src/gallium/drivers/r300/r300_emit.h @@ -92,6 +92,8 @@ void r300_emit_vertex_shader(struct r300_context* r300, void r300_emit_viewport_state(struct r300_context* r300, struct r300_viewport_state* viewport); +void r300_emit_texture_count(struct r300_context* r300); + void r300_flush_textures(struct r300_context* r300); /* Emit all dirty state. */ -- cgit v1.2.3 From b7078a88119e248b0196f7446abe029c22f1ee28 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Sat, 14 Nov 2009 23:27:20 +0100 Subject: r300g: add texture lod clamping These now work: piglit/lodclamp piglit/levelclamp --- src/gallium/drivers/r300/r300_context.h | 4 ++++ src/gallium/drivers/r300/r300_emit.c | 11 ++++++++++- src/gallium/drivers/r300/r300_reg.h | 5 +++-- src/gallium/drivers/r300/r300_state.c | 5 +++++ src/gallium/drivers/r300/r300_texture.c | 3 +-- 5 files changed, 23 insertions(+), 5 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index f954ba7f9aa..60ef415caa9 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -92,6 +92,10 @@ struct r300_sampler_state { uint32_t filter0; /* R300_TX_FILTER0: 0x4400 */ uint32_t filter1; /* R300_TX_FILTER1: 0x4440 */ uint32_t border_color; /* R300_TX_BORDER_COLOR: 0x45c0 */ + + /* Min/max LOD must be clamped to [0, last_level], thus + * it's dependent on a currently bound texture */ + unsigned min_lod, max_lod; }; struct r300_scissor_state { diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 6d702c00272..ad7dff36bea 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -583,6 +583,8 @@ void r300_emit_texture(struct r300_context* r300, unsigned offset) { uint32_t filter0 = sampler->filter0; + uint32_t format0 = tex->state.format0; + unsigned min_level, max_level; CS_LOCALS(r300); /* to emulate 1D textures through 2D ones correctly */ @@ -591,13 +593,20 @@ void r300_emit_texture(struct r300_context* r300, filter0 |= R300_TX_WRAP_T(R300_TX_CLAMP_TO_EDGE); } + /* determine min/max levels */ + /* the MAX_MIP level is the largest (finest) one */ + max_level = MIN2(sampler->max_lod, tex->tex.last_level); + min_level = MIN2(sampler->min_lod, max_level); + format0 |= R300_TX_NUM_LEVELS(max_level); + filter0 |= R300_TX_MAX_MIP_LEVEL(min_level); + BEGIN_CS(16); OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), filter0 | (offset << 28)); OUT_CS_REG(R300_TX_FILTER1_0 + (offset * 4), sampler->filter1); OUT_CS_REG(R300_TX_BORDER_COLOR_0 + (offset * 4), sampler->border_color); - OUT_CS_REG(R300_TX_FORMAT0_0 + (offset * 4), tex->state.format0); + OUT_CS_REG(R300_TX_FORMAT0_0 + (offset * 4), format0); OUT_CS_REG(R300_TX_FORMAT1_0 + (offset * 4), tex->state.format1); OUT_CS_REG(R300_TX_FORMAT2_0 + (offset * 4), tex->state.format2); OUT_CS_REG_SEQ(R300_TX_OFFSET_0 + (offset * 4), 1); diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h index 8ca785cb587..66fdada2210 100644 --- a/src/gallium/drivers/r300/r300_reg.h +++ b/src/gallium/drivers/r300/r300_reg.h @@ -1463,6 +1463,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_TX_MIN_FILTER_MIP_NEAREST (1 << 13) # define R300_TX_MIN_FILTER_MIP_LINEAR (2 << 13) # define R300_TX_MIN_FILTER_MIP_MASK (3 << 13) +# define R300_TX_MAX_MIP_LEVEL_SHIFT 17 +# define R300_TX_MAX_MIP_LEVEL_MASK (0xf << 17) # define R300_TX_MAX_ANISO_1_TO_1 (0 << 21) # define R300_TX_MAX_ANISO_2_TO_1 (1 << 21) # define R300_TX_MAX_ANISO_4_TO_1 (2 << 21) @@ -1471,6 +1473,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_TX_MAX_ANISO_MASK (7 << 21) # define R300_TX_WRAP_S(x) ((x) << 0) # define R300_TX_WRAP_T(x) ((x) << 3) +# define R300_TX_MAX_MIP_LEVEL(x) ((x) << 17) #define R300_TX_FILTER1_0 0x4440 # define R300_CHROMA_KEY_MODE_DISABLE 0 @@ -1500,8 +1503,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. # define R300_TX_HEIGHTMASK_MASK (2047 << 11) # define R300_TX_DEPTHMASK_SHIFT 22 # define R300_TX_DEPTHMASK_MASK (0xf << 22) -# define R300_TX_MAX_MIP_LEVEL_SHIFT 26 -# define R300_TX_MAX_MIP_LEVEL_MASK (0xf << 26) # define R300_TX_SIZE_PROJECTED (1 << 30) # define R300_TX_PITCH_EN (1 << 31) # define R300_TX_WIDTH(x) ((x) << 0) diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index 5422a2cc9c8..f2867675f0f 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -523,6 +523,11 @@ static void* state->mag_img_filter, state->min_mip_filter); + /* Unfortunately, r300-r500 don't support floating-point mipmap lods. */ + /* We must pass these to the emit function to clamp them properly. */ + sampler->min_lod = MAX2((unsigned)state->min_lod, 0); + sampler->max_lod = MAX2((unsigned)ceilf(state->max_lod), 0); + lod_bias = CLAMP((int)(state->lod_bias * 32), -(1 << 9), (1 << 9) - 1); sampler->filter1 |= lod_bias << R300_LOD_BIAS_SHIFT; diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c index aea25cf71dd..d13aa8f036b 100644 --- a/src/gallium/drivers/r300/r300_texture.c +++ b/src/gallium/drivers/r300/r300_texture.c @@ -43,8 +43,7 @@ static void r300_setup_texture_state(struct r300_texture* tex, boolean is_r500) state->format2 = (tex->pitch[0] - 1) & 0x1fff; } else { /* power of two textures (3D, mipmaps, and no pitch) */ - state->format0 |= R300_TX_DEPTH(util_logbase2(pt->depth[0]) & 0xf) | - R300_TX_NUM_LEVELS(pt->last_level & 0xf); + state->format0 |= R300_TX_DEPTH(util_logbase2(pt->depth[0]) & 0xf); } state->format1 = r300_translate_texformat(pt->format); -- cgit v1.2.3 From 8d80b5400a1bbf4e959cd8257d11dfe0483e93db Mon Sep 17 00:00:00 2001 From: Brian Paul Date: Mon, 23 Nov 2009 18:06:19 -0700 Subject: r300g: use util_bitcount() --- src/gallium/drivers/r300/r300_emit.c | 19 ++----------------- 1 file changed, 2 insertions(+), 17 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index ad7dff36bea..e6ab8e4af14 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -721,21 +721,6 @@ void r300_emit_vertex_format_state(struct r300_context* r300) END_CS; } -/* XXX This should go to util ... */ -/* Return the number of bits set in the given number. */ -static unsigned bitcount(unsigned n) -{ - unsigned bits = 0; - - while (n) { - if (n & 1) { - bits++; - } - n >>= 1; - } - - return bits; -} void r300_emit_vertex_program_code(struct r300_context* r300, struct r300_vertex_program_code* code) @@ -745,8 +730,8 @@ void r300_emit_vertex_program_code(struct r300_context* r300, unsigned instruction_count = code->length / 4; int vtx_mem_size = r300screen->caps->is_r500 ? 128 : 72; - int input_count = MAX2(bitcount(code->InputsRead), 1); - int output_count = MAX2(bitcount(code->OutputsWritten), 1); + int input_count = MAX2(util_bitcount(code->InputsRead), 1); + int output_count = MAX2(util_bitcount(code->OutputsWritten), 1); int temp_count = MAX2(code->num_temporaries, 1); int pvs_num_slots = MIN3(vtx_mem_size / input_count, vtx_mem_size / output_count, 10); -- cgit v1.2.3 From b748a9f574361273df6b05c06c647ac4fd9b3e41 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Fri, 27 Nov 2009 17:40:24 +0100 Subject: r300g,llvmpipe: fix some more merge problems --- src/gallium/drivers/llvmpipe/lp_texture.c | 2 +- src/gallium/drivers/r300/r300_emit.c | 4 ++-- 2 files changed, 3 insertions(+), 3 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/llvmpipe/lp_texture.c b/src/gallium/drivers/llvmpipe/lp_texture.c index 0a0f31f8a38..65d62fd0723 100644 --- a/src/gallium/drivers/llvmpipe/lp_texture.c +++ b/src/gallium/drivers/llvmpipe/lp_texture.c @@ -169,7 +169,7 @@ llvmpipe_texture_blanket(struct pipe_screen * screen, /* Only supports one type */ if (base->target != PIPE_TEXTURE_2D || base->last_level != 0 || - base->depth[0] != 1) { + base->depth0 != 1) { return NULL; } diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index e6ab8e4af14..98a39390bf9 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -145,8 +145,8 @@ static const float * get_shader_constant( * normalized coords. Should only show up on non-r500. */ case RC_STATE_R300_TEXRECT_FACTOR: tex = &r300->textures[constant->u.State[1]]->tex; - vec[0] = 1.0 / tex->width[0]; - vec[1] = 1.0 / tex->height[0]; + vec[0] = 1.0 / tex->width0; + vec[1] = 1.0 / tex->height0; break; default: -- cgit v1.2.3 From c78748a5274e58bcbb122923edf81065be9bbe16 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Wed, 2 Dec 2009 02:08:26 +0100 Subject: gallium: adapt drivers to interface cleanups --- src/gallium/drivers/cell/ppu/cell_texture.c | 30 +++--- src/gallium/drivers/i915/i915_surface.c | 21 +++-- src/gallium/drivers/i915/i915_texture.c | 125 ++++++++++--------------- src/gallium/drivers/llvmpipe/lp_setup.c | 2 +- src/gallium/drivers/llvmpipe/lp_tex_cache.c | 2 +- src/gallium/drivers/llvmpipe/lp_texture.c | 48 +++++----- src/gallium/drivers/llvmpipe/lp_tile_cache.c | 6 +- src/gallium/drivers/r300/r300_emit.c | 12 +-- src/gallium/drivers/r300/r300_screen.c | 9 +- src/gallium/drivers/r300/r300_texture.c | 12 +-- src/gallium/drivers/svga/svga_screen_texture.c | 37 ++++---- src/gallium/drivers/svga/svga_state_vs.c | 2 +- 12 files changed, 138 insertions(+), 168 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index e6b8a870452..77a57aef147 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -65,14 +65,11 @@ cell_texture_layout(struct cell_texture *ct) w_tile = align(width, TILE_SIZE); h_tile = align(height, TILE_SIZE); - pt->nblocksx[level] = pf_get_nblocksx(&pt->block, w_tile); - pt->nblocksy[level] = pf_get_nblocksy(&pt->block, h_tile); - - ct->stride[level] = pt->nblocksx[level] * pt->block.size; + ct->stride[level] = pf_get_stride(pt->format, w_tile); ct->level_offset[level] = ct->buffer_size; - size = pt->nblocksx[level] * pt->nblocksy[level] * pt->block.size; + size = ct->stride[level] * pf_get_nblocksy(pt->format, h_tile); if (pt->target == PIPE_TEXTURE_CUBE) size *= 6; else @@ -283,10 +280,12 @@ cell_get_tex_surface(struct pipe_screen *screen, ps->zslice = zslice; if (pt->target == PIPE_TEXTURE_CUBE) { - ps->offset += face * pt->nblocksy[level] * ct->stride[level]; + unsigned h_tile = align(ps->height, TILE_SIZE); + ps->offset += face * pf_get_nblocksy(ps->format, h_tile) * ct->stride[level]; } else if (pt->target == PIPE_TEXTURE_3D) { - ps->offset += zslice * pt->nblocksy[level] * ct->stride[level]; + unsigned h_tile = align(ps->height, TILE_SIZE); + ps->offset += zslice * pf_get_nblocksy(ps->format, h_tile) * ct->stride[level]; } else { assert(face == 0); @@ -327,14 +326,10 @@ cell_get_tex_transfer(struct pipe_screen *screen, if (ctrans) { struct pipe_transfer *pt = &ctrans->base; pipe_texture_reference(&pt->texture, texture); - pt->format = texture->format; - pt->block = texture->block; pt->x = x; pt->y = y; pt->width = w; pt->height = h; - pt->nblocksx = texture->nblocksx[level]; - pt->nblocksy = texture->nblocksy[level]; pt->stride = ct->stride[level]; pt->usage = usage; pt->face = face; @@ -344,10 +339,12 @@ cell_get_tex_transfer(struct pipe_screen *screen, ctrans->offset = ct->level_offset[level]; if (texture->target == PIPE_TEXTURE_CUBE) { - ctrans->offset += face * pt->nblocksy * pt->stride; + unsigned h_tile = align(u_minify(texture->height0, level), TILE_SIZE); + ctrans->offset += face * pf_get_nblocksy(texture->format, h_tile) * pt->stride; } else if (texture->target == PIPE_TEXTURE_3D) { - ctrans->offset += zslice * pt->nblocksy * pt->stride; + unsigned h_tile = align(u_minify(texture->height0, level), TILE_SIZE); + ctrans->offset += zslice * pf_get_nblocksy(texture->format, h_tile) * pt->stride; } else { assert(face == 0); @@ -400,7 +397,8 @@ cell_transfer_map(struct pipe_screen *screen, struct pipe_transfer *transfer) * Create a buffer of ordinary memory for the linear texture. * This is the memory that the user will read/write. */ - size = pt->nblocksx[level] * pt->nblocksy[level] * pt->block.size; + size = pf_get_stride(pt->format, align(texWidth, TILE_SIZE)) * + pf_get_nblocksy(pt->format, align(texHeight, TILE_SIZE)); ctrans->map = align_malloc(size, 16); if (!ctrans->map) @@ -408,7 +406,7 @@ cell_transfer_map(struct pipe_screen *screen, struct pipe_transfer *transfer) if (transfer->usage & PIPE_TRANSFER_READ) { /* need to untwiddle the texture to make a linear version */ - const uint bpp = pf_get_size(ct->base.format); + const uint bpp = pf_get_blocksize(ct->base.format); if (bpp == 4) { const uint *src = (uint *) (ct->mapped + ctrans->offset); uint *dst = ctrans->map; @@ -451,7 +449,7 @@ cell_transfer_unmap(struct pipe_screen *screen, /* The user wrote new texture data into the mapped buffer. * We need to convert the new linear data into the twiddled/tiled format. */ - const uint bpp = pf_get_size(ct->base.format); + const uint bpp = pf_get_blocksize(ct->base.format); if (bpp == 4) { const uint *src = ctrans->map; uint *dst = (uint *) (ct->mapped + ctrans->offset); diff --git a/src/gallium/drivers/i915/i915_surface.c b/src/gallium/drivers/i915/i915_surface.c index ab8331f3e64..24e1024aaa3 100644 --- a/src/gallium/drivers/i915/i915_surface.c +++ b/src/gallium/drivers/i915/i915_surface.c @@ -48,17 +48,19 @@ i915_surface_copy(struct pipe_context *pipe, { struct i915_texture *dst_tex = (struct i915_texture *)dst->texture; struct i915_texture *src_tex = (struct i915_texture *)src->texture; + struct pipe_texture *dpt = &dst_tex->base; + struct pipe_texture *spt = &src_tex->base; assert( dst != src ); - assert( dst_tex->base.block.size == src_tex->base.block.size ); - assert( dst_tex->base.block.width == src_tex->base.block.height ); - assert( dst_tex->base.block.height == src_tex->base.block.height ); - assert( dst_tex->base.block.width == 1 ); - assert( dst_tex->base.block.height == 1 ); + assert( pf_get_blocksize(dpt->format) == pf_get_blocksize(spt->format) ); + assert( pf_get_blockwidth(dpt->format) == pf_get_blockwidth(spt->format) ); + assert( pf_get_blockheight(dpt->format) == pf_get_blockheight(spt->format) ); + assert( pf_get_blockwidth(dpt->format) == 1 ); + assert( pf_get_blockheight(dpt->format) == 1 ); i915_copy_blit( i915_context(pipe), FALSE, - dst_tex->base.block.size, + pf_get_blocksize(dpt->format), (unsigned short) src_tex->stride, src_tex->buffer, src->offset, (unsigned short) dst_tex->stride, dst_tex->buffer, dst->offset, (short) srcx, (short) srcy, (short) dstx, (short) dsty, (short) width, (short) height ); @@ -72,12 +74,13 @@ i915_surface_fill(struct pipe_context *pipe, unsigned width, unsigned height, unsigned value) { struct i915_texture *tex = (struct i915_texture *)dst->texture; + struct pipe_texture *pt = &tex->base; - assert(tex->base.block.width == 1); - assert(tex->base.block.height == 1); + assert(pf_get_blockwidth(pt->format) == 1); + assert(pf_get_blockheight(pt->format) == 1); i915_fill_blit( i915_context(pipe), - tex->base.block.size, + pf_get_blocksize(pt->format), (unsigned short) tex->stride, tex->buffer, dst->offset, (short) dstx, (short) dsty, diff --git a/src/gallium/drivers/i915/i915_texture.c b/src/gallium/drivers/i915/i915_texture.c index c7b86dd4c57..b28b413771d 100644 --- a/src/gallium/drivers/i915/i915_texture.c +++ b/src/gallium/drivers/i915/i915_texture.c @@ -74,6 +74,9 @@ static const int step_offsets[6][2] = { {-1, 1} }; +/* XXX really need twice the size if x is already pot? + Otherwise just use util_next_power_of_two? +*/ static unsigned power_of_two(unsigned x) { @@ -83,13 +86,6 @@ power_of_two(unsigned x) return value; } -static unsigned -round_up(unsigned n, unsigned multiple) -{ - return (n + multiple - 1) & ~(multiple - 1); -} - - /* * More advanced helper funcs */ @@ -101,13 +97,8 @@ i915_miptree_set_level_info(struct i915_texture *tex, unsigned nr_images, unsigned w, unsigned h, unsigned d) { - struct pipe_texture *pt = &tex->base; - assert(level < PIPE_MAX_TEXTURE_LEVELS); - pt->nblocksx[level] = pf_get_nblocksx(&pt->block, w); - pt->nblocksy[level] = pf_get_nblocksy(&pt->block, h); - tex->nr_images[level] = nr_images; /* @@ -138,7 +129,7 @@ i915_miptree_set_image_offset(struct i915_texture *tex, assert(img < tex->nr_images[level]); - tex->image_offset[level][img] = y * tex->stride + x * tex->base.block.size; + tex->image_offset[level][img] = y * tex->stride + x * pf_get_blocksize(tex->base.format); /* printf("%s level %d img %d pos %d,%d image_offset %x\n", @@ -160,28 +151,28 @@ i915_scanout_layout(struct i915_texture *tex) { struct pipe_texture *pt = &tex->base; - if (pt->last_level > 0 || pt->block.size != 4) + if (pt->last_level > 0 || pf_get_blocksize(pt->format) != 4) return FALSE; i915_miptree_set_level_info(tex, 0, 1, - tex->base.width0, - tex->base.height0, + pt->width0, + pt->height0, 1); i915_miptree_set_image_offset(tex, 0, 0, 0, 0); - if (tex->base.width0 >= 240) { - tex->stride = power_of_two(tex->base.nblocksx[0] * pt->block.size); - tex->total_nblocksy = round_up(tex->base.nblocksy[0], 8); + if (pt->width0 >= 240) { + tex->stride = power_of_two(pf_get_stride(pt->format, pt->width0)); + tex->total_nblocksy = align(pf_get_nblocksy(pt->format, pt->height0), 8); tex->hw_tiled = INTEL_TILE_X; - } else if (tex->base.width0 == 64 && tex->base.height0 == 64) { - tex->stride = power_of_two(tex->base.nblocksx[0] * pt->block.size); - tex->total_nblocksy = round_up(tex->base.nblocksy[0], 8); + } else if (pt->width0 == 64 && pt->height0 == 64) { + tex->stride = power_of_two(pf_get_stride(pt->format, pt->width0)); + tex->total_nblocksy = align(pf_get_nblocksy(pt->format, pt->height0), 8); } else { return FALSE; } debug_printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - tex->base.width0, tex->base.height0, pt->block.size, + pt->width0, pt->height0, pf_get_blocksize(pt->format), tex->stride, tex->total_nblocksy, tex->stride * tex->total_nblocksy); return TRUE; @@ -195,25 +186,25 @@ i915_display_target_layout(struct i915_texture *tex) { struct pipe_texture *pt = &tex->base; - if (pt->last_level > 0 || pt->block.size != 4) + if (pt->last_level > 0 || pf_get_blocksize(pt->format) != 4) return FALSE; /* fallback to normal textures for small textures */ - if (tex->base.width0 < 240) + if (pt->width0 < 240) return FALSE; i915_miptree_set_level_info(tex, 0, 1, - tex->base.width0, - tex->base.height0, + pt->width0, + pt->height0, 1); i915_miptree_set_image_offset(tex, 0, 0, 0, 0); - tex->stride = power_of_two(tex->base.nblocksx[0] * pt->block.size); - tex->total_nblocksy = round_up(tex->base.nblocksy[0], 8); + tex->stride = power_of_two(pf_get_stride(pt->format, pt->width0)); + tex->total_nblocksy = align(pf_get_nblocksy(pt->format, pt->height0), 8); tex->hw_tiled = INTEL_TILE_X; debug_printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - tex->base.width0, tex->base.height0, pt->block.size, + pt->width0, pt->height0, pf_get_blocksize(pt->format), tex->stride, tex->total_nblocksy, tex->stride * tex->total_nblocksy); return TRUE; @@ -226,34 +217,32 @@ i915_miptree_layout_2d(struct i915_texture *tex) unsigned level; unsigned width = pt->width0; unsigned height = pt->height0; - unsigned nblocksx = pt->nblocksx[0]; - unsigned nblocksy = pt->nblocksy[0]; + unsigned nblocksy = pf_get_nblocksy(pt->format, pt->width0); /* used for scanouts that need special layouts */ - if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) + if (pt->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) if (i915_scanout_layout(tex)) return; /* for shared buffers we use some very like scanout */ - if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) + if (pt->tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) if (i915_display_target_layout(tex)) return; - tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4); + tex->stride = align(pf_get_stride(pt->format, pt->width0), 4); tex->total_nblocksy = 0; for (level = 0; level <= pt->last_level; level++) { i915_miptree_set_level_info(tex, level, 1, width, height, 1); i915_miptree_set_image_offset(tex, level, 0, 0, tex->total_nblocksy); - nblocksy = round_up(MAX2(2, nblocksy), 2); + nblocksy = align(MAX2(2, nblocksy), 2); tex->total_nblocksy += nblocksy; width = u_minify(width, 1); height = u_minify(height, 1); - nblocksx = pf_get_nblocksx(&pt->block, width); - nblocksy = pf_get_nblocksy(&pt->block, height); + nblocksy = pf_get_nblocksy(pt->format, height); } } @@ -266,13 +255,12 @@ i915_miptree_layout_3d(struct i915_texture *tex) unsigned width = pt->width0; unsigned height = pt->height0; unsigned depth = pt->depth0; - unsigned nblocksx = pt->nblocksx[0]; - unsigned nblocksy = pt->nblocksy[0]; + unsigned nblocksy = pf_get_nblocksy(pt->format, pt->height0); unsigned stack_nblocksy = 0; /* Calculate the size of a single slice. */ - tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4); + tex->stride = align(pf_get_stride(pt->format, pt->width0), 4); /* XXX: hardware expects/requires 9 levels at minimum. */ @@ -283,8 +271,7 @@ i915_miptree_layout_3d(struct i915_texture *tex) width = u_minify(width, 1); height = u_minify(height, 1); - nblocksx = pf_get_nblocksx(&pt->block, width); - nblocksy = pf_get_nblocksy(&pt->block, height); + nblocksy = pf_get_nblocksy(pt->format, height); } /* Fixup depth image_offsets: @@ -309,14 +296,14 @@ i915_miptree_layout_cube(struct i915_texture *tex) { struct pipe_texture *pt = &tex->base; unsigned width = pt->width0, height = pt->height0; - const unsigned nblocks = pt->nblocksx[0]; + const unsigned nblocks = pf_get_nblocksx(pt->format, pt->width0); unsigned level; unsigned face; assert(width == height); /* cubemap images are square */ /* double pitch for cube layouts */ - tex->stride = round_up(nblocks * pt->block.size * 2, 4); + tex->stride = align(nblocks * pf_get_blocksize(pt->format) * 2, 4); tex->total_nblocksy = nblocks * 4; for (level = 0; level <= pt->last_level; level++) { @@ -379,8 +366,8 @@ i945_miptree_layout_2d(struct i915_texture *tex) unsigned y = 0; unsigned width = pt->width0; unsigned height = pt->height0; - unsigned nblocksx = pt->nblocksx[0]; - unsigned nblocksy = pt->nblocksy[0]; + unsigned nblocksx = pf_get_nblocksx(pt->format, pt->width0); + unsigned nblocksy = pf_get_nblocksy(pt->format, pt->height0); /* used for scanouts that need special layouts */ if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) @@ -392,7 +379,7 @@ i945_miptree_layout_2d(struct i915_texture *tex) if (i915_display_target_layout(tex)) return; - tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4); + tex->stride = align(pf_get_stride(pt->format, pt->width0), 4); /* May need to adjust pitch to accomodate the placement of * the 2nd mipmap level. This occurs when the alignment @@ -401,11 +388,11 @@ i945_miptree_layout_2d(struct i915_texture *tex) */ if (pt->last_level > 0) { unsigned mip1_nblocksx - = align(pf_get_nblocksx(&pt->block, u_minify(width, 1)), align_x) - + pf_get_nblocksx(&pt->block, u_minify(width, 2)); + = align(pf_get_nblocksx(pt->format, u_minify(width, 1)), align_x) + + pf_get_nblocksx(pt->format, u_minify(width, 2)); if (mip1_nblocksx > nblocksx) - tex->stride = mip1_nblocksx * pt->block.size; + tex->stride = mip1_nblocksx * pf_get_blocksize(pt->format); } /* Pitch must be a whole number of dwords @@ -435,8 +422,8 @@ i945_miptree_layout_2d(struct i915_texture *tex) width = u_minify(width, 1); height = u_minify(height, 1); - nblocksx = pf_get_nblocksx(&pt->block, width); - nblocksy = pf_get_nblocksy(&pt->block, height); + nblocksx = pf_get_nblocksx(pt->format, width); + nblocksy = pf_get_nblocksy(pt->format, height); } } @@ -447,17 +434,16 @@ i945_miptree_layout_3d(struct i915_texture *tex) unsigned width = pt->width0; unsigned height = pt->height0; unsigned depth = pt->depth0; - unsigned nblocksx = pt->nblocksx[0]; - unsigned nblocksy = pt->nblocksy[0]; + unsigned nblocksy = pf_get_nblocksy(pt->format, pt->width0); unsigned pack_x_pitch, pack_x_nr; unsigned pack_y_pitch; unsigned level; - tex->stride = round_up(pt->nblocksx[0] * pt->block.size, 4); + tex->stride = align(pf_get_stride(pt->format, pt->width0), 4); tex->total_nblocksy = 0; - pack_y_pitch = MAX2(pt->nblocksy[0], 2); - pack_x_pitch = tex->stride / pt->block.size; + pack_y_pitch = MAX2(nblocksy, 2); + pack_x_pitch = tex->stride / pf_get_blocksize(pt->format); pack_x_nr = 1; for (level = 0; level <= pt->last_level; level++) { @@ -482,7 +468,7 @@ i945_miptree_layout_3d(struct i915_texture *tex) if (pack_x_pitch > 4) { pack_x_pitch >>= 1; pack_x_nr <<= 1; - assert(pack_x_pitch * pack_x_nr * pt->block.size <= tex->stride); + assert(pack_x_pitch * pack_x_nr * pf_get_blocksize(pt->format) <= tex->stride); } if (pack_y_pitch > 2) { @@ -492,8 +478,7 @@ i945_miptree_layout_3d(struct i915_texture *tex) width = u_minify(width, 1); height = u_minify(height, 1); depth = u_minify(depth, 1); - nblocksx = pf_get_nblocksx(&pt->block, width); - nblocksy = pf_get_nblocksy(&pt->block, height); + nblocksy = pf_get_nblocksy(pt->format, height); } } @@ -503,7 +488,7 @@ i945_miptree_layout_cube(struct i915_texture *tex) struct pipe_texture *pt = &tex->base; unsigned level; - const unsigned nblocks = pt->nblocksx[0]; + const unsigned nblocks = pf_get_nblocksx(pt->format, pt->width0); unsigned face; unsigned width = pt->width0; unsigned height = pt->height0; @@ -523,9 +508,9 @@ i945_miptree_layout_cube(struct i915_texture *tex) * or the final row of 4x4, 2x2 and 1x1 faces below this. */ if (nblocks > 32) - tex->stride = round_up(nblocks * pt->block.size * 2, 4); + tex->stride = align(nblocks * pf_get_blocksize(pt->format) * 2, 4); else - tex->stride = 14 * 8 * pt->block.size; + tex->stride = 14 * 8 * pf_get_blocksize(pt->format); tex->total_nblocksy = nblocks * 4; @@ -645,9 +630,6 @@ i915_texture_create(struct pipe_screen *screen, pipe_reference_init(&tex->base.reference, 1); tex->base.screen = screen; - tex->base.nblocksx[0] = pf_get_nblocksx(&tex->base.block, tex->base.width0); - tex->base.nblocksy[0] = pf_get_nblocksy(&tex->base.block, tex->base.height0); - if (is->is_i945) { if (!i945_miptree_layout(tex)) goto fail; @@ -829,14 +811,10 @@ i915_get_tex_transfer(struct pipe_screen *screen, trans = CALLOC_STRUCT(i915_transfer); if (trans) { pipe_texture_reference(&trans->base.texture, texture); - trans->base.format = trans->base.format; trans->base.x = x; trans->base.y = y; trans->base.width = w; trans->base.height = h; - trans->base.block = texture->block; - trans->base.nblocksx = texture->nblocksx[level]; - trans->base.nblocksy = texture->nblocksy[level]; trans->base.stride = tex->stride; trans->offset = offset; trans->base.usage = usage; @@ -852,6 +830,7 @@ i915_transfer_map(struct pipe_screen *screen, struct intel_winsys *iws = i915_screen(tex->base.screen)->iws; char *map; boolean write = FALSE; + enum pipe_format format = tex->base.format; if (transfer->usage & PIPE_TRANSFER_WRITE) write = TRUE; @@ -861,8 +840,8 @@ i915_transfer_map(struct pipe_screen *screen, return NULL; return map + i915_transfer(transfer)->offset + - transfer->y / transfer->block.height * transfer->stride + - transfer->x / transfer->block.width * transfer->block.size; + transfer->y / pf_get_blockheight(format) * transfer->stride + + transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format); } static void diff --git a/src/gallium/drivers/llvmpipe/lp_setup.c b/src/gallium/drivers/llvmpipe/lp_setup.c index ffcbc9a379f..b4aabd4d7cc 100644 --- a/src/gallium/drivers/llvmpipe/lp_setup.c +++ b/src/gallium/drivers/llvmpipe/lp_setup.c @@ -166,7 +166,7 @@ shade_quads(struct llvmpipe_context *llvmpipe, assert((y % 2) == 0); depth = llvmpipe->zsbuf_map + y*llvmpipe->zsbuf_transfer->stride + - 2*x*llvmpipe->zsbuf_transfer->block.size; + 2*x*pf_get_blocksize(llvmpipe->zsbuf_transfer->texture->format); } else depth = NULL; diff --git a/src/gallium/drivers/llvmpipe/lp_tex_cache.c b/src/gallium/drivers/llvmpipe/lp_tex_cache.c index c7c4143bc62..5dbc597d2c9 100644 --- a/src/gallium/drivers/llvmpipe/lp_tex_cache.c +++ b/src/gallium/drivers/llvmpipe/lp_tex_cache.c @@ -291,7 +291,7 @@ lp_find_cached_tex_tile(struct llvmpipe_tex_tile_cache *tc, assert(0); } - util_format_read_4ub(tc->tex_trans->format, + util_format_read_4ub(tc->tex_trans->texture->format, (uint8_t *)tile->color, sizeof tile->color[0], tc->tex_trans_map, tc->tex_trans->stride, x, y, w, h); diff --git a/src/gallium/drivers/llvmpipe/lp_texture.c b/src/gallium/drivers/llvmpipe/lp_texture.c index 65d62fd0723..f099f903bd7 100644 --- a/src/gallium/drivers/llvmpipe/lp_texture.c +++ b/src/gallium/drivers/llvmpipe/lp_texture.c @@ -48,7 +48,6 @@ /* Simple, maximally packed layout. */ - /* Conventional allocation path for non-display textures: */ static boolean @@ -63,20 +62,15 @@ llvmpipe_texture_layout(struct llvmpipe_screen *screen, unsigned buffer_size = 0; - pf_get_block(lpt->base.format, &lpt->base.block); - for (level = 0; level <= pt->last_level; level++) { unsigned nblocksx, nblocksy; - pt->nblocksx[level] = pf_get_nblocksx(&pt->block, width); - pt->nblocksy[level] = pf_get_nblocksy(&pt->block, height); - /* Allocate storage for whole quads. This is particularly important * for depth surfaces, which are currently stored in a swizzled format. */ - nblocksx = pf_get_nblocksx(&pt->block, align(width, 2)); - nblocksy = pf_get_nblocksy(&pt->block, align(height, 2)); + nblocksx = pf_get_nblocksx(pt->format, align(width, 2)); + nblocksy = pf_get_nblocksy(pt->format, align(height, 2)); - lpt->stride[level] = align(nblocksx*pt->block.size, 16); + lpt->stride[level] = align(nblocksx * pf_get_blocksize(pt->format), 16); lpt->level_offset[level] = buffer_size; @@ -100,10 +94,6 @@ llvmpipe_displaytarget_layout(struct llvmpipe_screen *screen, { struct llvmpipe_winsys *winsys = screen->winsys; - pf_get_block(lpt->base.format, &lpt->base.block); - lpt->base.nblocksx[0] = pf_get_nblocksx(&lpt->base.block, lpt->base.width0); - lpt->base.nblocksy[0] = pf_get_nblocksy(&lpt->base.block, lpt->base.height0); - lpt->dt = winsys->displaytarget_create(winsys, lpt->base.format, lpt->base.width0, @@ -180,8 +170,6 @@ llvmpipe_texture_blanket(struct pipe_screen * screen, lpt->base = *base; pipe_reference_init(&lpt->base.reference, 1); lpt->base.screen = screen; - lpt->base.nblocksx[0] = pf_get_nblocksx(&lpt->base.block, lpt->base.width0); - lpt->base.nblocksy[0] = pf_get_nblocksy(&lpt->base.block, lpt->base.height0); lpt->stride[0] = stride[0]; pipe_buffer_reference(&lpt->buffer, buffer); @@ -255,11 +243,17 @@ llvmpipe_get_tex_surface(struct pipe_screen *screen, ps->level = level; ps->zslice = zslice; + /* XXX shouldn't that rather be + tex_height = align(ps->height, 2); + to account for alignment done in llvmpipe_texture_layout ? + */ if (pt->target == PIPE_TEXTURE_CUBE) { - ps->offset += face * pt->nblocksy[level] * lpt->stride[level]; + unsigned tex_height = ps->height; + ps->offset += face * pf_get_nblocksy(pt->format, tex_height) * lpt->stride[level]; } else if (pt->target == PIPE_TEXTURE_3D) { - ps->offset += zslice * pt->nblocksy[level] * lpt->stride[level]; + unsigned tex_height = ps->height; + ps->offset += zslice * pf_get_nblocksy(pt->format, tex_height) * lpt->stride[level]; } else { assert(face == 0); @@ -300,14 +294,10 @@ llvmpipe_get_tex_transfer(struct pipe_screen *screen, if (lpt) { struct pipe_transfer *pt = &lpt->base; pipe_texture_reference(&pt->texture, texture); - pt->format = texture->format; - pt->block = texture->block; pt->x = x; pt->y = y; pt->width = w; pt->height = h; - pt->nblocksx = texture->nblocksx[level]; - pt->nblocksy = texture->nblocksy[level]; pt->stride = lptex->stride[level]; pt->usage = usage; pt->face = face; @@ -316,11 +306,17 @@ llvmpipe_get_tex_transfer(struct pipe_screen *screen, lpt->offset = lptex->level_offset[level]; + /* XXX shouldn't that rather be + tex_height = align(u_minify(texture->height0, level), 2) + to account for alignment done in llvmpipe_texture_layout ? + */ if (texture->target == PIPE_TEXTURE_CUBE) { - lpt->offset += face * pt->nblocksy * pt->stride; + unsigned tex_height = u_minify(texture->height0, level); + lpt->offset += face * pf_get_nblocksy(texture->format, tex_height) * pt->stride; } else if (texture->target == PIPE_TEXTURE_3D) { - lpt->offset += zslice * pt->nblocksy * pt->stride; + unsigned tex_height = u_minify(texture->height0, level); + lpt->offset += zslice * pf_get_nblocksy(texture->format, tex_height) * pt->stride; } else { assert(face == 0); @@ -352,9 +348,11 @@ llvmpipe_transfer_map( struct pipe_screen *_screen, struct llvmpipe_screen *screen = llvmpipe_screen(_screen); ubyte *map, *xfer_map; struct llvmpipe_texture *lpt; + enum pipe_format format; assert(transfer->texture); lpt = llvmpipe_texture(transfer->texture); + format = lpt->base.format; if(lpt->dt) { struct llvmpipe_winsys *winsys = screen->winsys; @@ -379,8 +377,8 @@ llvmpipe_transfer_map( struct pipe_screen *_screen, } xfer_map = map + llvmpipe_transfer(transfer)->offset + - transfer->y / transfer->block.height * transfer->stride + - transfer->x / transfer->block.width * transfer->block.size; + transfer->y / pf_get_blockheight(format) * transfer->stride + + transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format); /*printf("map = %p xfer map = %p\n", map, xfer_map);*/ return xfer_map; } diff --git a/src/gallium/drivers/llvmpipe/lp_tile_cache.c b/src/gallium/drivers/llvmpipe/lp_tile_cache.c index ec3e002d628..50891c42271 100644 --- a/src/gallium/drivers/llvmpipe/lp_tile_cache.c +++ b/src/gallium/drivers/llvmpipe/lp_tile_cache.c @@ -252,13 +252,13 @@ lp_flush_tile_cache(struct llvmpipe_tile_cache *tc) case LP_TILE_STATUS_CLEAR: /* Actually clear the tiles which were flagged as being in a * clear state. */ - util_fill_rect(tc->transfer_map, &pt->block, pt->stride, + util_fill_rect(tc->transfer_map, pt->texture->format, pt->stride, x, y, w, h, tc->clear_val); break; case LP_TILE_STATUS_DEFINED: - lp_tile_write_4ub(pt->format, + lp_tile_write_4ub(pt->texture->format, tile->color, tc->transfer_map, pt->stride, x, y, w, h); @@ -306,7 +306,7 @@ lp_get_cached_tile(struct llvmpipe_tile_cache *tc, y &= ~(TILE_SIZE - 1); if (!pipe_clip_tile(x, y, &w, &h, tc->transfer)) - lp_tile_read_4ub(pt->format, + lp_tile_read_4ub(pt->texture->format, tile->color, tc->transfer_map, tc->transfer->stride, x, y, w, h); diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 98a39390bf9..a479842f9eb 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -631,10 +631,10 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset) for (i = 0; i < aos_count - 1; i += 2) { int buf_num1 = velem[i].vertex_buffer_index; int buf_num2 = velem[i+1].vertex_buffer_index; - assert(vbuf[buf_num1].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0); - assert(vbuf[buf_num2].stride % 4 == 0 && pf_get_size(velem[i+1].src_format) % 4 == 0); - OUT_CS((pf_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num1].stride << 6) | - (pf_get_size(velem[i+1].src_format) << 14) | (vbuf[buf_num2].stride << 22)); + assert(vbuf[buf_num1].stride % 4 == 0 && pf_get_blocksize(velem[i].src_format) % 4 == 0); + assert(vbuf[buf_num2].stride % 4 == 0 && pf_get_blocksize(velem[i+1].src_format) % 4 == 0); + OUT_CS((pf_get_blocksize(velem[i].src_format) >> 2) | (vbuf[buf_num1].stride << 6) | + (pf_get_blocksize(velem[i+1].src_format) << 14) | (vbuf[buf_num2].stride << 22)); OUT_CS(vbuf[buf_num1].buffer_offset + velem[i].src_offset + offset * vbuf[buf_num1].stride); OUT_CS(vbuf[buf_num2].buffer_offset + velem[i+1].src_offset + @@ -642,8 +642,8 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset) } if (aos_count & 1) { int buf_num = velem[i].vertex_buffer_index; - assert(vbuf[buf_num].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0); - OUT_CS((pf_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num].stride << 6)); + assert(vbuf[buf_num].stride % 4 == 0 && pf_get_blocksize(velem[i].src_format) % 4 == 0); + OUT_CS((pf_get_blocksize(velem[i].src_format) >> 2) | (vbuf[buf_num].stride << 6)); OUT_CS(vbuf[buf_num].buffer_offset + velem[i].src_offset + offset * vbuf[buf_num].stride); } diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c index 390b63007e5..032fa69ec0f 100644 --- a/src/gallium/drivers/r300/r300_screen.c +++ b/src/gallium/drivers/r300/r300_screen.c @@ -311,14 +311,10 @@ r300_get_tex_transfer(struct pipe_screen *screen, trans = CALLOC_STRUCT(r300_transfer); if (trans) { pipe_texture_reference(&trans->transfer.texture, texture); - trans->transfer.format = texture->format; trans->transfer.x = x; trans->transfer.y = y; trans->transfer.width = w; trans->transfer.height = h; - trans->transfer.block = texture->block; - trans->transfer.nblocksx = texture->nblocksx[level]; - trans->transfer.nblocksy = texture->nblocksy[level]; trans->transfer.stride = r300_texture_get_stride(tex, level); trans->transfer.usage = usage; @@ -344,6 +340,7 @@ static void* r300_transfer_map(struct pipe_screen* screen, { struct r300_texture* tex = (struct r300_texture*)transfer->texture; char* map; + enum pipe_format format = tex->tex.format; map = pipe_buffer_map(screen, tex->buffer, pipe_transfer_buffer_flags(transfer)); @@ -353,8 +350,8 @@ static void* r300_transfer_map(struct pipe_screen* screen, } return map + r300_transfer(transfer)->offset + - transfer->y / transfer->block.height * transfer->stride + - transfer->x / transfer->block.width * transfer->block.size; + transfer->y / pf_get_blockheight(format) * transfer->stride + + transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format); } static void r300_transfer_unmap(struct pipe_screen* screen, diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c index 093a21ebe24..63fc6a235ac 100644 --- a/src/gallium/drivers/r300/r300_texture.c +++ b/src/gallium/drivers/r300/r300_texture.c @@ -105,7 +105,7 @@ unsigned r300_texture_get_stride(struct r300_texture* tex, unsigned level) return 0; } - return align(pf_get_stride(&tex->tex.block, u_minify(tex->tex.width0, level)), 32); + return align(pf_get_stride(tex->tex.format, u_minify(tex->tex.width0, level)), 32); } static void r300_setup_miptree(struct r300_texture* tex) @@ -115,11 +115,10 @@ static void r300_setup_miptree(struct r300_texture* tex) int i; for (i = 0; i <= base->last_level; i++) { - base->nblocksx[i] = pf_get_nblocksx(&base->block, u_minify(base->width0, i)); - base->nblocksy[i] = pf_get_nblocksy(&base->block, u_minify(base->height0, i)); + unsigned nblocksy = pf_get_nblocksy(base->format, u_minify(base->height0, i)); stride = r300_texture_get_stride(tex, i); - layer_size = stride * base->nblocksy[i]; + layer_size = stride * nblocksy; if (base->target == PIPE_TEXTURE_CUBE) size = layer_size * 6; @@ -129,7 +128,7 @@ static void r300_setup_miptree(struct r300_texture* tex) tex->offset[i] = align(tex->size, 32); tex->size = tex->offset[i] + size; tex->layer_size[i] = layer_size; - tex->pitch[i] = stride / base->block.size; + tex->pitch[i] = stride / pf_get_blocksize(base->format); debug_printf("r300: Texture miptree: Level %d " "(%dx%dx%d px, pitch %d bytes)\n", @@ -245,7 +244,7 @@ static struct pipe_texture* tex->tex.screen = screen; tex->stride_override = *stride; - tex->pitch[0] = *stride / base->block.size; + tex->pitch[0] = *stride / pf_get_blocksize(base->format); r300_setup_flags(tex); r300_setup_texture_state(tex, r300_screen(screen)->caps->is_r500); @@ -283,7 +282,6 @@ r300_video_surface_create(struct pipe_screen *screen, template.width0 = util_next_power_of_two(width); template.height0 = util_next_power_of_two(height); template.depth0 = 1; - pf_get_block(template.format, &template.block); template.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER | PIPE_TEXTURE_USAGE_RENDER_TARGET; diff --git a/src/gallium/drivers/svga/svga_screen_texture.c b/src/gallium/drivers/svga/svga_screen_texture.c index fb11b80dcf9..410adf881b8 100644 --- a/src/gallium/drivers/svga/svga_screen_texture.c +++ b/src/gallium/drivers/svga/svga_screen_texture.c @@ -158,7 +158,8 @@ svga_transfer_dma_band(struct svga_transfer *st, st->base.x + st->base.width, y + h, st->base.zslice + 1, - texture->base.block.size*8/(texture->base.block.width*texture->base.block.height)); + pf_get_blocksize(texture->base.format)*8/ + (pf_get_blockwidth(texture->base.format)*pf_get_blockheight(texture->base.format))); box.x = st->base.x; box.y = y; @@ -208,7 +209,8 @@ svga_transfer_dma(struct svga_transfer *st, } else { unsigned y, h, srcy; - h = st->hw_nblocksy * st->base.block.height; + unsigned blockheight = pf_get_blockheight(st->base.texture->format); + h = st->hw_nblocksy * blockheight; srcy = 0; for(y = 0; y < st->base.height; y += h) { unsigned offset, length; @@ -218,11 +220,11 @@ svga_transfer_dma(struct svga_transfer *st, h = st->base.height - y; /* Transfer band must be aligned to pixel block boundaries */ - assert(y % st->base.block.height == 0); - assert(h % st->base.block.height == 0); + assert(y % blockheight == 0); + assert(h % blockheight == 0); - offset = y * st->base.stride / st->base.block.height; - length = h * st->base.stride / st->base.block.height; + offset = y * st->base.stride / blockheight; + length = h * st->base.stride / blockheight; sw = (uint8_t *)st->swbuf + offset; @@ -291,8 +293,6 @@ svga_texture_create(struct pipe_screen *screen, height = templat->height0; depth = templat->depth0; for(level = 0; level <= templat->last_level; ++level) { - tex->base.nblocksx[level] = pf_get_nblocksx(&tex->base.block, width); - tex->base.nblocksy[level] = pf_get_nblocksy(&tex->base.block, height); width = u_minify(width, 1); height = u_minify(height, 1); depth = u_minify(depth, 1); @@ -750,6 +750,8 @@ svga_get_tex_transfer(struct pipe_screen *screen, struct svga_screen *ss = svga_screen(screen); struct svga_winsys_screen *sws = ss->sws; struct svga_transfer *st; + unsigned nblocksx = pf_get_nblocksx(texture->format, w); + unsigned nblocksy = pf_get_nblocksy(texture->format, h); /* We can't map texture storage directly */ if (usage & PIPE_TRANSFER_MAP_DIRECTLY) @@ -759,21 +761,17 @@ svga_get_tex_transfer(struct pipe_screen *screen, if (!st) return NULL; - st->base.format = texture->format; - st->base.block = texture->block; st->base.x = x; st->base.y = y; st->base.width = w; st->base.height = h; - st->base.nblocksx = pf_get_nblocksx(&texture->block, w); - st->base.nblocksy = pf_get_nblocksy(&texture->block, h); - st->base.stride = st->base.nblocksx*st->base.block.size; + st->base.stride = nblocksx*pf_get_blocksize(texture->format); st->base.usage = usage; st->base.face = face; st->base.level = level; st->base.zslice = zslice; - st->hw_nblocksy = st->base.nblocksy; + st->hw_nblocksy = nblocksy; st->hwbuf = svga_winsys_buffer_create(ss, 1, @@ -789,15 +787,15 @@ svga_get_tex_transfer(struct pipe_screen *screen, if(!st->hwbuf) goto no_hwbuf; - if(st->hw_nblocksy < st->base.nblocksy) { + if(st->hw_nblocksy < nblocksy) { /* We couldn't allocate a hardware buffer big enough for the transfer, * so allocate regular malloc memory instead */ debug_printf("%s: failed to allocate %u KB of DMA, splitting into %u x %u KB DMA transfers\n", __FUNCTION__, - (st->base.nblocksy*st->base.stride + 1023)/1024, - (st->base.nblocksy + st->hw_nblocksy - 1)/st->hw_nblocksy, + (nblocksy*st->base.stride + 1023)/1024, + (nblocksy + st->hw_nblocksy - 1)/st->hw_nblocksy, (st->hw_nblocksy*st->base.stride + 1023)/1024); - st->swbuf = MALLOC(st->base.nblocksy*st->base.stride); + st->swbuf = MALLOC(nblocksy*st->base.stride); if(!st->swbuf) goto no_swbuf; } @@ -1046,8 +1044,7 @@ svga_screen_buffer_from_texture(struct pipe_texture *texture, svga_translate_format(texture->format), stex->handle); - *stride = pf_get_nblocksx(&texture->block, texture->width0) * - texture->block.size; + *stride = pf_get_stride(texture->format, texture->width0); return *buffer != NULL; } diff --git a/src/gallium/drivers/svga/svga_state_vs.c b/src/gallium/drivers/svga/svga_state_vs.c index a947745732c..f1b0daf9f6b 100644 --- a/src/gallium/drivers/svga/svga_state_vs.c +++ b/src/gallium/drivers/svga/svga_state_vs.c @@ -210,7 +210,7 @@ static int update_zero_stride( struct svga_context *svga, mapped_buffer = pipe_buffer_map_range(svga->pipe.screen, vbuffer->buffer, vel->src_offset, - pf_get_size(vel->src_format), + pf_get_blocksize(vel->src_format), PIPE_BUFFER_USAGE_CPU_READ); translate->set_buffer(translate, vel->vertex_buffer_index, mapped_buffer, -- cgit v1.2.3 From cceeab39ea541b1be1521114316d660a77769c2a Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 3 Dec 2009 11:17:37 +0100 Subject: Move pf_get_bits/size() to u_format auxiliary module. --- src/gallium/auxiliary/util/u_format.h | 29 ++++++++++++++++++++++ src/gallium/auxiliary/util/u_gen_mipmap.c | 5 ++-- src/gallium/drivers/cell/ppu/cell_texture.c | 6 +++-- src/gallium/drivers/r300/r300_emit.c | 13 +++++----- src/gallium/drivers/softpipe/sp_tile_cache.c | 3 ++- src/gallium/drivers/svga/svga_state_vs.c | 3 ++- src/gallium/include/pipe/p_format.h | 36 ---------------------------- src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c | 5 ++-- src/gallium/winsys/gdi/gdi_softpipe_winsys.c | 4 ++-- src/mesa/state_tracker/st_cb_texture.c | 2 +- 10 files changed, 53 insertions(+), 53 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/auxiliary/util/u_format.h b/src/gallium/auxiliary/util/u_format.h index 583b62e606b..3ac53840247 100644 --- a/src/gallium/auxiliary/util/u_format.h +++ b/src/gallium/auxiliary/util/u_format.h @@ -183,6 +183,35 @@ util_format_get_block(enum pipe_format format, block->height = desc->block.height; } +/** + * Return total bits needed for the pixel format. + */ +static INLINE uint +util_format_get_bits(enum pipe_format format) +{ + const struct util_format_description *desc = util_format_description(format); + + assert(format); + if (!format) { + return 0; + } + + return desc->block.bits / (desc->block.width * desc->block.height); +} + +/** + * Return bytes per pixel for the given format. + */ +static INLINE uint +util_format_get_size(enum pipe_format format) +{ + uint bits = util_format_get_bits(format); + + assert(bits % 8 == 0); + + return bits / 8; +} + /* * Format access functions. diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index f67f1e458d4..70ec925d152 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -41,6 +41,7 @@ #include "pipe/p_shader_tokens.h" #include "pipe/p_state.h" +#include "util/u_format.h" #include "util/u_memory.h" #include "util/u_draw_quad.h" #include "util/u_gen_mipmap.h" @@ -996,7 +997,7 @@ reduce_2d(enum pipe_format pformat, { enum dtype datatype; uint comps; - const int bpt = pf_get_size(pformat); + const int bpt = util_format_get_size(pformat); const ubyte *srcA, *srcB; ubyte *dst; int row; @@ -1035,7 +1036,7 @@ reduce_3d(enum pipe_format pformat, int dstWidth, int dstHeight, int dstDepth, int dstRowStride, ubyte *dstPtr) { - const int bpt = pf_get_size(pformat); + const int bpt = util_format_get_size(pformat); const int border = 0; int img, row; int bytesPerSrcImage, bytesPerDstImage; diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index e6b8a870452..605d53a9486 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -35,6 +35,8 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "pipe/internal/p_winsys_screen.h" + +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -408,7 +410,7 @@ cell_transfer_map(struct pipe_screen *screen, struct pipe_transfer *transfer) if (transfer->usage & PIPE_TRANSFER_READ) { /* need to untwiddle the texture to make a linear version */ - const uint bpp = pf_get_size(ct->base.format); + const uint bpp = util_format_get_size(ct->base.format); if (bpp == 4) { const uint *src = (uint *) (ct->mapped + ctrans->offset); uint *dst = ctrans->map; @@ -451,7 +453,7 @@ cell_transfer_unmap(struct pipe_screen *screen, /* The user wrote new texture data into the mapped buffer. * We need to convert the new linear data into the twiddled/tiled format. */ - const uint bpp = pf_get_size(ct->base.format); + const uint bpp = util_format_get_size(ct->base.format); if (bpp == 4) { const uint *src = ctrans->map; uint *dst = (uint *) (ct->mapped + ctrans->offset); diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 98a39390bf9..171859b8e43 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -22,6 +22,7 @@ /* r300_emit: Functions for emitting state. */ +#include "util/u_format.h" #include "util/u_math.h" #include "r300_context.h" @@ -631,10 +632,10 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset) for (i = 0; i < aos_count - 1; i += 2) { int buf_num1 = velem[i].vertex_buffer_index; int buf_num2 = velem[i+1].vertex_buffer_index; - assert(vbuf[buf_num1].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0); - assert(vbuf[buf_num2].stride % 4 == 0 && pf_get_size(velem[i+1].src_format) % 4 == 0); - OUT_CS((pf_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num1].stride << 6) | - (pf_get_size(velem[i+1].src_format) << 14) | (vbuf[buf_num2].stride << 22)); + assert(vbuf[buf_num1].stride % 4 == 0 && util_format_get_size(velem[i].src_format) % 4 == 0); + assert(vbuf[buf_num2].stride % 4 == 0 && util_format_get_size(velem[i+1].src_format) % 4 == 0); + OUT_CS((util_format_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num1].stride << 6) | + (util_format_get_size(velem[i+1].src_format) << 14) | (vbuf[buf_num2].stride << 22)); OUT_CS(vbuf[buf_num1].buffer_offset + velem[i].src_offset + offset * vbuf[buf_num1].stride); OUT_CS(vbuf[buf_num2].buffer_offset + velem[i+1].src_offset + @@ -642,8 +643,8 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset) } if (aos_count & 1) { int buf_num = velem[i].vertex_buffer_index; - assert(vbuf[buf_num].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0); - OUT_CS((pf_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num].stride << 6)); + assert(vbuf[buf_num].stride % 4 == 0 && util_format_get_size(velem[i].src_format) % 4 == 0); + OUT_CS((util_format_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num].stride << 6)); OUT_CS(vbuf[buf_num].buffer_offset + velem[i].src_offset + offset * vbuf[buf_num].stride); } diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index 65872cecc4f..cde22600732 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -33,6 +33,7 @@ */ #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_memory.h" #include "util/u_tile.h" #include "sp_tile_cache.h" @@ -238,7 +239,7 @@ clear_tile(struct softpipe_cached_tile *tile, { uint i, j; - switch (pf_get_size(format)) { + switch (util_format_get_size(format)) { case 1: memset(tile->data.any, clear_value, TILE_SIZE * TILE_SIZE); break; diff --git a/src/gallium/drivers/svga/svga_state_vs.c b/src/gallium/drivers/svga/svga_state_vs.c index a947745732c..9e339577c75 100644 --- a/src/gallium/drivers/svga/svga_state_vs.c +++ b/src/gallium/drivers/svga/svga_state_vs.c @@ -25,6 +25,7 @@ #include "pipe/p_inlines.h" #include "pipe/p_defines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "translate/translate.h" @@ -210,7 +211,7 @@ static int update_zero_stride( struct svga_context *svga, mapped_buffer = pipe_buffer_map_range(svga->pipe.screen, vbuffer->buffer, vel->src_offset, - pf_get_size(vel->src_format), + util_format_get_size(vel->src_format), PIPE_BUFFER_USAGE_CPU_READ); translate->set_buffer(translate, vel->vertex_buffer_index, mapped_buffer, diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index 3be5b18a258..5fd073c95fd 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -422,42 +422,6 @@ static INLINE uint pf_get_component_bits( enum pipe_format format, uint comp ) return size << (pf_mixed_scale8( format ) * 3); } -/** - * Return total bits needed for the pixel format. - */ -static INLINE uint pf_get_bits( enum pipe_format format ) -{ - switch (pf_layout(format)) { - case PIPE_FORMAT_LAYOUT_RGBAZS: - case PIPE_FORMAT_LAYOUT_MIXED: - return - pf_get_component_bits( format, PIPE_FORMAT_COMP_0 ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_1 ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_R ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_G ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_B ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_A ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_S ); - case PIPE_FORMAT_LAYOUT_YCBCR: - assert( format == PIPE_FORMAT_YCBCR || format == PIPE_FORMAT_YCBCR_REV ); - /* return effective bits per pixel */ - return 16; - default: - assert( 0 ); - return 0; - } -} - -/** - * Return bytes per pixel for the given format. - */ -static INLINE uint pf_get_size( enum pipe_format format ) -{ - assert(pf_get_bits(format) % 8 == 0); - return pf_get_bits(format) / 8; -} - /** * Describe accurately the pixel format. * diff --git a/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c b/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c index e8bc0f55ac4..9cb77f3ad32 100644 --- a/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c +++ b/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c @@ -39,6 +39,7 @@ #include "pipe/p_format.h" #include "pipe/p_context.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" #include "llvmpipe/lp_winsys.h" @@ -147,8 +148,8 @@ gdi_llvmpipe_displaytarget_create(struct llvmpipe_winsys *winsys, gdt->width = width; gdt->height = height; - bpp = pf_get_bits(format); - cpp = pf_get_size(format); + bpp = util_format_get_bits(format); + cpp = util_format_get_size(format); gdt->stride = round_up(width * cpp, alignment); gdt->size = gdt->stride * height; diff --git a/src/gallium/winsys/gdi/gdi_softpipe_winsys.c b/src/gallium/winsys/gdi/gdi_softpipe_winsys.c index d9fb2080a15..a58648ddb95 100644 --- a/src/gallium/winsys/gdi/gdi_softpipe_winsys.c +++ b/src/gallium/winsys/gdi/gdi_softpipe_winsys.c @@ -284,10 +284,10 @@ gdi_softpipe_present(struct pipe_screen *screen, memset(&bmi, 0, sizeof(BITMAPINFO)); bmi.bmiHeader.biSize = sizeof(BITMAPINFOHEADER); - bmi.bmiHeader.biWidth = texture->stride[surface->level] / pf_get_size(surface->format); + bmi.bmiHeader.biWidth = texture->stride[surface->level] / util_format_get_size(surface->format); bmi.bmiHeader.biHeight= -(long)surface->height; bmi.bmiHeader.biPlanes = 1; - bmi.bmiHeader.biBitCount = pf_get_bits(surface->format); + bmi.bmiHeader.biBitCount = util_format_get_bits(surface->format); bmi.bmiHeader.biCompression = BI_RGB; bmi.bmiHeader.biSizeImage = 0; bmi.bmiHeader.biXPelsPerMeter = 0; diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 6084ded72d1..b41b36f98fd 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -834,7 +834,7 @@ decompress_with_blit(GLcontext * ctx, GLenum target, GLint level, /* copy/pack data into user buffer */ if (st_equal_formats(stImage->pt->format, format, type)) { /* memcpy */ - const uint bytesPerRow = width * pf_get_size(stImage->pt->format); + const uint bytesPerRow = width * util_format_get_size(stImage->pt->format); ubyte *map = screen->transfer_map(screen, tex_xfer); GLuint row; for (row = 0; row < height; row++) { -- cgit v1.2.3 From 7679447b5835fd73ab44b3d77b12a034c95af5c5 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Wed, 2 Dec 2009 17:15:27 +0100 Subject: r300g, radeong: fix the CS overflow --- src/gallium/drivers/r300/r300_cs.h | 2 +- src/gallium/drivers/r300/r300_emit.c | 9 ++++++++- src/gallium/winsys/drm/radeon/core/radeon_r300.c | 5 +++-- 3 files changed, 12 insertions(+), 4 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_cs.h b/src/gallium/drivers/r300/r300_cs.h index 8b100375fdf..9fcf3ab538c 100644 --- a/src/gallium/drivers/r300/r300_cs.h +++ b/src/gallium/drivers/r300/r300_cs.h @@ -55,7 +55,7 @@ int cs_count = 0; #define CHECK_CS(size) \ - cs_winsys->check_cs(cs_winsys, (size)) + assert(cs_winsys->check_cs(cs_winsys, (size))) #define BEGIN_CS(size) do { \ CHECK_CS(size); \ diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index a479842f9eb..3bb42f9e434 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -871,10 +871,17 @@ void r300_emit_dirty_state(struct r300_context* r300) return; } + /* Check size of CS. */ + /* Make sure we have at least 8*1024 spare dwords. */ + /* XXX It would be nice to know the number of dwords we really need to + * XXX emit. */ + if (!r300->winsys->check_cs(r300->winsys, 8*1024)) { + r300->context.flush(&r300->context, 0, NULL); + } + /* Clean out BOs. */ r300->winsys->reset_bos(r300->winsys); - /* XXX check size */ validate: /* Color buffers... */ for (i = 0; i < r300->framebuffer_state.nr_cbufs; i++) { diff --git a/src/gallium/winsys/drm/radeon/core/radeon_r300.c b/src/gallium/winsys/drm/radeon/core/radeon_r300.c index 7362279b77a..ba0596c30dc 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_r300.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_r300.c @@ -52,8 +52,9 @@ static boolean radeon_validate(struct radeon_winsys* winsys) static boolean radeon_check_cs(struct radeon_winsys* winsys, int size) { - /* XXX check size here, lazy ass! */ - return radeon_validate(winsys); + struct radeon_cs* cs = winsys->priv->cs; + + return radeon_validate(winsys) && cs->cdw + size <= cs->ndw; } static void radeon_begin_cs(struct radeon_winsys* winsys, -- cgit v1.2.3 From 042b524d48ebb15215430149b9b1653f4b46dee3 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Fri, 4 Dec 2009 15:54:29 +0100 Subject: radeong: flush CS if a buffer being mapped is referenced by it Also, overlapping occlusion queries seems to work now. --- src/gallium/drivers/r300/r300_emit.c | 2 -- src/gallium/winsys/drm/radeon/core/radeon_buffer.c | 5 +++++ 2 files changed, 5 insertions(+), 2 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 3bb42f9e434..60be03f54fc 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -382,8 +382,6 @@ static void r300_emit_query_start(struct r300_context *r300) if (!query) return; - /* XXX This will almost certainly not return good results - * for overlapping queries. */ BEGIN_CS(4); if (caps->family == CHIP_FAMILY_RV530) { OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL); diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c index 0ca7b392554..2a8daed051d 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c @@ -140,10 +140,15 @@ static void *radeon_buffer_map(struct pipe_winsys *ws, struct pipe_buffer *buffer, unsigned flags) { + struct radeon_winsys_priv *priv = ((struct radeon_winsys *)ws)->priv; struct radeon_pipe_buffer *radeon_buffer = (struct radeon_pipe_buffer*)buffer; int write = 0; + if (radeon_bo_is_referenced_by_cs(radeon_buffer->bo, priv->cs)) { + priv->cs->space_flush_fn(priv->cs->space_flush_data); + } + if (flags & PIPE_BUFFER_USAGE_DONTBLOCK) { uint32_t domain; -- cgit v1.2.3 From 56a4342a0493ad1d502d4791ab941ef171d36e60 Mon Sep 17 00:00:00 2001 From: Michel Dänzer Date: Sat, 5 Dec 2009 17:48:00 +0100 Subject: r300g: Need to emit a hardware scissor rectangle even if scissor is disabled. Just make it cover the whole framebuffer in that case. Otherwise the kernel CS checker may complain, e.g. running progs/demos/gearbox. That runs fast now here, but doesn't look right yet. --- src/gallium/drivers/r300/r300_context.h | 2 ++ src/gallium/drivers/r300/r300_emit.c | 9 +++++++-- src/gallium/drivers/r300/r300_state.c | 19 +++++++++++++++++++ 3 files changed, 28 insertions(+), 2 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index dd3f6ac1432..11cd9f855f3 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -101,6 +101,8 @@ struct r300_sampler_state { struct r300_scissor_state { uint32_t scissor_top_left; /* R300_SC_SCISSORS_TL: 0x43e0 */ uint32_t scissor_bottom_right; /* R300_SC_SCISSORS_BR: 0x43e4 */ + uint32_t no_scissor_top_left; /* R300_SC_SCISSORS_TL: 0x43e0 */ + uint32_t no_scissor_bottom_right; /* R300_SC_SCISSORS_BR: 0x43e4 */ }; struct r300_texture_state { diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 60be03f54fc..04dca292167 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -570,8 +570,13 @@ void r300_emit_scissor_state(struct r300_context* r300, BEGIN_CS(3); OUT_CS_REG_SEQ(R300_SC_SCISSORS_TL, 2); - OUT_CS(scissor->scissor_top_left); - OUT_CS(scissor->scissor_bottom_right); + if (r300->rs_state->rs.scissor) { + OUT_CS(scissor->scissor_top_left); + OUT_CS(scissor->scissor_bottom_right); + } else { + OUT_CS(scissor->no_scissor_top_left); + OUT_CS(scissor->no_scissor_bottom_right); + } END_CS; } diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index 442af70e143..2bc2b79c021 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -302,6 +302,25 @@ static void r300->framebuffer_state = *state; r300->dirty_state |= R300_NEW_FRAMEBUFFERS; + + if (r300_screen(r300->context.screen)->caps->is_r500) { + r300->scissor_state->no_scissor_top_left = + (0 << R300_SCISSORS_X_SHIFT) | + (0 << R300_SCISSORS_Y_SHIFT); + r300->scissor_state->no_scissor_bottom_right = + ((state->width - 1) << R300_SCISSORS_X_SHIFT) | + ((state->height - 1) << R300_SCISSORS_Y_SHIFT); + } else { + /* Offset of 1440 in non-R500 chipsets. */ + r300->scissor_state->no_scissor_top_left = + ((0 + 1440) << R300_SCISSORS_X_SHIFT) | + ((0 + 1440) << R300_SCISSORS_Y_SHIFT); + r300->scissor_state->no_scissor_bottom_right = + (((state->width - 1) + 1440) << R300_SCISSORS_X_SHIFT) | + (((state->height - 1) + 1440) << R300_SCISSORS_Y_SHIFT); + } + + r300->dirty_state |= R300_NEW_SCISSOR; } /* Create fragment shader state. */ -- cgit v1.2.3 From e1380cae885df37d4a211d0271f59487d9f2db78 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Sat, 5 Dec 2009 19:17:20 +0100 Subject: r300g: remove redundant code and clean up --- src/gallium/drivers/r300/r300_context.h | 11 +++-- src/gallium/drivers/r300/r300_emit.c | 19 +++++---- src/gallium/drivers/r300/r300_state.c | 73 +++++++++++++++++---------------- 3 files changed, 57 insertions(+), 46 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index 11cd9f855f3..23ea32c57ed 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -98,11 +98,14 @@ struct r300_sampler_state { unsigned min_lod, max_lod; }; +struct r300_scissor_regs { + uint32_t top_left; /* R300_SC_SCISSORS_TL: 0x43e0 */ + uint32_t bottom_right; /* R300_SC_SCISSORS_BR: 0x43e4 */ +}; + struct r300_scissor_state { - uint32_t scissor_top_left; /* R300_SC_SCISSORS_TL: 0x43e0 */ - uint32_t scissor_bottom_right; /* R300_SC_SCISSORS_BR: 0x43e4 */ - uint32_t no_scissor_top_left; /* R300_SC_SCISSORS_TL: 0x43e0 */ - uint32_t no_scissor_bottom_right; /* R300_SC_SCISSORS_BR: 0x43e4 */ + struct r300_scissor_regs framebuffer; + struct r300_scissor_regs scissor; }; struct r300_texture_state { diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 04dca292167..dbf316a9b57 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -563,21 +563,26 @@ void r300_emit_rs_block_state(struct r300_context* r300, END_CS; } -void r300_emit_scissor_state(struct r300_context* r300, - struct r300_scissor_state* scissor) +static void r300_emit_scissor_regs(struct r300_context* r300, + struct r300_scissor_regs* scissor) { CS_LOCALS(r300); BEGIN_CS(3); OUT_CS_REG_SEQ(R300_SC_SCISSORS_TL, 2); + OUT_CS(scissor->top_left); + OUT_CS(scissor->bottom_right); + END_CS; +} + +void r300_emit_scissor_state(struct r300_context* r300, + struct r300_scissor_state* scissor) +{ if (r300->rs_state->rs.scissor) { - OUT_CS(scissor->scissor_top_left); - OUT_CS(scissor->scissor_bottom_right); + r300_emit_scissor_regs(r300, &scissor->scissor); } else { - OUT_CS(scissor->no_scissor_top_left); - OUT_CS(scissor->no_scissor_bottom_right); + r300_emit_scissor_regs(r300, &scissor->framebuffer); } - END_CS; } void r300_emit_texture(struct r300_context* r300, diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index 2bc2b79c021..d3233557cea 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -289,11 +289,34 @@ static void r300_set_edgeflags(struct pipe_context* pipe, /* XXX and even worse, I have no idea WTF the bitfield is */ } +static void r300_set_scissor_regs(const struct pipe_scissor_state* state, + struct r300_scissor_regs *scissor, + boolean is_r500) +{ + if (is_r500) { + scissor->top_left = + (state->minx << R300_SCISSORS_X_SHIFT) | + (state->miny << R300_SCISSORS_Y_SHIFT); + scissor->bottom_right = + ((state->maxx - 1) << R300_SCISSORS_X_SHIFT) | + ((state->maxy - 1) << R300_SCISSORS_Y_SHIFT); + } else { + /* Offset of 1440 in non-R500 chipsets. */ + scissor->top_left = + ((state->minx + 1440) << R300_SCISSORS_X_SHIFT) | + ((state->miny + 1440) << R300_SCISSORS_Y_SHIFT); + scissor->bottom_right = + (((state->maxx - 1) + 1440) << R300_SCISSORS_X_SHIFT) | + (((state->maxy - 1) + 1440) << R300_SCISSORS_Y_SHIFT); + } +} + static void r300_set_framebuffer_state(struct pipe_context* pipe, const struct pipe_framebuffer_state* state) { struct r300_context* r300 = r300_context(pipe); + struct pipe_scissor_state scissor; if (r300->draw) { draw_flush(r300->draw); @@ -301,26 +324,17 @@ static void r300->framebuffer_state = *state; - r300->dirty_state |= R300_NEW_FRAMEBUFFERS; + scissor.minx = scissor.miny = 0; + scissor.maxx = state->width; + scissor.maxy = state->height; + r300_set_scissor_regs(&scissor, &r300->scissor_state->framebuffer, + r300_screen(r300->context.screen)->caps->is_r500); - if (r300_screen(r300->context.screen)->caps->is_r500) { - r300->scissor_state->no_scissor_top_left = - (0 << R300_SCISSORS_X_SHIFT) | - (0 << R300_SCISSORS_Y_SHIFT); - r300->scissor_state->no_scissor_bottom_right = - ((state->width - 1) << R300_SCISSORS_X_SHIFT) | - ((state->height - 1) << R300_SCISSORS_Y_SHIFT); - } else { - /* Offset of 1440 in non-R500 chipsets. */ - r300->scissor_state->no_scissor_top_left = - ((0 + 1440) << R300_SCISSORS_X_SHIFT) | - ((0 + 1440) << R300_SCISSORS_Y_SHIFT); - r300->scissor_state->no_scissor_bottom_right = - (((state->width - 1) + 1440) << R300_SCISSORS_X_SHIFT) | - (((state->height - 1) + 1440) << R300_SCISSORS_Y_SHIFT); + /* Don't rely on the order of states being set for the first time. */ + if (!r300->rs_state || !r300->rs_state->rs.scissor) { + r300->dirty_state |= R300_NEW_SCISSOR; } - - r300->dirty_state |= R300_NEW_SCISSOR; + r300->dirty_state |= R300_NEW_FRAMEBUFFERS; } /* Create fragment shader state. */ @@ -642,24 +656,13 @@ static void r300_set_scissor_state(struct pipe_context* pipe, { struct r300_context* r300 = r300_context(pipe); - if (r300_screen(r300->context.screen)->caps->is_r500) { - r300->scissor_state->scissor_top_left = - (state->minx << R300_SCISSORS_X_SHIFT) | - (state->miny << R300_SCISSORS_Y_SHIFT); - r300->scissor_state->scissor_bottom_right = - ((state->maxx - 1) << R300_SCISSORS_X_SHIFT) | - ((state->maxy - 1) << R300_SCISSORS_Y_SHIFT); - } else { - /* Offset of 1440 in non-R500 chipsets. */ - r300->scissor_state->scissor_top_left = - ((state->minx + 1440) << R300_SCISSORS_X_SHIFT) | - ((state->miny + 1440) << R300_SCISSORS_Y_SHIFT); - r300->scissor_state->scissor_bottom_right = - (((state->maxx - 1) + 1440) << R300_SCISSORS_X_SHIFT) | - (((state->maxy - 1) + 1440) << R300_SCISSORS_Y_SHIFT); - } + r300_set_scissor_regs(state, &r300->scissor_state->scissor, + r300_screen(r300->context.screen)->caps->is_r500); - r300->dirty_state |= R300_NEW_SCISSOR; + /* Don't rely on the order of states being set for the first time. */ + if (!r300->rs_state || r300->rs_state->rs.scissor) { + r300->dirty_state |= R300_NEW_SCISSOR; + } } static void r300_set_viewport_state(struct pipe_context* pipe, -- cgit v1.2.3 From 8cc570a48c2e8e18622027cbd76f16a746b430bc Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Wed, 9 Dec 2009 00:55:51 +0100 Subject: r300g: clean up r300_emit_aos --- src/gallium/drivers/r300/r300_cs.h | 9 +++++ src/gallium/drivers/r300/r300_emit.c | 71 +++++++++++++++++++++++------------- src/gallium/drivers/r300/r300_reg.h | 5 +++ 3 files changed, 59 insertions(+), 26 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_cs.h b/src/gallium/drivers/r300/r300_cs.h index 9fcf3ab538c..d142fee0502 100644 --- a/src/gallium/drivers/r300/r300_cs.h +++ b/src/gallium/drivers/r300/r300_cs.h @@ -115,6 +115,15 @@ cs_count -= 3; \ } while (0) +#define OUT_CS_RELOC_NO_OFFSET(bo, rd, wd, flags) do { \ + DBG(cs_context_copy, DBG_CS, "r300: writing relocation for buffer %p, " \ + "domains (%d, %d, %d)\n", \ + bo, rd, wd, flags); \ + assert(bo); \ + cs_winsys->write_cs_reloc(cs_winsys, bo, rd, wd, flags); \ + cs_count -= 2; \ +} while (0) + #define END_CS do { \ if (VERY_VERBOSE_CS) { \ DBG(cs_context_copy, DBG_CS, "r300: END_CS in %s (%s:%d)\n", __FUNCTION__, \ diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index dbf316a9b57..7620c73cac5 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -1,5 +1,6 @@ /* * Copyright 2008 Corbin Simpson + * Copyright 2009 Marek Olšák * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -623,50 +624,68 @@ void r300_emit_texture(struct r300_context* r300, END_CS; } -/* XXX I can't read this and that's not good */ -void r300_emit_aos(struct r300_context* r300, unsigned offset) +static boolean r300_validate_aos(struct r300_context *r300) { struct pipe_vertex_buffer *vbuf = r300->vertex_buffer; struct pipe_vertex_element *velem = r300->vertex_element; - CS_LOCALS(r300); int i; - unsigned aos_count = r300->vertex_element_count; + /* Check if formats and strides are aligned to the size of DWORD. */ + for (i = 0; i < r300->vertex_element_count; i++) { + if (vbuf[velem[i].vertex_buffer_index].stride % 4 != 0 || + pf_get_blocksize(velem[i].src_format) % 4 != 0) { + return FALSE; + } + } + return TRUE; +} + +void r300_emit_aos(struct r300_context* r300, unsigned offset) +{ + struct pipe_vertex_buffer *vb1, *vb2, *vbuf = r300->vertex_buffer; + struct pipe_vertex_element *velem = r300->vertex_element; + int i; + unsigned size1, size2, aos_count = r300->vertex_element_count; unsigned packet_size = (aos_count * 3 + 1) / 2; + CS_LOCALS(r300); + + /* XXX Move this checking to a more approriate place. */ + if (!r300_validate_aos(r300)) { + /* XXX We should fallback using Draw. */ + assert(0); + } + BEGIN_CS(2 + packet_size + aos_count * 2); OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, packet_size); OUT_CS(aos_count); + for (i = 0; i < aos_count - 1; i += 2) { - int buf_num1 = velem[i].vertex_buffer_index; - int buf_num2 = velem[i+1].vertex_buffer_index; - assert(vbuf[buf_num1].stride % 4 == 0 && pf_get_blocksize(velem[i].src_format) % 4 == 0); - assert(vbuf[buf_num2].stride % 4 == 0 && pf_get_blocksize(velem[i+1].src_format) % 4 == 0); - OUT_CS((pf_get_blocksize(velem[i].src_format) >> 2) | (vbuf[buf_num1].stride << 6) | - (pf_get_blocksize(velem[i+1].src_format) << 14) | (vbuf[buf_num2].stride << 22)); - OUT_CS(vbuf[buf_num1].buffer_offset + velem[i].src_offset + - offset * vbuf[buf_num1].stride); - OUT_CS(vbuf[buf_num2].buffer_offset + velem[i+1].src_offset + - offset * vbuf[buf_num2].stride); + vb1 = &vbuf[velem[i].vertex_buffer_index]; + vb2 = &vbuf[velem[i+1].vertex_buffer_index]; + size1 = pf_get_blocksize(velem[i].src_format); + size2 = pf_get_blocksize(velem[i+1].src_format); + + OUT_CS(R300_VBPNTR_SIZE0(size1) | R300_VBPNTR_STRIDE0(vb1->stride) | + R300_VBPNTR_SIZE1(size2) | R300_VBPNTR_STRIDE1(vb2->stride)); + OUT_CS(vb1->buffer_offset + velem[i].src_offset + offset * vb1->stride); + OUT_CS(vb2->buffer_offset + velem[i+1].src_offset + offset * vb2->stride); } + if (aos_count & 1) { - int buf_num = velem[i].vertex_buffer_index; - assert(vbuf[buf_num].stride % 4 == 0 && pf_get_blocksize(velem[i].src_format) % 4 == 0); - OUT_CS((pf_get_blocksize(velem[i].src_format) >> 2) | (vbuf[buf_num].stride << 6)); - OUT_CS(vbuf[buf_num].buffer_offset + velem[i].src_offset + - offset * vbuf[buf_num].stride); + vb1 = &vbuf[velem[i].vertex_buffer_index]; + size1 = pf_get_blocksize(velem[i].src_format); + + OUT_CS(R300_VBPNTR_SIZE0(size1) | R300_VBPNTR_STRIDE0(vb1->stride)); + OUT_CS(vb1->buffer_offset + velem[i].src_offset + offset * vb1->stride); } - /* XXX bare CS reloc */ for (i = 0; i < aos_count; i++) { - cs_winsys->write_cs_reloc(cs_winsys, - vbuf[velem[i].vertex_buffer_index].buffer, - RADEON_GEM_DOMAIN_GTT, - 0, - 0); - cs_count -= 2; + OUT_CS_RELOC_NO_OFFSET(vbuf[velem[i].vertex_buffer_index].buffer, + RADEON_GEM_DOMAIN_GTT, 0, 0); } END_CS; } + #if 0 void r300_emit_draw_packet(struct r300_context* r300) { diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h index 85b1ea568a3..c1ea87d11e9 100644 --- a/src/gallium/drivers/r300/r300_reg.h +++ b/src/gallium/drivers/r300/r300_reg.h @@ -3293,6 +3293,11 @@ enum { */ #define R300_PACKET3_3D_LOAD_VBPNTR 0x00002F00 +# define R300_VBPNTR_SIZE0(x) ((x) >> 2) +# define R300_VBPNTR_STRIDE0(x) (((x) >> 2) << 8) +# define R300_VBPNTR_SIZE1(x) (((x) >> 2) << 16) +# define R300_VBPNTR_STRIDE1(x) (((x) >> 2) << 24) + #define R300_PACKET3_INDX_BUFFER 0x00003300 # define R300_INDX_BUFFER_DST_SHIFT 0 # define R300_INDX_BUFFER_SKIP_SHIFT 16 -- cgit v1.2.3 From 87b822e024797ef2fdb51ec9364f21eeb4d07161 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Tue, 8 Dec 2009 04:55:32 +0100 Subject: r300g: make pow(0,0) return 1 instead of NaN in the R500 fragment shader Unfortunately we can't fix this easily in the R300 fragment shader, and it's probably not worth the effort. --- src/gallium/drivers/r300/r300_emit.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 7620c73cac5..55bc2b3528a 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -277,7 +277,7 @@ void r500_emit_fragment_program_code(struct r300_context* r300, BEGIN_CS(13 + ((code->inst_end + 1) * 6)); - OUT_CS_REG(R500_US_CONFIG, 0); + OUT_CS_REG(R500_US_CONFIG, R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO); OUT_CS_REG(R500_US_PIXSIZE, code->max_temp_idx); OUT_CS_REG(R500_US_CODE_RANGE, R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(code->inst_end)); -- cgit v1.2.3 From 6de7ac73bf027b9ace6f5f0c8063cbf724d95cee Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Tue, 8 Dec 2009 21:53:19 +0100 Subject: r300g: always disable unused colorbuffers --- src/gallium/drivers/r300/r300_emit.c | 15 ++++++++++++++- 1 file changed, 14 insertions(+), 1 deletion(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 55bc2b3528a..f784e1fa8e5 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -331,7 +331,13 @@ void r300_emit_fb_state(struct r300_context* r300, int i; CS_LOCALS(r300); - BEGIN_CS((10 * fb->nr_cbufs) + (fb->zsbuf ? 10 : 0) + 4); + /* Shouldn't fail unless there is a bug in the state tracker. */ + assert(fb->nr_cbufs <= 4); + + BEGIN_CS((10 * fb->nr_cbufs) + (2 * (4 - fb->nr_cbufs)) + + (fb->zsbuf ? 10 : 0) + 4); + + /* Flush and free renderbuffer caches. */ OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS | R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D); @@ -339,6 +345,7 @@ void r300_emit_fb_state(struct r300_context* r300, R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE | R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE); + /* Set up colorbuffers. */ for (i = 0; i < fb->nr_cbufs; i++) { surf = fb->cbufs[i]; tex = (struct r300_texture*)surf->texture; @@ -356,6 +363,12 @@ void r300_emit_fb_state(struct r300_context* r300, r300_translate_out_fmt(surf->format)); } + /* Disable unused colorbuffers. */ + for (; i < 4; i++) { + OUT_CS_REG(R300_US_OUT_FMT_0 + (4 * i), R300_US_OUT_FMT_UNUSED); + } + + /* Set up a zbuffer. */ if (fb->zsbuf) { surf = fb->zsbuf; tex = (struct r300_texture*)surf->texture; -- cgit v1.2.3 From cf85bf9cd0c168caed6210a896df285c3d86db03 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Mon, 14 Dec 2009 05:47:54 +0100 Subject: r300g: set the number of colorbuffers in RB3D_CCTL --- src/gallium/drivers/r300/r300_emit.c | 5 ++++- src/gallium/drivers/r300/r300_reg.h | 1 + src/gallium/drivers/r300/r300_state_invariant.c | 3 +-- 3 files changed, 6 insertions(+), 3 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index f784e1fa8e5..9644efb7174 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -335,7 +335,7 @@ void r300_emit_fb_state(struct r300_context* r300, assert(fb->nr_cbufs <= 4); BEGIN_CS((10 * fb->nr_cbufs) + (2 * (4 - fb->nr_cbufs)) + - (fb->zsbuf ? 10 : 0) + 4); + (fb->zsbuf ? 10 : 0) + 6); /* Flush and free renderbuffer caches. */ OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT, @@ -345,6 +345,9 @@ void r300_emit_fb_state(struct r300_context* r300, R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE | R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE); + /* Set the number of colorbuffers. */ + OUT_CS_REG(R300_RB3D_CCTL, R300_RB3D_CCTL_NUM_MULTIWRITES(fb->nr_cbufs)); + /* Set up colorbuffers. */ for (i = 0; i < fb->nr_cbufs; i++) { surf = fb->cbufs[i]; diff --git a/src/gallium/drivers/r300/r300_reg.h b/src/gallium/drivers/r300/r300_reg.h index c1ea87d11e9..d8d08fbe264 100644 --- a/src/gallium/drivers/r300/r300_reg.h +++ b/src/gallium/drivers/r300/r300_reg.h @@ -2145,6 +2145,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE. /* Unpipelined. */ #define R300_RB3D_CCTL 0x4e00 +# define R300_RB3D_CCTL_NUM_MULTIWRITES(x) (MAX2(((x)-1), 0) << 5) # define R300_RB3D_CCTL_NUM_MULTIWRITES_1_BUFFER (0 << 5) # define R300_RB3D_CCTL_NUM_MULTIWRITES_2_BUFFERS (1 << 5) # define R300_RB3D_CCTL_NUM_MULTIWRITES_3_BUFFERS (2 << 5) diff --git a/src/gallium/drivers/r300/r300_state_invariant.c b/src/gallium/drivers/r300/r300_state_invariant.c index 46d1cb39b54..3320d43b275 100644 --- a/src/gallium/drivers/r300/r300_state_invariant.c +++ b/src/gallium/drivers/r300/r300_state_invariant.c @@ -84,7 +84,7 @@ void r300_emit_invariant_state(struct r300_context* r300) END_CS; /* XXX unsorted stuff from surface_fill */ - BEGIN_CS(56 + (caps->has_tcl ? 5 : 0) + (caps->is_r500 ? 4 : 0)); + BEGIN_CS(54 + (caps->has_tcl ? 5 : 0) + (caps->is_r500 ? 4 : 0)); /* Flush PVS. */ OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0); @@ -123,7 +123,6 @@ void r300_emit_invariant_state(struct r300_context* r300) OUT_CS_REG(R300_SU_DEPTH_OFFSET, 0x00000000); OUT_CS_REG(R300_SC_HYPERZ, 0x0000001C); OUT_CS_REG(R300_SC_EDGERULE, 0x2DA49525); - OUT_CS_REG(R300_RB3D_CCTL, 0x00000000); OUT_CS_REG(R300_RB3D_AARESOLVE_CTL, 0x00000000); if (caps->is_r500) { OUT_CS_REG(R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 0x00000000); -- cgit v1.2.3 From a6d701d1c6ed8e0a649d62104aeded8fb25c66d8 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Mon, 14 Dec 2009 05:29:12 +0100 Subject: r300g: if no colorbuffers are set, disable blending and set the color mask to 0 This seems to be the only way to disable the first colorbuffer. --- src/gallium/drivers/r300/r300_emit.c | 13 ++++++++++--- src/gallium/drivers/r300/r300_state.c | 1 + 2 files changed, 11 insertions(+), 3 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 9644efb7174..55e4f94afe2 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -41,9 +41,16 @@ void r300_emit_blend_state(struct r300_context* r300, CS_LOCALS(r300); BEGIN_CS(8); OUT_CS_REG_SEQ(R300_RB3D_CBLEND, 3); - OUT_CS(blend->blend_control); - OUT_CS(blend->alpha_blend_control); - OUT_CS(blend->color_channel_mask); + if (r300->framebuffer_state.nr_cbufs) { + OUT_CS(blend->blend_control); + OUT_CS(blend->alpha_blend_control); + OUT_CS(blend->color_channel_mask); + } else { + OUT_CS(0); + OUT_CS(0); + OUT_CS(0); + /* XXX also disable fastfill here once it's supported */ + } OUT_CS_REG(R300_RB3D_ROPCNTL, blend->rop); OUT_CS_REG(R300_RB3D_DITHER_CTL, blend->dither); END_CS; diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index 3cfa2e63f90..91cf972edee 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -339,6 +339,7 @@ static void r300->dirty_state |= R300_NEW_SCISSOR; } r300->dirty_state |= R300_NEW_FRAMEBUFFERS; + r300->dirty_state |= R300_NEW_BLEND; } /* Create fragment shader state. */ -- cgit v1.2.3 From 2ddee2cfbcef59128b2a251d5391ddc2b4aea4bb Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Tue, 15 Dec 2009 05:35:03 +0100 Subject: r300g: fix emission of which textures are enabled It fixes most of the "Bad CS" issues in piglit/texCombine and piglit/fbo. Some other issues of this kind will get fixed in the kernel soon (depth-only rendering, S3TC, and RGTC). --- src/gallium/drivers/r300/r300_emit.c | 13 ++++++++++++- 1 file changed, 12 insertions(+), 1 deletion(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 55e4f94afe2..55c8aa07bdf 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -883,10 +883,21 @@ void r300_emit_viewport_state(struct r300_context* r300, void r300_emit_texture_count(struct r300_context* r300) { + uint32_t tx_enable = 0; + int i; CS_LOCALS(r300); + /* Notice that texture_count and sampler_count are just sizes + * of the respective arrays. We still have to check for the individual + * elements. */ + for (i = 0; i < MIN2(r300->sampler_count, r300->texture_count); i++) { + if (r300->textures[i]) { + tx_enable |= 1 << i; + } + } + BEGIN_CS(2); - OUT_CS_REG(R300_TX_ENABLE, (1 << r300->texture_count) - 1); + OUT_CS_REG(R300_TX_ENABLE, tx_enable); END_CS; } -- cgit v1.2.3 From b1ed72ebe2599ec178f51d86fd42f26486b9a19b Mon Sep 17 00:00:00 2001 From: Michal Krol Date: Thu, 17 Dec 2009 23:41:57 +0100 Subject: Move the remaining format pf_get_* functions to u_format.h. Previously they depended on format blocks, but after removing those they started depending on format encoding. --- progs/rbug/bin_to_bmp.c | 13 +- src/gallium/auxiliary/util/u_blitter.c | 4 +- src/gallium/auxiliary/util/u_debug.c | 7 +- src/gallium/auxiliary/util/u_format.h | 83 ++++++++++++- src/gallium/auxiliary/util/u_gen_mipmap.c | 12 +- src/gallium/auxiliary/util/u_rect.c | 21 ++-- src/gallium/auxiliary/util/u_tile.c | 9 +- src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c | 4 +- src/gallium/drivers/cell/ppu/cell_texture.c | 20 +-- src/gallium/drivers/i915/i915_surface.c | 19 +-- src/gallium/drivers/i915/i915_texture.c | 75 ++++++------ src/gallium/drivers/llvmpipe/lp_setup.c | 3 +- src/gallium/drivers/llvmpipe/lp_texture.c | 18 +-- src/gallium/drivers/nv04/nv04_surface_2d.c | 13 +- src/gallium/drivers/nv04/nv04_transfer.c | 3 +- src/gallium/drivers/nv10/nv10_miptree.c | 5 +- src/gallium/drivers/nv10/nv10_transfer.c | 3 +- src/gallium/drivers/nv20/nv20_miptree.c | 5 +- src/gallium/drivers/nv20/nv20_transfer.c | 3 +- src/gallium/drivers/nv30/nv30_miptree.c | 5 +- src/gallium/drivers/nv30/nv30_transfer.c | 3 +- src/gallium/drivers/nv40/nv40_miptree.c | 5 +- src/gallium/drivers/nv40/nv40_transfer.c | 3 +- src/gallium/drivers/nv50/nv50_miptree.c | 9 +- src/gallium/drivers/nv50/nv50_transfer.c | 23 ++-- src/gallium/drivers/r300/r300_emit.c | 8 +- src/gallium/drivers/r300/r300_screen.c | 5 +- src/gallium/drivers/r300/r300_texture.c | 8 +- src/gallium/drivers/softpipe/sp_texture.c | 14 +-- src/gallium/drivers/softpipe/sp_tile_cache.c | 2 +- src/gallium/drivers/svga/svga_screen_texture.c | 14 +-- src/gallium/drivers/svga/svga_state_vs.c | 2 +- src/gallium/drivers/trace/tr_rbug.c | 15 +-- src/gallium/drivers/trace/tr_screen.c | 3 +- src/gallium/include/pipe/p_format.h | 134 --------------------- src/gallium/state_trackers/python/st_sample.c | 15 +-- .../state_trackers/python/st_softpipe_winsys.c | 4 +- .../winsys/drm/nouveau/drm/nouveau_drm_api.c | 2 +- src/gallium/winsys/drm/radeon/core/radeon_buffer.c | 5 +- src/gallium/winsys/egl_xlib/sw_winsys.c | 4 +- src/gallium/winsys/g3dvl/xlib/xsp_winsys.c | 4 +- src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c | 4 +- src/gallium/winsys/gdi/gdi_softpipe_winsys.c | 8 +- src/gallium/winsys/xlib/xlib_cell.c | 4 +- src/gallium/winsys/xlib/xlib_llvmpipe.c | 10 +- src/gallium/winsys/xlib/xlib_softpipe.c | 10 +- src/mesa/state_tracker/st_cb_drawpixels.c | 4 +- src/mesa/state_tracker/st_cb_fbo.c | 4 +- src/mesa/state_tracker/st_cb_texture.c | 26 ++-- src/mesa/state_tracker/st_gen_mipmap.c | 5 +- 50 files changed, 322 insertions(+), 360 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/progs/rbug/bin_to_bmp.c b/progs/rbug/bin_to_bmp.c index cdae3486ce0..03ff622fee7 100644 --- a/progs/rbug/bin_to_bmp.c +++ b/progs/rbug/bin_to_bmp.c @@ -25,6 +25,7 @@ #include "pipe/p_compiler.h" #include "pipe/p_format.h" #include "pipe/p_state.h" +#include "util/u_format.h" #include "util/u_memory.h" #include "util/u_debug.h" #include "util/u_network.h" @@ -54,10 +55,7 @@ static void dump(unsigned width, unsigned height, unsigned src_stride, enum pipe_format src_format, uint8_t *data, unsigned src_size) { - struct pipe_format_block src_block; - enum pipe_format dst_format = PIPE_FORMAT_R32G32B32A32_FLOAT; - struct pipe_format_block dst_block; unsigned dst_stride; unsigned dst_size; float *rgba; @@ -65,14 +63,11 @@ static void dump(unsigned width, unsigned height, char filename[512]; { - pf_get_block(src_format, &src_block); - assert(src_stride >= pf_get_stride(&src_block, width)); - assert(src_size >= pf_get_2d_size(&src_block, src_stride, width)); + assert(src_stride >= util_format_get_stride(src_format, width)); } { - pf_get_block(dst_format, &dst_block); - dst_stride = pf_get_stride(&dst_block, width); - dst_size = pf_get_2d_size(&dst_block, dst_stride, width); + dst_stride = util_format_get_stride(dst_format, width); + dst_size = util_format_get_2d_size(dst_format, dst_stride, width); rgba = MALLOC(dst_size); } diff --git a/src/gallium/auxiliary/util/u_blitter.c b/src/gallium/auxiliary/util/u_blitter.c index 895af2c8d00..0242b796155 100644 --- a/src/gallium/auxiliary/util/u_blitter.c +++ b/src/gallium/auxiliary/util/u_blitter.c @@ -579,8 +579,8 @@ void util_blitter_copy(struct blitter_context *blitter, if (!dst->texture || !src->texture) return; - is_depth = pf_get_component_bits(src->format, PIPE_FORMAT_COMP_Z) != 0; - is_stencil = pf_get_component_bits(src->format, PIPE_FORMAT_COMP_S) != 0; + is_depth = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 0) != 0; + is_stencil = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 1) != 0; dst_tex_usage = is_depth || is_stencil ? PIPE_TEXTURE_USAGE_DEPTH_STENCIL : PIPE_TEXTURE_USAGE_RENDER_TARGET; diff --git a/src/gallium/auxiliary/util/u_debug.c b/src/gallium/auxiliary/util/u_debug.c index 40633574b08..27e0b0d1595 100644 --- a/src/gallium/auxiliary/util/u_debug.c +++ b/src/gallium/auxiliary/util/u_debug.c @@ -64,6 +64,7 @@ #include "pipe/p_format.h" #include "pipe/p_state.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_memory.h" #include "util/u_string.h" #include "util/u_stream.h" @@ -670,9 +671,9 @@ void debug_dump_surface(const char *prefix, debug_dump_image(prefix, texture->format, - pf_get_blocksize(texture->format), - pf_get_nblocksx(texture->format, transfer->width), - pf_get_nblocksy(texture->format, transfer->height), + util_format_get_blocksize(texture->format), + util_format_get_nblocksx(texture->format, transfer->width), + util_format_get_nblocksy(texture->format, transfer->height), transfer->stride, data); diff --git a/src/gallium/auxiliary/util/u_format.h b/src/gallium/auxiliary/util/u_format.h index 72da2a44c74..97e4d959bc7 100644 --- a/src/gallium/auxiliary/util/u_format.h +++ b/src/gallium/auxiliary/util/u_format.h @@ -200,7 +200,7 @@ util_format_is_depth_and_stencil(enum pipe_format format) * Return total bits needed for the pixel format. */ static INLINE uint -util_format_get_bits(enum pipe_format format) +util_format_get_blocksizebits(enum pipe_format format) { const struct util_format_description *desc = util_format_description(format); @@ -216,15 +216,92 @@ util_format_get_bits(enum pipe_format format) * Return bytes per pixel for the given format. */ static INLINE uint -util_format_get_size(enum pipe_format format) +util_format_get_blocksize(enum pipe_format format) { - uint bits = util_format_get_bits(format); + uint bits = util_format_get_blocksizebits(format); assert(bits % 8 == 0); return bits / 8; } +static INLINE uint +util_format_get_blockwidth(enum pipe_format format) +{ + const struct util_format_description *desc = util_format_description(format); + + assert(format); + if (!format) { + return 1; + } + + switch (desc->layout) { + case UTIL_FORMAT_LAYOUT_YUV: + return 2; + case UTIL_FORMAT_LAYOUT_DXT: + return 4; + default: + return 1; + } +} + +static INLINE uint +util_format_get_blockheight(enum pipe_format format) +{ + const struct util_format_description *desc = util_format_description(format); + + assert(format); + if (!format) { + return 1; + } + + switch (desc->layout) { + case UTIL_FORMAT_LAYOUT_DXT: + return 4; + default: + return 1; + } +} + +static INLINE unsigned +util_format_get_nblocksx(enum pipe_format format, + unsigned x) +{ + unsigned blockwidth = util_format_get_blockwidth(format); + return (x + blockwidth - 1) / blockwidth; +} + +static INLINE unsigned +util_format_get_nblocksy(enum pipe_format format, + unsigned y) +{ + unsigned blockheight = util_format_get_blockheight(format); + return (y + blockheight - 1) / blockheight; +} + +static INLINE unsigned +util_format_get_nblocks(enum pipe_format format, + unsigned width, + unsigned height) +{ + return util_format_get_nblocksx(format, width) * util_format_get_nblocksy(format, height); +} + +static INLINE size_t +util_format_get_stride(enum pipe_format format, + unsigned width) +{ + return util_format_get_nblocksx(format, width) * util_format_get_blocksize(format); +} + +static INLINE size_t +util_format_get_2d_size(enum pipe_format format, + size_t stride, + unsigned height) +{ + return util_format_get_nblocksy(format, height) * stride; +} + static INLINE uint util_format_get_component_bits(enum pipe_format format, enum util_format_colorspace colorspace, diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c index 2931dfac478..0dad6ccbc0a 100644 --- a/src/gallium/auxiliary/util/u_gen_mipmap.c +++ b/src/gallium/auxiliary/util/u_gen_mipmap.c @@ -998,7 +998,7 @@ reduce_2d(enum pipe_format pformat, { enum dtype datatype; uint comps; - const int bpt = util_format_get_size(pformat); + const int bpt = util_format_get_blocksize(pformat); const ubyte *srcA, *srcB; ubyte *dst; int row; @@ -1037,7 +1037,7 @@ reduce_3d(enum pipe_format pformat, int dstWidth, int dstHeight, int dstDepth, int dstRowStride, ubyte *dstPtr) { - const int bpt = util_format_get_size(pformat); + const int bpt = util_format_get_blocksize(pformat); const int border = 0; int img, row; int bytesPerSrcImage, bytesPerDstImage; @@ -1161,8 +1161,8 @@ make_2d_mipmap(struct gen_mipmap_state *ctx, const uint zslice = 0; uint dstLevel; - assert(pf_get_blockwidth(pt->format) == 1); - assert(pf_get_blockheight(pt->format) == 1); + assert(util_format_get_blockwidth(pt->format) == 1); + assert(util_format_get_blockheight(pt->format) == 1); for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { const uint srcLevel = dstLevel - 1; @@ -1206,8 +1206,8 @@ make_3d_mipmap(struct gen_mipmap_state *ctx, struct pipe_screen *screen = pipe->screen; uint dstLevel, zslice = 0; - assert(pf_get_blockwidth(pt->format) == 1); - assert(pf_get_blockheight(pt->format) == 1); + assert(util_format_get_blockwidth(pt->format) == 1); + assert(util_format_get_blockheight(pt->format) == 1); for (dstLevel = baseLevel + 1; dstLevel <= lastLevel; dstLevel++) { const uint srcLevel = dstLevel - 1; diff --git a/src/gallium/auxiliary/util/u_rect.c b/src/gallium/auxiliary/util/u_rect.c index 72725b59d2c..298fbacecba 100644 --- a/src/gallium/auxiliary/util/u_rect.c +++ b/src/gallium/auxiliary/util/u_rect.c @@ -34,6 +34,7 @@ #include "pipe/p_format.h" #include "pipe/p_context.h" #include "pipe/p_screen.h" +#include "util/u_format.h" #include "util/u_rect.h" @@ -57,9 +58,9 @@ util_copy_rect(ubyte * dst, { unsigned i; int src_stride_pos = src_stride < 0 ? -src_stride : src_stride; - int blocksize = pf_get_blocksize(format); - int blockwidth = pf_get_blockwidth(format); - int blockheight = pf_get_blockheight(format); + int blocksize = util_format_get_blocksize(format); + int blockwidth = util_format_get_blockwidth(format); + int blockheight = util_format_get_blockheight(format); assert(blocksize > 0); assert(blockwidth > 0); @@ -105,9 +106,9 @@ util_fill_rect(ubyte * dst, { unsigned i, j; unsigned width_size; - int blocksize = pf_get_blocksize(format); - int blockwidth = pf_get_blockwidth(format); - int blockheight = pf_get_blockheight(format); + int blocksize = util_format_get_blocksize(format); + int blockwidth = util_format_get_blockwidth(format); + int blockheight = util_format_get_blockheight(format); assert(blocksize > 0); assert(blockwidth > 0); @@ -203,9 +204,9 @@ util_surface_copy(struct pipe_context *pipe, PIPE_TRANSFER_WRITE, dst_x, dst_y, w, h); - assert(pf_get_blocksize(dst_format) == pf_get_blocksize(src_format)); - assert(pf_get_blockwidth(dst_format) == pf_get_blockwidth(src_format)); - assert(pf_get_blockheight(dst_format) == pf_get_blockheight(src_format)); + assert(util_format_get_blocksize(dst_format) == util_format_get_blocksize(src_format)); + assert(util_format_get_blockwidth(dst_format) == util_format_get_blockwidth(src_format)); + assert(util_format_get_blockheight(dst_format) == util_format_get_blockheight(src_format)); src_map = pipe->screen->transfer_map(screen, src_trans); dst_map = pipe->screen->transfer_map(screen, dst_trans); @@ -270,7 +271,7 @@ util_surface_fill(struct pipe_context *pipe, if (dst_map) { assert(dst_trans->stride > 0); - switch (pf_get_blocksize(dst_trans->texture->format)) { + switch (util_format_get_blocksize(dst_trans->texture->format)) { case 1: case 2: case 4: diff --git a/src/gallium/auxiliary/util/u_tile.c b/src/gallium/auxiliary/util/u_tile.c index 88c9a1f0977..5b8dd1abb94 100644 --- a/src/gallium/auxiliary/util/u_tile.c +++ b/src/gallium/auxiliary/util/u_tile.c @@ -34,6 +34,7 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" #include "util/u_rect.h" @@ -52,7 +53,7 @@ pipe_get_tile_raw(struct pipe_transfer *pt, const void *src; if (dst_stride == 0) - dst_stride = pf_get_stride(pt->texture->format, w); + dst_stride = util_format_get_stride(pt->texture->format, w); if (pipe_clip_tile(x, y, &w, &h, pt)) return; @@ -81,7 +82,7 @@ pipe_put_tile_raw(struct pipe_transfer *pt, enum pipe_format format = pt->texture->format; if (src_stride == 0) - src_stride = pf_get_stride(format, w); + src_stride = util_format_get_stride(format, w); if (pipe_clip_tile(x, y, &w, &h, pt)) return; @@ -1275,7 +1276,7 @@ pipe_get_tile_rgba(struct pipe_transfer *pt, if (pipe_clip_tile(x, y, &w, &h, pt)) return; - packed = MALLOC(pf_get_nblocks(format, w, h) * pf_get_blocksize(format)); + packed = MALLOC(util_format_get_nblocks(format, w, h) * util_format_get_blocksize(format)); if (!packed) return; @@ -1303,7 +1304,7 @@ pipe_put_tile_rgba(struct pipe_transfer *pt, if (pipe_clip_tile(x, y, &w, &h, pt)) return; - packed = MALLOC(pf_get_nblocks(format, w, h) * pf_get_blocksize(format)); + packed = MALLOC(util_format_get_nblocks(format, w, h) * util_format_get_blocksize(format)); if (!packed) return; diff --git a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c index 8f0185b42ab..ab196c21f87 100644 --- a/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c +++ b/src/gallium/auxiliary/vl/vl_mpeg12_mc_renderer.c @@ -1444,7 +1444,7 @@ grab_blocks(struct vl_mpeg12_mc_renderer *r, unsigned mbx, unsigned mby, assert(r); assert(blocks); - tex_pitch = r->tex_transfer[0]->stride / pf_get_blocksize(r->tex_transfer[0]->texture->format); + tex_pitch = r->tex_transfer[0]->stride / util_format_get_blocksize(r->tex_transfer[0]->texture->format); texels = r->texels[0] + mbpy * tex_pitch + mbpx; for (y = 0; y < 2; ++y) { @@ -1483,7 +1483,7 @@ grab_blocks(struct vl_mpeg12_mc_renderer *r, unsigned mbx, unsigned mby, mbpy /= 2; for (tb = 0; tb < 2; ++tb) { - tex_pitch = r->tex_transfer[tb + 1]->stride / pf_get_blocksize(r->tex_transfer[tb + 1]->texture->format); + tex_pitch = r->tex_transfer[tb + 1]->stride / util_format_get_blocksize(r->tex_transfer[tb + 1]->texture->format); texels = r->texels[tb + 1] + mbpy * tex_pitch + mbpx; if ((cbp >> (1 - tb)) & 1) { diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c index bc2e625f54b..998944f77a3 100644 --- a/src/gallium/drivers/cell/ppu/cell_texture.c +++ b/src/gallium/drivers/cell/ppu/cell_texture.c @@ -67,11 +67,11 @@ cell_texture_layout(struct cell_texture *ct) w_tile = align(width, TILE_SIZE); h_tile = align(height, TILE_SIZE); - ct->stride[level] = pf_get_stride(pt->format, w_tile); + ct->stride[level] = util_format_get_stride(pt->format, w_tile); ct->level_offset[level] = ct->buffer_size; - size = ct->stride[level] * pf_get_nblocksy(pt->format, h_tile); + size = ct->stride[level] * util_format_get_nblocksy(pt->format, h_tile); if (pt->target == PIPE_TEXTURE_CUBE) size *= 6; else @@ -283,11 +283,11 @@ cell_get_tex_surface(struct pipe_screen *screen, if (pt->target == PIPE_TEXTURE_CUBE) { unsigned h_tile = align(ps->height, TILE_SIZE); - ps->offset += face * pf_get_nblocksy(ps->format, h_tile) * ct->stride[level]; + ps->offset += face * util_format_get_nblocksy(ps->format, h_tile) * ct->stride[level]; } else if (pt->target == PIPE_TEXTURE_3D) { unsigned h_tile = align(ps->height, TILE_SIZE); - ps->offset += zslice * pf_get_nblocksy(ps->format, h_tile) * ct->stride[level]; + ps->offset += zslice * util_format_get_nblocksy(ps->format, h_tile) * ct->stride[level]; } else { assert(face == 0); @@ -342,11 +342,11 @@ cell_get_tex_transfer(struct pipe_screen *screen, if (texture->target == PIPE_TEXTURE_CUBE) { unsigned h_tile = align(u_minify(texture->height0, level), TILE_SIZE); - ctrans->offset += face * pf_get_nblocksy(texture->format, h_tile) * pt->stride; + ctrans->offset += face * util_format_get_nblocksy(texture->format, h_tile) * pt->stride; } else if (texture->target == PIPE_TEXTURE_3D) { unsigned h_tile = align(u_minify(texture->height0, level), TILE_SIZE); - ctrans->offset += zslice * pf_get_nblocksy(texture->format, h_tile) * pt->stride; + ctrans->offset += zslice * util_format_get_nblocksy(texture->format, h_tile) * pt->stride; } else { assert(face == 0); @@ -399,8 +399,8 @@ cell_transfer_map(struct pipe_screen *screen, struct pipe_transfer *transfer) * Create a buffer of ordinary memory for the linear texture. * This is the memory that the user will read/write. */ - size = pf_get_stride(pt->format, align(texWidth, TILE_SIZE)) * - pf_get_nblocksy(pt->format, align(texHeight, TILE_SIZE)); + size = util_format_get_stride(pt->format, align(texWidth, TILE_SIZE)) * + util_format_get_nblocksy(pt->format, align(texHeight, TILE_SIZE)); ctrans->map = align_malloc(size, 16); if (!ctrans->map) @@ -408,7 +408,7 @@ cell_transfer_map(struct pipe_screen *screen, struct pipe_transfer *transfer) if (transfer->usage & PIPE_TRANSFER_READ) { /* need to untwiddle the texture to make a linear version */ - const uint bpp = util_format_get_size(ct->base.format); + const uint bpp = util_format_get_blocksize(ct->base.format); if (bpp == 4) { const uint *src = (uint *) (ct->mapped + ctrans->offset); uint *dst = ctrans->map; @@ -451,7 +451,7 @@ cell_transfer_unmap(struct pipe_screen *screen, /* The user wrote new texture data into the mapped buffer. * We need to convert the new linear data into the twiddled/tiled format. */ - const uint bpp = util_format_get_size(ct->base.format); + const uint bpp = util_format_get_blocksize(ct->base.format); if (bpp == 4) { const uint *src = ctrans->map; uint *dst = (uint *) (ct->mapped + ctrans->offset); diff --git a/src/gallium/drivers/i915/i915_surface.c b/src/gallium/drivers/i915/i915_surface.c index 24e1024aaa3..c693eb30e87 100644 --- a/src/gallium/drivers/i915/i915_surface.c +++ b/src/gallium/drivers/i915/i915_surface.c @@ -32,6 +32,7 @@ #include "pipe/p_inlines.h" #include "pipe/p_inlines.h" #include "pipe/internal/p_winsys_screen.h" +#include "util/u_format.h" #include "util/u_tile.h" #include "util/u_rect.h" @@ -52,15 +53,15 @@ i915_surface_copy(struct pipe_context *pipe, struct pipe_texture *spt = &src_tex->base; assert( dst != src ); - assert( pf_get_blocksize(dpt->format) == pf_get_blocksize(spt->format) ); - assert( pf_get_blockwidth(dpt->format) == pf_get_blockwidth(spt->format) ); - assert( pf_get_blockheight(dpt->format) == pf_get_blockheight(spt->format) ); - assert( pf_get_blockwidth(dpt->format) == 1 ); - assert( pf_get_blockheight(dpt->format) == 1 ); + assert( util_format_get_blocksize(dpt->format) == util_format_get_blocksize(spt->format) ); + assert( util_format_get_blockwidth(dpt->format) == util_format_get_blockwidth(spt->format) ); + assert( util_format_get_blockheight(dpt->format) == util_format_get_blockheight(spt->format) ); + assert( util_format_get_blockwidth(dpt->format) == 1 ); + assert( util_format_get_blockheight(dpt->format) == 1 ); i915_copy_blit( i915_context(pipe), FALSE, - pf_get_blocksize(dpt->format), + util_format_get_blocksize(dpt->format), (unsigned short) src_tex->stride, src_tex->buffer, src->offset, (unsigned short) dst_tex->stride, dst_tex->buffer, dst->offset, (short) srcx, (short) srcy, (short) dstx, (short) dsty, (short) width, (short) height ); @@ -76,11 +77,11 @@ i915_surface_fill(struct pipe_context *pipe, struct i915_texture *tex = (struct i915_texture *)dst->texture; struct pipe_texture *pt = &tex->base; - assert(pf_get_blockwidth(pt->format) == 1); - assert(pf_get_blockheight(pt->format) == 1); + assert(util_format_get_blockwidth(pt->format) == 1); + assert(util_format_get_blockheight(pt->format) == 1); i915_fill_blit( i915_context(pipe), - pf_get_blocksize(pt->format), + util_format_get_blocksize(pt->format), (unsigned short) tex->stride, tex->buffer, dst->offset, (short) dstx, (short) dsty, diff --git a/src/gallium/drivers/i915/i915_texture.c b/src/gallium/drivers/i915/i915_texture.c index b28b413771d..50a9e19094b 100644 --- a/src/gallium/drivers/i915/i915_texture.c +++ b/src/gallium/drivers/i915/i915_texture.c @@ -35,6 +35,7 @@ #include "pipe/p_defines.h" #include "pipe/p_inlines.h" #include "pipe/internal/p_winsys_screen.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -129,7 +130,7 @@ i915_miptree_set_image_offset(struct i915_texture *tex, assert(img < tex->nr_images[level]); - tex->image_offset[level][img] = y * tex->stride + x * pf_get_blocksize(tex->base.format); + tex->image_offset[level][img] = y * tex->stride + x * util_format_get_blocksize(tex->base.format); /* printf("%s level %d img %d pos %d,%d image_offset %x\n", @@ -151,7 +152,7 @@ i915_scanout_layout(struct i915_texture *tex) { struct pipe_texture *pt = &tex->base; - if (pt->last_level > 0 || pf_get_blocksize(pt->format) != 4) + if (pt->last_level > 0 || util_format_get_blocksize(pt->format) != 4) return FALSE; i915_miptree_set_level_info(tex, 0, 1, @@ -161,18 +162,18 @@ i915_scanout_layout(struct i915_texture *tex) i915_miptree_set_image_offset(tex, 0, 0, 0, 0); if (pt->width0 >= 240) { - tex->stride = power_of_two(pf_get_stride(pt->format, pt->width0)); - tex->total_nblocksy = align(pf_get_nblocksy(pt->format, pt->height0), 8); + tex->stride = power_of_two(util_format_get_stride(pt->format, pt->width0)); + tex->total_nblocksy = align(util_format_get_nblocksy(pt->format, pt->height0), 8); tex->hw_tiled = INTEL_TILE_X; } else if (pt->width0 == 64 && pt->height0 == 64) { - tex->stride = power_of_two(pf_get_stride(pt->format, pt->width0)); - tex->total_nblocksy = align(pf_get_nblocksy(pt->format, pt->height0), 8); + tex->stride = power_of_two(util_format_get_stride(pt->format, pt->width0)); + tex->total_nblocksy = align(util_format_get_nblocksy(pt->format, pt->height0), 8); } else { return FALSE; } debug_printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - pt->width0, pt->height0, pf_get_blocksize(pt->format), + pt->width0, pt->height0, util_format_get_blocksize(pt->format), tex->stride, tex->total_nblocksy, tex->stride * tex->total_nblocksy); return TRUE; @@ -186,7 +187,7 @@ i915_display_target_layout(struct i915_texture *tex) { struct pipe_texture *pt = &tex->base; - if (pt->last_level > 0 || pf_get_blocksize(pt->format) != 4) + if (pt->last_level > 0 || util_format_get_blocksize(pt->format) != 4) return FALSE; /* fallback to normal textures for small textures */ @@ -199,12 +200,12 @@ i915_display_target_layout(struct i915_texture *tex) 1); i915_miptree_set_image_offset(tex, 0, 0, 0, 0); - tex->stride = power_of_two(pf_get_stride(pt->format, pt->width0)); - tex->total_nblocksy = align(pf_get_nblocksy(pt->format, pt->height0), 8); + tex->stride = power_of_two(util_format_get_stride(pt->format, pt->width0)); + tex->total_nblocksy = align(util_format_get_nblocksy(pt->format, pt->height0), 8); tex->hw_tiled = INTEL_TILE_X; debug_printf("%s size: %d,%d,%d offset %d,%d (0x%x)\n", __FUNCTION__, - pt->width0, pt->height0, pf_get_blocksize(pt->format), + pt->width0, pt->height0, util_format_get_blocksize(pt->format), tex->stride, tex->total_nblocksy, tex->stride * tex->total_nblocksy); return TRUE; @@ -217,7 +218,7 @@ i915_miptree_layout_2d(struct i915_texture *tex) unsigned level; unsigned width = pt->width0; unsigned height = pt->height0; - unsigned nblocksy = pf_get_nblocksy(pt->format, pt->width0); + unsigned nblocksy = util_format_get_nblocksy(pt->format, pt->width0); /* used for scanouts that need special layouts */ if (pt->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) @@ -229,7 +230,7 @@ i915_miptree_layout_2d(struct i915_texture *tex) if (i915_display_target_layout(tex)) return; - tex->stride = align(pf_get_stride(pt->format, pt->width0), 4); + tex->stride = align(util_format_get_stride(pt->format, pt->width0), 4); tex->total_nblocksy = 0; for (level = 0; level <= pt->last_level; level++) { @@ -242,7 +243,7 @@ i915_miptree_layout_2d(struct i915_texture *tex) width = u_minify(width, 1); height = u_minify(height, 1); - nblocksy = pf_get_nblocksy(pt->format, height); + nblocksy = util_format_get_nblocksy(pt->format, height); } } @@ -255,12 +256,12 @@ i915_miptree_layout_3d(struct i915_texture *tex) unsigned width = pt->width0; unsigned height = pt->height0; unsigned depth = pt->depth0; - unsigned nblocksy = pf_get_nblocksy(pt->format, pt->height0); + unsigned nblocksy = util_format_get_nblocksy(pt->format, pt->height0); unsigned stack_nblocksy = 0; /* Calculate the size of a single slice. */ - tex->stride = align(pf_get_stride(pt->format, pt->width0), 4); + tex->stride = align(util_format_get_stride(pt->format, pt->width0), 4); /* XXX: hardware expects/requires 9 levels at minimum. */ @@ -271,7 +272,7 @@ i915_miptree_layout_3d(struct i915_texture *tex) width = u_minify(width, 1); height = u_minify(height, 1); - nblocksy = pf_get_nblocksy(pt->format, height); + nblocksy = util_format_get_nblocksy(pt->format, height); } /* Fixup depth image_offsets: @@ -296,14 +297,14 @@ i915_miptree_layout_cube(struct i915_texture *tex) { struct pipe_texture *pt = &tex->base; unsigned width = pt->width0, height = pt->height0; - const unsigned nblocks = pf_get_nblocksx(pt->format, pt->width0); + const unsigned nblocks = util_format_get_nblocksx(pt->format, pt->width0); unsigned level; unsigned face; assert(width == height); /* cubemap images are square */ /* double pitch for cube layouts */ - tex->stride = align(nblocks * pf_get_blocksize(pt->format) * 2, 4); + tex->stride = align(nblocks * util_format_get_blocksize(pt->format) * 2, 4); tex->total_nblocksy = nblocks * 4; for (level = 0; level <= pt->last_level; level++) { @@ -366,8 +367,8 @@ i945_miptree_layout_2d(struct i915_texture *tex) unsigned y = 0; unsigned width = pt->width0; unsigned height = pt->height0; - unsigned nblocksx = pf_get_nblocksx(pt->format, pt->width0); - unsigned nblocksy = pf_get_nblocksy(pt->format, pt->height0); + unsigned nblocksx = util_format_get_nblocksx(pt->format, pt->width0); + unsigned nblocksy = util_format_get_nblocksy(pt->format, pt->height0); /* used for scanouts that need special layouts */ if (tex->base.tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) @@ -379,7 +380,7 @@ i945_miptree_layout_2d(struct i915_texture *tex) if (i915_display_target_layout(tex)) return; - tex->stride = align(pf_get_stride(pt->format, pt->width0), 4); + tex->stride = align(util_format_get_stride(pt->format, pt->width0), 4); /* May need to adjust pitch to accomodate the placement of * the 2nd mipmap level. This occurs when the alignment @@ -388,11 +389,11 @@ i945_miptree_layout_2d(struct i915_texture *tex) */ if (pt->last_level > 0) { unsigned mip1_nblocksx - = align(pf_get_nblocksx(pt->format, u_minify(width, 1)), align_x) - + pf_get_nblocksx(pt->format, u_minify(width, 2)); + = align(util_format_get_nblocksx(pt->format, u_minify(width, 1)), align_x) + + util_format_get_nblocksx(pt->format, u_minify(width, 2)); if (mip1_nblocksx > nblocksx) - tex->stride = mip1_nblocksx * pf_get_blocksize(pt->format); + tex->stride = mip1_nblocksx * util_format_get_blocksize(pt->format); } /* Pitch must be a whole number of dwords @@ -422,8 +423,8 @@ i945_miptree_layout_2d(struct i915_texture *tex) width = u_minify(width, 1); height = u_minify(height, 1); - nblocksx = pf_get_nblocksx(pt->format, width); - nblocksy = pf_get_nblocksy(pt->format, height); + nblocksx = util_format_get_nblocksx(pt->format, width); + nblocksy = util_format_get_nblocksy(pt->format, height); } } @@ -434,16 +435,16 @@ i945_miptree_layout_3d(struct i915_texture *tex) unsigned width = pt->width0; unsigned height = pt->height0; unsigned depth = pt->depth0; - unsigned nblocksy = pf_get_nblocksy(pt->format, pt->width0); + unsigned nblocksy = util_format_get_nblocksy(pt->format, pt->width0); unsigned pack_x_pitch, pack_x_nr; unsigned pack_y_pitch; unsigned level; - tex->stride = align(pf_get_stride(pt->format, pt->width0), 4); + tex->stride = align(util_format_get_stride(pt->format, pt->width0), 4); tex->total_nblocksy = 0; pack_y_pitch = MAX2(nblocksy, 2); - pack_x_pitch = tex->stride / pf_get_blocksize(pt->format); + pack_x_pitch = tex->stride / util_format_get_blocksize(pt->format); pack_x_nr = 1; for (level = 0; level <= pt->last_level; level++) { @@ -468,7 +469,7 @@ i945_miptree_layout_3d(struct i915_texture *tex) if (pack_x_pitch > 4) { pack_x_pitch >>= 1; pack_x_nr <<= 1; - assert(pack_x_pitch * pack_x_nr * pf_get_blocksize(pt->format) <= tex->stride); + assert(pack_x_pitch * pack_x_nr * util_format_get_blocksize(pt->format) <= tex->stride); } if (pack_y_pitch > 2) { @@ -478,7 +479,7 @@ i945_miptree_layout_3d(struct i915_texture *tex) width = u_minify(width, 1); height = u_minify(height, 1); depth = u_minify(depth, 1); - nblocksy = pf_get_nblocksy(pt->format, height); + nblocksy = util_format_get_nblocksy(pt->format, height); } } @@ -488,7 +489,7 @@ i945_miptree_layout_cube(struct i915_texture *tex) struct pipe_texture *pt = &tex->base; unsigned level; - const unsigned nblocks = pf_get_nblocksx(pt->format, pt->width0); + const unsigned nblocks = util_format_get_nblocksx(pt->format, pt->width0); unsigned face; unsigned width = pt->width0; unsigned height = pt->height0; @@ -508,9 +509,9 @@ i945_miptree_layout_cube(struct i915_texture *tex) * or the final row of 4x4, 2x2 and 1x1 faces below this. */ if (nblocks > 32) - tex->stride = align(nblocks * pf_get_blocksize(pt->format) * 2, 4); + tex->stride = align(nblocks * util_format_get_blocksize(pt->format) * 2, 4); else - tex->stride = 14 * 8 * pf_get_blocksize(pt->format); + tex->stride = 14 * 8 * util_format_get_blocksize(pt->format); tex->total_nblocksy = nblocks * 4; @@ -840,8 +841,8 @@ i915_transfer_map(struct pipe_screen *screen, return NULL; return map + i915_transfer(transfer)->offset + - transfer->y / pf_get_blockheight(format) * transfer->stride + - transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format); + transfer->y / util_format_get_blockheight(format) * transfer->stride + + transfer->x / util_format_get_blockwidth(format) * util_format_get_blocksize(format); } static void diff --git a/src/gallium/drivers/llvmpipe/lp_setup.c b/src/gallium/drivers/llvmpipe/lp_setup.c index b4aabd4d7cc..b18f17c0cd3 100644 --- a/src/gallium/drivers/llvmpipe/lp_setup.c +++ b/src/gallium/drivers/llvmpipe/lp_setup.c @@ -41,6 +41,7 @@ #include "draw/draw_vertex.h" #include "pipe/p_shader_tokens.h" #include "pipe/p_thread.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" #include "lp_bld_debug.h" @@ -166,7 +167,7 @@ shade_quads(struct llvmpipe_context *llvmpipe, assert((y % 2) == 0); depth = llvmpipe->zsbuf_map + y*llvmpipe->zsbuf_transfer->stride + - 2*x*pf_get_blocksize(llvmpipe->zsbuf_transfer->texture->format); + 2*x*util_format_get_blocksize(llvmpipe->zsbuf_transfer->texture->format); } else depth = NULL; diff --git a/src/gallium/drivers/llvmpipe/lp_texture.c b/src/gallium/drivers/llvmpipe/lp_texture.c index 9e41bc40745..2c135029ea2 100644 --- a/src/gallium/drivers/llvmpipe/lp_texture.c +++ b/src/gallium/drivers/llvmpipe/lp_texture.c @@ -69,10 +69,10 @@ llvmpipe_texture_layout(struct llvmpipe_screen *screen, /* Allocate storage for whole quads. This is particularly important * for depth surfaces, which are currently stored in a swizzled format. */ - nblocksx = pf_get_nblocksx(pt->format, align(width, 2)); - nblocksy = pf_get_nblocksy(pt->format, align(height, 2)); + nblocksx = util_format_get_nblocksx(pt->format, align(width, 2)); + nblocksy = util_format_get_nblocksy(pt->format, align(height, 2)); - lpt->stride[level] = align(nblocksx * pf_get_blocksize(pt->format), 16); + lpt->stride[level] = align(nblocksx * util_format_get_blocksize(pt->format), 16); lpt->level_offset[level] = buffer_size; @@ -251,11 +251,11 @@ llvmpipe_get_tex_surface(struct pipe_screen *screen, */ if (pt->target == PIPE_TEXTURE_CUBE) { unsigned tex_height = ps->height; - ps->offset += face * pf_get_nblocksy(pt->format, tex_height) * lpt->stride[level]; + ps->offset += face * util_format_get_nblocksy(pt->format, tex_height) * lpt->stride[level]; } else if (pt->target == PIPE_TEXTURE_3D) { unsigned tex_height = ps->height; - ps->offset += zslice * pf_get_nblocksy(pt->format, tex_height) * lpt->stride[level]; + ps->offset += zslice * util_format_get_nblocksy(pt->format, tex_height) * lpt->stride[level]; } else { assert(face == 0); @@ -314,11 +314,11 @@ llvmpipe_get_tex_transfer(struct pipe_screen *screen, */ if (texture->target == PIPE_TEXTURE_CUBE) { unsigned tex_height = u_minify(texture->height0, level); - lpt->offset += face * pf_get_nblocksy(texture->format, tex_height) * pt->stride; + lpt->offset += face * util_format_get_nblocksy(texture->format, tex_height) * pt->stride; } else if (texture->target == PIPE_TEXTURE_3D) { unsigned tex_height = u_minify(texture->height0, level); - lpt->offset += zslice * pf_get_nblocksy(texture->format, tex_height) * pt->stride; + lpt->offset += zslice * util_format_get_nblocksy(texture->format, tex_height) * pt->stride; } else { assert(face == 0); @@ -379,8 +379,8 @@ llvmpipe_transfer_map( struct pipe_screen *_screen, } xfer_map = map + llvmpipe_transfer(transfer)->offset + - transfer->y / pf_get_blockheight(format) * transfer->stride + - transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format); + transfer->y / util_format_get_blockheight(format) * transfer->stride + + transfer->x / util_format_get_blockwidth(format) * util_format_get_blocksize(format); /*printf("map = %p xfer map = %p\n", map, xfer_map);*/ return xfer_map; } diff --git a/src/gallium/drivers/nv04/nv04_surface_2d.c b/src/gallium/drivers/nv04/nv04_surface_2d.c index 3020806c5d6..12df7fd1997 100644 --- a/src/gallium/drivers/nv04/nv04_surface_2d.c +++ b/src/gallium/drivers/nv04/nv04_surface_2d.c @@ -1,5 +1,6 @@ #include "pipe/p_context.h" #include "pipe/p_format.h" +#include "util/u_format.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -158,10 +159,10 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx, sub_w = MIN2(sub_w, w - x); /* Must be 64-byte aligned */ - assert(!((dst->offset + nv04_swizzle_bits(dx+x, dy+y) * pf_get_blocksize(dst->texture->format)) & 63)); + assert(!((dst->offset + nv04_swizzle_bits(dx+x, dy+y) * util_format_get_blocksize(dst->texture->format)) & 63)); BEGIN_RING(chan, swzsurf, NV04_SWIZZLED_SURFACE_OFFSET, 1); - OUT_RELOCl(chan, dst_bo, dst->offset + nv04_swizzle_bits(dx+x, dy+y) * pf_get_blocksize(dst->texture->format), + OUT_RELOCl(chan, dst_bo, dst->offset + nv04_swizzle_bits(dx+x, dy+y) * util_format_get_blocksize(dst->texture->format), NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_WR); BEGIN_RING(chan, sifm, NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_CONVERSION, 9); @@ -180,7 +181,7 @@ nv04_surface_copy_swizzle(struct nv04_surface_2d *ctx, OUT_RING (chan, src_pitch | NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_ORIGIN_CENTER | NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT_FILTER_POINT_SAMPLE); - OUT_RELOCl(chan, src_bo, src->offset + (sy+y) * src_pitch + (sx+x) * pf_get_blocksize(src->texture->format), + OUT_RELOCl(chan, src_bo, src->offset + (sy+y) * src_pitch + (sx+x) * util_format_get_blocksize(src->texture->format), NOUVEAU_BO_GART | NOUVEAU_BO_VRAM | NOUVEAU_BO_RD); OUT_RING (chan, 0); } @@ -201,9 +202,9 @@ nv04_surface_copy_m2mf(struct nv04_surface_2d *ctx, unsigned src_pitch = ((struct nv04_surface *)src)->pitch; unsigned dst_pitch = ((struct nv04_surface *)dst)->pitch; unsigned dst_offset = dst->offset + dy * dst_pitch + - dx * pf_get_blocksize(dst->texture->format); + dx * util_format_get_blocksize(dst->texture->format); unsigned src_offset = src->offset + sy * src_pitch + - sx * pf_get_blocksize(src->texture->format); + sx * util_format_get_blocksize(src->texture->format); MARK_RING (chan, 3 + ((h / 2047) + 1) * 9, 2 + ((h / 2047) + 1) * 2); BEGIN_RING(chan, m2mf, NV04_MEMORY_TO_MEMORY_FORMAT_DMA_BUFFER_IN, 2); @@ -222,7 +223,7 @@ nv04_surface_copy_m2mf(struct nv04_surface_2d *ctx, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART | NOUVEAU_BO_WR); OUT_RING (chan, src_pitch); OUT_RING (chan, dst_pitch); - OUT_RING (chan, w * pf_get_blocksize(src->texture->format)); + OUT_RING (chan, w * util_format_get_blocksize(src->texture->format)); OUT_RING (chan, count); OUT_RING (chan, 0x0101); OUT_RING (chan, 0); diff --git a/src/gallium/drivers/nv04/nv04_transfer.c b/src/gallium/drivers/nv04/nv04_transfer.c index d66d6c6346c..8446073ae80 100644 --- a/src/gallium/drivers/nv04/nv04_transfer.c +++ b/src/gallium/drivers/nv04/nv04_transfer.c @@ -1,6 +1,7 @@ #include #include #include +#include #include #include #include @@ -151,7 +152,7 @@ nv04_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx) pipe_transfer_buffer_flags(ptx)); return map + ns->base.offset + - ptx->y * ns->pitch + ptx->x * pf_get_blocksize(ptx->texture->format); + ptx->y * ns->pitch + ptx->x * util_format_get_blocksize(ptx->texture->format); } static void diff --git a/src/gallium/drivers/nv10/nv10_miptree.c b/src/gallium/drivers/nv10/nv10_miptree.c index 6a52b6af362..908482ad854 100644 --- a/src/gallium/drivers/nv10/nv10_miptree.c +++ b/src/gallium/drivers/nv10/nv10_miptree.c @@ -1,6 +1,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "nv10_context.h" @@ -23,9 +24,9 @@ nv10_miptree_layout(struct nv10_miptree *nv10mt) for (l = 0; l <= pt->last_level; l++) { if (swizzled) - nv10mt->level[l].pitch = pf_get_stride(pt->format, width); + nv10mt->level[l].pitch = util_format_get_stride(pt->format, width); else - nv10mt->level[l].pitch = pf_get_stride(pt->format, pt->width0); + nv10mt->level[l].pitch = util_format_get_stride(pt->format, pt->width0); nv10mt->level[l].pitch = (nv10mt->level[l].pitch + 63) & ~63; nv10mt->level[l].image_offset = diff --git a/src/gallium/drivers/nv10/nv10_transfer.c b/src/gallium/drivers/nv10/nv10_transfer.c index 06bb5134173..c664973e904 100644 --- a/src/gallium/drivers/nv10/nv10_transfer.c +++ b/src/gallium/drivers/nv10/nv10_transfer.c @@ -1,6 +1,7 @@ #include #include #include +#include #include #include #include @@ -151,7 +152,7 @@ nv10_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx) pipe_transfer_buffer_flags(ptx)); return map + ns->base.offset + - ptx->y * ns->pitch + ptx->x * pf_get_blocksize(ptx->texture->format); + ptx->y * ns->pitch + ptx->x * util_format_get_blocksize(ptx->texture->format); } static void diff --git a/src/gallium/drivers/nv20/nv20_miptree.c b/src/gallium/drivers/nv20/nv20_miptree.c index e2e01bd849b..d1291a92e0a 100644 --- a/src/gallium/drivers/nv20/nv20_miptree.c +++ b/src/gallium/drivers/nv20/nv20_miptree.c @@ -1,6 +1,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "nv20_context.h" @@ -27,9 +28,9 @@ nv20_miptree_layout(struct nv20_miptree *nv20mt) for (l = 0; l <= pt->last_level; l++) { if (wide_pitch && (pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) - nv20mt->level[l].pitch = align(pf_get_stride(pt->format, pt->width0), 64); + nv20mt->level[l].pitch = align(util_format_get_stride(pt->format, pt->width0), 64); else - nv20mt->level[l].pitch = pf_get_stride(pt->format, width); + nv20mt->level[l].pitch = util_format_get_stride(pt->format, width); nv20mt->level[l].image_offset = CALLOC(nr_faces, sizeof(unsigned)); diff --git a/src/gallium/drivers/nv20/nv20_transfer.c b/src/gallium/drivers/nv20/nv20_transfer.c index 26a73c5143a..69b79c809f4 100644 --- a/src/gallium/drivers/nv20/nv20_transfer.c +++ b/src/gallium/drivers/nv20/nv20_transfer.c @@ -1,6 +1,7 @@ #include #include #include +#include #include #include #include @@ -151,7 +152,7 @@ nv20_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx) pipe_transfer_buffer_flags(ptx)); return map + ns->base.offset + - ptx->y * ns->pitch + ptx->x * pf_get_blocksize(ptx->texture->format); + ptx->y * ns->pitch + ptx->x * util_format_get_blocksize(ptx->texture->format); } static void diff --git a/src/gallium/drivers/nv30/nv30_miptree.c b/src/gallium/drivers/nv30/nv30_miptree.c index 920fe64c32f..ce95d9700f6 100644 --- a/src/gallium/drivers/nv30/nv30_miptree.c +++ b/src/gallium/drivers/nv30/nv30_miptree.c @@ -1,6 +1,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "nv30_context.h" @@ -29,9 +30,9 @@ nv30_miptree_layout(struct nv30_miptree *nv30mt) for (l = 0; l <= pt->last_level; l++) { if (wide_pitch && (pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) - nv30mt->level[l].pitch = align(pf_get_stride(pt->format, pt->width0), 64); + nv30mt->level[l].pitch = align(util_format_get_stride(pt->format, pt->width0), 64); else - nv30mt->level[l].pitch = pf_get_stride(pt->format, width); + nv30mt->level[l].pitch = util_format_get_stride(pt->format, width); nv30mt->level[l].image_offset = CALLOC(nr_faces, sizeof(unsigned)); diff --git a/src/gallium/drivers/nv30/nv30_transfer.c b/src/gallium/drivers/nv30/nv30_transfer.c index e29bfbd3efd..2255a02caed 100644 --- a/src/gallium/drivers/nv30/nv30_transfer.c +++ b/src/gallium/drivers/nv30/nv30_transfer.c @@ -1,6 +1,7 @@ #include #include #include +#include #include #include #include @@ -151,7 +152,7 @@ nv30_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx) pipe_transfer_buffer_flags(ptx)); return map + ns->base.offset + - ptx->y * ns->pitch + ptx->x * pf_get_blocksize(ptx->texture->format); + ptx->y * ns->pitch + ptx->x * util_format_get_blocksize(ptx->texture->format); } static void diff --git a/src/gallium/drivers/nv40/nv40_miptree.c b/src/gallium/drivers/nv40/nv40_miptree.c index 89ddf373e9e..b974e68a077 100644 --- a/src/gallium/drivers/nv40/nv40_miptree.c +++ b/src/gallium/drivers/nv40/nv40_miptree.c @@ -1,6 +1,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "nv40_context.h" @@ -31,9 +32,9 @@ nv40_miptree_layout(struct nv40_miptree *mt) for (l = 0; l <= pt->last_level; l++) { if (wide_pitch && (pt->tex_usage & NOUVEAU_TEXTURE_USAGE_LINEAR)) - mt->level[l].pitch = align(pf_get_stride(pt->format, pt->width0), 64); + mt->level[l].pitch = align(util_format_get_stride(pt->format, pt->width0), 64); else - mt->level[l].pitch = pf_get_stride(pt->format, width); + mt->level[l].pitch = util_format_get_stride(pt->format, width); mt->level[l].image_offset = CALLOC(nr_faces, sizeof(unsigned)); diff --git a/src/gallium/drivers/nv40/nv40_transfer.c b/src/gallium/drivers/nv40/nv40_transfer.c index ed5be1cf879..b084a38b482 100644 --- a/src/gallium/drivers/nv40/nv40_transfer.c +++ b/src/gallium/drivers/nv40/nv40_transfer.c @@ -1,6 +1,7 @@ #include #include #include +#include #include #include #include @@ -151,7 +152,7 @@ nv40_transfer_map(struct pipe_screen *pscreen, struct pipe_transfer *ptx) pipe_transfer_buffer_flags(ptx)); return map + ns->base.offset + - ptx->y * ns->pitch + ptx->x * pf_get_blocksize(ptx->texture->format); + ptx->y * ns->pitch + ptx->x * util_format_get_blocksize(ptx->texture->format); } static void diff --git a/src/gallium/drivers/nv50/nv50_miptree.c b/src/gallium/drivers/nv50/nv50_miptree.c index 9e083b662dd..3f1edf0a139 100644 --- a/src/gallium/drivers/nv50/nv50_miptree.c +++ b/src/gallium/drivers/nv50/nv50_miptree.c @@ -23,6 +23,7 @@ #include "pipe/p_state.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "nv50_context.h" @@ -105,10 +106,10 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) for (l = 0; l <= pt->last_level; l++) { struct nv50_miptree_level *lvl = &mt->level[l]; - unsigned nblocksy = pf_get_nblocksy(pt->format, height); + unsigned nblocksy = util_format_get_nblocksy(pt->format, height); lvl->image_offset = CALLOC(mt->image_nr, sizeof(int)); - lvl->pitch = align(pf_get_stride(pt->format, width), 64); + lvl->pitch = align(util_format_get_stride(pt->format, width), 64); lvl->tile_mode = get_tile_mode(nblocksy, depth); width = u_minify(width, 1); @@ -130,7 +131,7 @@ nv50_miptree_create(struct pipe_screen *pscreen, const struct pipe_texture *tmp) unsigned tile_d = get_tile_depth(lvl->tile_mode); size = lvl->pitch; - size *= align(pf_get_nblocksy(pt->format, u_minify(pt->height0, l)), tile_h); + size *= align(util_format_get_nblocksy(pt->format, u_minify(pt->height0, l)), tile_h); size *= align(u_minify(pt->depth0, l), tile_d); lvl->image_offset[i] = mt->total_size; @@ -222,7 +223,7 @@ nv50_miptree_surface_new(struct pipe_screen *pscreen, struct pipe_texture *pt, ps->offset = lvl->image_offset[img]; if (pt->target == PIPE_TEXTURE_3D) { - unsigned nb_h = pf_get_nblocksy(pt->format, ps->height); + unsigned nb_h = util_format_get_nblocksy(pt->format, ps->height); ps->offset += get_zslice_offset(lvl->tile_mode, zslice, lvl->pitch, nb_h); } diff --git a/src/gallium/drivers/nv50/nv50_transfer.c b/src/gallium/drivers/nv50/nv50_transfer.c index 6240a0c757a..4d9afa6fedc 100644 --- a/src/gallium/drivers/nv50/nv50_transfer.c +++ b/src/gallium/drivers/nv50/nv50_transfer.c @@ -1,6 +1,7 @@ #include "pipe/p_context.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_math.h" #include "nv50_context.h" @@ -140,11 +141,11 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, return NULL; pipe_texture_reference(&tx->base.texture, pt); - tx->nblocksx = pf_get_nblocksx(pt->format, u_minify(pt->width0, level)); - tx->nblocksy = pf_get_nblocksy(pt->format, u_minify(pt->height0, level)); + tx->nblocksx = util_format_get_nblocksx(pt->format, u_minify(pt->width0, level)); + tx->nblocksy = util_format_get_nblocksy(pt->format, u_minify(pt->height0, level)); tx->base.width = w; tx->base.height = h; - tx->base.stride = tx->nblocksx * pf_get_blocksize(pt->format); + tx->base.stride = tx->nblocksx * util_format_get_blocksize(pt->format); tx->base.usage = usage; tx->level_pitch = lvl->pitch; @@ -154,8 +155,8 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, tx->level_offset = lvl->image_offset[image]; tx->level_tiling = lvl->tile_mode; tx->level_z = zslice; - tx->level_x = pf_get_nblocksx(pt->format, x); - tx->level_y = pf_get_nblocksy(pt->format, y); + tx->level_x = util_format_get_nblocksx(pt->format, x); + tx->level_y = util_format_get_nblocksy(pt->format, y); ret = nouveau_bo_new(dev, NOUVEAU_BO_GART | NOUVEAU_BO_MAP, 0, tx->nblocksy * tx->base.stride, &tx->bo); if (ret) { @@ -164,8 +165,8 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, } if (usage & PIPE_TRANSFER_READ) { - nx = pf_get_nblocksx(pt->format, tx->base.width); - ny = pf_get_nblocksy(pt->format, tx->base.height); + nx = util_format_get_nblocksx(pt->format, tx->base.width); + ny = util_format_get_nblocksy(pt->format, tx->base.height); nv50_transfer_rect_m2mf(pscreen, mt->base.bo, tx->level_offset, tx->level_pitch, tx->level_tiling, @@ -176,7 +177,7 @@ nv50_transfer_new(struct pipe_screen *pscreen, struct pipe_texture *pt, tx->base.stride, tx->bo->tile_mode, 0, 0, 0, tx->nblocksx, tx->nblocksy, 1, - pf_get_blocksize(pt->format), nx, ny, + util_format_get_blocksize(pt->format), nx, ny, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART, NOUVEAU_BO_GART); } @@ -191,8 +192,8 @@ nv50_transfer_del(struct pipe_transfer *ptx) struct nv50_miptree *mt = nv50_miptree(ptx->texture); struct pipe_texture *pt = ptx->texture; - unsigned nx = pf_get_nblocksx(pt->format, tx->base.width); - unsigned ny = pf_get_nblocksy(pt->format, tx->base.height); + unsigned nx = util_format_get_nblocksx(pt->format, tx->base.width); + unsigned ny = util_format_get_nblocksy(pt->format, tx->base.height); if (ptx->usage & PIPE_TRANSFER_WRITE) { struct pipe_screen *pscreen = pt->screen; @@ -206,7 +207,7 @@ nv50_transfer_del(struct pipe_transfer *ptx) tx->level_x, tx->level_y, tx->level_z, tx->nblocksx, tx->nblocksy, tx->level_depth, - pf_get_blocksize(pt->format), nx, ny, + util_format_get_blocksize(pt->format), nx, ny, NOUVEAU_BO_GART, NOUVEAU_BO_VRAM | NOUVEAU_BO_GART); } diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 52a8388ead3..d7b6511d6d5 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -657,7 +657,7 @@ static boolean r300_validate_aos(struct r300_context *r300) /* Check if formats and strides are aligned to the size of DWORD. */ for (i = 0; i < r300->vertex_element_count; i++) { if (vbuf[velem[i].vertex_buffer_index].stride % 4 != 0 || - pf_get_blocksize(velem[i].src_format) % 4 != 0) { + util_format_get_blocksize(velem[i].src_format) % 4 != 0) { return FALSE; } } @@ -686,8 +686,8 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset) for (i = 0; i < aos_count - 1; i += 2) { vb1 = &vbuf[velem[i].vertex_buffer_index]; vb2 = &vbuf[velem[i+1].vertex_buffer_index]; - size1 = util_format_get_size(velem[i].src_format); - size2 = util_format_get_size(velem[i+1].src_format); + size1 = util_format_get_blocksize(velem[i].src_format); + size2 = util_format_get_blocksize(velem[i+1].src_format); OUT_CS(R300_VBPNTR_SIZE0(size1) | R300_VBPNTR_STRIDE0(vb1->stride) | R300_VBPNTR_SIZE1(size2) | R300_VBPNTR_STRIDE1(vb2->stride)); @@ -697,7 +697,7 @@ void r300_emit_aos(struct r300_context* r300, unsigned offset) if (aos_count & 1) { vb1 = &vbuf[velem[i].vertex_buffer_index]; - size1 = util_format_get_size(velem[i].src_format); + size1 = util_format_get_blocksize(velem[i].src_format); OUT_CS(R300_VBPNTR_SIZE0(size1) | R300_VBPNTR_STRIDE0(vb1->stride)); OUT_CS(vb1->buffer_offset + velem[i].src_offset + offset * vb1->stride); diff --git a/src/gallium/drivers/r300/r300_screen.c b/src/gallium/drivers/r300/r300_screen.c index feb571a23dd..a7ef3dbcc29 100644 --- a/src/gallium/drivers/r300/r300_screen.c +++ b/src/gallium/drivers/r300/r300_screen.c @@ -21,6 +21,7 @@ * USE OR OTHER DEALINGS IN THE SOFTWARE. */ #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_memory.h" #include "util/u_simple_screen.h" @@ -357,8 +358,8 @@ static void* r300_transfer_map(struct pipe_screen* screen, } return map + r300_transfer(transfer)->offset + - transfer->y / pf_get_blockheight(format) * transfer->stride + - transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format); + transfer->y / util_format_get_blockheight(format) * transfer->stride + + transfer->x / util_format_get_blockwidth(format) * util_format_get_blocksize(format); } static void r300_transfer_unmap(struct pipe_screen* screen, diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c index 170483b9bb0..9a96206a4dc 100644 --- a/src/gallium/drivers/r300/r300_texture.c +++ b/src/gallium/drivers/r300/r300_texture.c @@ -106,7 +106,7 @@ unsigned r300_texture_get_stride(struct r300_texture* tex, unsigned level) return 0; } - return align(pf_get_stride(tex->tex.format, u_minify(tex->tex.width0, level)), 32); + return align(util_format_get_stride(tex->tex.format, u_minify(tex->tex.width0, level)), 32); } static void r300_setup_miptree(struct r300_texture* tex) @@ -116,7 +116,7 @@ static void r300_setup_miptree(struct r300_texture* tex) int i; for (i = 0; i <= base->last_level; i++) { - unsigned nblocksy = pf_get_nblocksy(base->format, u_minify(base->height0, i)); + unsigned nblocksy = util_format_get_nblocksy(base->format, u_minify(base->height0, i)); stride = r300_texture_get_stride(tex, i); layer_size = stride * nblocksy; @@ -129,7 +129,7 @@ static void r300_setup_miptree(struct r300_texture* tex) tex->offset[i] = align(tex->size, 32); tex->size = tex->offset[i] + size; tex->layer_size[i] = layer_size; - tex->pitch[i] = stride / pf_get_blocksize(base->format); + tex->pitch[i] = stride / util_format_get_blocksize(base->format); debug_printf("r300: Texture miptree: Level %d " "(%dx%dx%d px, pitch %d bytes)\n", @@ -245,7 +245,7 @@ static struct pipe_texture* tex->tex.screen = screen; tex->stride_override = *stride; - tex->pitch[0] = *stride / pf_get_blocksize(base->format); + tex->pitch[0] = *stride / util_format_get_blocksize(base->format); r300_setup_flags(tex); r300_setup_texture_state(tex, r300_screen(screen)->caps->is_r500); diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c index 9e83410fcd0..a9436a33942 100644 --- a/src/gallium/drivers/softpipe/sp_texture.c +++ b/src/gallium/drivers/softpipe/sp_texture.c @@ -65,11 +65,11 @@ softpipe_texture_layout(struct pipe_screen *screen, pt->depth0 = depth; for (level = 0; level <= pt->last_level; level++) { - spt->stride[level] = pf_get_stride(pt->format, width); + spt->stride[level] = util_format_get_stride(pt->format, width); spt->level_offset[level] = buffer_size; - buffer_size += (pf_get_nblocksy(pt->format, height) * + buffer_size += (util_format_get_nblocksy(pt->format, height) * ((pt->target == PIPE_TEXTURE_CUBE) ? 6 : depth) * spt->stride[level]); @@ -239,11 +239,11 @@ softpipe_get_tex_surface(struct pipe_screen *screen, ps->zslice = zslice; if (pt->target == PIPE_TEXTURE_CUBE) { - ps->offset += face * pf_get_nblocksy(pt->format, u_minify(pt->height0, level)) * + ps->offset += face * util_format_get_nblocksy(pt->format, u_minify(pt->height0, level)) * spt->stride[level]; } else if (pt->target == PIPE_TEXTURE_3D) { - ps->offset += zslice * pf_get_nblocksy(pt->format, u_minify(pt->height0, level)) * + ps->offset += zslice * util_format_get_nblocksy(pt->format, u_minify(pt->height0, level)) * spt->stride[level]; } else { @@ -299,7 +299,7 @@ softpipe_get_tex_transfer(struct pipe_screen *screen, spt = CALLOC_STRUCT(softpipe_transfer); if (spt) { struct pipe_transfer *pt = &spt->base; - int nblocksy = pf_get_nblocksy(texture->format, u_minify(texture->height0, level)); + int nblocksy = util_format_get_nblocksy(texture->format, u_minify(texture->height0, level)); pipe_texture_reference(&pt->texture, texture); pt->x = x; pt->y = y; @@ -376,8 +376,8 @@ softpipe_transfer_map( struct pipe_screen *screen, } xfer_map = map + softpipe_transfer(transfer)->offset + - transfer->y / pf_get_blockheight(format) * transfer->stride + - transfer->x / pf_get_blockwidth(format) * pf_get_blocksize(format); + transfer->y / util_format_get_blockheight(format) * transfer->stride + + transfer->x / util_format_get_blockwidth(format) * util_format_get_blocksize(format); /*printf("map = %p xfer map = %p\n", map, xfer_map);*/ return xfer_map; } diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c index 49e1ab0ca71..112a6fe0cf3 100644 --- a/src/gallium/drivers/softpipe/sp_tile_cache.c +++ b/src/gallium/drivers/softpipe/sp_tile_cache.c @@ -239,7 +239,7 @@ clear_tile(struct softpipe_cached_tile *tile, { uint i, j; - switch (util_format_get_size(format)) { + switch (util_format_get_blocksize(format)) { case 1: memset(tile->data.any, clear_value, TILE_SIZE * TILE_SIZE); break; diff --git a/src/gallium/drivers/svga/svga_screen_texture.c b/src/gallium/drivers/svga/svga_screen_texture.c index ab93dab2239..2224c2d3945 100644 --- a/src/gallium/drivers/svga/svga_screen_texture.c +++ b/src/gallium/drivers/svga/svga_screen_texture.c @@ -159,8 +159,8 @@ svga_transfer_dma_band(struct svga_transfer *st, st->base.x + st->base.width, y + h, st->base.zslice + 1, - pf_get_blocksize(texture->base.format)*8/ - (pf_get_blockwidth(texture->base.format)*pf_get_blockheight(texture->base.format))); + util_format_get_blocksize(texture->base.format)*8/ + (util_format_get_blockwidth(texture->base.format)*util_format_get_blockheight(texture->base.format))); box.x = st->base.x; box.y = y; @@ -210,7 +210,7 @@ svga_transfer_dma(struct svga_transfer *st, } else { unsigned y, h, srcy; - unsigned blockheight = pf_get_blockheight(st->base.texture->format); + unsigned blockheight = util_format_get_blockheight(st->base.texture->format); h = st->hw_nblocksy * blockheight; srcy = 0; for(y = 0; y < st->base.height; y += h) { @@ -772,8 +772,8 @@ svga_get_tex_transfer(struct pipe_screen *screen, struct svga_screen *ss = svga_screen(screen); struct svga_winsys_screen *sws = ss->sws; struct svga_transfer *st; - unsigned nblocksx = pf_get_nblocksx(texture->format, w); - unsigned nblocksy = pf_get_nblocksy(texture->format, h); + unsigned nblocksx = util_format_get_nblocksx(texture->format, w); + unsigned nblocksy = util_format_get_nblocksy(texture->format, h); /* We can't map texture storage directly */ if (usage & PIPE_TRANSFER_MAP_DIRECTLY) @@ -787,7 +787,7 @@ svga_get_tex_transfer(struct pipe_screen *screen, st->base.y = y; st->base.width = w; st->base.height = h; - st->base.stride = nblocksx*pf_get_blocksize(texture->format); + st->base.stride = nblocksx*util_format_get_blocksize(texture->format); st->base.usage = usage; st->base.face = face; st->base.level = level; @@ -1071,7 +1071,7 @@ svga_screen_buffer_from_texture(struct pipe_texture *texture, svga_translate_format(texture->format), stex->handle); - *stride = pf_get_stride(texture->format, texture->width0); + *stride = util_format_get_stride(texture->format, texture->width0); return *buffer != NULL; } diff --git a/src/gallium/drivers/svga/svga_state_vs.c b/src/gallium/drivers/svga/svga_state_vs.c index 9e339577c75..44b7ceb4fa4 100644 --- a/src/gallium/drivers/svga/svga_state_vs.c +++ b/src/gallium/drivers/svga/svga_state_vs.c @@ -211,7 +211,7 @@ static int update_zero_stride( struct svga_context *svga, mapped_buffer = pipe_buffer_map_range(svga->pipe.screen, vbuffer->buffer, vel->src_offset, - util_format_get_size(vel->src_format), + util_format_get_blocksize(vel->src_format), PIPE_BUFFER_USAGE_CPU_READ); translate->set_buffer(translate, vel->vertex_buffer_index, mapped_buffer, diff --git a/src/gallium/drivers/trace/tr_rbug.c b/src/gallium/drivers/trace/tr_rbug.c index af1d7f3224e..c31b1d86986 100644 --- a/src/gallium/drivers/trace/tr_rbug.c +++ b/src/gallium/drivers/trace/tr_rbug.c @@ -26,6 +26,7 @@ **************************************************************************/ +#include "util/u_format.h" #include "util/u_string.h" #include "util/u_memory.h" #include "util/u_simple_list.h" @@ -203,9 +204,9 @@ trace_rbug_texture_info(struct trace_rbug *tr_rbug, struct rbug_header *header, &t->width0, 1, &t->height0, 1, &t->depth0, 1, - pf_get_blockwidth(t->format), - pf_get_blockheight(t->format), - pf_get_blocksize(t->format), + util_format_get_blockwidth(t->format), + util_format_get_blockheight(t->format), + util_format_get_blocksize(t->format), t->last_level, t->nr_samples, t->tex_usage, @@ -254,11 +255,11 @@ trace_rbug_texture_read(struct trace_rbug *tr_rbug, struct rbug_header *header, rbug_send_texture_read_reply(tr_rbug->con, serial, t->texture->format, - pf_get_blockwidth(t->texture->format), - pf_get_blockheight(t->texture->format), - pf_get_blocksize(t->texture->format), + util_format_get_blockwidth(t->texture->format), + util_format_get_blockheight(t->texture->format), + util_format_get_blocksize(t->texture->format), (uint8_t*)map, - t->stride * pf_get_nblocksy(t->texture->format, t->height), + t->stride * util_format_get_nblocksy(t->texture->format, t->height), t->stride, NULL); diff --git a/src/gallium/drivers/trace/tr_screen.c b/src/gallium/drivers/trace/tr_screen.c index f69f7da000d..ac20a47af1e 100644 --- a/src/gallium/drivers/trace/tr_screen.c +++ b/src/gallium/drivers/trace/tr_screen.c @@ -25,6 +25,7 @@ * **************************************************************************/ +#include "util/u_format.h" #include "util/u_memory.h" #include "util/u_simple_list.h" @@ -425,7 +426,7 @@ trace_screen_transfer_unmap(struct pipe_screen *_screen, struct pipe_transfer *transfer = tr_trans->transfer; if(tr_trans->map) { - size_t size = pf_get_nblocksy(transfer->texture->format, transfer->width) * transfer->stride; + size_t size = util_format_get_nblocksy(transfer->texture->format, transfer->width) * transfer->stride; trace_dump_call_begin("pipe_screen", "transfer_write"); diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h index d334114d62a..6bfff1cc59c 100644 --- a/src/gallium/include/pipe/p_format.h +++ b/src/gallium/include/pipe/p_format.h @@ -174,140 +174,6 @@ enum pipe_format { */ extern const char *pf_name( enum pipe_format format ); -/** - * Return bits for a particular component. - * \param comp component index, starting at 0 - */ -static INLINE uint pf_get_component_bits( enum pipe_format format, uint comp ) -{ - uint size; - - if (pf_swizzle_x(format) == comp) { - size = pf_size_x(format); - } - else if (pf_swizzle_y(format) == comp) { - size = pf_size_y(format); - } - else if (pf_swizzle_z(format) == comp) { - size = pf_size_z(format); - } - else if (pf_swizzle_w(format) == comp) { - size = pf_size_w(format); - } - else { - size = 0; - } - if (pf_layout( format ) == PIPE_FORMAT_LAYOUT_RGBAZS) - return size << pf_exp2( format ); - return size << (pf_mixed_scale8( format ) * 3); -} - - -/** - * Return total bits needed for the pixel format per block. - */ -static INLINE uint pf_get_blocksizebits( enum pipe_format format ) -{ - switch (pf_layout(format)) { - case PIPE_FORMAT_LAYOUT_RGBAZS: - case PIPE_FORMAT_LAYOUT_MIXED: - return - pf_get_component_bits( format, PIPE_FORMAT_COMP_0 ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_1 ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_R ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_G ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_B ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_A ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_Z ) + - pf_get_component_bits( format, PIPE_FORMAT_COMP_S ); - case PIPE_FORMAT_LAYOUT_YCBCR: - assert( format == PIPE_FORMAT_YCBCR || format == PIPE_FORMAT_YCBCR_REV ); - return 32; - case PIPE_FORMAT_LAYOUT_DXT: - switch(format) { - case PIPE_FORMAT_DXT1_RGBA: - case PIPE_FORMAT_DXT1_RGB: - case PIPE_FORMAT_DXT1_SRGBA: - case PIPE_FORMAT_DXT1_SRGB: - return 64; - case PIPE_FORMAT_DXT3_RGBA: - case PIPE_FORMAT_DXT5_RGBA: - case PIPE_FORMAT_DXT3_SRGBA: - case PIPE_FORMAT_DXT5_SRGBA: - return 128; - default: - assert( 0 ); - return 0; - } - - default: - assert( 0 ); - return 0; - } -} - -/** - * Return bytes per element for the given format. - */ -static INLINE uint pf_get_blocksize( enum pipe_format format ) -{ - assert(pf_get_blocksizebits(format) % 8 == 0); - return pf_get_blocksizebits(format) / 8; -} - -static INLINE uint pf_get_blockwidth( enum pipe_format format ) -{ - switch (pf_layout(format)) { - case PIPE_FORMAT_LAYOUT_YCBCR: - return 2; - case PIPE_FORMAT_LAYOUT_DXT: - return 4; - default: - return 1; - } -} - -static INLINE uint pf_get_blockheight( enum pipe_format format ) -{ - switch (pf_layout(format)) { - case PIPE_FORMAT_LAYOUT_DXT: - return 4; - default: - return 1; - } -} - -static INLINE unsigned -pf_get_nblocksx(enum pipe_format format, unsigned x) -{ - unsigned blockwidth = pf_get_blockwidth(format); - return (x + blockwidth - 1) / blockwidth; -} - -static INLINE unsigned -pf_get_nblocksy(enum pipe_format format, unsigned y) -{ - unsigned blockheight = pf_get_blockheight(format); - return (y + blockheight - 1) / blockheight; -} - -static INLINE unsigned -pf_get_nblocks(enum pipe_format format, unsigned width, unsigned height) -{ - return pf_get_nblocksx(format, width) * pf_get_nblocksy(format, height); -} - -static INLINE size_t -pf_get_stride(enum pipe_format format, unsigned width) -{ - return pf_get_nblocksx(format, width) * pf_get_blocksize(format); -} - -static INLINE size_t -pf_get_2d_size(enum pipe_format format, size_t stride, unsigned height) -{ - return pf_get_nblocksy(format, height) * stride; -} enum pipe_video_chroma_format { diff --git a/src/gallium/state_trackers/python/st_sample.c b/src/gallium/state_trackers/python/st_sample.c index 97ca2afc543..96377414211 100644 --- a/src/gallium/state_trackers/python/st_sample.c +++ b/src/gallium/state_trackers/python/st_sample.c @@ -30,6 +30,7 @@ #include "pipe/p_format.h" #include "pipe/p_state.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_tile.h" #include "util/u_math.h" #include "util/u_memory.h" @@ -461,7 +462,7 @@ st_sample_dxt_pixel_block(enum pipe_format format, for(ch = 0; ch < 4; ++ch) rgba[y*rgba_stride + x*4 + ch] = (float)(data[i].rgba[y*4*4 + x*4 + ch])/255.0f; - memcpy(raw, data[i].raw, pf_get_blocksize(format)); + memcpy(raw, data[i].raw, util_format_get_blocksize(format)); } @@ -473,7 +474,7 @@ st_sample_generic_pixel_block(enum pipe_format format, { unsigned i; unsigned x, y, ch; - int blocksize = pf_get_blocksize(format); + int blocksize = util_format_get_blocksize(format); for(i = 0; i < blocksize; ++i) raw[i] = (uint8_t)st_random(); @@ -548,11 +549,11 @@ st_sample_surface(struct st_surface *surface, float *rgba) if (raw) { enum pipe_format format = texture->format; uint x, y; - int nblocksx = pf_get_nblocksx(format, width); - int nblocksy = pf_get_nblocksy(format, height); - int blockwidth = pf_get_blockwidth(format); - int blockheight = pf_get_blockheight(format); - int blocksize = pf_get_blocksize(format); + int nblocksx = util_format_get_nblocksx(format, width); + int nblocksy = util_format_get_nblocksy(format, height); + int blockwidth = util_format_get_blockwidth(format); + int blockheight = util_format_get_blockheight(format); + int blocksize = util_format_get_blocksize(format); for (y = 0; y < nblocksy; ++y) { diff --git a/src/gallium/state_trackers/python/st_softpipe_winsys.c b/src/gallium/state_trackers/python/st_softpipe_winsys.c index b8535a4217b..a3294e877a6 100644 --- a/src/gallium/state_trackers/python/st_softpipe_winsys.c +++ b/src/gallium/state_trackers/python/st_softpipe_winsys.c @@ -169,8 +169,8 @@ st_softpipe_surface_buffer_create(struct pipe_winsys *winsys, const unsigned alignment = 64; unsigned nblocksy; - nblocksy = pf_get_nblocksy(format, height); - *stride = align(pf_get_stride(format, width), alignment); + nblocksy = util_format_get_nblocksy(format, height); + *stride = align(util_format_get_stride(format, width), alignment); return winsys->buffer_create(winsys, alignment, usage, diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c index 6fd402cee4a..7106a06492d 100644 --- a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c +++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c @@ -247,7 +247,7 @@ nouveau_drm_handle_from_pt(struct drm_api *api, struct pipe_screen *pscreen, return false; *handle = mt->bo->handle; - *stride = pf_get_stride(mt->base.format, mt->base.width0); + *stride = util_format_get_stride(mt->base.format, mt->base.width0); return true; } diff --git a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c index 76acc99ad73..dfecb8a728f 100644 --- a/src/gallium/winsys/drm/radeon/core/radeon_buffer.c +++ b/src/gallium/winsys/drm/radeon/core/radeon_buffer.c @@ -35,6 +35,7 @@ #include "radeon_bo_gem.h" #include "softpipe/sp_texture.h" #include "r300_context.h" +#include "util/u_format.h" #include "util/u_math.h" #include @@ -121,8 +122,8 @@ static struct pipe_buffer *radeon_surface_buffer_create(struct pipe_winsys *ws, const unsigned alignment = 64; unsigned nblocksy, size; - nblocksy = pf_get_nblocksy(format, height); - *stride = align(pf_get_stride(format, width), alignment); + nblocksy = util_format_get_nblocksy(format, height); + *stride = align(util_format_get_stride(format, width), alignment); size = *stride * nblocksy; return radeon_buffer_create(ws, 64, usage, size); diff --git a/src/gallium/winsys/egl_xlib/sw_winsys.c b/src/gallium/winsys/egl_xlib/sw_winsys.c index a36a9e52a65..6ee3ede38cb 100644 --- a/src/gallium/winsys/egl_xlib/sw_winsys.c +++ b/src/gallium/winsys/egl_xlib/sw_winsys.c @@ -163,8 +163,8 @@ surface_buffer_create(struct pipe_winsys *winsys, const unsigned alignment = 64; unsigned nblocksy; - nblocksy = pf_get_nblocksy(format, height); - *stride = align(pf_get_stride(format, width), alignment); + nblocksy = util_format_get_nblocksy(format, height); + *stride = align(util_format_get_stride(format, width), alignment); return winsys->buffer_create(winsys, alignment, usage, diff --git a/src/gallium/winsys/g3dvl/xlib/xsp_winsys.c b/src/gallium/winsys/g3dvl/xlib/xsp_winsys.c index 3fc44925a8f..f15bcd37b50 100644 --- a/src/gallium/winsys/g3dvl/xlib/xsp_winsys.c +++ b/src/gallium/winsys/g3dvl/xlib/xsp_winsys.c @@ -141,8 +141,8 @@ static struct pipe_buffer* xsp_surface_buffer_create const unsigned int ALIGNMENT = 1; unsigned nblocksy; - nblocksy = pf_get_nblocksy(format, height); - *stride = align(pf_get_stride(format, width), ALIGNMENT); + nblocksy = util_format_get_nblocksy(format, height); + *stride = align(util_format_get_stride(format, width), ALIGNMENT); return pws->buffer_create(pws, ALIGNMENT, usage, *stride * nblocksy); diff --git a/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c b/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c index 71223650279..7d076be3a31 100644 --- a/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c +++ b/src/gallium/winsys/gdi/gdi_llvmpipe_winsys.c @@ -137,8 +137,8 @@ gdi_llvmpipe_displaytarget_create(struct llvmpipe_winsys *winsys, gdt->width = width; gdt->height = height; - bpp = util_format_get_bits(format); - cpp = util_format_get_size(format); + bpp = util_format_get_blocksizebits(format); + cpp = util_format_get_blocksize(format); gdt->stride = align(width * cpp, alignment); gdt->size = gdt->stride * height; diff --git a/src/gallium/winsys/gdi/gdi_softpipe_winsys.c b/src/gallium/winsys/gdi/gdi_softpipe_winsys.c index 40f612e8931..2ad794c3f0f 100644 --- a/src/gallium/winsys/gdi/gdi_softpipe_winsys.c +++ b/src/gallium/winsys/gdi/gdi_softpipe_winsys.c @@ -163,8 +163,8 @@ gdi_softpipe_surface_buffer_create(struct pipe_winsys *winsys, const unsigned alignment = 64; unsigned nblocksy; - nblocksy = pf_get_nblocksy(format, height); - *stride = align(pf_get_stride(format, width), alignment); + nblocksy = util_format_get_nblocksy(format, height); + *stride = align(util_format_get_stride(format, width), alignment); return winsys->buffer_create(winsys, alignment, usage, @@ -271,10 +271,10 @@ gdi_softpipe_present(struct pipe_screen *screen, memset(&bmi, 0, sizeof(BITMAPINFO)); bmi.bmiHeader.biSize = sizeof(BITMAPINFOHEADER); - bmi.bmiHeader.biWidth = texture->stride[surface->level] / util_format_get_size(surface->format); + bmi.bmiHeader.biWidth = texture->stride[surface->level] / util_format_get_blocksize(surface->format); bmi.bmiHeader.biHeight= -(long)surface->height; bmi.bmiHeader.biPlanes = 1; - bmi.bmiHeader.biBitCount = util_format_get_bits(surface->format); + bmi.bmiHeader.biBitCount = util_format_get_blocksizebits(surface->format); bmi.bmiHeader.biCompression = BI_RGB; bmi.bmiHeader.biSizeImage = 0; bmi.bmiHeader.biXPelsPerMeter = 0; diff --git a/src/gallium/winsys/xlib/xlib_cell.c b/src/gallium/winsys/xlib/xlib_cell.c index cd838e73227..47ae0519a4b 100644 --- a/src/gallium/winsys/xlib/xlib_cell.c +++ b/src/gallium/winsys/xlib/xlib_cell.c @@ -289,8 +289,8 @@ xm_surface_buffer_create(struct pipe_winsys *winsys, const unsigned alignment = 64; unsigned nblocksy; - nblocksy = pf_get_nblocksy(format, height); - *stride = align(pf_get_stride(format, width), alignment); + nblocksy = util_format_get_nblocksy(format, height); + *stride = align(util_format_get_stride(format, width), alignment); return winsys->buffer_create(winsys, alignment, usage, diff --git a/src/gallium/winsys/xlib/xlib_llvmpipe.c b/src/gallium/winsys/xlib/xlib_llvmpipe.c index d96311e4520..2a434b5fd21 100644 --- a/src/gallium/winsys/xlib/xlib_llvmpipe.c +++ b/src/gallium/winsys/xlib/xlib_llvmpipe.c @@ -262,10 +262,10 @@ xm_llvmpipe_display(struct xmesa_buffer *xm_buffer, { if (xm_dt->tempImage == NULL) { - assert(pf_get_blockwidth(xm_dt->format) == 1); - assert(pf_get_blockheight(xm_dt->format) == 1); + assert(util_format_get_blockwidth(xm_dt->format) == 1); + assert(util_format_get_blockheight(xm_dt->format) == 1); alloc_shm_ximage(xm_dt, xm_buffer, - xm_dt->stride / pf_get_blocksize(xm_dt->format), + xm_dt->stride / util_format_get_blocksize(xm_dt->format), xm_dt->height); } @@ -331,8 +331,8 @@ xm_displaytarget_create(struct llvmpipe_winsys *winsys, xm_dt->width = width; xm_dt->height = height; - nblocksy = pf_get_nblocksy(format, height); - xm_dt->stride = align(pf_get_stride(format, width), alignment); + nblocksy = util_format_get_nblocksy(format, height); + xm_dt->stride = align(util_format_get_stride(format, width), alignment); size = xm_dt->stride * nblocksy; #ifdef USE_XSHM diff --git a/src/gallium/winsys/xlib/xlib_softpipe.c b/src/gallium/winsys/xlib/xlib_softpipe.c index 48cee0a41dc..f7c0099584e 100644 --- a/src/gallium/winsys/xlib/xlib_softpipe.c +++ b/src/gallium/winsys/xlib/xlib_softpipe.c @@ -255,10 +255,10 @@ xlib_softpipe_display_surface(struct xmesa_buffer *b, { if (xm_buf->tempImage == NULL) { - assert(pf_get_blockwidth(surf->texture->format) == 1); - assert(pf_get_blockheight(surf->texture->format) == 1); + assert(util_format_get_blockwidth(surf->texture->format) == 1); + assert(util_format_get_blockheight(surf->texture->format) == 1); alloc_shm_ximage(xm_buf, b, spt->stride[surf->level] / - pf_get_blocksize(surf->texture->format), surf->height); + util_format_get_blocksize(surf->texture->format), surf->height); } ximage = xm_buf->tempImage; @@ -363,8 +363,8 @@ xm_surface_buffer_create(struct pipe_winsys *winsys, const unsigned alignment = 64; unsigned nblocksy, size; - nblocksy = pf_get_nblocksy(format, height); - *stride = align(pf_get_stride(format, width), alignment); + nblocksy = util_format_get_nblocksy(format, height); + *stride = align(util_format_get_stride(format, width), alignment); size = *stride * nblocksy; #ifdef USE_XSHM diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c index c61c77479e0..7c664267d4e 100644 --- a/src/mesa/state_tracker/st_cb_drawpixels.c +++ b/src/mesa/state_tracker/st_cb_drawpixels.c @@ -857,8 +857,8 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy, usage, dstx, dsty, width, height); - assert(pf_get_blockwidth(ptDraw->texture->format) == 1); - assert(pf_get_blockheight(ptDraw->texture->format) == 1); + assert(util_format_get_blockwidth(ptDraw->texture->format) == 1); + assert(util_format_get_blockheight(ptDraw->texture->format) == 1); /* map the stencil buffer */ drawMap = screen->transfer_map(screen, ptDraw); diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c index 78aed07bf91..45ce34a85f7 100644 --- a/src/mesa/state_tracker/st_cb_fbo.c +++ b/src/mesa/state_tracker/st_cb_fbo.c @@ -105,8 +105,8 @@ st_renderbuffer_alloc_storage(GLcontext * ctx, struct gl_renderbuffer *rb, assert(strb->format != PIPE_FORMAT_NONE); - strb->stride = pf_get_stride(strb->format, width); - size = pf_get_2d_size(strb->format, strb->stride, height); + strb->stride = util_format_get_stride(strb->format, width); + size = util_format_get_2d_size(strb->format, strb->stride, height); strb->data = _mesa_malloc(size); diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c index 000e6eb2a5d..6e1ecb1c502 100644 --- a/src/mesa/state_tracker/st_cb_texture.c +++ b/src/mesa/state_tracker/st_cb_texture.c @@ -833,7 +833,7 @@ decompress_with_blit(GLcontext * ctx, GLenum target, GLint level, /* copy/pack data into user buffer */ if (st_equal_formats(stImage->pt->format, format, type)) { /* memcpy */ - const uint bytesPerRow = width * util_format_get_size(stImage->pt->format); + const uint bytesPerRow = width * util_format_get_blocksize(stImage->pt->format); ubyte *map = screen->transfer_map(screen, tex_xfer); GLuint row; for (row = 0; row < height; row++) { @@ -915,7 +915,7 @@ st_get_tex_image(GLcontext * ctx, GLenum target, GLint level, PIPE_TRANSFER_READ, 0, 0, stImage->base.Width, stImage->base.Height); - texImage->RowStride = stImage->transfer->stride / pf_get_blocksize(stImage->pt->format); + texImage->RowStride = stImage->transfer->stride / util_format_get_blocksize(stImage->pt->format); } else { /* Otherwise, the image should actually be stored in @@ -1178,7 +1178,7 @@ st_CompressedTexSubImage2D(GLcontext *ctx, GLenum target, GLint level, xoffset, yoffset, width, height); - srcBlockStride = pf_get_stride(pformat, width); + srcBlockStride = util_format_get_stride(pformat, width); dstBlockStride = stImage->transfer->stride; } else { assert(stImage->pt); @@ -1192,16 +1192,16 @@ st_CompressedTexSubImage2D(GLcontext *ctx, GLenum target, GLint level, return; } - assert(xoffset % pf_get_blockwidth(pformat) == 0); - assert(yoffset % pf_get_blockheight(pformat) == 0); - assert(width % pf_get_blockwidth(pformat) == 0); - assert(height % pf_get_blockheight(pformat) == 0); + assert(xoffset % util_format_get_blockwidth(pformat) == 0); + assert(yoffset % util_format_get_blockheight(pformat) == 0); + assert(width % util_format_get_blockwidth(pformat) == 0); + assert(height % util_format_get_blockheight(pformat) == 0); - for (y = 0; y < height; y += pf_get_blockheight(pformat)) { + for (y = 0; y < height; y += util_format_get_blockheight(pformat)) { /* don't need to adjust for xoffset and yoffset as st_texture_image_map does that */ - const char *src = (const char*)data + srcBlockStride * pf_get_nblocksy(pformat, y); - char *dst = (char*)texImage->Data + dstBlockStride * pf_get_nblocksy(pformat, y); - memcpy(dst, src, pf_get_stride(pformat, width)); + const char *src = (const char*)data + srcBlockStride * util_format_get_nblocksy(pformat, y); + char *dst = (char*)texImage->Data + dstBlockStride * util_format_get_nblocksy(pformat, y); + memcpy(dst, src, util_format_get_stride(pformat, width)); } if (stImage->pt) { @@ -1691,10 +1691,10 @@ copy_image_data_to_texture(struct st_context *st, dstLevel, stImage->base.Data, stImage->base.RowStride * - pf_get_blocksize(stObj->pt->format), + util_format_get_blocksize(stObj->pt->format), stImage->base.RowStride * stImage->base.Height * - pf_get_blocksize(stObj->pt->format)); + util_format_get_blocksize(stObj->pt->format)); _mesa_align_free(stImage->base.Data); stImage->base.Data = NULL; } diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c index 77005518302..5a433dd7b90 100644 --- a/src/mesa/state_tracker/st_gen_mipmap.c +++ b/src/mesa/state_tracker/st_gen_mipmap.c @@ -37,6 +37,7 @@ #include "pipe/p_context.h" #include "pipe/p_defines.h" #include "pipe/p_inlines.h" +#include "util/u_format.h" #include "util/u_gen_mipmap.h" #include "util/u_math.h" @@ -146,8 +147,8 @@ fallback_generate_mipmap(GLcontext *ctx, GLenum target, srcData = (ubyte *) screen->transfer_map(screen, srcTrans); dstData = (ubyte *) screen->transfer_map(screen, dstTrans); - srcStride = srcTrans->stride / pf_get_blocksize(srcTrans->texture->format); - dstStride = dstTrans->stride / pf_get_blocksize(dstTrans->texture->format); + srcStride = srcTrans->stride / util_format_get_blocksize(srcTrans->texture->format); + dstStride = dstTrans->stride / util_format_get_blocksize(dstTrans->texture->format); _mesa_generate_mipmap_level(target, datatype, comps, 0 /*border*/, -- cgit v1.2.3 From 6ba83cd63f84a4d74dd679d62662d59533fd1bdb Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Wed, 16 Dec 2009 06:07:39 +0100 Subject: r300g: add texture compare modes --- src/gallium/drivers/r300/r300_context.h | 2 + src/gallium/drivers/r300/r300_emit.c | 17 +++-- src/gallium/drivers/r300/r300_fs.c | 102 +++++++++++++++++++++++------ src/gallium/drivers/r300/r300_fs.h | 32 ++++++--- src/gallium/drivers/r300/r300_state.c | 23 ++++++- src/gallium/drivers/r300/r300_tgsi_to_rc.c | 9 ++- 6 files changed, 148 insertions(+), 37 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h index 6bd2766730f..232530b7dc4 100644 --- a/src/gallium/drivers/r300/r300_context.h +++ b/src/gallium/drivers/r300/r300_context.h @@ -91,6 +91,8 @@ struct r300_rs_block { }; struct r300_sampler_state { + struct pipe_sampler_state state; + uint32_t filter0; /* R300_TX_FILTER0: 0x4400 */ uint32_t filter1; /* R300_TX_FILTER1: 0x4440 */ uint32_t border_color; /* R300_TX_BORDER_COLOR: 0x45c0 */ diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index d7b6511d6d5..199ce3a945d 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -158,6 +158,13 @@ static const float * get_shader_constant( vec[1] = 1.0 / tex->height0; break; + /* Texture compare-fail value. */ + /* XXX Since Gallium doesn't support GL_ARB_shadow_ambient, + * this is always (0,0,0,0). */ + case RC_STATE_SHADOW_AMBIENT: + vec[3] = 0; + break; + default: debug_printf("r300: Implementation error: " "Unknown RC_CONSTANT type %d\n", constant->u.State[0]); @@ -1030,18 +1037,20 @@ validate: if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER) { if (r300screen->caps->is_r500) { - r500_emit_fragment_program_code(r300, &r300->fs->code); + r500_emit_fragment_program_code(r300, &r300->fs->shader->code); } else { - r300_emit_fragment_program_code(r300, &r300->fs->code); + r300_emit_fragment_program_code(r300, &r300->fs->shader->code); } r300->dirty_state &= ~R300_NEW_FRAGMENT_SHADER; } if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER_CONSTANTS) { if (r300screen->caps->is_r500) { - r500_emit_fs_constant_buffer(r300, &r300->fs->code.constants); + r500_emit_fs_constant_buffer(r300, + &r300->fs->shader->code.constants); } else { - r300_emit_fs_constant_buffer(r300, &r300->fs->code.constants); + r300_emit_fs_constant_buffer(r300, + &r300->fs->shader->code.constants); } r300->dirty_state &= ~R300_NEW_FRAGMENT_SHADER_CONSTANTS; } diff --git a/src/gallium/drivers/r300/r300_fs.c b/src/gallium/drivers/r300/r300_fs.c index 79b01bb4dc2..4e1b61ca409 100644 --- a/src/gallium/drivers/r300/r300_fs.c +++ b/src/gallium/drivers/r300/r300_fs.c @@ -22,6 +22,9 @@ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE * USE OR OTHER DEALINGS IN THE SOFTWARE. */ +#include "util/u_math.h" +#include "util/u_memory.h" + #include "tgsi/tgsi_dump.h" #include "r300_context.h" @@ -33,8 +36,8 @@ #include "radeon_compiler.h" /* Convert info about FS input semantics to r300_shader_semantics. */ -static void r300_shader_read_fs_inputs(struct tgsi_shader_info* info, - struct r300_shader_semantics* fs_inputs) +void r300_shader_read_fs_inputs(struct tgsi_shader_info* info, + struct r300_shader_semantics* fs_inputs) { int i; unsigned index; @@ -66,7 +69,6 @@ static void r300_shader_read_fs_inputs(struct tgsi_shader_info* info, } } - static void find_output_registers(struct r300_fragment_program_compiler * compiler, struct r300_fragment_shader * fs) { @@ -95,7 +97,7 @@ static void allocate_hardware_inputs( void * mydata) { struct r300_shader_semantics* inputs = - &((struct r300_fragment_shader*)c->UserData)->inputs; + (struct r300_shader_semantics*)c->UserData; int i, reg = 0; /* Allocate input registers. */ @@ -114,31 +116,45 @@ static void allocate_hardware_inputs( } } -void r300_translate_fragment_shader(struct r300_context* r300, - struct r300_fragment_shader* fs) +static void get_compare_state( + struct r300_context* r300, + struct r300_fragment_program_external_state* state, + unsigned shadow_samplers) +{ + memset(state, 0, sizeof(*state)); + + for (int i = 0; i < r300->sampler_count; i++) { + struct r300_sampler_state* s = r300->sampler_states[i]; + + if (s && s->state.compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) { + /* XXX Gallium doesn't provide us with any information regarding + * this mode, so we are screwed. I'm setting 0 = LUMINANCE. */ + state->unit[i].depth_texture_mode = 0; + + /* Fortunately, no need to translate this. */ + state->unit[i].texture_compare_func = s->state.compare_func; + } + } +} + +static void r300_translate_fragment_shader( + struct r300_context* r300, + struct r300_fragment_shader_code* shader) { + struct r300_fragment_shader* fs = r300->fs; struct r300_fragment_program_compiler compiler; struct tgsi_to_rc ttr; - /* Initialize. */ - r300_shader_read_fs_inputs(&fs->info, &fs->inputs); - /* Setup the compiler. */ memset(&compiler, 0, sizeof(compiler)); rc_init(&compiler.Base); compiler.Base.Debug = DBG_ON(r300, DBG_FP); - compiler.code = &fs->code; + compiler.code = &shader->code; + compiler.state = shader->compare_state; compiler.is_r500 = r300_screen(r300->context.screen)->caps->is_r500; compiler.AllocateHwInputs = &allocate_hardware_inputs; - compiler.UserData = fs; - - /* XXX: Program compilation depends on texture compare modes, - * which are sampler state. Therefore, programs need to be recompiled - * depending on this state as in the classic Mesa driver. - * - * This is not yet handled correctly. - */ + compiler.UserData = &fs->inputs; find_output_registers(&compiler, fs); @@ -153,6 +169,8 @@ void r300_translate_fragment_shader(struct r300_context* r300, r300_tgsi_to_rc(&ttr, fs->state.tokens); + fs->shadow_samplers = compiler.Base.Program.ShadowSamplers; + /* Invoke the compiler */ r3xx_compile_fragment_program(&compiler); if (compiler.Base.Error) { @@ -164,5 +182,51 @@ void r300_translate_fragment_shader(struct r300_context* r300, /* And, finally... */ rc_destroy(&compiler.Base); - fs->translated = TRUE; +} + +boolean r300_pick_fragment_shader(struct r300_context* r300) +{ + struct r300_fragment_shader* fs = r300->fs; + struct r300_fragment_program_external_state state; + struct r300_fragment_shader_code* ptr; + + if (!fs->first) { + /* Build the fragment shader for the first time. */ + fs->first = fs->shader = CALLOC_STRUCT(r300_fragment_shader_code); + + /* BTW shadow samplers will be known after the first translation, + * therefore we set ~0, which means it should look at all sampler + * states. This choice doesn't have any impact on the correctness. */ + get_compare_state(r300, &fs->shader->compare_state, ~0); + r300_translate_fragment_shader(r300, fs->shader); + return TRUE; + + } else if (fs->shadow_samplers) { + get_compare_state(r300, &state, fs->shadow_samplers); + + /* Check if the currently-bound shader has been compiled + * with the texture-compare state we need. */ + if (memcmp(&fs->shader->compare_state, &state, sizeof(state)) != 0) { + /* Search for the right shader. */ + ptr = fs->first; + while (ptr) { + if (memcmp(&ptr->compare_state, &state, sizeof(state)) == 0) { + fs->shader = ptr; + return TRUE; + } + ptr = ptr->next; + } + + /* Not found, gotta compile a new one. */ + ptr = CALLOC_STRUCT(r300_fragment_shader_code); + ptr->next = fs->first; + fs->first = fs->shader = ptr; + + ptr->compare_state = state; + r300_translate_fragment_shader(r300, ptr); + return TRUE; + } + } + + return FALSE; } diff --git a/src/gallium/drivers/r300/r300_fs.h b/src/gallium/drivers/r300/r300_fs.h index 630e2d0c8a5..40ce874353c 100644 --- a/src/gallium/drivers/r300/r300_fs.h +++ b/src/gallium/drivers/r300/r300_fs.h @@ -30,6 +30,13 @@ #include "radeon_code.h" #include "r300_shader_semantics.h" +struct r300_fragment_shader_code { + struct r300_fragment_program_external_state compare_state; + struct rX00_fragment_program_code code; + + struct r300_fragment_shader_code* next; +}; + struct r300_fragment_shader { /* Parent class */ struct pipe_shader_state state; @@ -37,21 +44,28 @@ struct r300_fragment_shader { struct tgsi_shader_info info; struct r300_shader_semantics inputs; - /* Has this shader been translated yet? */ - boolean translated; + /* Bits 0-15: TRUE if it's a shadow sampler, FALSE otherwise. */ + unsigned shadow_samplers; - /* Compiled code */ - struct rX00_fragment_program_code code; + /* Currently-bound fragment shader. */ + struct r300_fragment_shader_code* shader; + + /* List of the same shaders compiled with different texture-compare + * states. */ + struct r300_fragment_shader_code* first; }; +void r300_shader_read_fs_inputs(struct tgsi_shader_info* info, + struct r300_shader_semantics* fs_inputs); -void r300_translate_fragment_shader(struct r300_context* r300, - struct r300_fragment_shader* fs); +/* Return TRUE if the shader was switched and should be re-emitted. */ +boolean r300_pick_fragment_shader(struct r300_context* r300); -static inline boolean r300_fragment_shader_writes_depth(struct r300_fragment_shader *fs) +static INLINE boolean r300_fragment_shader_writes_depth(struct r300_fragment_shader *fs) { if (!fs) - return FALSE; - return (fs->code.writes_depth) ? TRUE : FALSE; + return FALSE; + return (fs->shader->code.writes_depth) ? TRUE : FALSE; } + #endif /* R300_FS_H */ diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index 5f332b2e0c2..8bcd6c50608 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -355,6 +355,7 @@ static void* r300_create_fs_state(struct pipe_context* pipe, fs->state.tokens = tgsi_dup_tokens(shader->tokens); tgsi_scan_shader(shader->tokens, &fs->info); + r300_shader_read_fs_inputs(&fs->info, &fs->inputs); return (void*)fs; } @@ -368,11 +369,10 @@ static void r300_bind_fs_state(struct pipe_context* pipe, void* shader) if (fs == NULL) { r300->fs = NULL; return; - } else if (!fs->translated) { - r300_translate_fragment_shader(r300, fs); } r300->fs = fs; + r300_pick_fragment_shader(r300); r300->dirty_state |= R300_NEW_FRAGMENT_SHADER | R300_NEW_FRAGMENT_SHADER_CONSTANTS; } @@ -381,7 +381,14 @@ static void r300_bind_fs_state(struct pipe_context* pipe, void* shader) static void r300_delete_fs_state(struct pipe_context* pipe, void* shader) { struct r300_fragment_shader* fs = (struct r300_fragment_shader*)shader; - rc_constants_destroy(&fs->code.constants); + struct r300_fragment_shader_code *tmp, *ptr = fs->first; + + while (ptr) { + tmp = ptr; + ptr = ptr->next; + rc_constants_destroy(&tmp->code.constants); + FREE(tmp); + } FREE((void*)fs->state.tokens); FREE(shader); } @@ -547,6 +554,8 @@ static void* int lod_bias; union util_color uc; + sampler->state = *state; + sampler->filter0 |= (r300_translate_wrap(state->wrap_s) << R300_TX_WRAP_S_SHIFT) | (r300_translate_wrap(state->wrap_t) << R300_TX_WRAP_T_SHIFT) | @@ -597,6 +606,14 @@ static void r300_bind_sampler_states(struct pipe_context* pipe, } r300->sampler_count = count; + + /* Pick a fragment shader based on the texture compare state. */ + if (r300->fs && (r300->dirty_state & R300_ANY_NEW_SAMPLERS)) { + if (r300_pick_fragment_shader(r300)) { + r300->dirty_state |= R300_NEW_FRAGMENT_SHADER | + R300_NEW_FRAGMENT_SHADER_CONSTANTS; + } + } } static void r300_lacks_vertex_textures(struct pipe_context* pipe, diff --git a/src/gallium/drivers/r300/r300_tgsi_to_rc.c b/src/gallium/drivers/r300/r300_tgsi_to_rc.c index 9fb2de24032..096cdb20bbe 100644 --- a/src/gallium/drivers/r300/r300_tgsi_to_rc.c +++ b/src/gallium/drivers/r300/r300_tgsi_to_rc.c @@ -212,7 +212,8 @@ static void transform_srcreg( dst->Negate = src->Register.Negate ? RC_MASK_XYZW : 0; } -static void transform_texture(struct rc_instruction * dst, struct tgsi_instruction_texture src) +static void transform_texture(struct rc_instruction * dst, struct tgsi_instruction_texture src, + uint32_t *shadowSamplers) { switch(src.Texture) { case TGSI_TEXTURE_1D: @@ -233,14 +234,17 @@ static void transform_texture(struct rc_instruction * dst, struct tgsi_instructi case TGSI_TEXTURE_SHADOW1D: dst->U.I.TexSrcTarget = RC_TEXTURE_1D; dst->U.I.TexShadow = 1; + *shadowSamplers |= 1 << dst->U.I.TexSrcUnit; break; case TGSI_TEXTURE_SHADOW2D: dst->U.I.TexSrcTarget = RC_TEXTURE_2D; dst->U.I.TexShadow = 1; + *shadowSamplers |= 1 << dst->U.I.TexSrcUnit; break; case TGSI_TEXTURE_SHADOWRECT: dst->U.I.TexSrcTarget = RC_TEXTURE_RECT; dst->U.I.TexShadow = 1; + *shadowSamplers |= 1 << dst->U.I.TexSrcUnit; break; } } @@ -269,7 +273,8 @@ static void transform_instruction(struct tgsi_to_rc * ttr, struct tgsi_full_inst /* Texturing. */ if (src->Instruction.Texture) - transform_texture(dst, src->Texture); + transform_texture(dst, src->Texture, + &ttr->compiler->Program.ShadowSamplers); } static void handle_immediate(struct tgsi_to_rc * ttr, struct tgsi_full_immediate * imm) -- cgit v1.2.3 From be6fb5c08350138ac85c1bf727553aaa9f4583e6 Mon Sep 17 00:00:00 2001 From: Vinson Lee Date: Sat, 2 Jan 2010 20:17:43 -0800 Subject: r300g: Use C-style comments. --- src/gallium/drivers/r300/r300_emit.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index eeb97a2d370..b44c7bdbb3d 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -871,7 +871,7 @@ validate: goto validate; } } else { - // debug_printf("No VBO while emitting dirty state!\n"); + /* debug_printf("No VBO while emitting dirty state!\n"); */ } if (!r300->winsys->validate(r300->winsys)) { r300->context.flush(&r300->context, 0, NULL); @@ -999,7 +999,7 @@ validate: */ /* Finally, emit the VBO. */ - //r300_emit_vertex_buffer(r300); + /* r300_emit_vertex_buffer(r300); */ r300->dirty_hw++; } -- cgit v1.2.3 From e78648db0a91269ed83238bfa3ced5d4d7dae2e0 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Thu, 24 Dec 2009 17:40:53 +0100 Subject: r300g: fix outputting depth in the FS --- src/gallium/drivers/r300/r300_emit.c | 17 +++++++++++++++++ src/gallium/drivers/r300/r300_state_invariant.c | 4 +--- 2 files changed, 18 insertions(+), 3 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 1dc9216a7b2..596ec21bc0e 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -283,6 +283,22 @@ void r300_emit_fs_constant_buffer(struct r300_context* r300, END_CS; } +static void r300_emit_fragment_depth_config(struct r300_context* r300, + struct r300_fragment_shader* fs) +{ + CS_LOCALS(r300); + + BEGIN_CS(4); + if (r300_fragment_shader_writes_depth(fs)) { + OUT_CS_REG(R300_FG_DEPTH_SRC, R300_FG_DEPTH_SRC_SHADER); + OUT_CS_REG(R300_US_W_FMT, R300_W_FMT_W24 | R300_W_SRC_US); + } else { + OUT_CS_REG(R300_FG_DEPTH_SRC, R300_FG_DEPTH_SRC_SCAN); + OUT_CS_REG(R300_US_W_FMT, R300_W_FMT_W0 | R300_W_SRC_US); + } + END_CS; +} + void r500_emit_fragment_program_code(struct r300_context* r300, struct rX00_fragment_program_code* generic_code) { @@ -1036,6 +1052,7 @@ validate: } if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER) { + r300_emit_fragment_depth_config(r300, r300->fs); if (r300screen->caps->is_r500) { r500_emit_fragment_program_code(r300, &r300->fs->shader->code); } else { diff --git a/src/gallium/drivers/r300/r300_state_invariant.c b/src/gallium/drivers/r300/r300_state_invariant.c index 64dc14ea767..f25f3ca217d 100644 --- a/src/gallium/drivers/r300/r300_state_invariant.c +++ b/src/gallium/drivers/r300/r300_state_invariant.c @@ -43,7 +43,7 @@ void r300_emit_invariant_state(struct r300_context* r300) struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps; CS_LOCALS(r300); - BEGIN_CS(20 + (caps->has_tcl ? 2: 0)); + BEGIN_CS(16 + (caps->has_tcl ? 2: 0)); /*** Graphics Backend (GB) ***/ /* Various GB enables */ @@ -66,8 +66,6 @@ void r300_emit_invariant_state(struct r300_context* r300) OUT_CS_REG(R300_FG_FOG_COLOR_R, 0x0); OUT_CS_REG(R300_FG_FOG_COLOR_G, 0x0); OUT_CS_REG(R300_FG_FOG_COLOR_B, 0x0); - OUT_CS_REG(R300_FG_DEPTH_SRC, 0x0); - OUT_CS_REG(R300_US_W_FMT, 0x0); /*** VAP ***/ /* Sign/normalize control */ -- cgit v1.2.3 From bf60eb3fec844a7c3793aba0c70da56b74a17344 Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Thu, 24 Dec 2009 03:10:33 +0100 Subject: r300g: add WPOS --- src/gallium/drivers/r300/r300_emit.c | 22 ++++++++++ src/gallium/drivers/r300/r300_fs.c | 21 +++++++++ src/gallium/drivers/r300/r300_shader_semantics.h | 2 + src/gallium/drivers/r300/r300_state.c | 6 +++ src/gallium/drivers/r300/r300_state_derived.c | 20 +++++++++ src/gallium/drivers/r300/r300_vs.c | 56 ++++++++++++++++++++++-- 6 files changed, 124 insertions(+), 3 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 596ec21bc0e..30a1000c294 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -165,6 +165,28 @@ static const float * get_shader_constant( vec[3] = 0; break; + case RC_STATE_R300_VIEWPORT_SCALE: + if (r300->rs_state->enable_vte) { + vec[0] = r300->viewport_state->xscale; + vec[1] = r300->viewport_state->yscale; + vec[2] = r300->viewport_state->zscale; + } else { + vec[0] = 1; + vec[1] = 1; + vec[2] = 1; + } + break; + + case RC_STATE_R300_VIEWPORT_OFFSET: + if (r300->rs_state->enable_vte) { + vec[0] = r300->viewport_state->xoffset; + vec[1] = r300->viewport_state->yoffset; + vec[2] = r300->viewport_state->zoffset; + } else { + /* Zeros. */ + } + break; + default: debug_printf("r300: Implementation error: " "Unknown RC_CONSTANT type %d\n", constant->u.State[0]); diff --git a/src/gallium/drivers/r300/r300_fs.c b/src/gallium/drivers/r300/r300_fs.c index 4e1b61ca409..60ea9c171d5 100644 --- a/src/gallium/drivers/r300/r300_fs.c +++ b/src/gallium/drivers/r300/r300_fs.c @@ -63,6 +63,11 @@ void r300_shader_read_fs_inputs(struct tgsi_shader_info* info, fs_inputs->fog = i; break; + case TGSI_SEMANTIC_POSITION: + assert(index == 0); + fs_inputs->wpos = i; + break; + default: assert(0); } @@ -114,6 +119,9 @@ static void allocate_hardware_inputs( if (inputs->fog != ATTR_UNUSED) { allocate(mydata, inputs->fog, reg++); } + if (inputs->wpos != ATTR_UNUSED) { + allocate(mydata, inputs->wpos, reg++); + } } static void get_compare_state( @@ -144,6 +152,7 @@ static void r300_translate_fragment_shader( struct r300_fragment_shader* fs = r300->fs; struct r300_fragment_program_compiler compiler; struct tgsi_to_rc ttr; + int wpos = fs->inputs.wpos; /* Setup the compiler. */ memset(&compiler, 0, sizeof(compiler)); @@ -171,6 +180,18 @@ static void r300_translate_fragment_shader( fs->shadow_samplers = compiler.Base.Program.ShadowSamplers; + /** + * Transform the program to support WPOS. + * + * Introduce a small fragment at the start of the program that will be + * the only code that directly reads the WPOS input. + * All other code pieces that reference that input will be rewritten + * to read from a newly allocated temporary. */ + if (wpos != ATTR_UNUSED) { + /* Moving the input to some other reg is not really necessary. */ + rc_transform_fragment_wpos(&compiler.Base, wpos, wpos, TRUE); + } + /* Invoke the compiler */ r3xx_compile_fragment_program(&compiler); if (compiler.Base.Error) { diff --git a/src/gallium/drivers/r300/r300_shader_semantics.h b/src/gallium/drivers/r300/r300_shader_semantics.h index 85184e2cfd7..6796841b29b 100644 --- a/src/gallium/drivers/r300/r300_shader_semantics.h +++ b/src/gallium/drivers/r300/r300_shader_semantics.h @@ -40,6 +40,7 @@ struct r300_shader_semantics { int bcolor[ATTR_COLOR_COUNT]; int generic[ATTR_GENERIC_COUNT]; int fog; + int wpos; }; static INLINE void r300_shader_semantics_reset( @@ -50,6 +51,7 @@ static INLINE void r300_shader_semantics_reset( info->pos = ATTR_UNUSED; info->psize = ATTR_UNUSED; info->fog = ATTR_UNUSED; + info->wpos = ATTR_UNUSED; for (i = 0; i < ATTR_COLOR_COUNT; i++) { info->color[i] = ATTR_UNUSED; diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index 872a393321d..ad8e210dcd8 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -719,6 +719,9 @@ static void r300_bind_rs_state(struct pipe_context* pipe, void* state) r300->dirty_state |= R300_NEW_RS_BLOCK; r300->dirty_state |= R300_NEW_SCISSOR; r300->dirty_state |= R300_NEW_VIEWPORT; + if (r300->fs && r300->fs->inputs.wpos != ATTR_UNUSED) { + r300->dirty_state |= R300_NEW_FRAGMENT_SHADER_CONSTANTS; + } } /* Free rasterizer state. */ @@ -897,6 +900,9 @@ static void r300_set_viewport_state(struct pipe_context* pipe, } r300->dirty_state |= R300_NEW_VIEWPORT; + if (r300->fs && r300->fs->inputs.wpos != ATTR_UNUSED) { + r300->dirty_state |= R300_NEW_FRAGMENT_SHADER_CONSTANTS; + } } static void r300_set_vertex_buffers(struct pipe_context* pipe, diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c index 727ae7ade6d..242716fcccb 100644 --- a/src/gallium/drivers/r300/r300_state_derived.c +++ b/src/gallium/drivers/r300/r300_state_derived.c @@ -410,6 +410,26 @@ static void r300_update_rs_block(struct r300_context* r300, } } + /* Rasterize WPOS. */ + if (vs_outputs->wpos != ATTR_UNUSED) { + /* Always rasterize if it's written by the VS, + * otherwise it locks up. */ + rX00_rs_tex(rs, tex_count, tex_count, FALSE); + + /* Write it to the FS input register if it's used by the FS. */ + if (fs_inputs->wpos != ATTR_UNUSED) { + rX00_rs_tex_write(rs, tex_count, fp_offset); + fp_offset++; + } + tex_count++; + } else { + /* Skip the FS input register, leave it uninitialized. */ + /* If we try to set it to (0,0,0,1), it will lock up. */ + if (fs_inputs->wpos != ATTR_UNUSED) { + fp_offset++; + } + } + /* Rasterize at least one color, or bad things happen. */ if (col_count == 0 && tex_count == 0) { rX00_rs_col(rs, 0, 0, TRUE); diff --git a/src/gallium/drivers/r300/r300_vs.c b/src/gallium/drivers/r300/r300_vs.c index c4ed0d712f9..57531fc554f 100644 --- a/src/gallium/drivers/r300/r300_vs.c +++ b/src/gallium/drivers/r300/r300_vs.c @@ -33,6 +33,8 @@ #include "radeon_compiler.h" +#include "util/u_math.h" + /* Convert info about VS output semantics into r300_shader_semantics. */ static void r300_shader_read_vs_outputs( struct tgsi_shader_info* info, @@ -144,6 +146,13 @@ static void r300_shader_vap_output_fmt( gen_count++; } + /* WPOS. */ + if (vs_outputs->wpos != ATTR_UNUSED) { + hwfmt[1] |= (R300_INPUT_CNTL_TC0 << gen_count); + hwfmt[3] |= (4 << (3 * gen_count)); + gen_count++; + } + /* XXX magic */ assert(gen_count <= 8); } @@ -195,6 +204,13 @@ static void r300_stream_locations_notcl( gen_count++; } + /* WPOS. */ + if (vs_outputs->wpos != ATTR_UNUSED) { + assert(tabi < 16); + stream_loc[tabi++] = 6 + gen_count; + gen_count++; + } + /* XXX magic */ assert(gen_count <= 8); @@ -246,6 +262,33 @@ static void set_vertex_inputs_outputs(struct r300_vertex_program_compiler * c) if (outputs->fog != ATTR_UNUSED) { c->code->outputs[outputs->fog] = reg++; } + + /* WPOS. */ + if (outputs->wpos != ATTR_UNUSED) { + c->code->outputs[outputs->wpos] = reg++; + } +} + +static void r300_insert_wpos(struct r300_vertex_program_compiler* c, + struct r300_shader_semantics* outputs) +{ + int i, lastOutput = 0; + + /* Find the max output index. */ + lastOutput = MAX2(lastOutput, outputs->psize); + for (i = 0; i < ATTR_COLOR_COUNT; i++) { + lastOutput = MAX2(lastOutput, outputs->color[i]); + lastOutput = MAX2(lastOutput, outputs->bcolor[i]); + } + for (i = 0; i < ATTR_GENERIC_COUNT; i++) { + lastOutput = MAX2(lastOutput, outputs->generic[i]); + } + lastOutput = MAX2(lastOutput, outputs->fog); + + /* Set WPOS after the last output. */ + lastOutput++; + rc_copy_output(&c->Base, 0, lastOutput); /* out[lastOutput] = out[0]; */ + outputs->wpos = lastOutput; } void r300_translate_vertex_shader(struct r300_context* r300, @@ -253,11 +296,10 @@ void r300_translate_vertex_shader(struct r300_context* r300, { struct r300_vertex_program_compiler compiler; struct tgsi_to_rc ttr; + boolean use_wpos = TRUE; /* Initialize. */ r300_shader_read_vs_outputs(&vs->info, &vs->outputs); - r300_shader_vap_output_fmt(&vs->outputs, vs->hwfmt); - r300_stream_locations_notcl(&vs->outputs, vs->stream_loc_notcl); /* Setup the compiler */ rc_init(&compiler.Base); @@ -277,9 +319,17 @@ void r300_translate_vertex_shader(struct r300_context* r300, r300_tgsi_to_rc(&ttr, vs->state.tokens); - compiler.RequiredOutputs = ~(~0 << vs->info.num_outputs); + compiler.RequiredOutputs = ~(~0 << (vs->info.num_outputs+use_wpos)); compiler.SetHwInputOutput = &set_vertex_inputs_outputs; + /* Insert the WPOS output. */ + if (use_wpos) { + r300_insert_wpos(&compiler, &vs->outputs); + } + + r300_shader_vap_output_fmt(&vs->outputs, vs->hwfmt); + r300_stream_locations_notcl(&vs->outputs, vs->stream_loc_notcl); + /* Invoke the compiler */ r3xx_compile_vertex_program(&compiler); if (compiler.Base.Error) { -- cgit v1.2.3 From 1a4f242be60237fb1f1acf346b1e641167bc6cee Mon Sep 17 00:00:00 2001 From: Marek Olšák Date: Thu, 24 Dec 2009 14:23:57 +0100 Subject: r300g: if no DS buffer is set, disable reading from and writing to it It fixes another "Bad CS" issue. --- src/gallium/drivers/r300/r300_emit.c | 11 +++++++++-- src/gallium/drivers/r300/r300_state.c | 1 + 2 files changed, 10 insertions(+), 2 deletions(-) (limited to 'src/gallium/drivers/r300/r300_emit.c') diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c index 30a1000c294..f8bfa714fef 100644 --- a/src/gallium/drivers/r300/r300_emit.c +++ b/src/gallium/drivers/r300/r300_emit.c @@ -121,8 +121,15 @@ void r300_emit_dsa_state(struct r300_context* r300, }*/ OUT_CS_REG_SEQ(R300_ZB_CNTL, 3); - OUT_CS(dsa->z_buffer_control); - OUT_CS(dsa->z_stencil_control); + + if (r300->framebuffer_state.zsbuf) { + OUT_CS(dsa->z_buffer_control); + OUT_CS(dsa->z_stencil_control); + } else { + OUT_CS(0); + OUT_CS(0); + } + OUT_CS(dsa->stencil_ref_mask); OUT_CS_REG(R300_ZB_ZTOP, r300->ztop_state.z_buffer_top); diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c index affd0b45910..a145a7f18a5 100644 --- a/src/gallium/drivers/r300/r300_state.c +++ b/src/gallium/drivers/r300/r300_state.c @@ -522,6 +522,7 @@ static void } r300->dirty_state |= R300_NEW_FRAMEBUFFERS; r300->dirty_state |= R300_NEW_BLEND; + r300->dirty_state |= R300_NEW_DSA; } /* Create fragment shader state. */ -- cgit v1.2.3