diff options
Diffstat (limited to 'src/gallium/drivers/r600')
24 files changed, 1578 insertions, 875 deletions
diff --git a/src/gallium/drivers/r600/Makefile b/src/gallium/drivers/r600/Makefile index 0dc9663de6a..436de9c4dbd 100644 --- a/src/gallium/drivers/r600/Makefile +++ b/src/gallium/drivers/r600/Makefile @@ -20,6 +20,8 @@ C_SOURCES = \ r600_video_context.c \ r700_asm.c \ evergreen_state.c \ - eg_asm.c + eg_asm.c \ + r600_translate.c \ + r600_state_common.c include ../../Makefile.template diff --git a/src/gallium/drivers/r600/SConscript b/src/gallium/drivers/r600/SConscript index b6d9b7d8d22..bf0ad8571ba 100644 --- a/src/gallium/drivers/r600/SConscript +++ b/src/gallium/drivers/r600/SConscript @@ -18,6 +18,7 @@ r600 = env.ConvenienceLibrary( source = [ 'r600_asm.c', 'r600_buffer.c', + 'r600_blit.c', 'r600_helper.c', 'r600_pipe.c', 'r600_query.c', diff --git a/src/gallium/drivers/r600/eg_asm.c b/src/gallium/drivers/r600/eg_asm.c index 52b7189e9e5..c30f09c394b 100644 --- a/src/gallium/drivers/r600/eg_asm.c +++ b/src/gallium/drivers/r600/eg_asm.c @@ -36,8 +36,13 @@ int eg_bc_cf_build(struct r600_bc *bc, struct r600_bc_cf *cf) case (EG_V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU << 3): case (EG_V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_PUSH_BEFORE << 3): bc->bytecode[id++] = S_SQ_CF_ALU_WORD0_ADDR(cf->addr >> 1) | - S_SQ_CF_ALU_WORD0_KCACHE_MODE0(cf->kcache0_mode); + S_SQ_CF_ALU_WORD0_KCACHE_MODE0(cf->kcache0_mode) | + S_SQ_CF_ALU_WORD0_KCACHE_BANK0(cf->kcache0_bank) | + S_SQ_CF_ALU_WORD0_KCACHE_BANK1(cf->kcache1_bank); bc->bytecode[id++] = S_SQ_CF_ALU_WORD1_CF_INST(cf->inst >> 3) | + S_SQ_CF_ALU_WORD1_KCACHE_MODE1(cf->kcache1_mode) | + S_SQ_CF_ALU_WORD1_KCACHE_ADDR0(cf->kcache0_addr) | + S_SQ_CF_ALU_WORD1_KCACHE_ADDR1(cf->kcache1_addr) | S_SQ_CF_ALU_WORD1_BARRIER(1) | S_SQ_CF_ALU_WORD1_COUNT((cf->ndw / 2) - 1); break; diff --git a/src/gallium/drivers/r600/eg_state_inlines.h b/src/gallium/drivers/r600/eg_state_inlines.h index 5a4e00aac38..be81c28b43f 100644 --- a/src/gallium/drivers/r600/eg_state_inlines.h +++ b/src/gallium/drivers/r600/eg_state_inlines.h @@ -25,6 +25,7 @@ #include "util/u_format.h" #include "evergreend.h" +#include "r600_formats.h" static INLINE uint32_t r600_translate_blend_function(int blend_func) { @@ -523,32 +524,32 @@ static INLINE uint32_t r600_translate_vertex_data_type(enum pipe_format format) case 16: switch (desc->nr_channels) { case 1: - result = V_030008_FMT_16_FLOAT; + result = FMT_16_FLOAT; break; case 2: - result = V_030008_FMT_16_16_FLOAT; + result = FMT_16_16_FLOAT; break; case 3: - result = V_030008_FMT_16_16_16_FLOAT; + result = FMT_16_16_16_FLOAT; break; case 4: - result = V_030008_FMT_16_16_16_16_FLOAT; + result = FMT_16_16_16_16_FLOAT; break; } break; case 32: switch (desc->nr_channels) { case 1: - result = V_030008_FMT_32_FLOAT; + result = FMT_32_FLOAT; break; case 2: - result = V_030008_FMT_32_32_FLOAT; + result = FMT_32_32_FLOAT; break; case 3: - result = V_030008_FMT_32_32_32_FLOAT; + result = FMT_32_32_32_FLOAT; break; case 4: - result = V_030008_FMT_32_32_32_32_FLOAT; + result = FMT_32_32_32_32_FLOAT; break; } break; @@ -564,48 +565,48 @@ static INLINE uint32_t r600_translate_vertex_data_type(enum pipe_format format) case 8: switch (desc->nr_channels) { case 1: - result = V_030008_FMT_8; + result = FMT_8; break; case 2: - result = V_030008_FMT_8_8; + result = FMT_8_8; break; case 3: // result = V_038008_FMT_8_8_8; /* fails piglit draw-vertices test */ // break; case 4: - result = V_030008_FMT_8_8_8_8; + result = FMT_8_8_8_8; break; } break; case 16: switch (desc->nr_channels) { case 1: - result = V_030008_FMT_16; + result = FMT_16; break; case 2: - result = V_030008_FMT_16_16; + result = FMT_16_16; break; case 3: // result = V_038008_FMT_16_16_16; /* fails piglit draw-vertices test */ // break; case 4: - result = V_030008_FMT_16_16_16_16; + result = FMT_16_16_16_16; break; } break; case 32: switch (desc->nr_channels) { case 1: - result = V_030008_FMT_32; + result = FMT_32; break; case 2: - result = V_030008_FMT_32_32; + result = FMT_32_32; break; case 3: - result = V_030008_FMT_32_32_32; + result = FMT_32_32_32; break; case 4: - result = V_030008_FMT_32_32_32_32; + result = FMT_32_32_32_32; break; } break; diff --git a/src/gallium/drivers/r600/evergreen_state.c b/src/gallium/drivers/r600/evergreen_state.c index 0fd1d399518..72223485067 100644 --- a/src/gallium/drivers/r600/evergreen_state.c +++ b/src/gallium/drivers/r600/evergreen_state.c @@ -39,6 +39,7 @@ #include <util/u_pack_color.h> #include <util/u_memory.h> #include <util/u_inlines.h> +#include <util/u_framebuffer.h> #include <pipebuffer/pb_buffer.h> #include "r600.h" #include "evergreend.h" @@ -136,20 +137,6 @@ static void *evergreen_create_blend_state(struct pipe_context *ctx, return rstate; } -static void evergreen_bind_blend_state(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - struct r600_pipe_blend *blend = (struct r600_pipe_blend *)state; - struct r600_pipe_state *rstate; - - if (state == NULL) - return; - rstate = &blend->rstate; - rctx->states[rstate->id] = rstate; - rctx->cb_target_mask = blend->cb_target_mask; - r600_context_pipe_state_set(&rctx->ctx, rstate); -} - static void *evergreen_create_dsa_state(struct pipe_context *ctx, const struct pipe_depth_stencil_alpha_state *state) { @@ -241,6 +228,7 @@ static void *evergreen_create_rs_state(struct pipe_context *ctx, struct r600_pipe_state *rstate; unsigned tmp; unsigned prov_vtx = 1, polygon_dual_mode; + unsigned clip_rule; if (rs == NULL) { return NULL; @@ -250,6 +238,8 @@ static void *evergreen_create_rs_state(struct pipe_context *ctx, rs->flatshade = state->flatshade; rs->sprite_coord_enable = state->sprite_coord_enable; + clip_rule = state->scissor ? 0xAAAA : 0xFFFF; + /* offset */ rs->offset_units = state->offset_units; rs->offset_scale = state->offset_scale * 12.0f; @@ -257,7 +247,7 @@ static void *evergreen_create_rs_state(struct pipe_context *ctx, rstate->id = R600_PIPE_STATE_RASTERIZER; if (state->flatshade_first) prov_vtx = 0; - tmp = 0x00000001; + tmp = S_0286D4_FLAT_SHADE_ENA(1); if (state->sprite_coord_enable) { tmp |= S_0286D4_PNT_SPRITE_ENA(1) | S_0286D4_PNT_SPRITE_OVRD_X(2) | @@ -299,39 +289,10 @@ static void *evergreen_create_rs_state(struct pipe_context *ctx, r600_pipe_state_add_reg(rstate, R_028C18_PA_CL_GB_HORZ_DISC_ADJ, 0x3F800000, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_028B7C_PA_SU_POLY_OFFSET_CLAMP, 0x0, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_028C08_PA_SU_VTX_CNTL, 0x00000005, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_02820C_PA_SC_CLIPRECT_RULE, clip_rule, 0xFFFFFFFF, NULL); return rstate; } -static void evergreen_bind_rs_state(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_rasterizer *rs = (struct r600_pipe_rasterizer *)state; - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - - if (state == NULL) - return; - - rctx->flatshade = rs->flatshade; - rctx->sprite_coord_enable = rs->sprite_coord_enable; - rctx->rasterizer = rs; - - rctx->states[rs->rstate.id] = &rs->rstate; - r600_context_pipe_state_set(&rctx->ctx, &rs->rstate); -} - -static void evergreen_delete_rs_state(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - struct r600_pipe_rasterizer *rs = (struct r600_pipe_rasterizer *)state; - - if (rctx->rasterizer == rs) { - rctx->rasterizer = NULL; - } - if (rctx->states[rs->rstate.id] == &rs->rstate) { - rctx->states[rs->rstate.id] = NULL; - } - free(rs); -} - static void *evergreen_create_sampler_state(struct pipe_context *ctx, const struct pipe_sampler_state *state) { @@ -372,28 +333,6 @@ static void *evergreen_create_sampler_state(struct pipe_context *ctx, return rstate; } -static void *evergreen_create_vertex_elements(struct pipe_context *ctx, - unsigned count, - const struct pipe_vertex_element *elements) -{ - struct r600_vertex_element *v = CALLOC_STRUCT(r600_vertex_element); - - assert(count < 32); - v->count = count; - v->refcount = 1; - memcpy(v->elements, elements, count * sizeof(struct pipe_vertex_element)); - return v; -} - -static void evergreen_sampler_view_destroy(struct pipe_context *ctx, - struct pipe_sampler_view *state) -{ - struct r600_pipe_sampler_view *resource = (struct r600_pipe_sampler_view *)state; - - pipe_resource_reference(&state->texture, NULL); - FREE(resource); -} - static struct pipe_sampler_view *evergreen_create_sampler_view(struct pipe_context *ctx, struct pipe_resource *texture, const struct pipe_sampler_view *state) @@ -446,7 +385,7 @@ static struct pipe_sampler_view *evergreen_create_sampler_view(struct pipe_conte bo[0] = rbuffer->bo; bo[1] = rbuffer->bo; } - pitch = align(tmp->pitch[0] / tmp->bpt, 8); + pitch = align(tmp->pitch_in_pixels[0], 8); /* FIXME properly handle first level != 0 */ r600_pipe_state_add_reg(rstate, R_030000_RESOURCE0_WORD0, @@ -464,7 +403,6 @@ static struct pipe_sampler_view *evergreen_create_sampler_view(struct pipe_conte r600_pipe_state_add_reg(rstate, R_030010_RESOURCE0_WORD4, word4 | S_030010_NUM_FORMAT_ALL(V_030010_SQ_NUM_FORMAT_NORM) | S_030010_SRF_MODE_ALL(V_030010_SFR_MODE_NO_ZERO) | - S_030010_REQUEST_SIZE(1) | S_030010_BASE_LEVEL(state->first_level), 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_030014_RESOURCE0_WORD5, S_030014_LAST_LEVEL(state->last_level) | @@ -481,32 +419,42 @@ static struct pipe_sampler_view *evergreen_create_sampler_view(struct pipe_conte static void evergreen_set_vs_sampler_view(struct pipe_context *ctx, unsigned count, struct pipe_sampler_view **views) { - /* TODO */ - assert(1); -} - -static void evergreen_set_ps_sampler_view(struct pipe_context *ctx, unsigned count, - struct pipe_sampler_view **views) -{ struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; struct r600_pipe_sampler_view **resource = (struct r600_pipe_sampler_view **)views; for (int i = 0; i < count; i++) { if (resource[i]) { - evergreen_context_pipe_state_set_ps_resource(&rctx->ctx, &resource[i]->state, i); + evergreen_context_pipe_state_set_vs_resource(&rctx->ctx, &resource[i]->state, i + PIPE_MAX_ATTRIBS); } } } -static void evergreen_bind_state(struct pipe_context *ctx, void *state) +static void evergreen_set_ps_sampler_view(struct pipe_context *ctx, unsigned count, + struct pipe_sampler_view **views) { struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - struct r600_pipe_state *rstate = (struct r600_pipe_state *)state; - - if (state == NULL) - return; - rctx->states[rstate->id] = rstate; - r600_context_pipe_state_set(&rctx->ctx, rstate); + struct r600_pipe_sampler_view **resource = (struct r600_pipe_sampler_view **)views; + int i; + + for (i = 0; i < count; i++) { + if (&rctx->ps_samplers.views[i]->base != views[i]) { + if (resource[i]) + evergreen_context_pipe_state_set_ps_resource(&rctx->ctx, &resource[i]->state, i); + else + evergreen_context_pipe_state_set_ps_resource(&rctx->ctx, NULL, i); + + pipe_sampler_view_reference( + (struct pipe_sampler_view **)&rctx->ps_samplers.views[i], + views[i]); + } + } + for (i = count; i < NUM_TEX_UNITS; i++) { + if (rctx->ps_samplers.views[i]) { + evergreen_context_pipe_state_set_ps_resource(&rctx->ctx, NULL, i); + pipe_sampler_view_reference((struct pipe_sampler_view **)&rctx->ps_samplers.views[i], NULL); + } + } + rctx->ps_samplers.n_views = count; } static void evergreen_bind_ps_sampler(struct pipe_context *ctx, unsigned count, void **states) @@ -514,6 +462,10 @@ static void evergreen_bind_ps_sampler(struct pipe_context *ctx, unsigned count, struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; struct r600_pipe_state **rstates = (struct r600_pipe_state **)states; + + memcpy(rctx->ps_samplers.samplers, states, sizeof(void*) * count); + rctx->ps_samplers.n_samplers = count; + for (int i = 0; i < count; i++) { evergreen_context_pipe_state_set_ps_sampler(&rctx->ctx, rstates[i], i); } @@ -524,37 +476,11 @@ static void evergreen_bind_vs_sampler(struct pipe_context *ctx, unsigned count, struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; struct r600_pipe_state **rstates = (struct r600_pipe_state **)states; - /* TODO implement */ for (int i = 0; i < count; i++) { evergreen_context_pipe_state_set_vs_sampler(&rctx->ctx, rstates[i], i); } } -static void evergreen_delete_state(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - struct r600_pipe_state *rstate = (struct r600_pipe_state *)state; - - if (rctx->states[rstate->id] == rstate) { - rctx->states[rstate->id] = NULL; - } - for (int i = 0; i < rstate->nregs; i++) { - r600_bo_reference(rctx->radeon, &rstate->regs[i].bo, NULL); - } - free(rstate); -} - -static void evergreen_delete_vertex_element(struct pipe_context *ctx, void *state) -{ - struct r600_vertex_element *v = (struct r600_vertex_element*)state; - - if (v == NULL) - return; - if (--v->refcount) - return; - free(v); -} - static void evergreen_set_clip_state(struct pipe_context *ctx, const struct pipe_clip_state *state) { @@ -590,19 +516,6 @@ static void evergreen_set_clip_state(struct pipe_context *ctx, r600_context_pipe_state_set(&rctx->ctx, rstate); } -static void evergreen_bind_vertex_elements(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - struct r600_vertex_element *v = (struct r600_vertex_element*)state; - - evergreen_delete_vertex_element(ctx, rctx->vertex_elements); - rctx->vertex_elements = v; - if (v) { - v->refcount++; -// rctx->vs_rebuild = TRUE; - } -} - static void evergreen_set_polygon_stipple(struct pipe_context *ctx, const struct pipe_poly_stipple *state) { @@ -626,18 +539,6 @@ static void evergreen_set_scissor_state(struct pipe_context *ctx, tl = S_028240_TL_X(state->minx) | S_028240_TL_Y(state->miny); br = S_028244_BR_X(state->maxx) | S_028244_BR_Y(state->maxy); r600_pipe_state_add_reg(rstate, - R_028030_PA_SC_SCREEN_SCISSOR_TL, tl, - 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, - R_028034_PA_SC_SCREEN_SCISSOR_BR, br, - 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, - R_028204_PA_SC_WINDOW_SCISSOR_TL, tl, - 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, - R_028208_PA_SC_WINDOW_SCISSOR_BR, br, - 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028210_PA_SC_CLIPRECT_0_TL, tl, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, @@ -661,15 +562,6 @@ static void evergreen_set_scissor_state(struct pipe_context *ctx, r600_pipe_state_add_reg(rstate, R_02822C_PA_SC_CLIPRECT_3_BR, br, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, - R_028200_PA_SC_WINDOW_OFFSET, 0x00000000, - 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, - R_02820C_PA_SC_CLIPRECT_RULE, 0x0000FFFF, - 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, - R_028230_PA_SC_EDGERULE, 0xAAAAAAAA, - 0xFFFFFFFF, NULL); free(rctx->states[R600_PIPE_STATE_SCISSOR]); rctx->states[R600_PIPE_STATE_SCISSOR] = rstate; @@ -733,6 +625,7 @@ static void evergreen_cb(struct r600_pipe_context *rctx, struct r600_pipe_state { struct r600_resource_texture *rtex; struct r600_resource *rbuffer; + struct r600_surface *surf; unsigned level = state->cbufs[cb]->level; unsigned pitch, slice; unsigned color_info; @@ -740,14 +633,15 @@ static void evergreen_cb(struct r600_pipe_context *rctx, struct r600_pipe_state const struct util_format_description *desc; struct r600_bo *bo[3]; + surf = (struct r600_surface *)state->cbufs[cb]; rtex = (struct r600_resource_texture*)state->cbufs[cb]->texture; rbuffer = &rtex->resource; bo[0] = rbuffer->bo; bo[1] = rbuffer->bo; bo[2] = rbuffer->bo; - pitch = (rtex->pitch[level] / rtex->bpt) / 8 - 1; - slice = (rtex->pitch[level] / rtex->bpt) * state->cbufs[cb]->height / 64 - 1; + pitch = rtex->pitch_in_pixels[level] / 8 - 1; + slice = rtex->pitch_in_pixels[level] * surf->aligned_height / 64 - 1; ntype = 0; desc = util_format_description(rtex->resource.base.b.format); if (desc->colorspace == UTIL_FORMAT_COLORSPACE_SRGB) @@ -794,22 +688,25 @@ static void evergreen_db(struct r600_pipe_context *rctx, struct r600_pipe_state { struct r600_resource_texture *rtex; struct r600_resource *rbuffer; + struct r600_surface *surf; unsigned level; unsigned pitch, slice, format, stencil_format; if (state->zsbuf == NULL) return; + level = state->zsbuf->level; + + surf = (struct r600_surface *)state->zsbuf; rtex = (struct r600_resource_texture*)state->zsbuf->texture; rtex->tiled = 1; - rtex->array_mode = 2; + rtex->array_mode[level] = 2; rtex->tile_type = 1; rtex->depth = 1; rbuffer = &rtex->resource; - level = state->zsbuf->level; - pitch = (rtex->pitch[level] / rtex->bpt) / 8 - 1; - slice = (rtex->pitch[level] / rtex->bpt) * state->zsbuf->height / 64 - 1; + pitch = rtex->pitch_in_pixels[level] / 8 - 1; + slice = rtex->pitch_in_pixels[level] * surf->aligned_height / 64 - 1; format = r600_translate_dbformat(state->zsbuf->texture->format); stencil_format = r600_translate_stencilformat(state->zsbuf->texture->format); @@ -821,7 +718,7 @@ static void evergreen_db(struct r600_pipe_context *rctx, struct r600_pipe_state if (stencil_format) { uint32_t stencil_offset; - stencil_offset = ((state->zsbuf->height * rtex->pitch[level]) + 255) & ~255; + stencil_offset = ((surf->aligned_height * rtex->pitch_in_bytes[level]) + 255) & ~255; r600_pipe_state_add_reg(rstate, R_02804C_DB_STENCIL_READ_BASE, (state->zsbuf->offset + stencil_offset + r600_bo_offset(rbuffer->bo)) >> 8, 0xFFFFFFFF, rbuffer->bo); r600_pipe_state_add_reg(rstate, R_028054_DB_STENCIL_WRITE_BASE, @@ -833,7 +730,7 @@ static void evergreen_db(struct r600_pipe_context *rctx, struct r600_pipe_state S_028044_FORMAT(stencil_format), 0xFFFFFFFF, rbuffer->bo); r600_pipe_state_add_reg(rstate, R_028040_DB_Z_INFO, - S_028040_ARRAY_MODE(rtex->array_mode) | S_028040_FORMAT(format), + S_028040_ARRAY_MODE(rtex->array_mode[level]) | S_028040_FORMAT(format), 0xFFFFFFFF, rbuffer->bo); r600_pipe_state_add_reg(rstate, R_028058_DB_DEPTH_SIZE, S_028058_PITCH_TILE_MAX(pitch), @@ -855,14 +752,9 @@ static void evergreen_set_framebuffer_state(struct pipe_context *ctx, /* unreference old buffer and reference new one */ rstate->id = R600_PIPE_STATE_FRAMEBUFFER; - for (int i = 0; i < rctx->framebuffer.nr_cbufs; i++) { - pipe_surface_reference(&rctx->framebuffer.cbufs[i], NULL); - } - for (int i = 0; i < state->nr_cbufs; i++) { - pipe_surface_reference(&rctx->framebuffer.cbufs[i], state->cbufs[i]); - } - pipe_surface_reference(&rctx->framebuffer.zsbuf, state->zsbuf); - rctx->framebuffer = *state; + + util_copy_framebuffer_state(&rctx->framebuffer, state); + rctx->pframebuffer = &rctx->framebuffer; /* build states */ @@ -895,6 +787,24 @@ static void evergreen_set_framebuffer_state(struct pipe_context *ctx, r600_pipe_state_add_reg(rstate, R_028254_PA_SC_VPORT_SCISSOR_0_BR, br, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, + R_028030_PA_SC_SCREEN_SCISSOR_TL, tl, + 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, + R_028034_PA_SC_SCREEN_SCISSOR_BR, br, + 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, + R_028204_PA_SC_WINDOW_SCISSOR_TL, tl, + 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, + R_028208_PA_SC_WINDOW_SCISSOR_BR, br, + 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, + R_028200_PA_SC_WINDOW_OFFSET, 0x00000000, + 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, + R_028230_PA_SC_EDGERULE, 0xAAAAAAAA, + 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_028238_CB_TARGET_MASK, 0x00000000, target_mask, NULL); @@ -910,40 +820,6 @@ static void evergreen_set_framebuffer_state(struct pipe_context *ctx, r600_context_pipe_state_set(&rctx->ctx, rstate); } -static void evergreen_set_index_buffer(struct pipe_context *ctx, - const struct pipe_index_buffer *ib) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - - if (ib) { - pipe_resource_reference(&rctx->index_buffer.buffer, ib->buffer); - memcpy(&rctx->index_buffer, ib, sizeof(rctx->index_buffer)); - } else { - pipe_resource_reference(&rctx->index_buffer.buffer, NULL); - memset(&rctx->index_buffer, 0, sizeof(rctx->index_buffer)); - } - - /* TODO make this more like a state */ -} - -static void evergreen_set_vertex_buffers(struct pipe_context *ctx, unsigned count, - const struct pipe_vertex_buffer *buffers) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - - for (int i = 0; i < rctx->nvertex_buffer; i++) { - pipe_resource_reference(&rctx->vertex_buffer[i].buffer, NULL); - } - memcpy(rctx->vertex_buffer, buffers, sizeof(struct pipe_vertex_buffer) * count); - for (int i = 0; i < count; i++) { - rctx->vertex_buffer[i].buffer = NULL; - if (r600_buffer_is_user_buffer(buffers[i].buffer)) - rctx->any_user_vbs = TRUE; - pipe_resource_reference(&rctx->vertex_buffer[i].buffer, buffers[i].buffer); - } - rctx->nvertex_buffer = count; -} - static void evergreen_set_constant_buffer(struct pipe_context *ctx, uint shader, uint index, struct pipe_resource *buffer) { @@ -979,84 +855,31 @@ static void evergreen_set_constant_buffer(struct pipe_context *ctx, uint shader, } } -static void *evergreen_create_shader_state(struct pipe_context *ctx, - const struct pipe_shader_state *state) -{ - struct r600_pipe_shader *shader = CALLOC_STRUCT(r600_pipe_shader); - int r; - - r = r600_pipe_shader_create(ctx, shader, state->tokens); - if (r) { - return NULL; - } - return shader; -} - -static void evergreen_bind_ps_shader(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - - /* TODO delete old shader */ - rctx->ps_shader = (struct r600_pipe_shader *)state; -} - -static void evergreen_bind_vs_shader(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - - /* TODO delete old shader */ - rctx->vs_shader = (struct r600_pipe_shader *)state; -} - -static void evergreen_delete_ps_shader(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - struct r600_pipe_shader *shader = (struct r600_pipe_shader *)state; - - if (rctx->ps_shader == shader) { - rctx->ps_shader = NULL; - } - /* TODO proper delete */ - free(shader); -} - -static void evergreen_delete_vs_shader(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - struct r600_pipe_shader *shader = (struct r600_pipe_shader *)state; - - if (rctx->vs_shader == shader) { - rctx->vs_shader = NULL; - } - /* TODO proper delete */ - free(shader); -} - void evergreen_init_state_functions(struct r600_pipe_context *rctx) { rctx->context.create_blend_state = evergreen_create_blend_state; rctx->context.create_depth_stencil_alpha_state = evergreen_create_dsa_state; - rctx->context.create_fs_state = evergreen_create_shader_state; + rctx->context.create_fs_state = r600_create_shader_state; rctx->context.create_rasterizer_state = evergreen_create_rs_state; rctx->context.create_sampler_state = evergreen_create_sampler_state; rctx->context.create_sampler_view = evergreen_create_sampler_view; - rctx->context.create_vertex_elements_state = evergreen_create_vertex_elements; - rctx->context.create_vs_state = evergreen_create_shader_state; - rctx->context.bind_blend_state = evergreen_bind_blend_state; - rctx->context.bind_depth_stencil_alpha_state = evergreen_bind_state; + rctx->context.create_vertex_elements_state = r600_create_vertex_elements; + rctx->context.create_vs_state = r600_create_shader_state; + rctx->context.bind_blend_state = r600_bind_blend_state; + rctx->context.bind_depth_stencil_alpha_state = r600_bind_state; rctx->context.bind_fragment_sampler_states = evergreen_bind_ps_sampler; - rctx->context.bind_fs_state = evergreen_bind_ps_shader; - rctx->context.bind_rasterizer_state = evergreen_bind_rs_state; - rctx->context.bind_vertex_elements_state = evergreen_bind_vertex_elements; + rctx->context.bind_fs_state = r600_bind_ps_shader; + rctx->context.bind_rasterizer_state = r600_bind_rs_state; + rctx->context.bind_vertex_elements_state = r600_bind_vertex_elements; rctx->context.bind_vertex_sampler_states = evergreen_bind_vs_sampler; - rctx->context.bind_vs_state = evergreen_bind_vs_shader; - rctx->context.delete_blend_state = evergreen_delete_state; - rctx->context.delete_depth_stencil_alpha_state = evergreen_delete_state; - rctx->context.delete_fs_state = evergreen_delete_ps_shader; - rctx->context.delete_rasterizer_state = evergreen_delete_rs_state; - rctx->context.delete_sampler_state = evergreen_delete_state; - rctx->context.delete_vertex_elements_state = evergreen_delete_vertex_element; - rctx->context.delete_vs_state = evergreen_delete_vs_shader; + rctx->context.bind_vs_state = r600_bind_vs_shader; + rctx->context.delete_blend_state = r600_delete_state; + rctx->context.delete_depth_stencil_alpha_state = r600_delete_state; + rctx->context.delete_fs_state = r600_delete_ps_shader; + rctx->context.delete_rasterizer_state = r600_delete_rs_state; + rctx->context.delete_sampler_state = r600_delete_state; + rctx->context.delete_vertex_elements_state = r600_delete_vertex_element; + rctx->context.delete_vs_state = r600_delete_vs_shader; rctx->context.set_blend_color = evergreen_set_blend_color; rctx->context.set_clip_state = evergreen_set_clip_state; rctx->context.set_constant_buffer = evergreen_set_constant_buffer; @@ -1066,11 +889,11 @@ void evergreen_init_state_functions(struct r600_pipe_context *rctx) rctx->context.set_sample_mask = evergreen_set_sample_mask; rctx->context.set_scissor_state = evergreen_set_scissor_state; rctx->context.set_stencil_ref = evergreen_set_stencil_ref; - rctx->context.set_vertex_buffers = evergreen_set_vertex_buffers; - rctx->context.set_index_buffer = evergreen_set_index_buffer; + rctx->context.set_vertex_buffers = r600_set_vertex_buffers; + rctx->context.set_index_buffer = r600_set_index_buffer; rctx->context.set_vertex_sampler_views = evergreen_set_vs_sampler_view; rctx->context.set_viewport_state = evergreen_set_viewport_state; - rctx->context.sampler_view_destroy = evergreen_sampler_view_destroy; + rctx->context.sampler_view_destroy = r600_sampler_view_destroy; } void evergreen_init_config(struct r600_pipe_context *rctx) @@ -1353,6 +1176,12 @@ void evergreen_draw(struct pipe_context *ctx, const struct pipe_draw_info *info) struct r600_draw rdraw; struct r600_pipe_state vgt; struct r600_drawl draw; + boolean translate = FALSE; + + if (rctx->vertex_elements->incompatible_layout) { + r600_begin_vertex_translate(rctx); + translate = TRUE; + } if (rctx->any_user_vbs) { r600_upload_user_buffers(rctx); @@ -1429,11 +1258,11 @@ void evergreen_draw(struct pipe_context *ctx, const struct pipe_draw_info *info) vertex_buffer->buffer_offset + r600_bo_offset(rbuffer->bo); - format = r600_translate_vertex_data_type(rctx->vertex_elements->elements[i].src_format); + format = r600_translate_vertex_data_type(rctx->vertex_elements->hw_format[i]); word2 = format | S_030008_STRIDE(vertex_buffer->stride); - word3 = r600_translate_vertex_data_swizzle(rctx->vertex_elements->elements[i].src_format); + word3 = r600_translate_vertex_data_swizzle(rctx->vertex_elements->hw_format[i]); r600_pipe_state_add_reg(rstate, R_030000_RESOURCE0_WORD0, offset, 0xFFFFFFFF, rbuffer->bo); r600_pipe_state_add_reg(rstate, R_030004_RESOURCE0_WORD1, rbuffer->size - offset - 1, 0xFFFFFFFF, NULL); @@ -1514,6 +1343,9 @@ void evergreen_draw(struct pipe_context *ctx, const struct pipe_draw_info *info) } evergreen_context_draw(&rctx->ctx, &rdraw); + if (translate) + r600_end_vertex_translate(rctx); + pipe_resource_reference(&draw.index_buffer, NULL); } @@ -1522,23 +1354,39 @@ void evergreen_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shader struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; struct r600_pipe_state *rstate = &shader->rstate; struct r600_shader *rshader = &shader->shader; - unsigned i, tmp, exports_ps, num_cout, spi_ps_in_control_0, spi_input_z; - boolean have_pos = FALSE, have_face = FALSE; + unsigned i, tmp, exports_ps, num_cout, spi_ps_in_control_0, spi_input_z, spi_ps_in_control_1; + int pos_index = -1, face_index = -1; + int ninterp = 0; + boolean have_linear = FALSE, have_centroid = FALSE, have_perspective = FALSE; + unsigned spi_baryc_cntl; /* clear previous register */ rstate->nregs = 0; for (i = 0; i < rshader->ninput; i++) { tmp = S_028644_SEMANTIC(r600_find_vs_semantic_index(&rctx->vs_shader->shader, rshader, i)); + /* evergreen NUM_INTERP only contains values interpolated into the LDS, + POSITION goes via GPRs from the SC so isn't counted */ if (rshader->input[i].name == TGSI_SEMANTIC_POSITION) - have_pos = TRUE; + pos_index = i; + else if (rshader->input[i].name == TGSI_SEMANTIC_FACE) + face_index = i; + else { + if (rshader->input[i].interpolate == TGSI_INTERPOLATE_LINEAR || + rshader->input[i].interpolate == TGSI_INTERPOLATE_PERSPECTIVE) + ninterp++; + if (rshader->input[i].interpolate == TGSI_INTERPOLATE_LINEAR) + have_linear = TRUE; + if (rshader->input[i].interpolate == TGSI_INTERPOLATE_PERSPECTIVE) + have_perspective = TRUE; + if (rshader->input[i].centroid) + have_centroid = TRUE; + } if (rshader->input[i].name == TGSI_SEMANTIC_COLOR || rshader->input[i].name == TGSI_SEMANTIC_BCOLOR || rshader->input[i].name == TGSI_SEMANTIC_POSITION) { tmp |= S_028644_FLAT_SHADE(rshader->flat_shade); } - if (rshader->input[i].name == TGSI_SEMANTIC_FACE) - have_face = TRUE; if (rshader->input[i].name == TGSI_SEMANTIC_GENERIC && rctx->sprite_coord_enable & (1 << rshader->input[i].sid)) { tmp |= S_028644_PT_SPRITE_TEX(1); @@ -1546,17 +1394,23 @@ void evergreen_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shader r600_pipe_state_add_reg(rstate, R_028644_SPI_PS_INPUT_CNTL_0 + i * 4, tmp, 0xFFFFFFFF, NULL); } for (i = 0; i < rshader->noutput; i++) { - if (rshader->input[i].name == TGSI_SEMANTIC_POSITION) + if (rshader->output[i].name == TGSI_SEMANTIC_POSITION) r600_pipe_state_add_reg(rstate, R_02880C_DB_SHADER_CONTROL, S_02880C_Z_EXPORT_ENABLE(1), S_02880C_Z_EXPORT_ENABLE(1), NULL); + if (rshader->output[i].name == TGSI_SEMANTIC_STENCIL) + r600_pipe_state_add_reg(rstate, + R_02880C_DB_SHADER_CONTROL, + S_02880C_STENCIL_EXPORT_ENABLE(1), + S_02880C_STENCIL_EXPORT_ENABLE(1), NULL); } exports_ps = 0; num_cout = 0; for (i = 0; i < rshader->noutput; i++) { - if (rshader->output[i].name == TGSI_SEMANTIC_POSITION) + if (rshader->output[i].name == TGSI_SEMANTIC_POSITION || + rshader->output[i].name == TGSI_SEMANTIC_STENCIL) exports_ps |= 1; else if (rshader->output[i].name == TGSI_SEMANTIC_COLOR) { num_cout++; @@ -1568,19 +1422,49 @@ void evergreen_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shader exports_ps = 2; } - spi_ps_in_control_0 = S_0286CC_NUM_INTERP(rshader->ninput) | - S_0286CC_PERSP_GRADIENT_ENA(1); + if (ninterp == 0) { + ninterp = 1; + have_perspective = TRUE; + } + + spi_ps_in_control_0 = S_0286CC_NUM_INTERP(ninterp) | + S_0286CC_PERSP_GRADIENT_ENA(have_perspective) | + S_0286CC_LINEAR_GRADIENT_ENA(have_linear); spi_input_z = 0; - if (have_pos) { - spi_ps_in_control_0 |= S_0286CC_POSITION_ENA(1); + if (pos_index != -1) { + spi_ps_in_control_0 |= S_0286CC_POSITION_ENA(1) | + S_0286CC_POSITION_CENTROID(rshader->input[pos_index].centroid) | + S_0286CC_POSITION_ADDR(rshader->input[pos_index].gpr); spi_input_z |= 1; } + + spi_ps_in_control_1 = 0; + if (face_index != -1) { + spi_ps_in_control_1 |= S_0286D0_FRONT_FACE_ENA(1) | + S_0286D0_FRONT_FACE_ADDR(rshader->input[face_index].gpr); + } + + spi_baryc_cntl = 0; + if (have_perspective) + spi_baryc_cntl |= S_0286E0_PERSP_CENTER_ENA(1) | + S_0286E0_PERSP_CENTROID_ENA(have_centroid); + if (have_linear) + spi_baryc_cntl |= S_0286E0_LINEAR_CENTER_ENA(1) | + S_0286E0_LINEAR_CENTROID_ENA(have_centroid); + r600_pipe_state_add_reg(rstate, R_0286CC_SPI_PS_IN_CONTROL_0, spi_ps_in_control_0, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_0286D0_SPI_PS_IN_CONTROL_1, - S_0286D0_FRONT_FACE_ENA(have_face), 0xFFFFFFFF, NULL); + spi_ps_in_control_1, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_0286E4_SPI_PS_IN_CONTROL_2, + 0, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_0286D8_SPI_INPUT_Z, spi_input_z, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, + R_0286E0_SPI_BARYC_CNTL, + spi_baryc_cntl, + 0xFFFFFFFF, NULL); + + r600_pipe_state_add_reg(rstate, R_028840_SQ_PGM_START_PS, (r600_bo_offset(shader->bo)) >> 8, 0xFFFFFFFF, shader->bo); r600_pipe_state_add_reg(rstate, @@ -1595,11 +1479,6 @@ void evergreen_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shader r600_pipe_state_add_reg(rstate, R_02884C_SQ_PGM_EXPORTS_PS, exports_ps, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, - R_0286E0_SPI_BARYC_CNTL, - S_0286E0_PERSP_CENTROID_ENA(1) | - S_0286E0_LINEAR_CENTROID_ENA(1), - 0xFFFFFFFF, NULL); if (rshader->uses_kill) { /* only set some bits here, the other bits are set in the dsa state */ diff --git a/src/gallium/drivers/r600/evergreend.h b/src/gallium/drivers/r600/evergreend.h index 9971dded782..8e96f9355e6 100644 --- a/src/gallium/drivers/r600/evergreend.h +++ b/src/gallium/drivers/r600/evergreend.h @@ -686,6 +686,9 @@ #define S_02880C_Z_EXPORT_ENABLE(x) (((x) & 0x1) << 0) #define G_02880C_Z_EXPORT_ENABLE(x) (((x) >> 0) & 0x1) #define C_02880C_Z_EXPORT_ENABLE 0xFFFFFFFE +#define S_02880C_STENCIL_EXPORT_ENABLE(x) (((x) & 0x1) << 1) +#define G_02880C_STENCIL_EXPORT_ENABLE(x) (((x) >> 1) & 0x1) +#define C_02880C_STENCIL_EXPORT_ENABLE 0xFFFFFFFD #define S_02880C_Z_ORDER(x) (((x) & 0x3) << 4) #define G_02880C_Z_ORDER(x) (((x) >> 4) & 0x3) #define C_02880C_Z_ORDER 0xFFFFFCFF @@ -984,9 +987,6 @@ #define S_030010_ENDIAN_SWAP(x) (((x) & 0x3) << 12) #define G_030010_ENDIAN_SWAP(x) (((x) >> 12) & 0x3) #define C_030010_ENDIAN_SWAP 0xFFFFCFFF -#define S_030010_REQUEST_SIZE(x) (((x) & 0x3) << 14) -#define G_030010_REQUEST_SIZE(x) (((x) >> 14) & 0x3) -#define C_030010_REQUEST_SIZE 0xFFFF3FFF #define S_030010_DST_SEL_X(x) (((x) & 0x7) << 16) #define G_030010_DST_SEL_X(x) (((x) >> 16) & 0x7) #define C_030010_DST_SEL_X 0xFFF8FFFF @@ -1050,45 +1050,6 @@ #define S_030008_DATA_FORMAT(x) (((x) & 0x3F) << 20) #define G_030008_DATA_FORMAT(x) (((x) >> 20) & 0x3F) #define C_030008_DATA_FORMAT 0xFC0FFFFF -#define V_030008_FMT_INVALID 0x00000000 -#define V_030008_FMT_8 0x00000001 -#define V_030008_FMT_4_4 0x00000002 -#define V_030008_FMT_3_3_2 0x00000003 -#define V_030008_FMT_16 0x00000005 -#define V_030008_FMT_16_FLOAT 0x00000006 -#define V_030008_FMT_8_8 0x00000007 -#define V_030008_FMT_5_6_5 0x00000008 -#define V_030008_FMT_6_5_5 0x00000009 -#define V_030008_FMT_1_5_5_5 0x0000000A -#define V_030008_FMT_4_4_4_4 0x0000000B -#define V_030008_FMT_5_5_5_1 0x0000000C -#define V_030008_FMT_32 0x0000000D -#define V_030008_FMT_32_FLOAT 0x0000000E -#define V_030008_FMT_16_16 0x0000000F -#define V_030008_FMT_16_16_FLOAT 0x00000010 -#define V_030008_FMT_8_24 0x00000011 -#define V_030008_FMT_8_24_FLOAT 0x00000012 -#define V_030008_FMT_24_8 0x00000013 -#define V_030008_FMT_24_8_FLOAT 0x00000014 -#define V_030008_FMT_10_11_11 0x00000015 -#define V_030008_FMT_10_11_11_FLOAT 0x00000016 -#define V_030008_FMT_11_11_10 0x00000017 -#define V_030008_FMT_11_11_10_FLOAT 0x00000018 -#define V_030008_FMT_2_10_10_10 0x00000019 -#define V_030008_FMT_8_8_8_8 0x0000001A -#define V_030008_FMT_10_10_10_2 0x0000001B -#define V_030008_FMT_X24_8_32_FLOAT 0x0000001C -#define V_030008_FMT_32_32 0x0000001D -#define V_030008_FMT_32_32_FLOAT 0x0000001E -#define V_030008_FMT_16_16_16_16 0x0000001F -#define V_030008_FMT_16_16_16_16_FLOAT 0x00000020 -#define V_030008_FMT_32_32_32_32 0x00000022 -#define V_030008_FMT_32_32_32_32_FLOAT 0x00000023 -#define V_030008_FMT_8_8_8 0x0000002c -#define V_030008_FMT_16_16_16 0x0000002d -#define V_030008_FMT_16_16_16_FLOAT 0x0000002e -#define V_030008_FMT_32_32_32 0x0000002f -#define V_030008_FMT_32_32_32_FLOAT 0x00000030 #define S_030008_NUM_FORMAT_ALL(x) (((x) & 0x3) << 26) #define G_030008_NUM_FORMAT_ALL(x) (((x) >> 26) & 0x3) #define C_030008_NUM_FORMAT_ALL 0xF3FFFFFF diff --git a/src/gallium/drivers/r600/r600.h b/src/gallium/drivers/r600/r600.h index 24e25cec0db..62d983269f5 100644 --- a/src/gallium/drivers/r600/r600.h +++ b/src/gallium/drivers/r600/r600.h @@ -99,15 +99,22 @@ enum chip_class { EVERGREEN, }; +struct r600_tiling_info { + unsigned num_channels; + unsigned num_banks; + unsigned group_bytes; +}; + enum radeon_family r600_get_family(struct radeon *rw); enum chip_class r600_get_family_class(struct radeon *radeon); +struct r600_tiling_info *r600_get_tiling_info(struct radeon *radeon); /* r600_bo.c */ struct r600_bo; struct r600_bo *r600_bo(struct radeon *radeon, unsigned size, unsigned alignment, unsigned usage); struct r600_bo *r600_bo_handle(struct radeon *radeon, - unsigned handle); + unsigned handle, unsigned *array_mode); void *r600_bo_map(struct radeon *radeon, struct r600_bo *bo, unsigned usage, void *ctx); void r600_bo_unmap(struct radeon *radeon, struct r600_bo *bo); void r600_bo_reference(struct radeon *radeon, struct r600_bo **dst, diff --git a/src/gallium/drivers/r600/r600_asm.c b/src/gallium/drivers/r600/r600_asm.c index d13da0ef638..8a7f3ce575c 100644 --- a/src/gallium/drivers/r600/r600_asm.c +++ b/src/gallium/drivers/r600/r600_asm.c @@ -701,9 +701,14 @@ static int r600_bc_cf_build(struct r600_bc *bc, struct r600_bc_cf *cf) case (V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU << 3): case (V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU_PUSH_BEFORE << 3): bc->bytecode[id++] = S_SQ_CF_ALU_WORD0_ADDR(cf->addr >> 1) | - S_SQ_CF_ALU_WORD0_KCACHE_MODE0(cf->kcache0_mode); + S_SQ_CF_ALU_WORD0_KCACHE_MODE0(cf->kcache0_mode) | + S_SQ_CF_ALU_WORD0_KCACHE_BANK0(cf->kcache0_bank) | + S_SQ_CF_ALU_WORD0_KCACHE_BANK1(cf->kcache1_bank); bc->bytecode[id++] = S_SQ_CF_ALU_WORD1_CF_INST(cf->inst >> 3) | + S_SQ_CF_ALU_WORD1_KCACHE_MODE1(cf->kcache1_mode) | + S_SQ_CF_ALU_WORD1_KCACHE_ADDR0(cf->kcache0_addr) | + S_SQ_CF_ALU_WORD1_KCACHE_ADDR1(cf->kcache1_addr) | S_SQ_CF_ALU_WORD1_BARRIER(1) | S_SQ_CF_ALU_WORD1_USES_WATERFALL(bc->chiprev == 0 ? cf->r6xx_uses_waterfall : 0) | S_SQ_CF_ALU_WORD1_COUNT((cf->ndw / 2) - 1); @@ -871,3 +876,39 @@ int r600_bc_build(struct r600_bc *bc) } return 0; } + +void r600_bc_clear(struct r600_bc *bc) +{ + struct r600_bc_cf *cf = NULL, *next_cf; + + free(bc->bytecode); + bc->bytecode = NULL; + + LIST_FOR_EACH_ENTRY_SAFE(cf, next_cf, &bc->cf, list) { + struct r600_bc_alu *alu = NULL, *next_alu; + struct r600_bc_tex *tex = NULL, *next_tex; + struct r600_bc_tex *vtx = NULL, *next_vtx; + + LIST_FOR_EACH_ENTRY_SAFE(alu, next_alu, &cf->alu, list) { + free(alu); + } + + LIST_INITHEAD(&cf->alu); + + LIST_FOR_EACH_ENTRY_SAFE(tex, next_tex, &cf->tex, list) { + free(tex); + } + + LIST_INITHEAD(&cf->tex); + + LIST_FOR_EACH_ENTRY_SAFE(vtx, next_vtx, &cf->vtx, list) { + free(vtx); + } + + LIST_INITHEAD(&cf->vtx); + + free(cf); + } + + LIST_INITHEAD(&cf->list); +} diff --git a/src/gallium/drivers/r600/r600_asm.h b/src/gallium/drivers/r600/r600_asm.h index bebc7c15b00..25cda16837d 100644 --- a/src/gallium/drivers/r600/r600_asm.h +++ b/src/gallium/drivers/r600/r600_asm.h @@ -132,6 +132,11 @@ struct r600_bc_cf { unsigned pop_count; unsigned cf_addr; /* control flow addr */ unsigned kcache0_mode; + unsigned kcache1_mode; + unsigned kcache0_addr; + unsigned kcache1_addr; + unsigned kcache0_bank; + unsigned kcache1_bank; unsigned r6xx_uses_waterfall; struct list_head alu; struct list_head tex; @@ -185,6 +190,7 @@ int eg_bc_cf_build(struct r600_bc *bc, struct r600_bc_cf *cf); /* r600_asm.c */ int r600_bc_init(struct r600_bc *bc, enum radeon_family family); +void r600_bc_clear(struct r600_bc *bc); int r600_bc_add_alu(struct r600_bc *bc, const struct r600_bc_alu *alu); int r600_bc_add_literal(struct r600_bc *bc, const u32 *value); int r600_bc_add_vtx(struct r600_bc *bc, const struct r600_bc_vtx *vtx); diff --git a/src/gallium/drivers/r600/r600_blit.c b/src/gallium/drivers/r600/r600_blit.c index 4bf44a171af..50d47060c1a 100644 --- a/src/gallium/drivers/r600/r600_blit.c +++ b/src/gallium/drivers/r600/r600_blit.c @@ -22,12 +22,22 @@ */ #include <util/u_surface.h> #include <util/u_blitter.h> +#include <util/u_format.h> #include "r600_pipe.h" -static void r600_blitter_save_states(struct pipe_context *ctx) +enum r600_blitter_op /* bitmask */ +{ + R600_CLEAR = 1, + R600_CLEAR_SURFACE = 2, + R600_COPY = 4 +}; + +static void r600_blitter_begin(struct pipe_context *ctx, enum r600_blitter_op op) { struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; + r600_context_queries_suspend(&rctx->ctx); + util_blitter_save_blend(rctx->blitter, rctx->states[R600_PIPE_STATE_BLEND]); util_blitter_save_depth_stencil_alpha(rctx->blitter, rctx->states[R600_PIPE_STATE_DSA]); if (rctx->states[R600_PIPE_STATE_STENCIL_REF]) { @@ -47,25 +57,34 @@ static void r600_blitter_save_states(struct pipe_context *ctx) rctx->vertex_elements = NULL; - /* TODO queries */ + if (op & (R600_CLEAR_SURFACE | R600_COPY)) + util_blitter_save_framebuffer(rctx->blitter, &rctx->framebuffer); + + if (op & R600_COPY) { + util_blitter_save_fragment_sampler_states( + rctx->blitter, rctx->ps_samplers.n_samplers, + (void**)rctx->ps_samplers.samplers); + + util_blitter_save_fragment_sampler_views( + rctx->blitter, rctx->ps_samplers.n_views, + (struct pipe_sampler_view**)rctx->ps_samplers.views); + } + +} + +static void r600_blitter_end(struct pipe_context *ctx) +{ + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; + r600_context_queries_resume(&rctx->ctx); } int r600_blit_uncompress_depth(struct pipe_context *ctx, struct r600_resource_texture *texture) { struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - struct pipe_framebuffer_state fb = *rctx->pframebuffer; struct pipe_surface *zsurf, *cbsurf; int level = 0; float depth = 1.0f; - r600_context_queries_suspend(&rctx->ctx); - for (int i = 0; i < fb.nr_cbufs; i++) { - fb.cbufs[i] = NULL; - pipe_surface_reference(&fb.cbufs[i], rctx->pframebuffer->cbufs[i]); - } - fb.zsbuf = NULL; - pipe_surface_reference(&fb.zsbuf, rctx->pframebuffer->zsbuf); - zsurf = ctx->screen->get_tex_surface(ctx->screen, &texture->resource.base.b, 0, level, 0, PIPE_BIND_DEPTH_STENCIL); @@ -73,22 +92,17 @@ int r600_blit_uncompress_depth(struct pipe_context *ctx, struct r600_resource_te (struct pipe_resource*)texture->flushed_depth_texture, 0, level, 0, PIPE_BIND_RENDER_TARGET); - r600_blitter_save_states(ctx); - util_blitter_save_framebuffer(rctx->blitter, &fb); - if (rctx->family == CHIP_RV610 || rctx->family == CHIP_RV630 || - rctx->family == CHIP_RV620 || rctx->family == CHIP_RV635) + rctx->family == CHIP_RV620 || rctx->family == CHIP_RV635) depth = 0.0f; + r600_blitter_begin(ctx, R600_CLEAR_SURFACE); util_blitter_custom_depth_stencil(rctx->blitter, zsurf, cbsurf, rctx->custom_dsa_flush, depth); + r600_blitter_end(ctx); pipe_surface_reference(&zsurf, NULL); pipe_surface_reference(&cbsurf, NULL); - for (int i = 0; i < fb.nr_cbufs; i++) { - pipe_surface_reference(&fb.cbufs[i], NULL); - } - pipe_surface_reference(&fb.zsbuf, NULL); - r600_context_queries_resume(&rctx->ctx); + return 0; } @@ -99,12 +113,11 @@ static void r600_clear(struct pipe_context *ctx, unsigned buffers, struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; struct pipe_framebuffer_state *fb = &rctx->framebuffer; - r600_context_queries_suspend(&rctx->ctx); - r600_blitter_save_states(ctx); + r600_blitter_begin(ctx, R600_CLEAR); util_blitter_clear(rctx->blitter, fb->width, fb->height, fb->nr_cbufs, buffers, rgba, depth, stencil); - r600_context_queries_resume(&rctx->ctx); + r600_blitter_end(ctx); } static void r600_clear_render_target(struct pipe_context *ctx, @@ -114,13 +127,11 @@ static void r600_clear_render_target(struct pipe_context *ctx, unsigned width, unsigned height) { struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - struct pipe_framebuffer_state *fb = &rctx->framebuffer; - r600_context_queries_suspend(&rctx->ctx); - util_blitter_save_framebuffer(rctx->blitter, fb); + r600_blitter_begin(ctx, R600_CLEAR_SURFACE); util_blitter_clear_render_target(rctx->blitter, dst, rgba, dstx, dsty, width, height); - r600_context_queries_resume(&rctx->ctx); + r600_blitter_end(ctx); } static void r600_clear_depth_stencil(struct pipe_context *ctx, @@ -132,16 +143,34 @@ static void r600_clear_depth_stencil(struct pipe_context *ctx, unsigned width, unsigned height) { struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - struct pipe_framebuffer_state *fb = &rctx->framebuffer; - r600_context_queries_suspend(&rctx->ctx); - util_blitter_save_framebuffer(rctx->blitter, fb); + r600_blitter_begin(ctx, R600_CLEAR_SURFACE); util_blitter_clear_depth_stencil(rctx->blitter, dst, clear_flags, depth, stencil, dstx, dsty, width, height); - r600_context_queries_resume(&rctx->ctx); + r600_blitter_end(ctx); } + +/* Copy a block of pixels from one surface to another using HW. */ +static void r600_hw_copy_region(struct pipe_context *ctx, + struct pipe_resource *dst, + struct pipe_subresource subdst, + unsigned dstx, unsigned dsty, unsigned dstz, + struct pipe_resource *src, + struct pipe_subresource subsrc, + unsigned srcx, unsigned srcy, unsigned srcz, + unsigned width, unsigned height) +{ + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; + + r600_blitter_begin(ctx, R600_COPY); + util_blitter_copy_region(rctx->blitter, dst, subdst, dstx, dsty, dstz, + src, subsrc, srcx, srcy, srcz, width, height, + TRUE); + r600_blitter_end(ctx); +} + static void r600_resource_copy_region(struct pipe_context *ctx, struct pipe_resource *dst, struct pipe_subresource subdst, @@ -151,8 +180,16 @@ static void r600_resource_copy_region(struct pipe_context *ctx, unsigned srcx, unsigned srcy, unsigned srcz, unsigned width, unsigned height) { - util_resource_copy_region(ctx, dst, subdst, dstx, dsty, dstz, - src, subsrc, srcx, srcy, srcz, width, height); + boolean is_depth; + /* there is something wrong with depth resource copies at the moment so avoid them for now */ + is_depth = util_format_get_component_bits(src->format, UTIL_FORMAT_COLORSPACE_ZS, 0) != 0; + if (is_depth) + util_resource_copy_region(ctx, dst, subdst, dstx, dsty, dstz, + src, subsrc, srcx, srcy, srcz, width, height); + else + r600_hw_copy_region(ctx, dst, subdst, dstx, dsty, dstz, + src, subsrc, srcx, srcy, srcz, width, height); + } void r600_init_blit_functions(struct r600_pipe_context *rctx) diff --git a/src/gallium/drivers/r600/r600_buffer.c b/src/gallium/drivers/r600/r600_buffer.c index 2bfa4e22fec..455aa2e81f6 100644 --- a/src/gallium/drivers/r600/r600_buffer.c +++ b/src/gallium/drivers/r600/r600_buffer.c @@ -227,7 +227,7 @@ struct pipe_resource *r600_buffer_from_handle(struct pipe_screen *screen, struct r600_resource *rbuffer; struct r600_bo *bo = NULL; - bo = r600_bo_handle(rw, whandle->handle); + bo = r600_bo_handle(rw, whandle->handle, NULL); if (bo == NULL) { return NULL; } diff --git a/src/gallium/drivers/r600/r600_formats.h b/src/gallium/drivers/r600/r600_formats.h new file mode 100644 index 00000000000..0c91a212384 --- /dev/null +++ b/src/gallium/drivers/r600/r600_formats.h @@ -0,0 +1,56 @@ +#ifndef R600_FORMATS_H +#define R600_FORMATS_H + +/* list of formats from R700 ISA document - apply across GPUs in different registers */ +#define FMT_INVALID 0x00000000 +#define FMT_8 0x00000001 +#define FMT_4_4 0x00000002 +#define FMT_3_3_2 0x00000003 +#define FMT_16 0x00000005 +#define FMT_16_FLOAT 0x00000006 +#define FMT_8_8 0x00000007 +#define FMT_5_6_5 0x00000008 +#define FMT_6_5_5 0x00000009 +#define FMT_1_5_5_5 0x0000000A +#define FMT_4_4_4_4 0x0000000B +#define FMT_5_5_5_1 0x0000000C +#define FMT_32 0x0000000D +#define FMT_32_FLOAT 0x0000000E +#define FMT_16_16 0x0000000F +#define FMT_16_16_FLOAT 0x00000010 +#define FMT_8_24 0x00000011 +#define FMT_8_24_FLOAT 0x00000012 +#define FMT_24_8 0x00000013 +#define FMT_24_8_FLOAT 0x00000014 +#define FMT_10_11_11 0x00000015 +#define FMT_10_11_11_FLOAT 0x00000016 +#define FMT_11_11_10 0x00000017 +#define FMT_11_11_10_FLOAT 0x00000018 +#define FMT_2_10_10_10 0x00000019 +#define FMT_8_8_8_8 0x0000001A +#define FMT_10_10_10_2 0x0000001B +#define FMT_X24_8_32_FLOAT 0x0000001C +#define FMT_32_32 0x0000001D +#define FMT_32_32_FLOAT 0x0000001E +#define FMT_16_16_16_16 0x0000001F +#define FMT_16_16_16_16_FLOAT 0x00000020 +#define FMT_32_32_32_32 0x00000022 +#define FMT_32_32_32_32_FLOAT 0x00000023 +#define FMT_1 0x00000025 +#define FMT_GB_GR 0x00000027 +#define FMT_BG_RG 0x00000028 +#define FMT_32_AS_8 0x00000029 +#define FMT_32_AS_8_8 0x0000002a +#define FMT_5_9_9_9_SHAREDEXP 0x0000002b +#define FMT_8_8_8 0x0000002c +#define FMT_16_16_16 0x0000002d +#define FMT_16_16_16_FLOAT 0x0000002e +#define FMT_32_32_32 0x0000002f +#define FMT_32_32_32_FLOAT 0x00000030 +#define FMT_BC1 0x00000031 +#define FMT_BC2 0x00000032 +#define FMT_BC3 0x00000033 +#define FMT_BC4 0x00000034 +#define FMT_BC5 0x00000035 + +#endif diff --git a/src/gallium/drivers/r600/r600_opcodes.h b/src/gallium/drivers/r600/r600_opcodes.h index 0cf9c1c401c..4f9b39a7fdc 100644 --- a/src/gallium/drivers/r600/r600_opcodes.h +++ b/src/gallium/drivers/r600/r600_opcodes.h @@ -233,12 +233,6 @@ #define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_CEIL 0x00000012 #define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RNDNE 0x00000013 #define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLOOR 0x00000014 -/* same up to here */ -/* -#define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA 0x00000015 -#define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA_FLOOR 0x00000016 -#define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA_INT 0x00000018 -*/ #define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_ASHR_INT 0x00000015 #define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LSHR_INT 0x00000016 #define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_LSHL_INT 0x00000017 @@ -336,9 +330,11 @@ #define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIPSQRT_CLAMPED_64 0x00000098 #define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_SQRT_64 0x00000099 /* TODO Fill in more ALU */ +#define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLT_TO_INT_FLOOR 0x000000B1 #define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4 0x000000BE #define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4_IEEE 0x000000BF #define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_CUBE 0x000000C0 +#define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA_INT 0x000000CC #define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INTERP_XY 0x000000D6 #define EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INTERP_ZW 0x000000D7 diff --git a/src/gallium/drivers/r600/r600_pipe.c b/src/gallium/drivers/r600/r600_pipe.c index a9fc31d9fbd..87ec8cd2e83 100644 --- a/src/gallium/drivers/r600/r600_pipe.c +++ b/src/gallium/drivers/r600/r600_pipe.c @@ -83,9 +83,16 @@ static void r600_destroy_context(struct pipe_context *context) free(rctx->states[i]); } + util_blitter_destroy(rctx->blitter); + u_upload_destroy(rctx->upload_vb); u_upload_destroy(rctx->upload_ib); + if (rctx->tran.translate_cache) + translate_cache_destroy(rctx->tran.translate_cache); + + FREE(rctx->ps_resource); + FREE(rctx->vs_resource); FREE(rctx); } @@ -172,6 +179,24 @@ static struct pipe_context *r600_create_context(struct pipe_screen *screen, void return NULL; } + rctx->tran.translate_cache = translate_cache_create(); + if (rctx->tran.translate_cache == NULL) { + FREE(rctx); + return NULL; + } + + rctx->vs_resource = CALLOC(R600_RESOURCE_ARRAY_SIZE, sizeof(struct r600_pipe_state)); + if (!rctx->vs_resource) { + FREE(rctx); + return NULL; + } + + rctx->ps_resource = CALLOC(R600_RESOURCE_ARRAY_SIZE, sizeof(struct r600_pipe_state)); + if (!rctx->ps_resource) { + FREE(rctx); + return NULL; + } + class = r600_get_family_class(rctx->radeon); if (class == R600 || class == R700) rctx->custom_dsa_flush = r600_create_db_flush_dsa(rctx); @@ -243,6 +268,7 @@ static int r600_get_param(struct pipe_screen* pscreen, enum pipe_cap param) case PIPE_CAP_INDEP_BLEND_ENABLE: case PIPE_CAP_DEPTHSTENCIL_CLEAR_SEPARATE: case PIPE_CAP_DEPTH_CLAMP: + case PIPE_CAP_SHADER_STENCIL_EXPORT: return 1; /* Unsupported features (boolean caps). */ @@ -258,7 +284,7 @@ static int r600_get_param(struct pipe_screen* pscreen, enum pipe_cap param) return 14; case PIPE_CAP_MAX_VERTEX_TEXTURE_UNITS: /* FIXME allow this once infrastructure is there */ - return 0; + return 16; case PIPE_CAP_MAX_TEXTURE_IMAGE_UNITS: case PIPE_CAP_MAX_COMBINED_SAMPLERS: return 16; @@ -430,5 +456,7 @@ struct pipe_screen *r600_screen_create(struct radeon *radeon) r600_init_screen_texture_functions(&rscreen->screen); r600_init_screen_resource_functions(&rscreen->screen); + rscreen->tiling_info = r600_get_tiling_info(radeon); + return &rscreen->screen; } diff --git a/src/gallium/drivers/r600/r600_pipe.h b/src/gallium/drivers/r600/r600_pipe.h index c46029a5617..1c691f6b764 100644 --- a/src/gallium/drivers/r600/r600_pipe.h +++ b/src/gallium/drivers/r600/r600_pipe.h @@ -30,6 +30,7 @@ #include <pipe/p_screen.h> #include <pipe/p_context.h> #include <util/u_math.h> +#include "translate/translate_cache.h" #include "r600.h" #include "r600_public.h" #include "r600_shader.h" @@ -58,6 +59,7 @@ enum r600_pipe_state_id { struct r600_screen { struct pipe_screen screen; struct radeon *radeon; + struct r600_tiling_info *tiling_info; }; struct r600_pipe_sampler_view { @@ -81,8 +83,10 @@ struct r600_pipe_blend { struct r600_vertex_element { unsigned count; - unsigned refcount; - struct pipe_vertex_element elements[32]; + struct pipe_vertex_element elements[PIPE_MAX_ATTRIBS]; + enum pipe_format hw_format[PIPE_MAX_ATTRIBS]; + unsigned hw_format_size[PIPE_MAX_ATTRIBS]; + boolean incompatible_layout; }; struct r600_pipe_shader { @@ -92,6 +96,29 @@ struct r600_pipe_shader { struct r600_vertex_element vertex_elements; }; +/* needed for blitter save */ +#define NUM_TEX_UNITS 16 + +struct r600_textures_info { + struct r600_pipe_sampler_view *views[NUM_TEX_UNITS]; + unsigned n_views; + void *samplers[NUM_TEX_UNITS]; + unsigned n_samplers; +}; + +struct r600_translate_context { + /* Translate cache for incompatible vertex offset/stride/format fallback. */ + struct translate_cache *translate_cache; + + /* The vertex buffer slot containing the translated buffer. */ + unsigned vb_slot; + /* Saved and new vertex element state. */ + void *saved_velems, *new_velems; +}; + +#define R600_CONSTANT_ARRAY_SIZE 256 +#define R600_RESOURCE_ARRAY_SIZE 160 + struct r600_pipe_context { struct pipe_context context; struct blitter_context *blitter; @@ -112,12 +139,8 @@ struct r600_pipe_context { struct pipe_stencil_ref stencil_ref; struct pipe_viewport_state viewport; struct pipe_clip_state clip; - unsigned vs_nconst; - unsigned ps_nconst; - struct r600_pipe_state vs_const[256]; - struct r600_pipe_state ps_const[256]; - struct r600_pipe_state vs_resource[160]; - struct r600_pipe_state ps_resource[160]; + struct r600_pipe_state *vs_resource; + struct r600_pipe_state *ps_resource; struct r600_pipe_state config; struct r600_pipe_shader *ps_shader; struct r600_pipe_shader *vs_shader; @@ -130,6 +153,11 @@ struct r600_pipe_context { struct u_upload_mgr *upload_vb; struct u_upload_mgr *upload_ib; unsigned any_user_vbs; + struct r600_textures_info ps_samplers; + + unsigned vb_max_index; + struct r600_translate_context tran; + }; struct r600_drawl { @@ -180,6 +208,7 @@ void r600_init_context_resource_functions(struct r600_pipe_context *r600); /* r600_shader.c */ int r600_pipe_shader_update(struct pipe_context *ctx, struct r600_pipe_shader *shader); int r600_pipe_shader_create(struct pipe_context *ctx, struct r600_pipe_shader *shader, const struct tgsi_token *tokens); +void r600_pipe_shader_destroy(struct pipe_context *ctx, struct r600_pipe_shader *shader); int r600_find_vs_semantic_index(struct r600_shader *vs, struct r600_shader *ps, int id); @@ -187,10 +216,6 @@ int r600_find_vs_semantic_index(struct r600_shader *vs, void r600_init_state_functions(struct r600_pipe_context *rctx); void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info); void r600_init_config(struct r600_pipe_context *rctx); -void r600_translate_index_buffer(struct r600_pipe_context *r600, - struct pipe_resource **index_buffer, - unsigned *index_size, - unsigned *start, unsigned count); void *r600_create_db_flush_dsa(struct r600_pipe_context *rctx); /* r600_helper.h */ int r600_conv_pipe_prim(unsigned pprim, unsigned *prim); @@ -201,6 +226,38 @@ uint32_t r600_translate_texformat(enum pipe_format format, const unsigned char *swizzle_view, uint32_t *word4_p, uint32_t *yuv_format_p); +/* r600_translate.c */ +void r600_begin_vertex_translate(struct r600_pipe_context *rctx); +void r600_end_vertex_translate(struct r600_pipe_context *rctx); +void r600_translate_index_buffer(struct r600_pipe_context *r600, + struct pipe_resource **index_buffer, + unsigned *index_size, + unsigned *start, unsigned count); + +/* r600_state_common.c */ +void r600_set_index_buffer(struct pipe_context *ctx, + const struct pipe_index_buffer *ib); +void r600_set_vertex_buffers(struct pipe_context *ctx, unsigned count, + const struct pipe_vertex_buffer *buffers); +void *r600_create_vertex_elements(struct pipe_context *ctx, + unsigned count, + const struct pipe_vertex_element *elements); +void r600_delete_vertex_element(struct pipe_context *ctx, void *state); +void r600_bind_blend_state(struct pipe_context *ctx, void *state); +void r600_bind_rs_state(struct pipe_context *ctx, void *state); +void r600_delete_rs_state(struct pipe_context *ctx, void *state); +void r600_sampler_view_destroy(struct pipe_context *ctx, + struct pipe_sampler_view *state); +void r600_bind_state(struct pipe_context *ctx, void *state); +void r600_delete_state(struct pipe_context *ctx, void *state); +void r600_bind_vertex_elements(struct pipe_context *ctx, void *state); + +void *r600_create_shader_state(struct pipe_context *ctx, + const struct pipe_shader_state *state); +void r600_bind_ps_shader(struct pipe_context *ctx, void *state); +void r600_bind_vs_shader(struct pipe_context *ctx, void *state); +void r600_delete_ps_shader(struct pipe_context *ctx, void *state); +void r600_delete_vs_shader(struct pipe_context *ctx, void *state); /* * common helpers */ diff --git a/src/gallium/drivers/r600/r600_resource.h b/src/gallium/drivers/r600/r600_resource.h index 323960960d6..d152285815c 100644 --- a/src/gallium/drivers/r600/r600_resource.h +++ b/src/gallium/drivers/r600/r600_resource.h @@ -25,6 +25,9 @@ #include "util/u_transfer.h" +/* flag to indicate a resource is to be used as a transfer so should not be tiled */ +#define R600_RESOURCE_FLAG_TRANSFER PIPE_RESOURCE_FLAG_DRV_PRIV + /* Texture transfer. */ struct r600_transfer { /* Base class. */ @@ -49,14 +52,14 @@ struct r600_resource { struct r600_resource_texture { struct r600_resource resource; - unsigned long offset[PIPE_MAX_TEXTURE_LEVELS]; - unsigned long pitch[PIPE_MAX_TEXTURE_LEVELS]; - unsigned long layer_size[PIPE_MAX_TEXTURE_LEVELS]; - unsigned long pitch_override; - unsigned long bpt; - unsigned long size; + unsigned offset[PIPE_MAX_TEXTURE_LEVELS]; + unsigned pitch_in_bytes[PIPE_MAX_TEXTURE_LEVELS]; + unsigned pitch_in_pixels[PIPE_MAX_TEXTURE_LEVELS]; + unsigned layer_size[PIPE_MAX_TEXTURE_LEVELS]; + unsigned array_mode[PIPE_MAX_TEXTURE_LEVELS]; + unsigned pitch_override; + unsigned size; unsigned tiled; - unsigned array_mode; unsigned tile_type; unsigned depth; unsigned dirty; @@ -124,4 +127,9 @@ void* r600_texture_transfer_map(struct pipe_context *ctx, void r600_texture_transfer_unmap(struct pipe_context *ctx, struct pipe_transfer* transfer); +struct r600_surface { + struct pipe_surface base; + unsigned aligned_height; +}; + #endif diff --git a/src/gallium/drivers/r600/r600_shader.c b/src/gallium/drivers/r600/r600_shader.c index 366d5d9c351..4106587398b 100644 --- a/src/gallium/drivers/r600/r600_shader.c +++ b/src/gallium/drivers/r600/r600_shader.c @@ -107,24 +107,28 @@ static void r600_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shade struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; struct r600_pipe_state *rstate = &shader->rstate; struct r600_shader *rshader = &shader->shader; - unsigned i, tmp, exports_ps, num_cout, spi_ps_in_control_0, spi_input_z; - boolean have_pos = FALSE, have_face = FALSE; + unsigned i, tmp, exports_ps, num_cout, spi_ps_in_control_0, spi_input_z, spi_ps_in_control_1; + int pos_index = -1, face_index = -1; /* clear previous register */ rstate->nregs = 0; for (i = 0; i < rshader->ninput; i++) { tmp = S_028644_SEMANTIC(r600_find_vs_semantic_index(&rctx->vs_shader->shader, rshader, i)); - tmp |= S_028644_SEL_CENTROID(1); + if (rshader->input[i].centroid) + tmp |= S_028644_SEL_CENTROID(1); + if (rshader->input[i].interpolate == TGSI_INTERPOLATE_LINEAR) + tmp |= S_028644_SEL_LINEAR(1); + if (rshader->input[i].name == TGSI_SEMANTIC_POSITION) - have_pos = TRUE; + pos_index = i; if (rshader->input[i].name == TGSI_SEMANTIC_COLOR || rshader->input[i].name == TGSI_SEMANTIC_BCOLOR || rshader->input[i].name == TGSI_SEMANTIC_POSITION) { tmp |= S_028644_FLAT_SHADE(rshader->flat_shade); } if (rshader->input[i].name == TGSI_SEMANTIC_FACE) - have_face = TRUE; + face_index = i; if (rshader->input[i].name == TGSI_SEMANTIC_GENERIC && rctx->sprite_coord_enable & (1 << rshader->input[i].sid)) { tmp |= S_028644_PT_SPRITE_TEX(1); @@ -137,12 +141,17 @@ static void r600_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shade R_02880C_DB_SHADER_CONTROL, S_02880C_Z_EXPORT_ENABLE(1), S_02880C_Z_EXPORT_ENABLE(1), NULL); + if (rshader->output[i].name == TGSI_SEMANTIC_STENCIL) + r600_pipe_state_add_reg(rstate, + R_02880C_DB_SHADER_CONTROL, + S_02880C_STENCIL_REF_EXPORT_ENABLE(1), + S_02880C_STENCIL_REF_EXPORT_ENABLE(1), NULL); } exports_ps = 0; num_cout = 0; for (i = 0; i < rshader->noutput; i++) { - if (rshader->output[i].name == TGSI_SEMANTIC_POSITION) + if (rshader->output[i].name == TGSI_SEMANTIC_POSITION || rshader->output[i].name == TGSI_SEMANTIC_STENCIL) exports_ps |= 1; else if (rshader->output[i].name == TGSI_SEMANTIC_COLOR) { num_cout++; @@ -157,13 +166,22 @@ static void r600_pipe_shader_ps(struct pipe_context *ctx, struct r600_pipe_shade spi_ps_in_control_0 = S_0286CC_NUM_INTERP(rshader->ninput) | S_0286CC_PERSP_GRADIENT_ENA(1); spi_input_z = 0; - if (have_pos) { - spi_ps_in_control_0 |= S_0286CC_POSITION_ENA(1) | - S_0286CC_BARYC_SAMPLE_CNTL(1); + if (pos_index != -1) { + spi_ps_in_control_0 |= (S_0286CC_POSITION_ENA(1) | + S_0286CC_POSITION_CENTROID(rshader->input[pos_index].centroid) | + S_0286CC_POSITION_ADDR(rshader->input[pos_index].gpr) | + S_0286CC_BARYC_SAMPLE_CNTL(1)); spi_input_z |= 1; } + + spi_ps_in_control_1 = 0; + if (face_index != -1) { + spi_ps_in_control_1 |= S_0286D0_FRONT_FACE_ENA(1) | + S_0286D0_FRONT_FACE_ADDR(rshader->input[face_index].gpr); + } + r600_pipe_state_add_reg(rstate, R_0286CC_SPI_PS_IN_CONTROL_0, spi_ps_in_control_0, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_0286D0_SPI_PS_IN_CONTROL_1, S_0286D0_FRONT_FACE_ENA(have_face), 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_0286D0_SPI_PS_IN_CONTROL_1, spi_ps_in_control_1, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_0286D8_SPI_INPUT_Z, spi_input_z, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_028840_SQ_PGM_START_PS, @@ -252,7 +270,7 @@ static int r600_shader_update(struct pipe_context *ctx, struct r600_pipe_shader } rshader->vertex_elements = *rctx->vertex_elements; for (i = 0; i < rctx->vertex_elements->count; i++) { - resource_format[nresources++] = rctx->vertex_elements->elements[i].src_format; + resource_format[nresources++] = rctx->vertex_elements->hw_format[i]; } r600_bo_reference(rctx->radeon, &rshader->bo, NULL); LIST_FOR_EACH_ENTRY(cf, &bc->cf, list) { @@ -320,6 +338,18 @@ int r600_pipe_shader_create(struct pipe_context *ctx, struct r600_pipe_shader *s return 0; } +void +r600_pipe_shader_destroy(struct pipe_context *ctx, struct r600_pipe_shader *shader) +{ + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; + + r600_bo_reference(rctx->radeon, &shader->bo, NULL); + + r600_bc_clear(&shader->shader.bc); + + /* FIXME: is there more stuff to free? */ +} + /* * tgsi -> r600 shader */ @@ -339,6 +369,11 @@ struct r600_shader_ctx { u32 *literals; u32 nliterals; u32 max_driver_temp_used; + /* needed for evergreen interpolation */ + boolean input_centroid; + boolean input_linear; + boolean input_perspective; + int num_interp_gpr; }; struct r600_shader_tgsi_instruction { @@ -371,11 +406,9 @@ static int tgsi_is_supported(struct r600_shader_ctx *ctx) } #endif for (j = 0; j < i->Instruction.NumSrcRegs; j++) { - if (i->Src[j].Register.Dimension || - i->Src[j].Register.Absolute) { - R600_ERR("unsupported src %d (dimension %d|absolute %d)\n", j, - i->Src[j].Register.Dimension, - i->Src[j].Register.Absolute); + if (i->Src[j].Register.Dimension) { + R600_ERR("unsupported src %d (dimension %d)\n", j, + i->Src[j].Register.Dimension); return -EINVAL; } } @@ -388,10 +421,33 @@ static int tgsi_is_supported(struct r600_shader_ctx *ctx) return 0; } -static int evergreen_interp_alu(struct r600_shader_ctx *ctx, int gpr) +static int evergreen_interp_alu(struct r600_shader_ctx *ctx, int input) { int i, r; struct r600_bc_alu alu; + int gpr = 0, base_chan = 0; + int ij_index = 0; + + if (ctx->shader->input[input].interpolate == TGSI_INTERPOLATE_PERSPECTIVE) { + ij_index = 0; + if (ctx->shader->input[input].centroid) + ij_index++; + } else if (ctx->shader->input[input].interpolate == TGSI_INTERPOLATE_LINEAR) { + ij_index = 0; + /* if we have perspective add one */ + if (ctx->input_perspective) { + ij_index++; + /* if we have perspective centroid */ + if (ctx->input_centroid) + ij_index++; + } + if (ctx->shader->input[input].centroid) + ij_index++; + } + + /* work out gpr and base_chan from index */ + gpr = ij_index / 2; + base_chan = (2 * (ij_index % 2)) + 1; for (i = 0; i < 8; i++) { memset(&alu, 0, sizeof(struct r600_bc_alu)); @@ -402,13 +458,16 @@ static int evergreen_interp_alu(struct r600_shader_ctx *ctx, int gpr) alu.inst = EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INTERP_XY; if ((i > 1) && (i < 6)) { - alu.dst.sel = ctx->shader->input[gpr].gpr; + alu.dst.sel = ctx->shader->input[input].gpr; alu.dst.write = 1; } alu.dst.chan = i % 4; - alu.src[0].chan = (1 - (i % 2)); - alu.src[1].sel = V_SQ_ALU_SRC_PARAM_BASE + gpr; + + alu.src[0].sel = gpr; + alu.src[0].chan = (base_chan - (i % 2)); + + alu.src[1].sel = V_SQ_ALU_SRC_PARAM_BASE + ctx->shader->input[input].lds_pos; alu.bank_swizzle_force = SQ_ALU_VEC_210; if ((i % 4) == 3) @@ -434,6 +493,7 @@ static int tgsi_declaration(struct r600_shader_ctx *ctx) ctx->shader->input[i].name = d->Semantic.Name; ctx->shader->input[i].sid = d->Semantic.Index; ctx->shader->input[i].interpolate = d->Declaration.Interpolate; + ctx->shader->input[i].centroid = d->Declaration.Centroid; ctx->shader->input[i].gpr = ctx->file_offset[TGSI_FILE_INPUT] + i; if (ctx->type == TGSI_PROCESSOR_VERTEX) { /* turn input into fetch */ @@ -457,7 +517,12 @@ static int tgsi_declaration(struct r600_shader_ctx *ctx) } if (ctx->type == TGSI_PROCESSOR_FRAGMENT && ctx->bc->chiprev == 2) { /* turn input into interpolate on EG */ - evergreen_interp_alu(ctx, i); + if (ctx->shader->input[i].name != TGSI_SEMANTIC_POSITION) { + if (ctx->shader->input[i].interpolate > 0) { + ctx->shader->input[i].lds_pos = ctx->shader->nlds++; + evergreen_interp_alu(ctx, i); + } + } } break; case TGSI_FILE_OUTPUT: @@ -484,6 +549,53 @@ static int r600_get_temp(struct r600_shader_ctx *ctx) return ctx->temp_reg + ctx->max_driver_temp_used++; } +/* + * for evergreen we need to scan the shader to find the number of GPRs we need to + * reserve for interpolation. + * + * we need to know if we are going to emit + * any centroid inputs + * if perspective and linear are required +*/ +static int evergreen_gpr_count(struct r600_shader_ctx *ctx) +{ + int i; + int num_baryc; + + ctx->input_linear = FALSE; + ctx->input_perspective = FALSE; + ctx->input_centroid = FALSE; + ctx->num_interp_gpr = 1; + + /* any centroid inputs */ + for (i = 0; i < ctx->info.num_inputs; i++) { + /* skip position/face */ + if (ctx->info.input_semantic_name[i] == TGSI_SEMANTIC_POSITION || + ctx->info.input_semantic_name[i] == TGSI_SEMANTIC_FACE) + continue; + if (ctx->info.input_interpolate[i] == TGSI_INTERPOLATE_LINEAR) + ctx->input_linear = TRUE; + if (ctx->info.input_interpolate[i] == TGSI_INTERPOLATE_PERSPECTIVE) + ctx->input_perspective = TRUE; + if (ctx->info.input_centroid[i]) + ctx->input_centroid = TRUE; + } + + num_baryc = 0; + /* ignoring sample for now */ + if (ctx->input_perspective) + num_baryc++; + if (ctx->input_linear) + num_baryc++; + if (ctx->input_centroid) + num_baryc *= 2; + + ctx->num_interp_gpr += (num_baryc + 1) >> 1; + + /* TODO PULL MODEL and LINE STIPPLE, FIXED PT POS */ + return ctx->num_interp_gpr; +} + int r600_shader_from_tgsi(const struct tgsi_token *tokens, struct r600_shader *shader) { struct tgsi_full_immediate *immediate; @@ -530,7 +642,7 @@ int r600_shader_from_tgsi(const struct tgsi_token *tokens, struct r600_shader *s ctx.file_offset[TGSI_FILE_INPUT] = 1; } if (ctx.type == TGSI_PROCESSOR_FRAGMENT && ctx.bc->chiprev == 2) { - ctx.file_offset[TGSI_FILE_INPUT] = 1; + ctx.file_offset[TGSI_FILE_INPUT] = evergreen_gpr_count(&ctx); } ctx.file_offset[TGSI_FILE_OUTPUT] = ctx.file_offset[TGSI_FILE_INPUT] + ctx.info.file_count[TGSI_FILE_INPUT]; @@ -628,7 +740,14 @@ int r600_shader_from_tgsi(const struct tgsi_token *tokens, struct r600_shader *s } else if (shader->output[i].name == TGSI_SEMANTIC_POSITION) { output[i].array_base = 61; output[i].swizzle_x = 2; - output[i].swizzle_y = output[i].swizzle_z = output[i].swizzle_w = 7; + output[i].swizzle_y = 7; + output[i].swizzle_z = output[i].swizzle_w = 7; + output[i].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL; + } else if (shader->output[i].name == TGSI_SEMANTIC_STENCIL) { + output[i].array_base = 61; + output[i].swizzle_x = 7; + output[i].swizzle_y = 1; + output[i].swizzle_z = output[i].swizzle_w = 7; output[i].type = V_SQ_CF_ALLOC_EXPORT_WORD0_SQ_EXPORT_PIXEL; } else { R600_ERR("unsupported fragment output name %d\n", shader->output[i].name); @@ -734,6 +853,7 @@ static int tgsi_src(struct r600_shader_ctx *ctx, if (tgsi_src->Register.Indirect) r600_src->rel = V_SQ_REL_RELATIVE; r600_src->neg = tgsi_src->Register.Negate; + r600_src->abs = tgsi_src->Register.Absolute; r600_src->sel += ctx->file_offset[tgsi_src->Register.File]; return 0; } @@ -796,6 +916,7 @@ static int tgsi_split_constant(struct r600_shader_ctx *ctx, struct r600_bc_alu_s alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV); alu.src[0].sel = r600_src[i].sel; alu.src[0].chan = k; + alu.src[0].rel = r600_src[i].rel; alu.dst.sel = treg; alu.dst.chan = k; alu.dst.write = 1; @@ -806,6 +927,7 @@ static int tgsi_split_constant(struct r600_shader_ctx *ctx, struct r600_bc_alu_s return r; } r600_src[i].sel = treg; + r600_src[i].rel =0; j--; } } @@ -1846,7 +1968,7 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) memset(&alu, 0, sizeof(struct r600_bc_alu)); alu.inst = CTX_INST(V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV); alu.src[0].sel = src_gpr; - alu.src[0].chan = i; + alu.src[0].chan = tgsi_chan(&inst->Src[0], i); alu.dst.sel = ctx->temp_reg; alu.dst.chan = i; if (i == 3) @@ -1866,8 +1988,10 @@ static int tgsi_tex(struct r600_shader_ctx *ctx) memset(&tex, 0, sizeof(struct r600_bc_tex)); tex.inst = opcode; - tex.resource_id = ctx->file_offset[inst->Src[1].Register.File] + inst->Src[1].Register.Index; - tex.sampler_id = tex.resource_id; + tex.sampler_id = ctx->file_offset[inst->Src[1].Register.File] + inst->Src[1].Register.Index; + tex.resource_id = tex.sampler_id; + if (ctx->shader->processor_type == TGSI_PROCESSOR_VERTEX) + tex.resource_id += PIPE_MAX_ATTRIBS; tex.src_gpr = src_gpr; tex.dst_gpr = ctx->file_offset[inst->Dst[0].Register.File] + inst->Dst[0].Register.Index; tex.dst_sel_x = (inst->Dst[0].Register.WriteMask & 1) ? 0 : 7; @@ -2175,7 +2299,7 @@ static int tgsi_xpd(struct r600_shader_ctx *ctx) static int tgsi_exp(struct r600_shader_ctx *ctx) { struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; - struct r600_bc_alu_src r600_src[3]; + struct r600_bc_alu_src r600_src[3] = { { 0 } }; struct r600_bc_alu alu; int r; @@ -2498,7 +2622,40 @@ static int tgsi_log(struct r600_shader_ctx *ctx) } /* r6/7 only for now */ -static int tgsi_arl(struct r600_shader_ctx *ctx) +static int tgsi_eg_arl(struct r600_shader_ctx *ctx) +{ + struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; + struct r600_bc_alu alu; + int r; + + memset(&alu, 0, sizeof(struct r600_bc_alu)); + + alu.inst = EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_FLT_TO_INT_FLOOR; + r = tgsi_src(ctx, &inst->Src[0], &alu.src[0]); + if (r) + return r; + alu.src[0].chan = tgsi_chan(&inst->Src[0], 0); + alu.last = 1; + alu.dst.chan = 0; + alu.dst.sel = ctx->temp_reg; + alu.dst.write = 1; + r = r600_bc_add_alu_type(ctx->bc, &alu, CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU)); + if (r) + return r; + memset(&alu, 0, sizeof(struct r600_bc_alu)); + alu.inst = EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOVA_INT; + r = tgsi_src(ctx, &inst->Src[0], &alu.src[0]); + if (r) + return r; + alu.src[0].sel = ctx->temp_reg; + alu.src[0].chan = 0; + alu.last = 1; + r = r600_bc_add_alu_type(ctx->bc, &alu, CTX_INST(V_SQ_CF_ALU_WORD1_SQ_CF_INST_ALU)); + if (r) + return r; + return 0; +} +static int tgsi_r600_arl(struct r600_shader_ctx *ctx) { /* TODO from r600c, ar values don't persist between clauses */ struct tgsi_full_instruction *inst = &ctx->parse.FullToken.FullInstruction; @@ -2843,7 +3000,7 @@ static int tgsi_loop_brk_cont(struct r600_shader_ctx *ctx) } static struct r600_shader_tgsi_instruction r600_shader_tgsi_instruction[] = { - {TGSI_OPCODE_ARL, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_arl}, + {TGSI_OPCODE_ARL, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_r600_arl}, {TGSI_OPCODE_MOV, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV, tgsi_op2}, {TGSI_OPCODE_LIT, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_lit}, @@ -2924,7 +3081,7 @@ static struct r600_shader_tgsi_instruction r600_shader_tgsi_instruction[] = { {TGSI_OPCODE_NRM, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, {TGSI_OPCODE_DIV, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, {TGSI_OPCODE_DP2, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4, tgsi_dp}, - {TGSI_OPCODE_TXL, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, + {TGSI_OPCODE_TXL, 0, SQ_TEX_INST_SAMPLE_L, tgsi_tex}, {TGSI_OPCODE_BRK, 0, V_SQ_CF_WORD1_SQ_CF_INST_LOOP_BREAK, tgsi_loop_brk_cont}, {TGSI_OPCODE_IF, 0, V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_if}, /* gap */ @@ -3007,7 +3164,7 @@ static struct r600_shader_tgsi_instruction r600_shader_tgsi_instruction[] = { }; static struct r600_shader_tgsi_instruction eg_shader_tgsi_instruction[] = { - {TGSI_OPCODE_ARL, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, + {TGSI_OPCODE_ARL, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_eg_arl}, {TGSI_OPCODE_MOV, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_MOV, tgsi_op2}, {TGSI_OPCODE_LIT, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_lit}, {TGSI_OPCODE_RCP, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_RECIP_IEEE, tgsi_trans_srcx_replicate}, @@ -3082,7 +3239,7 @@ static struct r600_shader_tgsi_instruction eg_shader_tgsi_instruction[] = { {TGSI_OPCODE_NRM, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, {TGSI_OPCODE_DIV, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, {TGSI_OPCODE_DP2, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_DOT4, tgsi_dp}, - {TGSI_OPCODE_TXL, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_unsupported}, + {TGSI_OPCODE_TXL, 0, SQ_TEX_INST_SAMPLE_L, tgsi_tex}, {TGSI_OPCODE_BRK, 0, EG_V_SQ_CF_WORD1_SQ_CF_INST_LOOP_BREAK, tgsi_loop_brk_cont}, {TGSI_OPCODE_IF, 0, EG_V_SQ_ALU_WORD1_OP2_SQ_OP2_INST_NOP, tgsi_if}, /* gap */ diff --git a/src/gallium/drivers/r600/r600_shader.h b/src/gallium/drivers/r600/r600_shader.h index 6e2620f2012..f8bc5951395 100644 --- a/src/gallium/drivers/r600/r600_shader.h +++ b/src/gallium/drivers/r600/r600_shader.h @@ -31,6 +31,8 @@ struct r600_shader_io { unsigned done; int sid; unsigned interpolate; + boolean centroid; + unsigned lds_pos; /* for evergreen */ }; struct r600_shader { @@ -39,6 +41,7 @@ struct r600_shader { boolean flat_shade; unsigned ninput; unsigned noutput; + unsigned nlds; struct r600_shader_io input[32]; struct r600_shader_io output[32]; enum radeon_family family; diff --git a/src/gallium/drivers/r600/r600_state.c b/src/gallium/drivers/r600/r600_state.c index 7ceedf6363b..b3e0d493217 100644 --- a/src/gallium/drivers/r600/r600_state.c +++ b/src/gallium/drivers/r600/r600_state.c @@ -37,7 +37,7 @@ #include <util/u_memory.h> #include <util/u_inlines.h> #include <util/u_upload_mgr.h> -#include <util/u_index_modify.h> +#include <util/u_framebuffer.h> #include <pipebuffer/pb_buffer.h> #include "r600.h" #include "r600d.h" @@ -98,7 +98,7 @@ static void r600_draw_common(struct r600_drawl *draw) vertex_buffer->buffer_offset + r600_bo_offset(rbuffer->bo); - format = r600_translate_vertex_data_type(rctx->vertex_elements->elements[i].src_format); + format = r600_translate_vertex_data_type(rctx->vertex_elements->hw_format[i]); word2 = format | S_038008_STRIDE(vertex_buffer->stride); @@ -181,40 +181,21 @@ static void r600_draw_common(struct r600_drawl *draw) r600_context_draw(&rctx->ctx, &rdraw); } -void r600_translate_index_buffer(struct r600_pipe_context *r600, - struct pipe_resource **index_buffer, - unsigned *index_size, - unsigned *start, unsigned count) -{ - switch (*index_size) { - case 1: - util_shorten_ubyte_elts(&r600->context, index_buffer, 0, *start, count); - *index_size = 2; - *start = 0; - break; - - case 2: - if (*start % 2 != 0) { - util_rebuild_ushort_elts(&r600->context, index_buffer, 0, *start, count); - *start = 0; - } - break; - - case 4: - break; - } -} - void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info) { struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; struct r600_drawl draw; + boolean translate = FALSE; + + if (rctx->vertex_elements->incompatible_layout) { + r600_begin_vertex_translate(rctx); + translate = TRUE; + } if (rctx->any_user_vbs) { r600_upload_user_buffers(rctx); rctx->any_user_vbs = FALSE; } - memset(&draw, 0, sizeof(struct r600_drawl)); draw.ctx = ctx; draw.mode = info->mode; @@ -245,6 +226,9 @@ void r600_draw_vbo(struct pipe_context *ctx, const struct pipe_draw_info *info) } r600_draw_common(&draw); + if (translate) + r600_end_vertex_translate(rctx); + pipe_resource_reference(&draw.index_buffer, NULL); } @@ -340,20 +324,6 @@ static void *r600_create_blend_state(struct pipe_context *ctx, return rstate; } -static void r600_bind_blend_state(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - struct r600_pipe_blend *blend = (struct r600_pipe_blend *)state; - struct r600_pipe_state *rstate; - - if (state == NULL) - return; - rstate = &blend->rstate; - rctx->states[rstate->id] = rstate; - rctx->cb_target_mask = blend->cb_target_mask; - r600_context_pipe_state_set(&rctx->ctx, rstate); -} - static void *r600_create_dsa_state(struct pipe_context *ctx, const struct pipe_depth_stencil_alpha_state *state) { @@ -446,6 +416,7 @@ static void *r600_create_rs_state(struct pipe_context *ctx, struct r600_pipe_state *rstate; unsigned tmp; unsigned prov_vtx = 1, polygon_dual_mode; + unsigned clip_rule; if (rs == NULL) { return NULL; @@ -455,6 +426,7 @@ static void *r600_create_rs_state(struct pipe_context *ctx, rs->flatshade = state->flatshade; rs->sprite_coord_enable = state->sprite_coord_enable; + clip_rule = state->scissor ? 0xAAAA : 0xFFFF; /* offset */ rs->offset_units = state->offset_units; rs->offset_scale = state->offset_scale * 12.0f; @@ -462,7 +434,7 @@ static void *r600_create_rs_state(struct pipe_context *ctx, rstate->id = R600_PIPE_STATE_RASTERIZER; if (state->flatshade_first) prov_vtx = 0; - tmp = 0x00000001; + tmp = S_0286D4_FLAT_SHADE_ENA(1); if (state->sprite_coord_enable) { tmp |= S_0286D4_PNT_SPRITE_ENA(1) | S_0286D4_PNT_SPRITE_OVRD_X(2) | @@ -496,7 +468,10 @@ static void *r600_create_rs_state(struct pipe_context *ctx, tmp = (unsigned)(state->point_size * 8.0); r600_pipe_state_add_reg(rstate, R_028A00_PA_SU_POINT_SIZE, S_028A00_HEIGHT(tmp) | S_028A00_WIDTH(tmp), 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_028A04_PA_SU_POINT_MINMAX, 0x80000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A08_PA_SU_LINE_CNTL, 0x00000008, 0xFFFFFFFF, NULL); + + tmp = (unsigned)(state->line_width * 8.0); + r600_pipe_state_add_reg(rstate, R_028A08_PA_SU_LINE_CNTL, S_028A08_WIDTH(tmp), 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028A0C_PA_SC_LINE_STIPPLE, 0x00000005, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_028A48_PA_SC_MPASS_PS_CNTL, 0x00000000, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_028C00_PA_SC_LINE_CNTL, 0x00000400, 0xFFFFFFFF, NULL); @@ -505,37 +480,9 @@ static void *r600_create_rs_state(struct pipe_context *ctx, r600_pipe_state_add_reg(rstate, R_028C14_PA_CL_GB_HORZ_CLIP_ADJ, 0x3F800000, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_028C18_PA_CL_GB_HORZ_DISC_ADJ, 0x3F800000, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_028DFC_PA_SU_POLY_OFFSET_CLAMP, 0x00000000, 0xFFFFFFFF, NULL); - return rstate; -} - -static void r600_bind_rs_state(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_rasterizer *rs = (struct r600_pipe_rasterizer *)state; - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - - if (state == NULL) - return; - - rctx->flatshade = rs->flatshade; - rctx->sprite_coord_enable = rs->sprite_coord_enable; - rctx->rasterizer = rs; + r600_pipe_state_add_reg(rstate, R_02820C_PA_SC_CLIPRECT_RULE, clip_rule, 0xFFFFFFFF, NULL); - rctx->states[rs->rstate.id] = &rs->rstate; - r600_context_pipe_state_set(&rctx->ctx, &rs->rstate); -} - -static void r600_delete_rs_state(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - struct r600_pipe_rasterizer *rs = (struct r600_pipe_rasterizer *)state; - - if (rctx->rasterizer == rs) { - rctx->rasterizer = NULL; - } - if (rctx->states[rs->rstate.id] == &rs->rstate) { - rctx->states[rs->rstate.id] = NULL; - } - free(rs); + return rstate; } static void *r600_create_sampler_state(struct pipe_context *ctx, @@ -574,28 +521,6 @@ static void *r600_create_sampler_state(struct pipe_context *ctx, return rstate; } -static void *r600_create_vertex_elements(struct pipe_context *ctx, - unsigned count, - const struct pipe_vertex_element *elements) -{ - struct r600_vertex_element *v = CALLOC_STRUCT(r600_vertex_element); - - assert(count < 32); - v->count = count; - v->refcount = 1; - memcpy(v->elements, elements, count * sizeof(struct pipe_vertex_element)); - return v; -} - -static void r600_sampler_view_destroy(struct pipe_context *ctx, - struct pipe_sampler_view *state) -{ - struct r600_pipe_sampler_view *resource = (struct r600_pipe_sampler_view *)state; - - pipe_resource_reference(&state->texture, NULL); - FREE(resource); -} - static struct pipe_sampler_view *r600_create_sampler_view(struct pipe_context *ctx, struct pipe_resource *texture, const struct pipe_sampler_view *state) @@ -648,7 +573,11 @@ static struct pipe_sampler_view *r600_create_sampler_view(struct pipe_context *c bo[0] = rbuffer->bo; bo[1] = rbuffer->bo; } - pitch = align(tmp->pitch[0] / tmp->bpt, 8); + pitch = align(tmp->pitch_in_pixels[0], 8); + if (tmp->tiled) { + array_mode = tmp->array_mode[0]; + tile_type = tmp->tile_type; + } /* FIXME properly handle first level != 0 */ r600_pipe_state_add_reg(rstate, R_038000_RESOURCE0_WORD0, @@ -683,32 +612,43 @@ static struct pipe_sampler_view *r600_create_sampler_view(struct pipe_context *c static void r600_set_vs_sampler_view(struct pipe_context *ctx, unsigned count, struct pipe_sampler_view **views) { - /* TODO */ - assert(1); -} - -static void r600_set_ps_sampler_view(struct pipe_context *ctx, unsigned count, - struct pipe_sampler_view **views) -{ struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; struct r600_pipe_sampler_view **resource = (struct r600_pipe_sampler_view **)views; for (int i = 0; i < count; i++) { if (resource[i]) { - r600_context_pipe_state_set_ps_resource(&rctx->ctx, &resource[i]->state, i); + r600_context_pipe_state_set_vs_resource(&rctx->ctx, &resource[i]->state, i + PIPE_MAX_ATTRIBS); } } } -static void r600_bind_state(struct pipe_context *ctx, void *state) +static void r600_set_ps_sampler_view(struct pipe_context *ctx, unsigned count, + struct pipe_sampler_view **views) { struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - struct r600_pipe_state *rstate = (struct r600_pipe_state *)state; + struct r600_pipe_sampler_view **resource = (struct r600_pipe_sampler_view **)views; + int i; - if (state == NULL) - return; - rctx->states[rstate->id] = rstate; - r600_context_pipe_state_set(&rctx->ctx, rstate); + for (i = 0; i < count; i++) { + if (&rctx->ps_samplers.views[i]->base != views[i]) { + if (resource[i]) + r600_context_pipe_state_set_ps_resource(&rctx->ctx, &resource[i]->state, i); + else + r600_context_pipe_state_set_ps_resource(&rctx->ctx, NULL, i); + + pipe_sampler_view_reference( + (struct pipe_sampler_view **)&rctx->ps_samplers.views[i], + views[i]); + + } + } + for (i = count; i < NUM_TEX_UNITS; i++) { + if (rctx->ps_samplers.views[i]) { + r600_context_pipe_state_set_ps_resource(&rctx->ctx, NULL, i); + pipe_sampler_view_reference((struct pipe_sampler_view **)&rctx->ps_samplers.views[i], NULL); + } + } + rctx->ps_samplers.n_views = count; } static void r600_bind_ps_sampler(struct pipe_context *ctx, unsigned count, void **states) @@ -716,6 +656,9 @@ static void r600_bind_ps_sampler(struct pipe_context *ctx, unsigned count, void struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; struct r600_pipe_state **rstates = (struct r600_pipe_state **)states; + memcpy(rctx->ps_samplers.samplers, states, sizeof(void*) * count); + rctx->ps_samplers.n_samplers = count; + for (int i = 0; i < count; i++) { r600_context_pipe_state_set_ps_sampler(&rctx->ctx, rstates[i], i); } @@ -726,37 +669,11 @@ static void r600_bind_vs_sampler(struct pipe_context *ctx, unsigned count, void struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; struct r600_pipe_state **rstates = (struct r600_pipe_state **)states; - /* TODO implement */ for (int i = 0; i < count; i++) { r600_context_pipe_state_set_vs_sampler(&rctx->ctx, rstates[i], i); } } -static void r600_delete_state(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - struct r600_pipe_state *rstate = (struct r600_pipe_state *)state; - - if (rctx->states[rstate->id] == rstate) { - rctx->states[rstate->id] = NULL; - } - for (int i = 0; i < rstate->nregs; i++) { - r600_bo_reference(rctx->radeon, &rstate->regs[i].bo, NULL); - } - free(rstate); -} - -static void r600_delete_vertex_element(struct pipe_context *ctx, void *state) -{ - struct r600_vertex_element *v = (struct r600_vertex_element*)state; - - if (v == NULL) - return; - if (--v->refcount) - return; - free(v); -} - static void r600_set_clip_state(struct pipe_context *ctx, const struct pipe_clip_state *state) { @@ -792,19 +709,6 @@ static void r600_set_clip_state(struct pipe_context *ctx, r600_context_pipe_state_set(&rctx->ctx, rstate); } -static void r600_bind_vertex_elements(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - struct r600_vertex_element *v = (struct r600_vertex_element*)state; - - r600_delete_vertex_element(ctx, rctx->vertex_elements); - rctx->vertex_elements = v; - if (v) { - v->refcount++; -// rctx->vs_rebuild = TRUE; - } -} - static void r600_set_polygon_stipple(struct pipe_context *ctx, const struct pipe_poly_stipple *state) { @@ -828,18 +732,6 @@ static void r600_set_scissor_state(struct pipe_context *ctx, tl = S_028240_TL_X(state->minx) | S_028240_TL_Y(state->miny) | S_028240_WINDOW_OFFSET_DISABLE(1); br = S_028244_BR_X(state->maxx) | S_028244_BR_Y(state->maxy); r600_pipe_state_add_reg(rstate, - R_028030_PA_SC_SCREEN_SCISSOR_TL, tl, - 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, - R_028034_PA_SC_SCREEN_SCISSOR_BR, br, - 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, - R_028204_PA_SC_WINDOW_SCISSOR_TL, tl, - 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, - R_028208_PA_SC_WINDOW_SCISSOR_BR, br, - 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028210_PA_SC_CLIPRECT_0_TL, tl, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, @@ -863,17 +755,6 @@ static void r600_set_scissor_state(struct pipe_context *ctx, r600_pipe_state_add_reg(rstate, R_02822C_PA_SC_CLIPRECT_3_BR, br, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, - R_028200_PA_SC_WINDOW_OFFSET, 0x00000000, - 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, - R_02820C_PA_SC_CLIPRECT_RULE, 0x0000FFFF, - 0xFFFFFFFF, NULL); - if (rctx->family >= CHIP_RV770) { - r600_pipe_state_add_reg(rstate, - R_028230_PA_SC_EDGERULE, 0xAAAAAAAA, - 0xFFFFFFFF, NULL); - } free(rctx->states[R600_PIPE_STATE_SCISSOR]); rctx->states[R600_PIPE_STATE_SCISSOR] = rstate; @@ -937,6 +818,7 @@ static void r600_cb(struct r600_pipe_context *rctx, struct r600_pipe_state *rsta { struct r600_resource_texture *rtex; struct r600_resource *rbuffer; + struct r600_surface *surf; unsigned level = state->cbufs[cb]->level; unsigned pitch, slice; unsigned color_info; @@ -944,14 +826,15 @@ static void r600_cb(struct r600_pipe_context *rctx, struct r600_pipe_state *rsta const struct util_format_description *desc; struct r600_bo *bo[3]; + surf = (struct r600_surface *)state->cbufs[cb]; rtex = (struct r600_resource_texture*)state->cbufs[cb]->texture; rbuffer = &rtex->resource; bo[0] = rbuffer->bo; bo[1] = rbuffer->bo; bo[2] = rbuffer->bo; - pitch = (rtex->pitch[level] / rtex->bpt) / 8 - 1; - slice = (rtex->pitch[level] / rtex->bpt) * state->cbufs[cb]->height / 64 - 1; + pitch = rtex->pitch_in_pixels[level] / 8 - 1; + slice = rtex->pitch_in_pixels[level] * surf->aligned_height / 64 - 1; ntype = 0; desc = util_format_description(rtex->resource.base.b.format); if (desc->colorspace == UTIL_FORMAT_COLORSPACE_SRGB) @@ -961,6 +844,7 @@ static void r600_cb(struct r600_pipe_context *rctx, struct r600_pipe_state *rsta swap = r600_translate_colorswap(rtex->resource.base.b.format); color_info = S_0280A0_FORMAT(format) | S_0280A0_COMP_SWAP(swap) | + S_0280A0_ARRAY_MODE(rtex->array_mode[level]) | S_0280A0_BLEND_CLAMP(1) | S_0280A0_NUMBER_TYPE(ntype); if (desc->colorspace != UTIL_FORMAT_COLORSPACE_ZS) @@ -996,22 +880,25 @@ static void r600_db(struct r600_pipe_context *rctx, struct r600_pipe_state *rsta { struct r600_resource_texture *rtex; struct r600_resource *rbuffer; + struct r600_surface *surf; unsigned level; unsigned pitch, slice, format; if (state->zsbuf == NULL) return; + level = state->zsbuf->level; + + surf = (struct r600_surface *)state->zsbuf; rtex = (struct r600_resource_texture*)state->zsbuf->texture; rtex->tiled = 1; - rtex->array_mode = 2; + rtex->array_mode[level] = 2; rtex->tile_type = 1; rtex->depth = 1; rbuffer = &rtex->resource; - level = state->zsbuf->level; - pitch = (rtex->pitch[level] / rtex->bpt) / 8 - 1; - slice = (rtex->pitch[level] / rtex->bpt) * state->zsbuf->height / 64 - 1; + pitch = rtex->pitch_in_pixels[level] / 8 - 1; + slice = rtex->pitch_in_pixels[level] * surf->aligned_height / 64 - 1; format = r600_translate_dbformat(state->zsbuf->texture->format); r600_pipe_state_add_reg(rstate, R_02800C_DB_DEPTH_BASE, @@ -1021,10 +908,10 @@ static void r600_db(struct r600_pipe_context *rctx, struct r600_pipe_state *rsta 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_028004_DB_DEPTH_VIEW, 0x00000000, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_028010_DB_DEPTH_INFO, - S_028010_ARRAY_MODE(rtex->array_mode) | S_028010_FORMAT(format), + S_028010_ARRAY_MODE(rtex->array_mode[level]) | S_028010_FORMAT(format), 0xFFFFFFFF, rbuffer->bo); r600_pipe_state_add_reg(rstate, R_028D34_DB_PREFETCH_LIMIT, - (state->zsbuf->height / 8) - 1, 0xFFFFFFFF, NULL); + (surf->aligned_height / 8) - 1, 0xFFFFFFFF, NULL); } static void r600_set_framebuffer_state(struct pipe_context *ctx, @@ -1039,14 +926,9 @@ static void r600_set_framebuffer_state(struct pipe_context *ctx, /* unreference old buffer and reference new one */ rstate->id = R600_PIPE_STATE_FRAMEBUFFER; - for (int i = 0; i < rctx->framebuffer.nr_cbufs; i++) { - pipe_surface_reference(&rctx->framebuffer.cbufs[i], NULL); - } - for (int i = 0; i < state->nr_cbufs; i++) { - pipe_surface_reference(&rctx->framebuffer.cbufs[i], state->cbufs[i]); - } - pipe_surface_reference(&rctx->framebuffer.zsbuf, state->zsbuf); - rctx->framebuffer = *state; + + util_copy_framebuffer_state(&rctx->framebuffer, state); + rctx->pframebuffer = &rctx->framebuffer; /* build states */ @@ -1070,6 +952,18 @@ static void r600_set_framebuffer_state(struct pipe_context *ctx, br = S_028244_BR_X(state->width) | S_028244_BR_Y(state->height); r600_pipe_state_add_reg(rstate, + R_028030_PA_SC_SCREEN_SCISSOR_TL, tl, + 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, + R_028034_PA_SC_SCREEN_SCISSOR_BR, br, + 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, + R_028204_PA_SC_WINDOW_SCISSOR_TL, tl, + 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, + R_028208_PA_SC_WINDOW_SCISSOR_BR, br, + 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028240_PA_SC_GENERIC_SCISSOR_TL, tl, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, @@ -1081,6 +975,14 @@ static void r600_set_framebuffer_state(struct pipe_context *ctx, r600_pipe_state_add_reg(rstate, R_028254_PA_SC_VPORT_SCISSOR_0_BR, br, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, + R_028200_PA_SC_WINDOW_OFFSET, 0x00000000, + 0xFFFFFFFF, NULL); + if (rctx->family >= CHIP_RV770) { + r600_pipe_state_add_reg(rstate, + R_028230_PA_SC_EDGERULE, 0xAAAAAAAA, + 0xFFFFFFFF, NULL); + } r600_pipe_state_add_reg(rstate, R_0287A0_CB_SHADER_CONTROL, shader_control, 0xFFFFFFFF, NULL); @@ -1110,40 +1012,6 @@ static void r600_set_framebuffer_state(struct pipe_context *ctx, r600_context_pipe_state_set(&rctx->ctx, rstate); } -static void r600_set_index_buffer(struct pipe_context *ctx, - const struct pipe_index_buffer *ib) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - - if (ib) { - pipe_resource_reference(&rctx->index_buffer.buffer, ib->buffer); - memcpy(&rctx->index_buffer, ib, sizeof(rctx->index_buffer)); - } else { - pipe_resource_reference(&rctx->index_buffer.buffer, NULL); - memset(&rctx->index_buffer, 0, sizeof(rctx->index_buffer)); - } - - /* TODO make this more like a state */ -} - -static void r600_set_vertex_buffers(struct pipe_context *ctx, unsigned count, - const struct pipe_vertex_buffer *buffers) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - - for (int i = 0; i < rctx->nvertex_buffer; i++) { - pipe_resource_reference(&rctx->vertex_buffer[i].buffer, NULL); - } - memcpy(rctx->vertex_buffer, buffers, sizeof(struct pipe_vertex_buffer) * count); - for (int i = 0; i < count; i++) { - rctx->vertex_buffer[i].buffer = NULL; - if (r600_buffer_is_user_buffer(buffers[i].buffer)) - rctx->any_user_vbs = TRUE; - pipe_resource_reference(&rctx->vertex_buffer[i].buffer, buffers[i].buffer); - } - rctx->nvertex_buffer = count; -} - static void r600_set_constant_buffer(struct pipe_context *ctx, uint shader, uint index, struct pipe_resource *buffer) { @@ -1179,59 +1047,6 @@ static void r600_set_constant_buffer(struct pipe_context *ctx, uint shader, uint } } -static void *r600_create_shader_state(struct pipe_context *ctx, - const struct pipe_shader_state *state) -{ - struct r600_pipe_shader *shader = CALLOC_STRUCT(r600_pipe_shader); - int r; - - r = r600_pipe_shader_create(ctx, shader, state->tokens); - if (r) { - return NULL; - } - return shader; -} - -static void r600_bind_ps_shader(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - - /* TODO delete old shader */ - rctx->ps_shader = (struct r600_pipe_shader *)state; -} - -static void r600_bind_vs_shader(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - - /* TODO delete old shader */ - rctx->vs_shader = (struct r600_pipe_shader *)state; -} - -static void r600_delete_ps_shader(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - struct r600_pipe_shader *shader = (struct r600_pipe_shader *)state; - - if (rctx->ps_shader == shader) { - rctx->ps_shader = NULL; - } - /* TODO proper delete */ - free(shader); -} - -static void r600_delete_vs_shader(struct pipe_context *ctx, void *state) -{ - struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; - struct r600_pipe_shader *shader = (struct r600_pipe_shader *)state; - - if (rctx->vs_shader == shader) { - rctx->vs_shader = NULL; - } - /* TODO proper delete */ - free(shader); -} - void r600_init_state_functions(struct r600_pipe_context *rctx) { rctx->context.create_blend_state = r600_create_blend_state; @@ -1480,14 +1295,14 @@ void r600_init_config(struct r600_pipe_context *rctx) r600_pipe_state_add_reg(rstate, R_009830_DB_DEBUG, 0x00000000, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_009838_DB_WATERMARKS, 0x00420204, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_0286C8_SPI_THREAD_GROUPING, 0x00000000, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A4C_PA_SC_MODE_CNTL, 0x00514000, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028A4C_PA_SC_MODE_CNTL, 0x00514002, 0xFFFFFFFF, NULL); } else { r600_pipe_state_add_reg(rstate, R_008D8C_SQ_DYN_GPR_CNTL_PS_FLUSH_REQ, 0x00000000, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_009508_TA_CNTL_AUX, 0x07000003, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_009830_DB_DEBUG, 0x82000000, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_009838_DB_WATERMARKS, 0x01020204, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_0286C8_SPI_THREAD_GROUPING, 0x00000001, 0xFFFFFFFF, NULL); - r600_pipe_state_add_reg(rstate, R_028A4C_PA_SC_MODE_CNTL, 0x00004010, 0xFFFFFFFF, NULL); + r600_pipe_state_add_reg(rstate, R_028A4C_PA_SC_MODE_CNTL, 0x00004012, 0xFFFFFFFF, NULL); } r600_pipe_state_add_reg(rstate, R_0288A8_SQ_ESGS_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL); r600_pipe_state_add_reg(rstate, R_0288AC_SQ_GSVS_RING_ITEMSIZE, 0x00000000, 0xFFFFFFFF, NULL); diff --git a/src/gallium/drivers/r600/r600_state_common.c b/src/gallium/drivers/r600/r600_state_common.c new file mode 100644 index 00000000000..210420e823b --- /dev/null +++ b/src/gallium/drivers/r600/r600_state_common.c @@ -0,0 +1,276 @@ +/* + * Copyright 2010 Red Hat Inc. + * 2010 Jerome Glisse + * + * 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. + * + * Authors: Dave Airlie <[email protected]> + * Jerome Glisse <[email protected]> + */ +#include <util/u_memory.h> +#include <util/u_format.h> +#include <pipebuffer/pb_buffer.h> +#include "r600_pipe.h" + +/* common state between evergreen and r600 */ +void r600_bind_blend_state(struct pipe_context *ctx, void *state) +{ + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; + struct r600_pipe_blend *blend = (struct r600_pipe_blend *)state; + struct r600_pipe_state *rstate; + + if (state == NULL) + return; + rstate = &blend->rstate; + rctx->states[rstate->id] = rstate; + rctx->cb_target_mask = blend->cb_target_mask; + r600_context_pipe_state_set(&rctx->ctx, rstate); +} + +void r600_bind_rs_state(struct pipe_context *ctx, void *state) +{ + struct r600_pipe_rasterizer *rs = (struct r600_pipe_rasterizer *)state; + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; + + if (state == NULL) + return; + + rctx->flatshade = rs->flatshade; + rctx->sprite_coord_enable = rs->sprite_coord_enable; + rctx->rasterizer = rs; + + rctx->states[rs->rstate.id] = &rs->rstate; + r600_context_pipe_state_set(&rctx->ctx, &rs->rstate); +} + +void r600_delete_rs_state(struct pipe_context *ctx, void *state) +{ + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; + struct r600_pipe_rasterizer *rs = (struct r600_pipe_rasterizer *)state; + + if (rctx->rasterizer == rs) { + rctx->rasterizer = NULL; + } + if (rctx->states[rs->rstate.id] == &rs->rstate) { + rctx->states[rs->rstate.id] = NULL; + } + free(rs); +} + +void r600_sampler_view_destroy(struct pipe_context *ctx, + struct pipe_sampler_view *state) +{ + struct r600_pipe_sampler_view *resource = (struct r600_pipe_sampler_view *)state; + + pipe_resource_reference(&state->texture, NULL); + FREE(resource); +} + +void r600_bind_state(struct pipe_context *ctx, void *state) +{ + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; + struct r600_pipe_state *rstate = (struct r600_pipe_state *)state; + + if (state == NULL) + return; + rctx->states[rstate->id] = rstate; + r600_context_pipe_state_set(&rctx->ctx, rstate); +} + +void r600_delete_state(struct pipe_context *ctx, void *state) +{ + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; + struct r600_pipe_state *rstate = (struct r600_pipe_state *)state; + + if (rctx->states[rstate->id] == rstate) { + rctx->states[rstate->id] = NULL; + } + for (int i = 0; i < rstate->nregs; i++) { + r600_bo_reference(rctx->radeon, &rstate->regs[i].bo, NULL); + } + free(rstate); +} + +void r600_bind_vertex_elements(struct pipe_context *ctx, void *state) +{ + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; + struct r600_vertex_element *v = (struct r600_vertex_element*)state; + + rctx->vertex_elements = v; + if (v) { +// rctx->vs_rebuild = TRUE; + } +} + +void r600_delete_vertex_element(struct pipe_context *ctx, void *state) +{ + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; + + FREE(state); + + if (rctx->vertex_elements == state) + rctx->vertex_elements = NULL; +} + + +void r600_set_index_buffer(struct pipe_context *ctx, + const struct pipe_index_buffer *ib) +{ + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; + + if (ib) { + pipe_resource_reference(&rctx->index_buffer.buffer, ib->buffer); + memcpy(&rctx->index_buffer, ib, sizeof(rctx->index_buffer)); + } else { + pipe_resource_reference(&rctx->index_buffer.buffer, NULL); + memset(&rctx->index_buffer, 0, sizeof(rctx->index_buffer)); + } + + /* TODO make this more like a state */ +} + +void r600_set_vertex_buffers(struct pipe_context *ctx, unsigned count, + const struct pipe_vertex_buffer *buffers) +{ + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; + struct pipe_vertex_buffer *vbo; + unsigned max_index = (unsigned)-1; + + for (int i = 0; i < rctx->nvertex_buffer; i++) { + pipe_resource_reference(&rctx->vertex_buffer[i].buffer, NULL); + } + memcpy(rctx->vertex_buffer, buffers, sizeof(struct pipe_vertex_buffer) * count); + + for (int i = 0; i < count; i++) { + vbo = (struct pipe_vertex_buffer*)&buffers[i]; + + rctx->vertex_buffer[i].buffer = NULL; + if (r600_buffer_is_user_buffer(buffers[i].buffer)) + rctx->any_user_vbs = TRUE; + pipe_resource_reference(&rctx->vertex_buffer[i].buffer, buffers[i].buffer); + + if (vbo->max_index == ~0) { + if (!vbo->stride) + vbo->max_index = 1; + else + vbo->max_index = (vbo->buffer->width0 - vbo->buffer_offset) / vbo->stride; + } + max_index = MIN2(vbo->max_index, max_index); + } + rctx->nvertex_buffer = count; + rctx->vb_max_index = max_index; +} + + +#define FORMAT_REPLACE(what, withwhat) \ + case PIPE_FORMAT_##what: *format = PIPE_FORMAT_##withwhat; break + +void *r600_create_vertex_elements(struct pipe_context *ctx, + unsigned count, + const struct pipe_vertex_element *elements) +{ + struct r600_vertex_element *v = CALLOC_STRUCT(r600_vertex_element); + int i; + enum pipe_format *format; + + assert(count < 32); + if (!v) + return NULL; + + v->count = count; + memcpy(v->elements, elements, count * sizeof(struct pipe_vertex_element)); + + for (i = 0; i < count; i++) { + v->hw_format[i] = v->elements[i].src_format; + format = &v->hw_format[i]; + + switch (*format) { + FORMAT_REPLACE(R64_FLOAT, R32_FLOAT); + FORMAT_REPLACE(R64G64_FLOAT, R32G32_FLOAT); + FORMAT_REPLACE(R64G64B64_FLOAT, R32G32B32_FLOAT); + FORMAT_REPLACE(R64G64B64A64_FLOAT, R32G32B32A32_FLOAT); + default:; + } + v->incompatible_layout = + v->incompatible_layout || + v->elements[i].src_format != v->hw_format[i] || + v->elements[i].src_offset % 4 != 0; + + v->hw_format_size[i] = + align(util_format_get_blocksize(v->hw_format[i]), 4); + } + + return v; +} + +void *r600_create_shader_state(struct pipe_context *ctx, + const struct pipe_shader_state *state) +{ + struct r600_pipe_shader *shader = CALLOC_STRUCT(r600_pipe_shader); + int r; + + r = r600_pipe_shader_create(ctx, shader, state->tokens); + if (r) { + return NULL; + } + return shader; +} + +void r600_bind_ps_shader(struct pipe_context *ctx, void *state) +{ + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; + + /* TODO delete old shader */ + rctx->ps_shader = (struct r600_pipe_shader *)state; +} + +void r600_bind_vs_shader(struct pipe_context *ctx, void *state) +{ + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; + + /* TODO delete old shader */ + rctx->vs_shader = (struct r600_pipe_shader *)state; +} + +void r600_delete_ps_shader(struct pipe_context *ctx, void *state) +{ + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; + struct r600_pipe_shader *shader = (struct r600_pipe_shader *)state; + + if (rctx->ps_shader == shader) { + rctx->ps_shader = NULL; + } + + r600_pipe_shader_destroy(ctx, shader); + free(shader); +} + +void r600_delete_vs_shader(struct pipe_context *ctx, void *state) +{ + struct r600_pipe_context *rctx = (struct r600_pipe_context *)ctx; + struct r600_pipe_shader *shader = (struct r600_pipe_shader *)state; + + if (rctx->vs_shader == shader) { + rctx->vs_shader = NULL; + } + + r600_pipe_shader_destroy(ctx, shader); + free(shader); +} diff --git a/src/gallium/drivers/r600/r600_state_inlines.h b/src/gallium/drivers/r600/r600_state_inlines.h index 29a29d87cd3..1c1978f8abb 100644 --- a/src/gallium/drivers/r600/r600_state_inlines.h +++ b/src/gallium/drivers/r600/r600_state_inlines.h @@ -25,6 +25,7 @@ #include "util/u_format.h" #include "r600d.h" +#include "r600_formats.h" static INLINE uint32_t r600_translate_blend_function(int blend_func) { @@ -352,13 +353,13 @@ static inline uint32_t r600_translate_colorswap(enum pipe_format format) /* 64-bit buffers. */ case PIPE_FORMAT_R16G16B16A16_UNORM: case PIPE_FORMAT_R16G16B16A16_SNORM: - // return V_0280A0_COLOR_16_16_16_16; + // return FMT_16_16_16_16; case PIPE_FORMAT_R16G16B16A16_FLOAT: - // return V_0280A0_COLOR_16_16_16_16_FLOAT; + // return FMT_16_16_16_16_FLOAT; /* 128-bit buffers. */ case PIPE_FORMAT_R32G32B32A32_FLOAT: - // return V_0280A0_COLOR_32_32_32_32_FLOAT; + // return FMT_32_32_32_32_FLOAT; return 0; default: R600_ERR("unsupported colorswap format %d\n", format); @@ -522,32 +523,32 @@ static INLINE uint32_t r600_translate_vertex_data_type(enum pipe_format format) case 16: switch (desc->nr_channels) { case 1: - result = V_038008_FMT_16_FLOAT; + result = FMT_16_FLOAT; break; case 2: - result = V_038008_FMT_16_16_FLOAT; + result = FMT_16_16_FLOAT; break; case 3: - result = V_038008_FMT_16_16_16_FLOAT; + result = FMT_16_16_16_FLOAT; break; case 4: - result = V_038008_FMT_16_16_16_16_FLOAT; + result = FMT_16_16_16_16_FLOAT; break; } break; case 32: switch (desc->nr_channels) { case 1: - result = V_038008_FMT_32_FLOAT; + result = FMT_32_FLOAT; break; case 2: - result = V_038008_FMT_32_32_FLOAT; + result = FMT_32_32_FLOAT; break; case 3: - result = V_038008_FMT_32_32_32_FLOAT; + result = FMT_32_32_32_FLOAT; break; case 4: - result = V_038008_FMT_32_32_32_32_FLOAT; + result = FMT_32_32_32_32_FLOAT; break; } break; @@ -563,48 +564,48 @@ static INLINE uint32_t r600_translate_vertex_data_type(enum pipe_format format) case 8: switch (desc->nr_channels) { case 1: - result = V_038008_FMT_8; + result = FMT_8; break; case 2: - result = V_038008_FMT_8_8; + result = FMT_8_8; break; case 3: - // result = V_038008_FMT_8_8_8; /* fails piglit draw-vertices test */ + // result = FMT_8_8_8; /* fails piglit draw-vertices test */ // break; case 4: - result = V_038008_FMT_8_8_8_8; + result = FMT_8_8_8_8; break; } break; case 16: switch (desc->nr_channels) { case 1: - result = V_038008_FMT_16; + result = FMT_16; break; case 2: - result = V_038008_FMT_16_16; + result = FMT_16_16; break; case 3: - // result = V_038008_FMT_16_16_16; /* fails piglit draw-vertices test */ + // result = FMT_16_16_16; /* fails piglit draw-vertices test */ // break; case 4: - result = V_038008_FMT_16_16_16_16; + result = FMT_16_16_16_16; break; } break; case 32: switch (desc->nr_channels) { case 1: - result = V_038008_FMT_32; + result = FMT_32; break; case 2: - result = V_038008_FMT_32_32; + result = FMT_32_32; break; case 3: - result = V_038008_FMT_32_32_32; + result = FMT_32_32_32; break; case 4: - result = V_038008_FMT_32_32_32_32; + result = FMT_32_32_32_32; break; } break; diff --git a/src/gallium/drivers/r600/r600_texture.c b/src/gallium/drivers/r600/r600_texture.c index db88346afbe..4ebd5b754b3 100644 --- a/src/gallium/drivers/r600/r600_texture.c +++ b/src/gallium/drivers/r600/r600_texture.c @@ -35,6 +35,7 @@ #include "r600_resource.h" #include "r600_state_inlines.h" #include "r600d.h" +#include "r600_formats.h" extern struct u_resource_vtbl r600_texture_vtbl; @@ -53,11 +54,30 @@ static void r600_copy_from_tiled_texture(struct pipe_context *ctx, struct r600_t transfer->box.width, transfer->box.height); } -static unsigned long r600_texture_get_offset(struct r600_resource_texture *rtex, + +/* Copy from a detiled texture to a tiled one. */ +static void r600_copy_into_tiled_texture(struct pipe_context *ctx, struct r600_transfer *rtransfer) +{ + struct pipe_transfer *transfer = (struct pipe_transfer*)rtransfer; + struct pipe_resource *texture = transfer->resource; + struct pipe_subresource subsrc; + + subsrc.face = 0; + subsrc.level = 0; + ctx->resource_copy_region(ctx, texture, transfer->sr, + transfer->box.x, transfer->box.y, transfer->box.z, + rtransfer->linear_texture, subsrc, + 0, 0, 0, + transfer->box.width, transfer->box.height); + + ctx->flush(ctx, 0, NULL); +} + +static unsigned r600_texture_get_offset(struct r600_resource_texture *rtex, unsigned level, unsigned zslice, unsigned face) { - unsigned long offset = rtex->offset[level]; + unsigned offset = rtex->offset[level]; switch (rtex->resource.base.b.target) { case PIPE_TEXTURE_3D: @@ -72,22 +92,158 @@ static unsigned long r600_texture_get_offset(struct r600_resource_texture *rtex, } } -static void r600_setup_miptree(struct r600_resource_texture *rtex, enum chip_class chipc) +static unsigned r600_get_pixel_alignment(struct pipe_screen *screen, + enum pipe_format format, + unsigned array_mode) +{ + struct r600_screen* rscreen = (struct r600_screen *)screen; + unsigned pixsize = util_format_get_blocksize(format); + int p_align; + + switch(array_mode) { + case V_038000_ARRAY_1D_TILED_THIN1: + p_align = MAX2(8, + ((rscreen->tiling_info->group_bytes / 8 / pixsize))); + break; + case V_038000_ARRAY_2D_TILED_THIN1: + p_align = MAX2(rscreen->tiling_info->num_banks, + (((rscreen->tiling_info->group_bytes / 8 / pixsize)) * + rscreen->tiling_info->num_banks)); + break; + case 0: + default: + p_align = 64; + break; + } + return p_align; +} + +static unsigned r600_get_height_alignment(struct pipe_screen *screen, + unsigned array_mode) +{ + struct r600_screen* rscreen = (struct r600_screen *)screen; + int h_align; + + switch (array_mode) { + case V_038000_ARRAY_2D_TILED_THIN1: + h_align = rscreen->tiling_info->num_channels * 8; + break; + case V_038000_ARRAY_1D_TILED_THIN1: + h_align = 8; + break; + default: + h_align = 1; + break; + } + return h_align; +} + +static unsigned mip_minify(unsigned size, unsigned level) +{ + unsigned val; + val = u_minify(size, level); + if (level > 0) + val = util_next_power_of_two(val); + return val; +} + +static unsigned r600_texture_get_stride(struct pipe_screen *screen, + struct r600_resource_texture *rtex, + unsigned level) +{ + struct pipe_resource *ptex = &rtex->resource.base.b; + struct radeon *radeon = (struct radeon *)screen->winsys; + enum chip_class chipc = r600_get_family_class(radeon); + unsigned width, stride, tile_width; + + if (rtex->pitch_override) + return rtex->pitch_override; + + width = mip_minify(ptex->width0, level); + if (util_format_is_plain(ptex->format)) { + tile_width = r600_get_pixel_alignment(screen, ptex->format, + rtex->array_mode[level]); + width = align(width, tile_width); + } + stride = util_format_get_stride(ptex->format, width); + if (chipc == EVERGREEN) + stride = align(stride, 512); + return stride; +} + +static unsigned r600_texture_get_nblocksy(struct pipe_screen *screen, + struct r600_resource_texture *rtex, + unsigned level) { struct pipe_resource *ptex = &rtex->resource.base.b; - unsigned long w, h, pitch, size, layer_size, i, offset; + unsigned height, tile_height; - rtex->bpt = util_format_get_blocksize(ptex->format); - for (i = 0, offset = 0; i <= ptex->last_level; i++) { - w = u_minify(ptex->width0, i); - h = u_minify(ptex->height0, i); - h = util_next_power_of_two(h); - pitch = util_format_get_stride(ptex->format, align(w, 64)); - if (chipc == EVERGREEN) - pitch = align(pitch, 512); + height = mip_minify(ptex->height0, level); + if (util_format_is_plain(ptex->format)) { + tile_height = r600_get_height_alignment(screen, + rtex->array_mode[level]); + height = align(height, tile_height); + } + return util_format_get_nblocksy(ptex->format, height); +} + +/* Get a width in pixels from a stride in bytes. */ +static unsigned pitch_to_width(enum pipe_format format, + unsigned pitch_in_bytes) +{ + return (pitch_in_bytes / util_format_get_blocksize(format)) * + util_format_get_blockwidth(format); +} + +static void r600_texture_set_array_mode(struct pipe_screen *screen, + struct r600_resource_texture *rtex, + unsigned level, unsigned array_mode) +{ + struct pipe_resource *ptex = &rtex->resource.base.b; + + switch (array_mode) { + case V_0280A0_ARRAY_LINEAR_GENERAL: + case V_0280A0_ARRAY_LINEAR_ALIGNED: + case V_0280A0_ARRAY_1D_TILED_THIN1: + default: + rtex->array_mode[level] = array_mode; + break; + case V_0280A0_ARRAY_2D_TILED_THIN1: + { + unsigned w, h, tile_height, tile_width; + + tile_height = r600_get_height_alignment(screen, array_mode); + tile_width = r600_get_pixel_alignment(screen, ptex->format, array_mode); + + w = mip_minify(ptex->width0, level); + h = mip_minify(ptex->height0, level); + if (w < tile_width || h < tile_height) + rtex->array_mode[level] = V_0280A0_ARRAY_1D_TILED_THIN1; else - pitch = align(pitch, 256); - layer_size = pitch * h; + rtex->array_mode[level] = array_mode; + } + break; + } +} + +static void r600_setup_miptree(struct pipe_screen *screen, + struct r600_resource_texture *rtex, + unsigned array_mode) +{ + struct pipe_resource *ptex = &rtex->resource.base.b; + struct radeon *radeon = (struct radeon *)screen->winsys; + enum chip_class chipc = r600_get_family_class(radeon); + unsigned pitch, size, layer_size, i, offset; + unsigned nblocksy; + + for (i = 0, offset = 0; i <= ptex->last_level; i++) { + r600_texture_set_array_mode(screen, rtex, i, array_mode); + + pitch = r600_texture_get_stride(screen, rtex, i); + nblocksy = r600_texture_get_nblocksy(screen, rtex, i); + + layer_size = pitch * nblocksy; + if (ptex->target == PIPE_TEXTURE_CUBE) { if (chipc >= R700) size = layer_size * 8; @@ -98,39 +254,69 @@ static void r600_setup_miptree(struct r600_resource_texture *rtex, enum chip_cla size = layer_size * u_minify(ptex->depth0, i); rtex->offset[i] = offset; rtex->layer_size[i] = layer_size; - rtex->pitch[i] = pitch; + rtex->pitch_in_bytes[i] = pitch; + rtex->pitch_in_pixels[i] = pitch_to_width(ptex->format, pitch); offset += size; } rtex->size = offset; } -struct pipe_resource *r600_texture_create(struct pipe_screen *screen, - const struct pipe_resource *templ) +static struct r600_resource_texture * +r600_texture_create_object(struct pipe_screen *screen, + const struct pipe_resource *base, + unsigned array_mode, + unsigned pitch_in_bytes_override, + unsigned max_buffer_size, + struct r600_bo *bo) { struct r600_resource_texture *rtex; struct r600_resource *resource; struct radeon *radeon = (struct radeon *)screen->winsys; rtex = CALLOC_STRUCT(r600_resource_texture); - if (!rtex) { + if (rtex == NULL) return NULL; - } + resource = &rtex->resource; - resource->base.b = *templ; + resource->base.b = *base; resource->base.vtbl = &r600_texture_vtbl; pipe_reference_init(&resource->base.b.reference, 1); resource->base.b.screen = screen; - r600_setup_miptree(rtex, r600_get_family_class(radeon)); - - /* FIXME alignment 4096 enought ? too much ? */ + resource->bo = bo; resource->domain = r600_domain_from_usage(resource->base.b.bind); + rtex->pitch_override = pitch_in_bytes_override; + + if (array_mode) + rtex->tiled = 1; + r600_setup_miptree(screen, rtex, array_mode); + resource->size = rtex->size; - resource->bo = r600_bo(radeon, rtex->size, 4096, 0); - if (resource->bo == NULL) { - FREE(rtex); - return NULL; + + if (!resource->bo) { + resource->bo = r600_bo(radeon, rtex->size, 4096, 0); + if (!resource->bo) { + FREE(rtex); + return NULL; + } + } + return rtex; +} + +struct pipe_resource *r600_texture_create(struct pipe_screen *screen, + const struct pipe_resource *templ) +{ + unsigned array_mode = 0; + + if (debug_get_bool_option("R600_FORCE_TILING", FALSE)) { + if (!(templ->flags & R600_RESOURCE_FLAG_TRANSFER) && + !(templ->bind & PIPE_BIND_SCANOUT)) { + array_mode = V_038000_ARRAY_2D_TILED_THIN1; + } } - return &resource->base.b; + + return (struct pipe_resource *)r600_texture_create_object(screen, templ, array_mode, + 0, 0, NULL); + } static void r600_texture_destroy(struct pipe_screen *screen, @@ -155,24 +341,27 @@ static struct pipe_surface *r600_get_tex_surface(struct pipe_screen *screen, unsigned zslice, unsigned flags) { struct r600_resource_texture *rtex = (struct r600_resource_texture*)texture; - struct pipe_surface *surface = CALLOC_STRUCT(pipe_surface); - unsigned long offset; + struct r600_surface *surface = CALLOC_STRUCT(r600_surface); + unsigned offset, tile_height; if (surface == NULL) return NULL; offset = r600_texture_get_offset(rtex, level, zslice, face); - pipe_reference_init(&surface->reference, 1); - pipe_resource_reference(&surface->texture, texture); - surface->format = texture->format; - surface->width = u_minify(texture->width0, level); - surface->height = u_minify(texture->height0, level); - surface->offset = offset; - surface->usage = flags; - surface->zslice = zslice; - surface->texture = texture; - surface->face = face; - surface->level = level; - return surface; + pipe_reference_init(&surface->base.reference, 1); + pipe_resource_reference(&surface->base.texture, texture); + surface->base.format = texture->format; + surface->base.width = mip_minify(texture->width0, level); + surface->base.height = mip_minify(texture->height0, level); + surface->base.offset = offset; + surface->base.usage = flags; + surface->base.zslice = zslice; + surface->base.texture = texture; + surface->base.face = face; + surface->base.level = level; + + tile_height = r600_get_height_alignment(screen, rtex->array_mode[level]); + surface->aligned_height = align(surface->base.height, tile_height); + return &surface->base; } static void r600_tex_surface_destroy(struct pipe_surface *surface) @@ -181,44 +370,29 @@ static void r600_tex_surface_destroy(struct pipe_surface *surface) FREE(surface); } + struct pipe_resource *r600_texture_from_handle(struct pipe_screen *screen, const struct pipe_resource *templ, struct winsys_handle *whandle) { struct radeon *rw = (struct radeon*)screen->winsys; - struct r600_resource_texture *rtex; - struct r600_resource *resource; struct r600_bo *bo = NULL; + unsigned array_mode = 0; /* Support only 2D textures without mipmaps */ if ((templ->target != PIPE_TEXTURE_2D && templ->target != PIPE_TEXTURE_RECT) || templ->depth0 != 1 || templ->last_level != 0) return NULL; - rtex = CALLOC_STRUCT(r600_resource_texture); - if (rtex == NULL) - return NULL; - - bo = r600_bo_handle(rw, whandle->handle); + bo = r600_bo_handle(rw, whandle->handle, &array_mode); if (bo == NULL) { - FREE(rtex); return NULL; } - resource = &rtex->resource; - resource->base.b = *templ; - resource->base.vtbl = &r600_texture_vtbl; - pipe_reference_init(&resource->base.b.reference, 1); - resource->base.b.screen = screen; - resource->bo = bo; - rtex->depth = 0; - rtex->pitch_override = whandle->stride; - rtex->bpt = util_format_get_blocksize(templ->format); - rtex->pitch[0] = whandle->stride; - rtex->offset[0] = 0; - rtex->size = align(rtex->pitch[0] * templ->height0, 64); - - return &resource->base.b; + return (struct pipe_resource *)r600_texture_create_object(screen, templ, array_mode, + whandle->stride, + 0, + bo); } static unsigned int r600_texture_is_referenced(struct pipe_context *context, @@ -244,14 +418,14 @@ int r600_texture_depth_flush(struct pipe_context *ctx, resource.format = texture->format; resource.width0 = texture->width0; resource.height0 = texture->height0; - resource.depth0 = 0; + resource.depth0 = 1; resource.last_level = 0; resource.nr_samples = 0; resource.usage = PIPE_USAGE_DYNAMIC; resource.bind = 0; - resource.flags = 0; + resource.flags = R600_RESOURCE_FLAG_TRANSFER; - resource.bind |= PIPE_BIND_RENDER_TARGET; + resource.bind |= PIPE_BIND_DEPTH_STENCIL; rtex->flushed_depth_texture = (struct r600_resource_texture *)ctx->screen->resource_create(ctx->screen, &resource); if (rtex->flushed_depth_texture == NULL) { @@ -282,8 +456,6 @@ struct pipe_transfer* r600_texture_get_transfer(struct pipe_context *ctx, trans->transfer.sr = sr; trans->transfer.usage = usage; trans->transfer.box = *box; - trans->transfer.stride = rtex->pitch[sr.level]; - trans->offset = r600_texture_get_offset(rtex, sr.level, box->z, sr.face); if (rtex->depth) { r = r600_texture_depth_flush(ctx, texture); if (r < 0) { @@ -297,12 +469,12 @@ struct pipe_transfer* r600_texture_get_transfer(struct pipe_context *ctx, resource.format = texture->format; resource.width0 = box->width; resource.height0 = box->height; - resource.depth0 = 0; + resource.depth0 = 1; resource.last_level = 0; resource.nr_samples = 0; resource.usage = PIPE_USAGE_DYNAMIC; resource.bind = 0; - resource.flags = 0; + resource.flags = R600_RESOURCE_FLAG_TRANSFER; /* For texture reading, the temporary (detiled) texture is used as * a render target when blitting from a tiled texture. */ if (usage & PIPE_TRANSFER_READ) { @@ -321,6 +493,9 @@ struct pipe_transfer* r600_texture_get_transfer(struct pipe_context *ctx, FREE(trans); return NULL; } + + trans->transfer.stride = + ((struct r600_resource_texture *)trans->linear_texture)->pitch_in_bytes[0]; if (usage & PIPE_TRANSFER_READ) { /* We cannot map a tiled texture directly because the data is * in a different order, therefore we do detiling using a blit. */ @@ -328,7 +503,10 @@ struct pipe_transfer* r600_texture_get_transfer(struct pipe_context *ctx, /* Always referenced in the blit. */ ctx->flush(ctx, 0, NULL); } + return &trans->transfer; } + trans->transfer.stride = rtex->pitch_in_bytes[sr.level]; + trans->offset = r600_texture_get_offset(rtex, sr.level, box->z, sr.face); return &trans->transfer; } @@ -339,12 +517,12 @@ void r600_texture_transfer_destroy(struct pipe_context *ctx, struct r600_resource_texture *rtex = (struct r600_resource_texture*)transfer->resource; if (rtransfer->linear_texture) { + if (transfer->usage & PIPE_TRANSFER_WRITE) { + r600_copy_into_tiled_texture(ctx, rtransfer); + } pipe_resource_reference(&rtransfer->linear_texture, NULL); } if (rtex->flushed_depth_texture) { - if (transfer->usage & PIPE_TRANSFER_WRITE) { - // TODO - } pipe_resource_reference((struct pipe_resource **)&rtex->flushed_depth_texture, NULL); } pipe_resource_reference(&transfer->resource, NULL); @@ -358,7 +536,7 @@ void* r600_texture_transfer_map(struct pipe_context *ctx, struct r600_bo *bo; enum pipe_format format = transfer->resource->format; struct radeon *radeon = (struct radeon *)ctx->screen->winsys; - unsigned long offset = 0; + unsigned offset = 0; char *map; if (rtransfer->linear_texture) { @@ -496,15 +674,23 @@ uint32_t r600_translate_texformat(enum pipe_format format, case UTIL_FORMAT_COLORSPACE_ZS: switch (format) { case PIPE_FORMAT_Z16_UNORM: - result = V_0280A0_COLOR_16; + result = FMT_16; goto out_word4; + case PIPE_FORMAT_X24S8_USCALED: + word4 |= S_038010_NUM_FORMAT_ALL(V_038010_SQ_NUM_FORMAT_INT); case PIPE_FORMAT_Z24X8_UNORM: case PIPE_FORMAT_Z24_UNORM_S8_USCALED: - result = V_0280A0_COLOR_8_24; + result = FMT_8_24; goto out_word4; + case PIPE_FORMAT_S8X24_USCALED: + word4 |= S_038010_NUM_FORMAT_ALL(V_038010_SQ_NUM_FORMAT_INT); case PIPE_FORMAT_X8Z24_UNORM: case PIPE_FORMAT_S8_USCALED_Z24_UNORM: - result = V_0280A0_COLOR_24_8; + result = FMT_24_8; + goto out_word4; + case PIPE_FORMAT_S8_USCALED: + result = V_0280A0_COLOR_8; + word4 |= S_038010_NUM_FORMAT_ALL(V_038010_SQ_NUM_FORMAT_INT); goto out_word4; default: goto out_unknown; @@ -558,7 +744,7 @@ uint32_t r600_translate_texformat(enum pipe_format format, if (desc->channel[0].size == 5 && desc->channel[1].size == 6 && desc->channel[2].size == 5) { - result = V_0280A0_COLOR_5_6_5; + result = FMT_5_6_5; goto out_word4; } goto out_unknown; @@ -567,14 +753,14 @@ uint32_t r600_translate_texformat(enum pipe_format format, desc->channel[1].size == 5 && desc->channel[2].size == 5 && desc->channel[3].size == 1) { - result = V_0280A0_COLOR_1_5_5_5; + result = FMT_1_5_5_5; goto out_word4; } if (desc->channel[0].size == 10 && desc->channel[1].size == 10 && desc->channel[2].size == 10 && desc->channel[3].size == 2) { - result = V_0280A0_COLOR_10_10_10_2; + result = FMT_10_10_10_2; goto out_word4; } goto out_unknown; @@ -605,36 +791,36 @@ uint32_t r600_translate_texformat(enum pipe_format format, case 4: switch (desc->nr_channels) { case 2: - result = V_0280A0_COLOR_4_4; + result = FMT_4_4; goto out_word4; case 4: - result = V_0280A0_COLOR_4_4_4_4; + result = FMT_4_4_4_4; goto out_word4; } goto out_unknown; case 8: switch (desc->nr_channels) { case 1: - result = V_0280A0_COLOR_8; + result = FMT_8; goto out_word4; case 2: - result = V_0280A0_COLOR_8_8; + result = FMT_8_8; goto out_word4; case 4: - result = V_0280A0_COLOR_8_8_8_8; + result = FMT_8_8_8_8; goto out_word4; } goto out_unknown; case 16: switch (desc->nr_channels) { case 1: - result = V_0280A0_COLOR_16; + result = FMT_16; goto out_word4; case 2: - result = V_0280A0_COLOR_16_16; + result = FMT_16_16; goto out_word4; case 4: - result = V_0280A0_COLOR_16_16_16_16; + result = FMT_16_16_16_16; goto out_word4; } } @@ -645,26 +831,26 @@ uint32_t r600_translate_texformat(enum pipe_format format, case 16: switch (desc->nr_channels) { case 1: - result = V_0280A0_COLOR_16_FLOAT; + result = FMT_16_FLOAT; goto out_word4; case 2: - result = V_0280A0_COLOR_16_16_FLOAT; + result = FMT_16_16_FLOAT; goto out_word4; case 4: - result = V_0280A0_COLOR_16_16_16_16_FLOAT; + result = FMT_16_16_16_16_FLOAT; goto out_word4; } goto out_unknown; case 32: switch (desc->nr_channels) { case 1: - result = V_0280A0_COLOR_32_FLOAT; + result = FMT_32_FLOAT; goto out_word4; case 2: - result = V_0280A0_COLOR_32_32_FLOAT; + result = FMT_32_32_FLOAT; goto out_word4; case 4: - result = V_0280A0_COLOR_32_32_32_32_FLOAT; + result = FMT_32_32_32_32_FLOAT; goto out_word4; } } diff --git a/src/gallium/drivers/r600/r600_translate.c b/src/gallium/drivers/r600/r600_translate.c new file mode 100644 index 00000000000..9a07cf2073f --- /dev/null +++ b/src/gallium/drivers/r600/r600_translate.c @@ -0,0 +1,211 @@ +/* + * Copyright 2010 Red Hat Inc. + * + * 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. + * + * Authors: Dave Airlie <[email protected]> + */ +#include "translate/translate_cache.h" +#include "translate/translate.h" +#include <pipebuffer/pb_buffer.h> +#include <util/u_index_modify.h> +#include "r600_pipe.h" + +void r600_begin_vertex_translate(struct r600_pipe_context *rctx) +{ + struct pipe_context *pipe = &rctx->context; + struct translate_key key = {0}; + struct translate_element *te; + unsigned tr_elem_index[PIPE_MAX_ATTRIBS] = {0}; + struct translate *tr; + struct r600_vertex_element *ve = rctx->vertex_elements; + boolean vb_translated[PIPE_MAX_ATTRIBS] = {0}; + void *vb_map[PIPE_MAX_ATTRIBS] = {0}, *out_map; + struct pipe_transfer *vb_transfer[PIPE_MAX_ATTRIBS] = {0}, *out_transfer; + struct pipe_resource *out_buffer; + unsigned i, num_verts; + + /* Initialize the translate key, i.e. the recipe how vertices should be + * translated. */ + for (i = 0; i < ve->count; i++) { + struct pipe_vertex_buffer *vb = + &rctx->vertex_buffer[ve->elements[i].vertex_buffer_index]; + enum pipe_format output_format = ve->hw_format[i]; + unsigned output_format_size = ve->hw_format_size[i]; + + /* Check for support. */ + if (ve->elements[i].src_format == ve->hw_format[i] && + (vb->buffer_offset + ve->elements[i].src_offset) % 4 == 0 && + vb->stride % 4 == 0) { + continue; + } + + /* Workaround for translate: output floats instead of halfs. */ + switch (output_format) { + case PIPE_FORMAT_R16_FLOAT: + output_format = PIPE_FORMAT_R32_FLOAT; + output_format_size = 4; + break; + case PIPE_FORMAT_R16G16_FLOAT: + output_format = PIPE_FORMAT_R32G32_FLOAT; + output_format_size = 8; + break; + case PIPE_FORMAT_R16G16B16_FLOAT: + output_format = PIPE_FORMAT_R32G32B32_FLOAT; + output_format_size = 12; + break; + case PIPE_FORMAT_R16G16B16A16_FLOAT: + output_format = PIPE_FORMAT_R32G32B32A32_FLOAT; + output_format_size = 16; + break; + default:; + } + + /* Add this vertex element. */ + te = &key.element[key.nr_elements]; + /*te->type; + te->instance_divisor;*/ + te->input_buffer = ve->elements[i].vertex_buffer_index; + te->input_format = ve->elements[i].src_format; + te->input_offset = vb->buffer_offset + ve->elements[i].src_offset; + te->output_format = output_format; + te->output_offset = key.output_stride; + + key.output_stride += output_format_size; + vb_translated[ve->elements[i].vertex_buffer_index] = TRUE; + tr_elem_index[i] = key.nr_elements; + key.nr_elements++; + } + + /* Get a translate object. */ + tr = translate_cache_find(rctx->tran.translate_cache, &key); + + /* Map buffers we want to translate. */ + for (i = 0; i < rctx->nvertex_buffer; i++) { + if (vb_translated[i]) { + struct pipe_vertex_buffer *vb = &rctx->vertex_buffer[i]; + + vb_map[i] = pipe_buffer_map(pipe, vb->buffer, + PIPE_TRANSFER_READ, &vb_transfer[i]); + + tr->set_buffer(tr, i, vb_map[i], vb->stride, vb->max_index); + } + } + + /* Create and map the output buffer. */ + num_verts = rctx->vb_max_index + 1; + + out_buffer = pipe_buffer_create(&rctx->screen->screen, + PIPE_BIND_VERTEX_BUFFER, + key.output_stride * num_verts); + + out_map = pipe_buffer_map(pipe, out_buffer, PIPE_TRANSFER_WRITE, + &out_transfer); + + /* Translate. */ + tr->run(tr, 0, num_verts, 0, out_map); + + /* Unmap all buffers. */ + for (i = 0; i < rctx->nvertex_buffer; i++) { + if (vb_translated[i]) { + pipe_buffer_unmap(pipe, rctx->vertex_buffer[i].buffer, + vb_transfer[i]); + } + } + + pipe_buffer_unmap(pipe, out_buffer, out_transfer); + + /* Setup the new vertex buffer in the first free slot. */ + for (i = 0; i < PIPE_MAX_ATTRIBS; i++) { + struct pipe_vertex_buffer *vb = &rctx->vertex_buffer[i]; + + if (!vb->buffer) { + pipe_resource_reference(&vb->buffer, out_buffer); + vb->buffer_offset = 0; + vb->max_index = num_verts - 1; + vb->stride = key.output_stride; + rctx->tran.vb_slot = i; + break; + } + } + + /* Save and replace vertex elements. */ + { + struct pipe_vertex_element new_velems[PIPE_MAX_ATTRIBS]; + + rctx->tran.saved_velems = rctx->vertex_elements; + + for (i = 0; i < ve->count; i++) { + if (vb_translated[ve->elements[i].vertex_buffer_index]) { + te = &key.element[tr_elem_index[i]]; + new_velems[i].instance_divisor = ve->elements[i].instance_divisor; + new_velems[i].src_format = te->output_format; + new_velems[i].src_offset = te->output_offset; + new_velems[i].vertex_buffer_index = rctx->tran.vb_slot; + } else { + memcpy(&new_velems[i], &ve->elements[i], + sizeof(struct pipe_vertex_element)); + } + } + + rctx->tran.new_velems = + pipe->create_vertex_elements_state(pipe, ve->count, new_velems); + pipe->bind_vertex_elements_state(pipe, rctx->tran.new_velems); + } + + pipe_resource_reference(&out_buffer, NULL); +} + +void r600_end_vertex_translate(struct r600_pipe_context *rctx) +{ + struct pipe_context *pipe = &rctx->context; + + /* Restore vertex elements. */ + pipe->bind_vertex_elements_state(pipe, rctx->tran.saved_velems); + pipe->delete_vertex_elements_state(pipe, rctx->tran.new_velems); + + /* Delete the now-unused VBO. */ + pipe_resource_reference(&rctx->vertex_buffer[rctx->tran.vb_slot].buffer, + NULL); +} + +void r600_translate_index_buffer(struct r600_pipe_context *r600, + struct pipe_resource **index_buffer, + unsigned *index_size, + unsigned *start, unsigned count) +{ + switch (*index_size) { + case 1: + util_shorten_ubyte_elts(&r600->context, index_buffer, 0, *start, count); + *index_size = 2; + *start = 0; + break; + + case 2: + if (*start % 2 != 0) { + util_rebuild_ushort_elts(&r600->context, index_buffer, 0, *start, count); + *start = 0; + } + break; + + case 4: + break; + } +} diff --git a/src/gallium/drivers/r600/r600d.h b/src/gallium/drivers/r600/r600d.h index a96d2ce26c1..a3cb5b86004 100644 --- a/src/gallium/drivers/r600/r600d.h +++ b/src/gallium/drivers/r600/r600d.h @@ -667,6 +667,9 @@ #define S_02880C_Z_EXPORT_ENABLE(x) (((x) & 0x1) << 0) #define G_02880C_Z_EXPORT_ENABLE(x) (((x) >> 0) & 0x1) #define C_02880C_Z_EXPORT_ENABLE 0xFFFFFFFE +#define S_02880C_STENCIL_REF_EXPORT_ENABLE(x) (((x) & 0x1) << 1) +#define G_02880C_STENCIL_REF_EXPORT_ENABLE(x) (((x) >> 1) & 0x1) +#define C_02880C_STENCIL_REF_EXPORT_ENABLE 0xFFFFFFFD #define S_02880C_Z_ORDER(x) (((x) & 0x3) << 4) #define G_02880C_Z_ORDER(x) (((x) >> 4) & 0x3) #define C_02880C_Z_ORDER 0xFFFFFCFF @@ -901,6 +904,10 @@ #define S_038000_TILE_MODE(x) (((x) & 0xF) << 3) #define G_038000_TILE_MODE(x) (((x) >> 3) & 0xF) #define C_038000_TILE_MODE 0xFFFFFF87 +#define V_038000_ARRAY_LINEAR_GENERAL 0x00000000 +#define V_038000_ARRAY_LINEAR_ALIGNED 0x00000001 +#define V_038000_ARRAY_1D_TILED_THIN1 0x00000002 +#define V_038000_ARRAY_2D_TILED_THIN1 0x00000004 #define S_038000_TILE_TYPE(x) (((x) & 0x1) << 7) #define G_038000_TILE_TYPE(x) (((x) >> 7) & 0x1) #define C_038000_TILE_TYPE 0xFFFFFF7F @@ -1025,45 +1032,7 @@ #define S_038008_DATA_FORMAT(x) (((x) & 0x3F) << 20) #define G_038008_DATA_FORMAT(x) (((x) >> 20) & 0x3F) #define C_038008_DATA_FORMAT 0xFC0FFFFF -#define V_038008_FMT_INVALID 0x00000000 -#define V_038008_FMT_8 0x00000001 -#define V_038008_FMT_4_4 0x00000002 -#define V_038008_FMT_3_3_2 0x00000003 -#define V_038008_FMT_16 0x00000005 -#define V_038008_FMT_16_FLOAT 0x00000006 -#define V_038008_FMT_8_8 0x00000007 -#define V_038008_FMT_5_6_5 0x00000008 -#define V_038008_FMT_6_5_5 0x00000009 -#define V_038008_FMT_1_5_5_5 0x0000000A -#define V_038008_FMT_4_4_4_4 0x0000000B -#define V_038008_FMT_5_5_5_1 0x0000000C -#define V_038008_FMT_32 0x0000000D -#define V_038008_FMT_32_FLOAT 0x0000000E -#define V_038008_FMT_16_16 0x0000000F -#define V_038008_FMT_16_16_FLOAT 0x00000010 -#define V_038008_FMT_8_24 0x00000011 -#define V_038008_FMT_8_24_FLOAT 0x00000012 -#define V_038008_FMT_24_8 0x00000013 -#define V_038008_FMT_24_8_FLOAT 0x00000014 -#define V_038008_FMT_10_11_11 0x00000015 -#define V_038008_FMT_10_11_11_FLOAT 0x00000016 -#define V_038008_FMT_11_11_10 0x00000017 -#define V_038008_FMT_11_11_10_FLOAT 0x00000018 -#define V_038008_FMT_2_10_10_10 0x00000019 -#define V_038008_FMT_8_8_8_8 0x0000001A -#define V_038008_FMT_10_10_10_2 0x0000001B -#define V_038008_FMT_X24_8_32_FLOAT 0x0000001C -#define V_038008_FMT_32_32 0x0000001D -#define V_038008_FMT_32_32_FLOAT 0x0000001E -#define V_038008_FMT_16_16_16_16 0x0000001F -#define V_038008_FMT_16_16_16_16_FLOAT 0x00000020 -#define V_038008_FMT_32_32_32_32 0x00000022 -#define V_038008_FMT_32_32_32_32_FLOAT 0x00000023 -#define V_038008_FMT_8_8_8 0x0000002c -#define V_038008_FMT_16_16_16 0x0000002d -#define V_038008_FMT_16_16_16_FLOAT 0x0000002e -#define V_038008_FMT_32_32_32 0x0000002f -#define V_038008_FMT_32_32_32_FLOAT 0x00000030 + #define S_038008_NUM_FORMAT_ALL(x) (((x) & 0x3) << 26) #define G_038008_NUM_FORMAT_ALL(x) (((x) >> 26) & 0x3) #define C_038008_NUM_FORMAT_ALL 0xF3FFFFFF |