diff options
Diffstat (limited to 'src/mesa/drivers/dri/i965')
25 files changed, 193 insertions, 266 deletions
diff --git a/src/mesa/drivers/dri/i965/brw_cc.c b/src/mesa/drivers/dri/i965/brw_cc.c index bad9c4a11ee..d662cf75211 100644 --- a/src/mesa/drivers/dri/i965/brw_cc.c +++ b/src/mesa/drivers/dri/i965/brw_cc.c @@ -37,7 +37,7 @@ #include "macros.h" #include "enums.h" -static int prepare_cc_vp( struct brw_context *brw ) +static void prepare_cc_vp( struct brw_context *brw ) { struct brw_cc_viewport ccv; @@ -48,7 +48,6 @@ static int prepare_cc_vp( struct brw_context *brw ) dri_bo_unreference(brw->cc.vp_bo); brw->cc.vp_bo = brw_cache_data( &brw->cache, BRW_CC_VP, &ccv, NULL, 0 ); - return dri_bufmgr_check_aperture_space(brw->cc.vp_bo); } const struct brw_tracked_state brw_cc_vp = { @@ -266,7 +265,7 @@ cc_unit_create_from_key(struct brw_context *brw, struct brw_cc_unit_key *key) return bo; } -static int prepare_cc_unit( struct brw_context *brw ) +static void prepare_cc_unit( struct brw_context *brw ) { struct brw_cc_unit_key key; @@ -280,7 +279,6 @@ static int prepare_cc_unit( struct brw_context *brw ) if (brw->cc.state_bo == NULL) brw->cc.state_bo = cc_unit_create_from_key(brw, &key); - return dri_bufmgr_check_aperture_space(brw->cc.state_bo); } const struct brw_tracked_state brw_cc_unit = { diff --git a/src/mesa/drivers/dri/i965/brw_clip.c b/src/mesa/drivers/dri/i965/brw_clip.c index 540108e5f42..22981fd2d92 100644 --- a/src/mesa/drivers/dri/i965/brw_clip.c +++ b/src/mesa/drivers/dri/i965/brw_clip.c @@ -131,7 +131,7 @@ static void compile_clip_prog( struct brw_context *brw, /* Calculate interpolants for triangle and line rasterization. */ -static int upload_clip_prog( struct brw_context *brw ) +static void upload_clip_prog(struct brw_context *brw) { GLcontext *ctx = &brw->intel.ctx; struct brw_clip_prog_key key; @@ -242,8 +242,6 @@ static int upload_clip_prog( struct brw_context *brw ) &brw->clip.prog_data); if (brw->clip.prog_bo == NULL) compile_clip_prog( brw, &key ); - - return dri_bufmgr_check_aperture_space(brw->clip.prog_bo); } diff --git a/src/mesa/drivers/dri/i965/brw_clip_state.c b/src/mesa/drivers/dri/i965/brw_clip_state.c index 974cb77cb8c..ae904c6253d 100644 --- a/src/mesa/drivers/dri/i965/brw_clip_state.c +++ b/src/mesa/drivers/dri/i965/brw_clip_state.c @@ -129,10 +129,9 @@ clip_unit_create_from_key(struct brw_context *brw, return bo; } -static int upload_clip_unit( struct brw_context *brw ) +static void upload_clip_unit( struct brw_context *brw ) { struct brw_clip_unit_key key; - int ret = 0; clip_unit_populate_key(brw, &key); @@ -144,9 +143,6 @@ static int upload_clip_unit( struct brw_context *brw ) if (brw->clip.state_bo == NULL) { brw->clip.state_bo = clip_unit_create_from_key(brw, &key); } - - ret = dri_bufmgr_check_aperture_space(brw->clip.state_bo); - return ret; } const struct brw_tracked_state brw_clip_unit = { diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h index 32e05542e0d..330d5714da2 100644 --- a/src/mesa/drivers/dri/i965/brw_context.h +++ b/src/mesa/drivers/dri/i965/brw_context.h @@ -135,6 +135,8 @@ struct brw_context; #define BRW_NEW_METAOPS 0x1000 #define BRW_NEW_FENCE 0x2000 #define BRW_NEW_LOCK 0x4000 +#define BRW_NEW_INDICES 0x8000 +#define BRW_NEW_VERTICES 0x10000 /** * Used for any batch entry with a relocated pointer that will be used * by any 3D rendering. @@ -332,7 +334,7 @@ struct brw_state_pointers { */ struct brw_tracked_state { struct brw_state_flags dirty; - int (*prepare)( struct brw_context *brw ); + void (*prepare)( struct brw_context *brw ); void (*emit)( struct brw_context *brw ); }; @@ -450,9 +452,22 @@ struct brw_context * for changes to this state: */ struct brw_vertex_info info; + unsigned int min_index, max_index; } vb; struct { + /** + * Index buffer for this draw_prims call. + * + * Updates are signaled by BRW_NEW_INDICES. + */ + const struct _mesa_index_buffer *ib; + + dri_bo *bo; + unsigned int offset; + } ib; + + struct { /* Will be allocated on demand if needed. */ struct brw_state_pointers attribs; @@ -641,7 +656,7 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis, /*====================================================================== * brw_state.c */ -int brw_validate_state( struct brw_context *brw ); +void brw_validate_state( struct brw_context *brw ); void brw_init_state( struct brw_context *brw ); void brw_destroy_state( struct brw_context *brw ); diff --git a/src/mesa/drivers/dri/i965/brw_curbe.c b/src/mesa/drivers/dri/i965/brw_curbe.c index 80479416028..0a3600193b2 100644 --- a/src/mesa/drivers/dri/i965/brw_curbe.c +++ b/src/mesa/drivers/dri/i965/brw_curbe.c @@ -46,7 +46,7 @@ /* Partition the CURBE between the various users of constant values: */ -static int calculate_curbe_offsets( struct brw_context *brw ) +static void calculate_curbe_offsets( struct brw_context *brw ) { /* CACHE_NEW_WM_PROG */ GLuint nr_fp_regs = (brw->wm.prog_data->nr_params + 15) / 16; @@ -117,7 +117,6 @@ static int calculate_curbe_offsets( struct brw_context *brw ) brw->state.dirty.brw |= BRW_NEW_CURBE_OFFSETS; } - return 0; } @@ -171,7 +170,7 @@ static GLfloat fixed_plane[6][4] = { * cache mechanism, but maybe would benefit from a comparison against * the current uploaded set of constants. */ -static int prepare_constant_buffer(struct brw_context *brw) +static void prepare_constant_buffer(struct brw_context *brw) { GLcontext *ctx = &brw->intel.ctx; struct brw_vertex_program *vp = (struct brw_vertex_program *)brw->vertex_program; @@ -195,8 +194,8 @@ static int prepare_constant_buffer(struct brw_context *brw) brw->curbe.last_buf = NULL; brw->curbe.last_bufsz = 0; } - - return 0; + + return; } buf = (GLfloat *)malloc(bufsz); @@ -321,9 +320,6 @@ static int prepare_constant_buffer(struct brw_context *brw) * flushes as necessary when doublebuffering of CURBEs isn't * possible. */ - - /* check aperture space for this bo */ - return dri_bufmgr_check_aperture_space(brw->curbe.curbe_bo); } @@ -331,6 +327,13 @@ static void emit_constant_buffer(struct brw_context *brw) { struct intel_context *intel = &brw->intel; GLuint sz = brw->curbe.total_size; + dri_bo *aper_array[] = { + brw->intel.batch->buf, + brw->curbe.curbe_bo, + }; + + if (dri_bufmgr_check_aperture_space(aper_array, ARRAY_SIZE(aper_array))) + intel_batchbuffer_flush(intel->batch); BEGIN_BATCH(2, IGNORE_CLIPRECTS); if (sz == 0) { diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c index f90c5f7b082..d43b52c2c7c 100644 --- a/src/mesa/drivers/dri/i965/brw_draw.c +++ b/src/mesa/drivers/dri/i965/brw_draw.c @@ -83,9 +83,8 @@ static const GLenum reduced_prim[GL_POLYGON+1] = { * programs be immune to the active primitive (ie. cope with all * possibilities). That may not be realistic however. */ -static GLuint brw_set_prim(struct brw_context *brw, GLenum prim, GLboolean *need_flush) +static GLuint brw_set_prim(struct brw_context *brw, GLenum prim) { - int ret; if (INTEL_DEBUG & DEBUG_PRIMS) _mesa_printf("PRIM: %s\n", _mesa_lookup_enum_by_nr(prim)); @@ -106,9 +105,7 @@ static GLuint brw_set_prim(struct brw_context *brw, GLenum prim, GLboolean *need brw->state.dirty.brw |= BRW_NEW_REDUCED_PRIMITIVE; } - ret = brw_validate_state(brw); - if (ret) - *need_flush = GL_TRUE; + brw_validate_state(brw); } return hw_prim[prim]; @@ -131,7 +128,6 @@ static void brw_emit_prim( struct brw_context *brw, { struct brw_3d_primitive prim_packet; - GLboolean need_flush = GL_FALSE; if (INTEL_DEBUG & DEBUG_PRIMS) _mesa_printf("PRIM: %s %d %d\n", _mesa_lookup_enum_by_nr(prim->mode), @@ -140,7 +136,7 @@ static void brw_emit_prim( struct brw_context *brw, prim_packet.header.opcode = CMD_3D_PRIM; prim_packet.header.length = sizeof(prim_packet)/4 - 2; prim_packet.header.pad = 0; - prim_packet.header.topology = brw_set_prim(brw, prim->mode, &need_flush); + prim_packet.header.topology = brw_set_prim(brw, prim->mode); prim_packet.header.indexed = prim->indexed; prim_packet.verts_per_instance = trim(prim->mode, prim->count); @@ -149,12 +145,13 @@ static void brw_emit_prim( struct brw_context *brw, prim_packet.start_instance_location = 0; prim_packet.base_vert_location = 0; + /* Can't wrap here, since we rely on the validated state. */ + brw->no_batch_wrap = GL_TRUE; if (prim_packet.verts_per_instance) { intel_batchbuffer_data( brw->intel.batch, &prim_packet, sizeof(prim_packet), LOOP_CLIPRECTS); } - - assert(need_flush == GL_FALSE); + brw->no_batch_wrap = GL_FALSE; } static void brw_merge_inputs( struct brw_context *brw, @@ -258,10 +255,6 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx, struct brw_context *brw = brw_context(ctx); GLboolean retval = GL_FALSE; GLuint i; - GLuint ib_offset; - dri_bo *ib_bo; - GLboolean force_flush = GL_FALSE; - int ret; if (ctx->NewState) _mesa_update_state( ctx ); @@ -271,7 +264,13 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx, /* Bind all inputs, derive varying and size information: */ brw_merge_inputs( brw, arrays ); - + + brw->ib.ib = ib; + brw->state.dirty.brw |= BRW_NEW_INDICES; + + brw->vb.min_index = min_index; + brw->vb.max_index = max_index; + brw->state.dirty.brw |= BRW_NEW_VERTICES; /* Have to validate state quite late. Will rebuild tnl_program, * which depends on varying information. * @@ -294,29 +293,18 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx, * an upper bound of how much we might emit in a single * brw_try_draw_prims(). */ - flush: - if (force_flush) - brw->no_batch_wrap = GL_FALSE; - if (intel->batch->ptr - intel->batch->map > intel->batch->size * 3 / 4 /* brw_emit_prim may change the cliprect_mode to LOOP_CLIPRECTS */ - || intel->batch->cliprect_mode != LOOP_CLIPRECTS || (force_flush == GL_TRUE)) + || intel->batch->cliprect_mode != LOOP_CLIPRECTS) intel_batchbuffer_flush(intel->batch); - force_flush = GL_FALSE; - brw->no_batch_wrap = GL_TRUE; - /* Set the first primitive early, ahead of validate_state: */ - brw_set_prim(brw, prim[0].mode, &force_flush); + brw_set_prim(brw, prim[0].mode); /* XXX: Need to separate validate and upload of state. */ - ret = brw_validate_state( brw ); - if (ret) { - force_flush = GL_TRUE; - goto flush; - } + brw_validate_state( brw ); /* Various fallback checks: */ @@ -326,31 +314,6 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx, if (check_fallbacks( brw, prim, nr_prims )) goto out; - /* need to account for index buffer and vertex buffer */ - if (ib) { - ret = brw_prepare_indices( brw, ib , &ib_bo, &ib_offset); - if (ret) { - force_flush = GL_TRUE; - goto flush; - } - } - - ret = brw_prepare_vertices( brw, min_index, max_index); - if (ret < 0) - goto out; - - if (ret > 0) { - force_flush = GL_TRUE; - goto flush; - } - - /* Upload index, vertex data: - */ - if (ib) - brw_emit_indices( brw, ib, ib_bo, ib_offset); - - brw_emit_vertices( brw, min_index, max_index); - for (i = 0; i < nr_prims; i++) { brw_emit_prim(brw, &prim[i]); } @@ -359,9 +322,6 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx, } out: - - brw->no_batch_wrap = GL_FALSE; - UNLOCK_HARDWARE(intel); if (!retval) diff --git a/src/mesa/drivers/dri/i965/brw_draw.h b/src/mesa/drivers/dri/i965/brw_draw.h index b3547400d40..2a3e0c1c5bf 100644 --- a/src/mesa/drivers/dri/i965/brw_draw.h +++ b/src/mesa/drivers/dri/i965/brw_draw.h @@ -51,27 +51,4 @@ void brw_draw_destroy( struct brw_context *brw ); void brw_init_current_values(GLcontext *ctx, struct gl_client_array *arrays); - -/* brw_draw_upload.c - */ -int brw_prepare_indices( struct brw_context *brw, - const struct _mesa_index_buffer *index_buffer, - dri_bo **bo_return, - GLuint *offset_return); - -void brw_emit_indices( struct brw_context *brw, - const struct _mesa_index_buffer *index_buffer, - dri_bo *bo, - GLuint offset); - -int brw_prepare_vertices( struct brw_context *brw, - GLuint min_index, - GLuint max_index ); - -void brw_emit_vertices( struct brw_context *brw, - GLuint min_index, - GLuint max_index ); - - - #endif diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c index 0181b06764a..9427131440c 100644 --- a/src/mesa/drivers/dri/i965/brw_draw_upload.c +++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c @@ -302,9 +302,7 @@ copy_array_to_vbo_array( struct brw_context *brw, dri_bo_unmap(element->bo); } -int brw_prepare_vertices( struct brw_context *brw, - GLuint min_index, - GLuint max_index ) +static void brw_prepare_vertices(struct brw_context *brw) { GLcontext *ctx = &brw->intel.ctx; struct intel_context *intel = intel_context(ctx); @@ -312,7 +310,8 @@ int brw_prepare_vertices( struct brw_context *brw, GLuint i; const unsigned char *ptr = NULL; GLuint interleave = 0; - int ret = 0; + unsigned int min_index = brw->vb.min_index; + unsigned int max_index = brw->vb.max_index; struct brw_vertex_element *enabled[VERT_ATTRIB_MAX]; GLuint nr_enabled = 0; @@ -340,8 +339,10 @@ int brw_prepare_vertices( struct brw_context *brw, * cases with > 17 vertex attributes enabled, so it probably * isn't an issue at this point. */ - if (nr_enabled >= BRW_VEP_MAX) - return -1; + if (nr_enabled >= BRW_VEP_MAX) { + intel->Fallback = 1; + return; + } for (i = 0; i < nr_enabled; i++) { struct brw_vertex_element *input = enabled[i]; @@ -359,8 +360,6 @@ int brw_prepare_vertices( struct brw_context *brw, dri_bo_reference(input->bo); input->offset = (unsigned long)input->glarray->Ptr; input->stride = input->glarray->StrideB; - - ret |= dri_bufmgr_check_aperture_space(input->bo); } else { /* Queue the buffer object up to be uploaded in the next pass, * when we've decided if we're doing interleaved or not. @@ -369,7 +368,7 @@ int brw_prepare_vertices( struct brw_context *brw, /* Position array not properly enabled: */ if (input->glarray->StrideB == 0) - return -1; + return; interleave = input->glarray->StrideB; ptr = input->glarray->Ptr; @@ -401,7 +400,6 @@ int brw_prepare_vertices( struct brw_context *brw, */ copy_array_to_vbo_array(brw, upload[0], interleave); - ret |= dri_bufmgr_check_aperture_space(upload[0]->bo); for (i = 1; i < nr_uploads; i++) { /* Then, just point upload[i] at upload[0]'s buffer. */ upload[i]->stride = interleave; @@ -415,23 +413,11 @@ int brw_prepare_vertices( struct brw_context *brw, /* Upload non-interleaved arrays */ for (i = 0; i < nr_uploads; i++) { copy_array_to_vbo_array(brw, upload[i], upload[i]->element_size); - if (upload[i]->bo) { - ret |= dri_bufmgr_check_aperture_space(upload[i]->bo); - } } } - - - if (ret) - return 1; - - - return 0; } -void brw_emit_vertices( struct brw_context *brw, - GLuint min_index, - GLuint max_index ) +static void brw_emit_vertices(struct brw_context *brw) { GLcontext *ctx = &brw->intel.ctx; struct intel_context *intel = intel_context(ctx); @@ -469,7 +455,7 @@ void brw_emit_vertices( struct brw_context *brw, OUT_RELOC(input->bo, I915_GEM_DOMAIN_VERTEX, 0, input->offset); - OUT_BATCH(max_index); + OUT_BATCH(brw->vb.max_index); OUT_BATCH(0); /* Instance data step rate */ /* Unreference the buffer so it can get freed, now that we won't @@ -513,18 +499,31 @@ void brw_emit_vertices( struct brw_context *brw, ADVANCE_BATCH(); } -int brw_prepare_indices( struct brw_context *brw, - const struct _mesa_index_buffer *index_buffer, - dri_bo **bo_return, - GLuint *offset_return) +const struct brw_tracked_state brw_vertices = { + .dirty = { + .mesa = 0, + .brw = BRW_NEW_BATCH | BRW_NEW_VERTICES, + .cache = 0, + }, + .prepare = brw_prepare_vertices, + .emit = brw_emit_vertices, +}; + +static void brw_prepare_indices(struct brw_context *brw) { GLcontext *ctx = &brw->intel.ctx; struct intel_context *intel = &brw->intel; - GLuint ib_size = get_size(index_buffer->type) * index_buffer->count; + const struct _mesa_index_buffer *index_buffer = brw->ib.ib; + GLuint ib_size; dri_bo *bo; - struct gl_buffer_object *bufferobj = index_buffer->obj; - GLuint offset = (GLuint)index_buffer->ptr; - int ret; + struct gl_buffer_object *bufferobj; + GLuint offset; + + if (index_buffer == NULL) + return; + + ib_size = get_size(index_buffer->type) * index_buffer->count; + bufferobj = index_buffer->obj;; /* Turn into a proper VBO: */ @@ -538,6 +537,8 @@ int brw_prepare_indices( struct brw_context *brw, */ dri_bo_subdata(bo, offset, ib_size, index_buffer->ptr); } else { + offset = (GLuint)index_buffer->ptr; + /* If the index buffer isn't aligned to its element size, we have to * rebase it into a temporary. */ @@ -560,19 +561,22 @@ int brw_prepare_indices( struct brw_context *brw, } } - *bo_return = bo; - *offset_return = offset; - ret = dri_bufmgr_check_aperture_space(bo); - return ret; + dri_bo_unreference(brw->ib.bo); + brw->ib.bo = bo; + brw->ib.offset = offset; } -void brw_emit_indices(struct brw_context *brw, - const struct _mesa_index_buffer *index_buffer, - dri_bo *bo, - GLuint offset) +static void brw_emit_indices(struct brw_context *brw) { struct intel_context *intel = &brw->intel; - GLuint ib_size = get_size(index_buffer->type) * index_buffer->count; + const struct _mesa_index_buffer *index_buffer = brw->ib.ib; + GLuint ib_size; + + if (index_buffer == NULL) + return; + + ib_size = get_size(index_buffer->type) * index_buffer->count; + /* Emit the indexbuffer packet: */ { @@ -588,16 +592,23 @@ void brw_emit_indices(struct brw_context *brw, BEGIN_BATCH(4, IGNORE_CLIPRECTS); OUT_BATCH( ib.header.dword ); - OUT_RELOC( bo, - I915_GEM_DOMAIN_VERTEX, 0, - offset); - OUT_RELOC( bo, - I915_GEM_DOMAIN_VERTEX, 0, - offset + ib_size); + OUT_RELOC(brw->ib.bo, + I915_GEM_DOMAIN_VERTEX, 0, + brw->ib.offset); + OUT_RELOC(brw->ib.bo, + I915_GEM_DOMAIN_VERTEX, 0, + brw->ib.offset + ib_size); OUT_BATCH( 0 ); ADVANCE_BATCH(); - - dri_bo_unreference(bo); } } +const struct brw_tracked_state brw_indices = { + .dirty = { + .mesa = 0, + .brw = BRW_NEW_BATCH | BRW_NEW_INDICES, + .cache = 0, + }, + .prepare = brw_prepare_indices, + .emit = brw_emit_indices, +}; diff --git a/src/mesa/drivers/dri/i965/brw_fallback.c b/src/mesa/drivers/dri/i965/brw_fallback.c index 8a8fb50cb99..693f68f32ac 100644 --- a/src/mesa/drivers/dri/i965/brw_fallback.c +++ b/src/mesa/drivers/dri/i965/brw_fallback.c @@ -95,10 +95,9 @@ static GLboolean do_check_fallback(struct brw_context *brw) return GL_FALSE; } -static int check_fallback(struct brw_context *brw) +static void check_fallback(struct brw_context *brw) { brw->intel.Fallback = do_check_fallback(brw); - return 0; } const struct brw_tracked_state brw_check_fallback = { diff --git a/src/mesa/drivers/dri/i965/brw_gs.c b/src/mesa/drivers/dri/i965/brw_gs.c index 9419315c7a3..2daef0093b5 100644 --- a/src/mesa/drivers/dri/i965/brw_gs.c +++ b/src/mesa/drivers/dri/i965/brw_gs.c @@ -162,10 +162,9 @@ static void populate_key( struct brw_context *brw, /* Calculate interpolants for triangle and line rasterization. */ -static int prepare_gs_prog( struct brw_context *brw ) +static void prepare_gs_prog(struct brw_context *brw) { struct brw_gs_prog_key key; - int ret = 0; /* Populate the key: */ populate_key(brw, &key); @@ -183,11 +182,7 @@ static int prepare_gs_prog( struct brw_context *brw ) &brw->gs.prog_data); if (brw->gs.prog_bo == NULL) compile_gs_prog( brw, &key ); - - ret |= dri_bufmgr_check_aperture_space(brw->gs.prog_bo); } - - return ret; } diff --git a/src/mesa/drivers/dri/i965/brw_gs_state.c b/src/mesa/drivers/dri/i965/brw_gs_state.c index ae6b48a5178..ff2e3ab0598 100644 --- a/src/mesa/drivers/dri/i965/brw_gs_state.c +++ b/src/mesa/drivers/dri/i965/brw_gs_state.c @@ -116,7 +116,7 @@ gs_unit_create_from_key(struct brw_context *brw, struct brw_gs_unit_key *key) return bo; } -static int prepare_gs_unit( struct brw_context *brw ) +static void prepare_gs_unit(struct brw_context *brw) { struct brw_gs_unit_key key; @@ -130,7 +130,6 @@ static int prepare_gs_unit( struct brw_context *brw ) if (brw->gs.state_bo == NULL) { brw->gs.state_bo = gs_unit_create_from_key(brw, &key); } - return dri_bufmgr_check_aperture_space(brw->gs.state_bo); } const struct brw_tracked_state brw_gs_unit = { diff --git a/src/mesa/drivers/dri/i965/brw_misc_state.c b/src/mesa/drivers/dri/i965/brw_misc_state.c index 9634d649dda..487c638ce21 100644 --- a/src/mesa/drivers/dri/i965/brw_misc_state.c +++ b/src/mesa/drivers/dri/i965/brw_misc_state.c @@ -81,6 +81,13 @@ const struct brw_tracked_state brw_blend_constant_color = { static void upload_binding_table_pointers(struct brw_context *brw) { struct intel_context *intel = &brw->intel; + dri_bo *aper_array[] = { + intel->batch->buf, + brw->wm.bind_bo, + }; + + if (dri_bufmgr_check_aperture_space(aper_array, ARRAY_SIZE(aper_array))) + intel_batchbuffer_flush(intel->batch); BEGIN_BATCH(6, IGNORE_CLIPRECTS); OUT_BATCH(CMD_BINDING_TABLE_PTRS << 16 | (6 - 2)); @@ -135,6 +142,19 @@ static void upload_pipelined_state_pointers(struct brw_context *brw ) static void upload_psp_urb_cbs(struct brw_context *brw ) { + struct intel_context *intel = &brw->intel; + dri_bo *aper_array[] = { + intel->batch->buf, + brw->vs.state_bo, + brw->gs.state_bo, + brw->clip.state_bo, + brw->wm.state_bo, + brw->cc.state_bo, + }; + + if (dri_bufmgr_check_aperture_space(aper_array, ARRAY_SIZE(aper_array))) + intel_batchbuffer_flush(intel->batch); + upload_pipelined_state_pointers(brw); brw_upload_urb_fence(brw); brw_upload_constant_buffer_state(brw); @@ -155,22 +175,6 @@ const struct brw_tracked_state brw_psp_urb_cbs = { .emit = upload_psp_urb_cbs, }; -/** - * Upload the depthbuffer offset and format. - * - * We have to do this per state validation as we need to emit the relocation - * in the batch buffer. - */ - -static int prepare_depthbuffer(struct brw_context *brw) -{ - struct intel_region *region = brw->state.depth_region; - - if (!region || !region->buffer) - return 0; - return dri_bufmgr_check_aperture_space(region->buffer); -} - static void emit_depthbuffer(struct brw_context *brw) { struct intel_context *intel = &brw->intel; @@ -192,6 +196,10 @@ static void emit_depthbuffer(struct brw_context *brw) ADVANCE_BATCH(); } else { unsigned int format; + dri_bo *aper_array[] = { + intel->batch->buf, + region->buffer + }; switch (region->cpp) { case 2: @@ -208,6 +216,9 @@ static void emit_depthbuffer(struct brw_context *brw) return; } + if (dri_bufmgr_check_aperture_space(aper_array, ARRAY_SIZE(aper_array))) + intel_batchbuffer_flush(intel->batch); + BEGIN_BATCH(len, IGNORE_CLIPRECTS); OUT_BATCH(CMD_DEPTH_BUFFER << 16 | (len - 2)); OUT_BATCH(((region->pitch * region->cpp) - 1) | @@ -236,7 +247,6 @@ const struct brw_tracked_state brw_depthbuffer = { .brw = BRW_NEW_DEPTH_BUFFER | BRW_NEW_BATCH, .cache = 0, }, - .prepare = prepare_depthbuffer, .emit = emit_depthbuffer, }; diff --git a/src/mesa/drivers/dri/i965/brw_sf.c b/src/mesa/drivers/dri/i965/brw_sf.c index 0b617483219..5bb9e11310b 100644 --- a/src/mesa/drivers/dri/i965/brw_sf.c +++ b/src/mesa/drivers/dri/i965/brw_sf.c @@ -125,7 +125,7 @@ static void compile_sf_prog( struct brw_context *brw, /* Calculate interpolants for triangle and line rasterization. */ -static int upload_sf_prog( struct brw_context *brw ) +static void upload_sf_prog(struct brw_context *brw) { struct brw_sf_prog_key key; @@ -174,7 +174,6 @@ static int upload_sf_prog( struct brw_context *brw ) &brw->sf.prog_data); if (brw->sf.prog_bo == NULL) compile_sf_prog( brw, &key ); - return dri_bufmgr_check_aperture_space(brw->sf.prog_bo); } diff --git a/src/mesa/drivers/dri/i965/brw_sf_state.c b/src/mesa/drivers/dri/i965/brw_sf_state.c index cbed301d314..2478872b827 100644 --- a/src/mesa/drivers/dri/i965/brw_sf_state.c +++ b/src/mesa/drivers/dri/i965/brw_sf_state.c @@ -37,7 +37,7 @@ #include "macros.h" #include "intel_fbo.h" -static int upload_sf_vp(struct brw_context *brw) +static void upload_sf_vp(struct brw_context *brw) { GLcontext *ctx = &brw->intel.ctx; const GLfloat depth_scale = 1.0F / ctx->DrawBuffer->_DepthMaxF; @@ -98,8 +98,6 @@ static int upload_sf_vp(struct brw_context *brw) dri_bo_unreference(brw->sf.vp_bo); brw->sf.vp_bo = brw_cache_data( &brw->cache, BRW_SF_VP, &sfv, NULL, 0 ); - - return dri_bufmgr_check_aperture_space(brw->sf.vp_bo); } const struct brw_tracked_state brw_sf_vp = { @@ -269,11 +267,10 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key, return bo; } -static int upload_sf_unit( struct brw_context *brw ) +static void upload_sf_unit( struct brw_context *brw ) { struct brw_sf_unit_key key; dri_bo *reloc_bufs[2]; - int ret = 0; sf_unit_populate_key(brw, &key); @@ -288,15 +285,6 @@ static int upload_sf_unit( struct brw_context *brw ) if (brw->sf.state_bo == NULL) { brw->sf.state_bo = sf_unit_create_from_key(brw, &key, reloc_bufs); } - - if (reloc_bufs[0]) - ret |= dri_bufmgr_check_aperture_space(reloc_bufs[0]); - - if (reloc_bufs[1]) - ret |= dri_bufmgr_check_aperture_space(reloc_bufs[1]); - - ret |= dri_bufmgr_check_aperture_space(brw->sf.state_bo); - return ret; } const struct brw_tracked_state brw_sf_unit = { diff --git a/src/mesa/drivers/dri/i965/brw_state.h b/src/mesa/drivers/dri/i965/brw_state.h index d1fca051ecc..3ea6151ae95 100644 --- a/src/mesa/drivers/dri/i965/brw_state.h +++ b/src/mesa/drivers/dri/i965/brw_state.h @@ -80,6 +80,9 @@ const struct brw_tracked_state brw_pipe_control; const struct brw_tracked_state brw_clear_surface_cache; const struct brw_tracked_state brw_clear_batch_cache; +const struct brw_tracked_state brw_indices; +const struct brw_tracked_state brw_vertices; + /*********************************************************************** * brw_state_cache.c */ diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c index b8dfcf5b031..d1d319d92e6 100644 --- a/src/mesa/drivers/dri/i965/brw_state_upload.c +++ b/src/mesa/drivers/dri/i965/brw_state_upload.c @@ -101,6 +101,8 @@ const struct brw_tracked_state *atoms[] = &brw_psp_urb_cbs, #endif + &brw_indices, + &brw_vertices, NULL, /* brw_constant_buffer */ }; @@ -172,10 +174,12 @@ static void xor_states( struct brw_state_flags *result, /*********************************************************************** * Emit all state: */ -int brw_validate_state( struct brw_context *brw ) +void brw_validate_state( struct brw_context *brw ) { + struct intel_context *intel = &brw->intel; struct brw_state_flags *state = &brw->state.dirty; - GLuint i, ret, count; + GLuint i, count, pass = 0; + dri_bo *last_batch_bo = NULL; state->mesa |= brw->intel.NewGLState; brw->intel.NewGLState = 0; @@ -201,7 +205,7 @@ int brw_validate_state( struct brw_context *brw ) if (state->mesa == 0 && state->cache == 0 && state->brw == 0) - return 0; + return; if (brw->state.dirty.brw & BRW_NEW_CONTEXT) brw_clear_batch_cache_flush(brw); @@ -219,15 +223,23 @@ int brw_validate_state( struct brw_context *brw ) if (check_state(state, &atom->dirty)) { if (atom->prepare) { - ret = atom->prepare(brw); - if (ret) - return ret; + atom->prepare(brw); } } } if (brw->intel.Fallback) - return 0; + return; + + /* We're about to try to set up a coherent state in the batchbuffer for + * the emission of primitives. If we exceed the aperture size in any of the + * emit() calls, we need to go back to square 1 and try setting up again. + */ +got_flushed: + dri_bo_unreference(last_batch_bo); + last_batch_bo = intel->batch->buf; + dri_bo_reference(last_batch_bo); + assert(pass++ <= 2); if (INTEL_DEBUG) { /* Debug version which enforces various sanity checks on the @@ -250,8 +262,11 @@ int brw_validate_state( struct brw_context *brw ) break; if (check_state(state, &atom->dirty)) { - if (atom->emit) + if (atom->emit) { atom->emit( brw ); + if (intel->batch->buf != last_batch_bo) + goto got_flushed; + } } accumulate_state(&examined, &atom->dirty); @@ -273,13 +288,17 @@ int brw_validate_state( struct brw_context *brw ) break; if (check_state(state, &atom->dirty)) { - if (atom->emit) + if (atom->emit) { atom->emit( brw ); + if (intel->batch->buf != last_batch_bo) + goto got_flushed; + } } } } + dri_bo_unreference(last_batch_bo); + if (!brw->intel.Fallback) memset(state, 0, sizeof(*state)); - return 0; } diff --git a/src/mesa/drivers/dri/i965/brw_urb.c b/src/mesa/drivers/dri/i965/brw_urb.c index 244c82169ae..1116ade0a47 100644 --- a/src/mesa/drivers/dri/i965/brw_urb.c +++ b/src/mesa/drivers/dri/i965/brw_urb.c @@ -74,7 +74,7 @@ static GLboolean check_urb_layout( struct brw_context *brw ) /* Most minimal update, forces re-emit of URB fence packet after GS * unit turned on/off. */ -static int recalculate_urb_fence( struct brw_context *brw ) +static void recalculate_urb_fence( struct brw_context *brw ) { GLuint csize = brw->curbe.total_size; GLuint vsize = brw->vs.prog_data->urb_entry_size; @@ -142,7 +142,6 @@ static int recalculate_urb_fence( struct brw_context *brw ) brw->state.dirty.brw |= BRW_NEW_URB_FENCE; } - return 0; } diff --git a/src/mesa/drivers/dri/i965/brw_vs.c b/src/mesa/drivers/dri/i965/brw_vs.c index f89b0e14a12..1db7ceebcfb 100644 --- a/src/mesa/drivers/dri/i965/brw_vs.c +++ b/src/mesa/drivers/dri/i965/brw_vs.c @@ -83,7 +83,7 @@ static void do_vs_prog( struct brw_context *brw, } -static int brw_upload_vs_prog( struct brw_context *brw ) +static void brw_upload_vs_prog(struct brw_context *brw) { struct brw_vs_prog_key key; struct brw_vertex_program *vp = @@ -115,7 +115,6 @@ static int brw_upload_vs_prog( struct brw_context *brw ) &brw->vs.prog_data); if (brw->vs.prog_bo == NULL) do_vs_prog(brw, vp, &key); - return dri_bufmgr_check_aperture_space(brw->vs.prog_bo); } diff --git a/src/mesa/drivers/dri/i965/brw_vs_constval.c b/src/mesa/drivers/dri/i965/brw_vs_constval.c index a0106b8975d..734a926e968 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_constval.c +++ b/src/mesa/drivers/dri/i965/brw_vs_constval.c @@ -166,7 +166,7 @@ static GLuint get_input_size(struct brw_context *brw, /* Calculate sizes of vertex program outputs. Size is the largest * component index which might vary from [0,0,0,1] */ -static int calc_wm_input_sizes( struct brw_context *brw ) +static void calc_wm_input_sizes( struct brw_context *brw ) { /* BRW_NEW_VERTEX_PROGRAM */ struct brw_vertex_program *vp = @@ -210,7 +210,6 @@ static int calc_wm_input_sizes( struct brw_context *brw ) memcpy(brw->wm.input_size_masks, t.size_masks, sizeof(t.size_masks)); brw->state.dirty.brw |= BRW_NEW_WM_INPUT_DIMENSIONS; } - return 0; } const struct brw_tracked_state brw_wm_input_sizes = { diff --git a/src/mesa/drivers/dri/i965/brw_vs_state.c b/src/mesa/drivers/dri/i965/brw_vs_state.c index e18cd42f4ea..909b942610d 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_state.c +++ b/src/mesa/drivers/dri/i965/brw_vs_state.c @@ -124,7 +124,7 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key) return bo; } -static int prepare_vs_unit( struct brw_context *brw ) +static void prepare_vs_unit(struct brw_context *brw) { struct brw_vs_unit_key key; @@ -138,7 +138,6 @@ static int prepare_vs_unit( struct brw_context *brw ) if (brw->vs.state_bo == NULL) { brw->vs.state_bo = vs_unit_create_from_key(brw, &key); } - return dri_bufmgr_check_aperture_space(brw->vs.state_bo); } const struct brw_tracked_state brw_vs_unit = { diff --git a/src/mesa/drivers/dri/i965/brw_vs_tnl.c b/src/mesa/drivers/dri/i965/brw_vs_tnl.c index e409620bbf2..2caa020e611 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_tnl.c +++ b/src/mesa/drivers/dri/i965/brw_vs_tnl.c @@ -1581,7 +1581,7 @@ static GLuint hash_key( struct state_key *key ) return hash; } -static int prepare_tnl_program( struct brw_context *brw ) +static void prepare_tnl_program( struct brw_context *brw ) { GLcontext *ctx = &brw->intel.ctx; struct state_key key; @@ -1590,7 +1590,7 @@ static int prepare_tnl_program( struct brw_context *brw ) /* _NEW_PROGRAM */ if (brw->attribs.VertexProgram->_Current) - return 0; + return; /* Grab all the relevent state and put it in a single structure: */ @@ -1623,7 +1623,7 @@ static int prepare_tnl_program( struct brw_context *brw ) if (old != brw->tnl_program) brw->state.dirty.brw |= BRW_NEW_TNL_PROGRAM; - return 0; + return; } /* Note: See brw_draw.c - the vertex program must not rely on @@ -1649,7 +1649,7 @@ const struct brw_tracked_state brw_tnl_vertprog = { -static int prepare_active_vertprog( struct brw_context *brw ) +static void prepare_active_vertprog( struct brw_context *brw ) { const struct gl_vertex_program *prev = brw->vertex_program; @@ -1664,8 +1664,6 @@ static int prepare_active_vertprog( struct brw_context *brw ) if (brw->vertex_program != prev) brw->state.dirty.brw |= BRW_NEW_VERTEX_PROGRAM; - - return 0; } diff --git a/src/mesa/drivers/dri/i965/brw_wm.c b/src/mesa/drivers/dri/i965/brw_wm.c index acbaf178d4a..93ae8dc693b 100644 --- a/src/mesa/drivers/dri/i965/brw_wm.c +++ b/src/mesa/drivers/dri/i965/brw_wm.c @@ -325,7 +325,7 @@ static void brw_wm_populate_key( struct brw_context *brw, } -static int brw_prepare_wm_prog( struct brw_context *brw ) +static void brw_prepare_wm_prog(struct brw_context *brw) { struct brw_wm_prog_key key; struct brw_fragment_program *fp = (struct brw_fragment_program *) @@ -342,8 +342,6 @@ static int brw_prepare_wm_prog( struct brw_context *brw ) &brw->wm.prog_data); if (brw->wm.prog_bo == NULL) do_wm_prog(brw, fp, &key); - - return dri_bufmgr_check_aperture_space(brw->wm.prog_bo); } diff --git a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c index 461f977aac7..08d01823dee 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c @@ -255,11 +255,10 @@ brw_wm_sampler_populate_key(struct brw_context *brw, * complicates various things. However, this is still too confusing - * FIXME: simplify all the different new texture state flags. */ -static int upload_wm_samplers( struct brw_context *brw ) +static void upload_wm_samplers( struct brw_context *brw ) { struct wm_sampler_key key; int i; - int ret = 0; brw_wm_sampler_populate_key(brw, &key); @@ -271,7 +270,7 @@ static int upload_wm_samplers( struct brw_context *brw ) dri_bo_unreference(brw->wm.sampler_bo); brw->wm.sampler_bo = NULL; if (brw->wm.sampler_count == 0) - return 0; + return; brw->wm.sampler_bo = brw_search_cache(&brw->cache, BRW_SAMPLER, &key, sizeof(key), @@ -304,7 +303,6 @@ static int upload_wm_samplers( struct brw_context *brw ) if (!brw->attribs.Texture->Unit[i]._ReallyEnabled) continue; - ret |= dri_bufmgr_check_aperture_space(brw->wm.sdc_bo[i]); intel_bo_emit_reloc(brw->wm.sampler_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0, @@ -313,10 +311,6 @@ static int upload_wm_samplers( struct brw_context *brw ) brw->wm.sdc_bo[i]); } } - - ret |= dri_bufmgr_check_aperture_space(brw->wm.sampler_bo); - return ret; - } const struct brw_tracked_state brw_wm_samplers = { diff --git a/src/mesa/drivers/dri/i965/brw_wm_state.c b/src/mesa/drivers/dri/i965/brw_wm_state.c index 6fe30f0a9ac..f97d0dc2854 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_state.c @@ -227,12 +227,11 @@ wm_unit_create_from_key(struct brw_context *brw, struct brw_wm_unit_key *key, } -static int upload_wm_unit( struct brw_context *brw ) +static void upload_wm_unit( struct brw_context *brw ) { struct intel_context *intel = &brw->intel; struct brw_wm_unit_key key; dri_bo *reloc_bufs[3]; - int ret = 0, i; wm_unit_populate_key(brw, &key); /* Allocate the necessary scratch space if we haven't already. Don't @@ -267,12 +266,6 @@ static int upload_wm_unit( struct brw_context *brw ) if (brw->wm.state_bo == NULL) { brw->wm.state_bo = wm_unit_create_from_key(brw, &key, reloc_bufs); } - - for (i = 0; i < 3; i++) - if (reloc_bufs[i]) - ret |= dri_bufmgr_check_aperture_space(reloc_bufs[i]); - ret |= dri_bufmgr_check_aperture_space(brw->wm.state_bo); - return ret; } const struct brw_tracked_state brw_wm_unit = { diff --git a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c index 761a5df33f9..0f5ba46a19a 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c @@ -229,7 +229,7 @@ brw_create_texture_surface( struct brw_context *brw, return bo; } -static int +static void brw_update_texture_surface( GLcontext *ctx, GLuint unit ) { struct brw_context *brw = brw_context(ctx); @@ -237,7 +237,6 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit ) struct intel_texture_object *intelObj = intel_texture_object(tObj); struct gl_texture_image *firstImage = tObj->Image[0][intelObj->firstLevel]; struct brw_wm_surface_key key; - int ret = 0; memset(&key, 0, sizeof(key)); key.target = tObj->Target; @@ -253,8 +252,6 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit ) key.depth = firstImage->Depth; key.tiling = intelObj->mt->region->tiling; - ret |= dri_bufmgr_check_aperture_space(key.bo); - dri_bo_unreference(brw->wm.surf_bo[unit + MAX_DRAW_BUFFERS]); brw->wm.surf_bo[unit + MAX_DRAW_BUFFERS] = brw_search_cache(&brw->cache, BRW_SS_SURFACE, &key, sizeof(key), @@ -263,9 +260,6 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit ) if (brw->wm.surf_bo[unit + MAX_DRAW_BUFFERS] == NULL) { brw->wm.surf_bo[unit + MAX_DRAW_BUFFERS] = brw_create_texture_surface(brw, &key); } - - ret |= dri_bufmgr_check_aperture_space(brw->wm.surf_bo[unit + MAX_DRAW_BUFFERS]); - return ret; } /** @@ -273,12 +267,11 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit ) * While it is only used for the front/back buffer currently, it should be * usable for further buffers when doing ARB_draw_buffer support. */ -static int +static void brw_update_region_surface(struct brw_context *brw, struct intel_region *region, unsigned int unit, GLboolean cached) { dri_bo *region_bo = NULL; - int ret = 0; struct { unsigned int surface_type; unsigned int surface_format; @@ -302,8 +295,6 @@ brw_update_region_surface(struct brw_context *brw, struct intel_region *region, key.width = region->pitch; /* XXX: not really! */ key.height = region->height; key.cpp = region->cpp; - - ret |= dri_bufmgr_check_aperture_space(region->buffer); } else { key.surface_type = BRW_SURFACE_NULL; key.surface_format = BRW_SURFACEFORMAT_B8G8R8A8_UNORM; @@ -367,10 +358,6 @@ brw_update_region_surface(struct brw_context *brw, struct intel_region *region, region_bo); } } - - ret |= dri_bufmgr_check_aperture_space(brw->wm.surf_bo[unit]); - - return ret; } @@ -422,23 +409,19 @@ brw_wm_get_binding_table(struct brw_context *brw) return bind_bo; } -static int prepare_wm_surfaces(struct brw_context *brw ) +static void prepare_wm_surfaces(struct brw_context *brw ) { GLcontext *ctx = &brw->intel.ctx; struct intel_context *intel = &brw->intel; - GLuint i, ret; + GLuint i; if (brw->state.nr_draw_regions > 1) { for (i = 0; i < brw->state.nr_draw_regions; i++) { - ret = brw_update_region_surface(brw, brw->state.draw_regions[i], i, - GL_FALSE); - if (ret) - return ret; + brw_update_region_surface(brw, brw->state.draw_regions[i], i, + GL_FALSE); } }else { - ret = brw_update_region_surface(brw, brw->state.draw_regions[0], 0, GL_TRUE); - if (ret) - return ret; + brw_update_region_surface(brw, brw->state.draw_regions[0], 0, GL_TRUE); } brw->wm.nr_surfaces = MAX_DRAW_BUFFERS; @@ -454,11 +437,8 @@ static int prepare_wm_surfaces(struct brw_context *brw ) dri_bo_reference(brw->wm.surf_bo[i+MAX_DRAW_BUFFERS]); brw->wm.nr_surfaces = i + MAX_DRAW_BUFFERS + 1; } else { - ret = brw_update_texture_surface(ctx, i); + brw_update_texture_surface(ctx, i); brw->wm.nr_surfaces = i + MAX_DRAW_BUFFERS + 1; - - if (ret) - return ret; } } else { dri_bo_unreference(brw->wm.surf_bo[i+MAX_DRAW_BUFFERS]); @@ -469,8 +449,6 @@ static int prepare_wm_surfaces(struct brw_context *brw ) dri_bo_unreference(brw->wm.bind_bo); brw->wm.bind_bo = brw_wm_get_binding_table(brw); - - return dri_bufmgr_check_aperture_space(brw->wm.bind_bo); } |