From ab12e764ba3f57ad9f0d7252262cfc6e07839928 Mon Sep 17 00:00:00 2001 From: Roland Scheidegger Date: Wed, 11 Nov 2009 17:57:56 -0800 Subject: i965: fix EXT_provoking_vertex support This didn't work for quad/quadstrips at all, and for all other primitive types it only worked when they were unclipped. Fix up the former in gs stage (could probably do without these changes and instead set QuadsFollowProvokingVertexConvention to false), and the rest in clip stage. --- src/mesa/drivers/dri/i965/brw_gs_emit.c | 38 +++++++++++++++++++++++---------- 1 file changed, 27 insertions(+), 11 deletions(-) (limited to 'src/mesa/drivers/dri/i965/brw_gs_emit.c') diff --git a/src/mesa/drivers/dri/i965/brw_gs_emit.c b/src/mesa/drivers/dri/i965/brw_gs_emit.c index a9b2aa2eac..0fc5b02c61 100644 --- a/src/mesa/drivers/dri/i965/brw_gs_emit.c +++ b/src/mesa/drivers/dri/i965/brw_gs_emit.c @@ -120,7 +120,7 @@ static void brw_gs_ff_sync(struct brw_gs_compile *c, int num_prim) } -void brw_gs_quads( struct brw_gs_compile *c ) +void brw_gs_quads( struct brw_gs_compile *c, struct brw_gs_prog_key *key ) { brw_gs_alloc_regs(c, 4); @@ -128,23 +128,39 @@ void brw_gs_quads( struct brw_gs_compile *c ) * is the PV for quads, but vertex 0 for polygons: */ if (c->need_ff_sync) - brw_gs_ff_sync(c, 1); - brw_gs_emit_vue(c, c->reg.vertex[3], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); - brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); - brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_POLYGON << 2)); - brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); + brw_gs_ff_sync(c, 1); + if (key->pv_first) { + brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[2], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[3], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); + } + else { + brw_gs_emit_vue(c, c->reg.vertex[3], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); + } } -void brw_gs_quad_strip( struct brw_gs_compile *c ) +void brw_gs_quad_strip( struct brw_gs_compile *c, struct brw_gs_prog_key *key ) { brw_gs_alloc_regs(c, 4); if (c->need_ff_sync) brw_gs_ff_sync(c, 1); - brw_gs_emit_vue(c, c->reg.vertex[2], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); - brw_gs_emit_vue(c, c->reg.vertex[3], 0, (_3DPRIM_POLYGON << 2)); - brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); - brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); + if (key->pv_first) { + brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[2], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[3], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); + } + else { + brw_gs_emit_vue(c, c->reg.vertex[2], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); + brw_gs_emit_vue(c, c->reg.vertex[3], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); + brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END)); + } } void brw_gs_tris( struct brw_gs_compile *c ) -- cgit v1.2.3 From 1c96e85c9d6b8c636b0636f3320d1057ab5357b3 Mon Sep 17 00:00:00 2001 From: Eric Anholt Date: Wed, 16 Dec 2009 14:45:50 -0800 Subject: intel: Replace IS_IGDNG checks with intel->is_ironlake or needs_ff_sync. Saves ~480 bytes of code. --- src/mesa/drivers/dri/i965/brw_clip.c | 11 ++++++----- src/mesa/drivers/dri/i965/brw_clip.h | 1 - src/mesa/drivers/dri/i965/brw_clip_line.c | 3 ++- src/mesa/drivers/dri/i965/brw_clip_state.c | 3 ++- src/mesa/drivers/dri/i965/brw_clip_tri.c | 5 +++-- src/mesa/drivers/dri/i965/brw_clip_util.c | 8 ++++++-- src/mesa/drivers/dri/i965/brw_draw_upload.c | 4 ++-- src/mesa/drivers/dri/i965/brw_eu_emit.c | 27 ++++++++++++++++++--------- src/mesa/drivers/dri/i965/brw_gs.c | 4 ++-- src/mesa/drivers/dri/i965/brw_gs.h | 1 - src/mesa/drivers/dri/i965/brw_gs_emit.c | 20 +++++++++++++++----- src/mesa/drivers/dri/i965/brw_gs_state.c | 3 ++- src/mesa/drivers/dri/i965/brw_misc_state.c | 8 ++++---- src/mesa/drivers/dri/i965/brw_sf_emit.c | 6 ++++-- src/mesa/drivers/dri/i965/brw_sf_state.c | 9 +++++---- src/mesa/drivers/dri/i965/brw_tex_layout.c | 2 +- src/mesa/drivers/dri/i965/brw_urb.c | 3 ++- src/mesa/drivers/dri/i965/brw_vs_emit.c | 13 ++++++++----- src/mesa/drivers/dri/i965/brw_vs_state.c | 7 ++++--- src/mesa/drivers/dri/i965/brw_wm_emit.c | 15 +++++++++------ src/mesa/drivers/dri/i965/brw_wm_glsl.c | 3 ++- src/mesa/drivers/dri/i965/brw_wm_state.c | 7 ++++--- src/mesa/drivers/dri/intel/intel_context.c | 5 +++++ src/mesa/drivers/dri/intel/intel_context.h | 2 ++ 24 files changed, 108 insertions(+), 62 deletions(-) (limited to 'src/mesa/drivers/dri/i965/brw_gs_emit.c') diff --git a/src/mesa/drivers/dri/i965/brw_clip.c b/src/mesa/drivers/dri/i965/brw_clip.c index dbd10a5297..af1d975de9 100644 --- a/src/mesa/drivers/dri/i965/brw_clip.c +++ b/src/mesa/drivers/dri/i965/brw_clip.c @@ -50,6 +50,7 @@ static void compile_clip_prog( struct brw_context *brw, struct brw_clip_prog_key *key ) { + struct intel_context *intel = &brw->intel; struct brw_clip_compile c; const GLuint *program; GLuint program_size; @@ -65,14 +66,13 @@ static void compile_clip_prog( struct brw_context *brw, c.func.single_program_flow = 1; c.key = *key; - c.need_ff_sync = BRW_IS_IGDNG(brw); /* Need to locate the two positions present in vertex + header. * These are currently hardcoded: */ c.header_position_offset = ATTR_SIZE; - if (BRW_IS_IGDNG(brw)) + if (intel->is_ironlake) delta = 3 * REG_SIZE; else delta = REG_SIZE; @@ -85,7 +85,7 @@ static void compile_clip_prog( struct brw_context *brw, c.nr_attrs = brw_count_bits(c.key.attrs); - if (BRW_IS_IGDNG(brw)) + if (intel->is_ironlake) c.nr_regs = (c.nr_attrs + 1) / 2 + 3; /* are vertices packed, or reg-aligned? */ else c.nr_regs = (c.nr_attrs + 1) / 2 + 1; /* are vertices packed, or reg-aligned? */ @@ -143,7 +143,8 @@ static void compile_clip_prog( struct brw_context *brw, */ static void upload_clip_prog(struct brw_context *brw) { - GLcontext *ctx = &brw->intel.ctx; + struct intel_context *intel = &brw->intel; + GLcontext *ctx = &intel->ctx; struct brw_clip_prog_key key; memset(&key, 0, sizeof(key)); @@ -160,7 +161,7 @@ static void upload_clip_prog(struct brw_context *brw) /* _NEW_TRANSFORM */ key.nr_userclip = brw_count_bits(ctx->Transform.ClipPlanesEnabled); - if (BRW_IS_IGDNG(brw)) + if (intel->is_ironlake) key.clip_mode = BRW_CLIPMODE_KERNEL_CLIP; else key.clip_mode = BRW_CLIPMODE_NORMAL; diff --git a/src/mesa/drivers/dri/i965/brw_clip.h b/src/mesa/drivers/dri/i965/brw_clip.h index 1c6825510a..d71bac7f61 100644 --- a/src/mesa/drivers/dri/i965/brw_clip.h +++ b/src/mesa/drivers/dri/i965/brw_clip.h @@ -118,7 +118,6 @@ struct brw_clip_compile { GLuint header_position_offset; GLuint offset[VERT_ATTRIB_MAX]; - GLboolean need_ff_sync; }; #define ATTR_SIZE (4*4) diff --git a/src/mesa/drivers/dri/i965/brw_clip_line.c b/src/mesa/drivers/dri/i965/brw_clip_line.c index fa9648f50f..59727af88d 100644 --- a/src/mesa/drivers/dri/i965/brw_clip_line.c +++ b/src/mesa/drivers/dri/i965/brw_clip_line.c @@ -46,6 +46,7 @@ static void brw_clip_line_alloc_regs( struct brw_clip_compile *c ) { + struct intel_context *intel = &c->func.brw->intel; GLuint i = 0,j; /* Register usage is static, precompute here: @@ -85,7 +86,7 @@ static void brw_clip_line_alloc_regs( struct brw_clip_compile *c ) i++; } - if (c->need_ff_sync) { + if (intel->needs_ff_sync) { c->reg.ff_sync = retype(brw_vec1_grf(i, 0), BRW_REGISTER_TYPE_UD); i++; } diff --git a/src/mesa/drivers/dri/i965/brw_clip_state.c b/src/mesa/drivers/dri/i965/brw_clip_state.c index 234b3744bf..28ed634631 100644 --- a/src/mesa/drivers/dri/i965/brw_clip_state.c +++ b/src/mesa/drivers/dri/i965/brw_clip_state.c @@ -74,6 +74,7 @@ static dri_bo * clip_unit_create_from_key(struct brw_context *brw, struct brw_clip_unit_key *key) { + struct intel_context *intel = &brw->intel; struct brw_clip_unit_state clip; dri_bo *bo; @@ -105,7 +106,7 @@ clip_unit_create_from_key(struct brw_context *brw, /* Although up to 16 concurrent Clip threads are allowed on IGDNG, * only 2 threads can output VUEs at a time. */ - if (BRW_IS_IGDNG(brw)) + if (intel->is_ironlake) clip.thread4.max_threads = 16 - 1; else clip.thread4.max_threads = 2 - 1; diff --git a/src/mesa/drivers/dri/i965/brw_clip_tri.c b/src/mesa/drivers/dri/i965/brw_clip_tri.c index cf79224be4..ec15e9a697 100644 --- a/src/mesa/drivers/dri/i965/brw_clip_tri.c +++ b/src/mesa/drivers/dri/i965/brw_clip_tri.c @@ -51,6 +51,7 @@ static void release_tmps( struct brw_clip_compile *c ) void brw_clip_tri_alloc_regs( struct brw_clip_compile *c, GLuint nr_verts ) { + struct intel_context *intel = &c->func.brw->intel; GLuint i = 0,j; /* Register usage is static, precompute here: @@ -78,7 +79,7 @@ void brw_clip_tri_alloc_regs( struct brw_clip_compile *c, for (j = 0; j < 3; j++) { GLuint delta = c->nr_attrs*16 + 32; - if (BRW_IS_IGDNG(c->func.brw)) + if (intel->is_ironlake) delta = c->nr_attrs * 16 + 32 * 3; brw_MOV(&c->func, byte_offset(c->reg.vertex[j], delta), brw_imm_f(0)); @@ -119,7 +120,7 @@ void brw_clip_tri_alloc_regs( struct brw_clip_compile *c, i++; } - if (c->need_ff_sync) { + if (intel->needs_ff_sync) { c->reg.ff_sync = retype(brw_vec1_grf(i, 0), BRW_REGISTER_TYPE_UD); i++; } diff --git a/src/mesa/drivers/dri/i965/brw_clip_util.c b/src/mesa/drivers/dri/i965/brw_clip_util.c index 5a73abdfee..091ebfc34c 100644 --- a/src/mesa/drivers/dri/i965/brw_clip_util.c +++ b/src/mesa/drivers/dri/i965/brw_clip_util.c @@ -359,7 +359,9 @@ void brw_clip_init_clipmask( struct brw_clip_compile *c ) void brw_clip_ff_sync(struct brw_clip_compile *c) { - if (c->need_ff_sync) { + struct intel_context *intel = &c->func.brw->intel; + + if (intel->needs_ff_sync) { struct brw_compile *p = &c->func; struct brw_instruction *need_ff_sync; @@ -388,7 +390,9 @@ void brw_clip_ff_sync(struct brw_clip_compile *c) void brw_clip_init_ff_sync(struct brw_clip_compile *c) { - if (c->need_ff_sync) { + struct intel_context *intel = &c->func.brw->intel; + + if (intel->needs_ff_sync) { struct brw_compile *p = &c->func; brw_MOV(p, c->reg.ff_sync, brw_imm_ud(0)); diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c index 7c796dae93..2c9902c90f 100644 --- a/src/mesa/drivers/dri/i965/brw_draw_upload.c +++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c @@ -527,7 +527,7 @@ static void brw_emit_vertices(struct brw_context *brw) OUT_RELOC(input->bo, I915_GEM_DOMAIN_VERTEX, 0, input->offset); - if (BRW_IS_IGDNG(brw)) { + if (intel->is_ironlake) { OUT_RELOC(input->bo, I915_GEM_DOMAIN_VERTEX, 0, input->bo->size - 1); @@ -563,7 +563,7 @@ static void brw_emit_vertices(struct brw_context *brw) (format << BRW_VE0_FORMAT_SHIFT) | (0 << BRW_VE0_SRC_OFFSET_SHIFT)); - if (BRW_IS_IGDNG(brw)) + if (intel->is_ironlake) OUT_BATCH((comp0 << BRW_VE1_COMPONENT_0_SHIFT) | (comp1 << BRW_VE1_COMPONENT_1_SHIFT) | (comp2 << BRW_VE1_COMPONENT_2_SHIFT) | diff --git a/src/mesa/drivers/dri/i965/brw_eu_emit.c b/src/mesa/drivers/dri/i965/brw_eu_emit.c index 7ceabba288..67673e9911 100644 --- a/src/mesa/drivers/dri/i965/brw_eu_emit.c +++ b/src/mesa/drivers/dri/i965/brw_eu_emit.c @@ -252,9 +252,10 @@ static void brw_set_math_message( struct brw_context *brw, GLboolean saturate, GLuint dataType ) { + struct intel_context *intel = &brw->intel; brw_set_src1(insn, brw_imm_d(0)); - if (BRW_IS_IGDNG(brw)) { + if (intel->is_ironlake) { insn->bits3.math_igdng.function = function; insn->bits3.math_igdng.int_type = integer_type; insn->bits3.math_igdng.precision = low_precision; @@ -319,9 +320,10 @@ static void brw_set_urb_message( struct brw_context *brw, GLuint offset, GLuint swizzle_control ) { + struct intel_context *intel = &brw->intel; brw_set_src1(insn, brw_imm_d(0)); - if (BRW_IS_IGDNG(brw)) { + if (intel->is_ironlake) { insn->bits3.urb_igdng.opcode = 0; /* ? */ insn->bits3.urb_igdng.offset = offset; insn->bits3.urb_igdng.swizzle_control = swizzle_control; @@ -358,9 +360,10 @@ static void brw_set_dp_write_message( struct brw_context *brw, GLuint response_length, GLuint end_of_thread ) { + struct intel_context *intel = &brw->intel; brw_set_src1(insn, brw_imm_d(0)); - if (BRW_IS_IGDNG(brw)) { + if (intel->is_ironlake) { insn->bits3.dp_write_igdng.binding_table_index = binding_table_index; insn->bits3.dp_write_igdng.msg_control = msg_control; insn->bits3.dp_write_igdng.pixel_scoreboard_clear = pixel_scoreboard_clear; @@ -395,9 +398,10 @@ static void brw_set_dp_read_message( struct brw_context *brw, GLuint response_length, GLuint end_of_thread ) { + struct intel_context *intel = &brw->intel; brw_set_src1(insn, brw_imm_d(0)); - if (BRW_IS_IGDNG(brw)) { + if (intel->is_ironlake) { insn->bits3.dp_read_igdng.binding_table_index = binding_table_index; insn->bits3.dp_read_igdng.msg_control = msg_control; insn->bits3.dp_read_igdng.msg_type = msg_type; @@ -433,10 +437,11 @@ static void brw_set_sampler_message(struct brw_context *brw, GLuint header_present, GLuint simd_mode) { + struct intel_context *intel = &brw->intel; assert(eot == 0); brw_set_src1(insn, brw_imm_d(0)); - if (BRW_IS_IGDNG(brw)) { + if (intel->is_ironlake) { insn->bits3.sampler_igdng.binding_table_index = binding_table_index; insn->bits3.sampler_igdng.sampler = sampler; insn->bits3.sampler_igdng.msg_type = msg_type; @@ -648,10 +653,11 @@ struct brw_instruction *brw_IF(struct brw_compile *p, GLuint execute_size) struct brw_instruction *brw_ELSE(struct brw_compile *p, struct brw_instruction *if_insn) { + struct intel_context *intel = &p->brw->intel; struct brw_instruction *insn; GLuint br = 1; - if (BRW_IS_IGDNG(p->brw)) + if (intel->is_ironlake) br = 2; if (p->single_program_flow) { @@ -690,9 +696,10 @@ struct brw_instruction *brw_ELSE(struct brw_compile *p, void brw_ENDIF(struct brw_compile *p, struct brw_instruction *patch_insn) { + struct intel_context *intel = &p->brw->intel; GLuint br = 1; - if (BRW_IS_IGDNG(p->brw)) + if (intel->is_ironlake) br = 2; if (p->single_program_flow) { @@ -803,10 +810,11 @@ struct brw_instruction *brw_DO(struct brw_compile *p, GLuint execute_size) struct brw_instruction *brw_WHILE(struct brw_compile *p, struct brw_instruction *do_insn) { + struct intel_context *intel = &p->brw->intel; struct brw_instruction *insn; GLuint br = 1; - if (BRW_IS_IGDNG(p->brw)) + if (intel->is_ironlake) br = 2; if (p->single_program_flow) @@ -846,10 +854,11 @@ struct brw_instruction *brw_WHILE(struct brw_compile *p, void brw_land_fwd_jump(struct brw_compile *p, struct brw_instruction *jmp_insn) { + struct intel_context *intel = &p->brw->intel; struct brw_instruction *landing = &p->store[p->nr_insn]; GLuint jmpi = 1; - if (BRW_IS_IGDNG(p->brw)) + if (intel->is_ironlake) jmpi = 2; assert(jmp_insn->header.opcode == BRW_OPCODE_JMPI); diff --git a/src/mesa/drivers/dri/i965/brw_gs.c b/src/mesa/drivers/dri/i965/brw_gs.c index 610b6c35e2..1bc3eccf49 100644 --- a/src/mesa/drivers/dri/i965/brw_gs.c +++ b/src/mesa/drivers/dri/i965/brw_gs.c @@ -47,6 +47,7 @@ static void compile_gs_prog( struct brw_context *brw, struct brw_gs_prog_key *key ) { + struct intel_context *intel = &brw->intel; struct brw_gs_compile c; const GLuint *program; GLuint program_size; @@ -54,13 +55,12 @@ static void compile_gs_prog( struct brw_context *brw, memset(&c, 0, sizeof(c)); c.key = *key; - c.need_ff_sync = BRW_IS_IGDNG(brw); /* Need to locate the two positions present in vertex + header. * These are currently hardcoded: */ c.nr_attrs = brw_count_bits(c.key.attrs); - if (BRW_IS_IGDNG(brw)) + if (intel->is_ironlake) c.nr_regs = (c.nr_attrs + 1) / 2 + 3; /* are vertices packed, or reg-aligned? */ else c.nr_regs = (c.nr_attrs + 1) / 2 + 1; /* are vertices packed, or reg-aligned? */ diff --git a/src/mesa/drivers/dri/i965/brw_gs.h b/src/mesa/drivers/dri/i965/brw_gs.h index 010c1c2352..813b8d447a 100644 --- a/src/mesa/drivers/dri/i965/brw_gs.h +++ b/src/mesa/drivers/dri/i965/brw_gs.h @@ -63,7 +63,6 @@ struct brw_gs_compile { GLuint nr_attrs; GLuint nr_regs; GLuint nr_bytes; - GLboolean need_ff_sync; }; #define ATTR_SIZE (4*4) diff --git a/src/mesa/drivers/dri/i965/brw_gs_emit.c b/src/mesa/drivers/dri/i965/brw_gs_emit.c index 0fc5b02c61..a81b972ef4 100644 --- a/src/mesa/drivers/dri/i965/brw_gs_emit.c +++ b/src/mesa/drivers/dri/i965/brw_gs_emit.c @@ -122,12 +122,14 @@ static void brw_gs_ff_sync(struct brw_gs_compile *c, int num_prim) void brw_gs_quads( struct brw_gs_compile *c, struct brw_gs_prog_key *key ) { + struct intel_context *intel = &c->func.brw->intel; + brw_gs_alloc_regs(c, 4); /* Use polygons for correct edgeflag behaviour. Note that vertex 3 * is the PV for quads, but vertex 0 for polygons: */ - if (c->need_ff_sync) + if (intel->needs_ff_sync) brw_gs_ff_sync(c, 1); if (key->pv_first) { brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); @@ -145,9 +147,11 @@ void brw_gs_quads( struct brw_gs_compile *c, struct brw_gs_prog_key *key ) void brw_gs_quad_strip( struct brw_gs_compile *c, struct brw_gs_prog_key *key ) { + struct intel_context *intel = &c->func.brw->intel; + brw_gs_alloc_regs(c, 4); - if (c->need_ff_sync) + if (intel->needs_ff_sync) brw_gs_ff_sync(c, 1); if (key->pv_first) { brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START)); @@ -165,9 +169,11 @@ void brw_gs_quad_strip( struct brw_gs_compile *c, struct brw_gs_prog_key *key ) void brw_gs_tris( struct brw_gs_compile *c ) { + struct intel_context *intel = &c->func.brw->intel; + brw_gs_alloc_regs(c, 3); - if (c->need_ff_sync) + if (intel->needs_ff_sync) brw_gs_ff_sync(c, 1); brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_TRILIST << 2) | R02_PRIM_START)); brw_gs_emit_vue(c, c->reg.vertex[1], 0, (_3DPRIM_TRILIST << 2)); @@ -176,9 +182,11 @@ void brw_gs_tris( struct brw_gs_compile *c ) void brw_gs_lines( struct brw_gs_compile *c ) { + struct intel_context *intel = &c->func.brw->intel; + brw_gs_alloc_regs(c, 2); - if (c->need_ff_sync) + if (intel->needs_ff_sync) brw_gs_ff_sync(c, 1); brw_gs_emit_vue(c, c->reg.vertex[0], 0, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_START)); brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_LINESTRIP << 2) | R02_PRIM_END)); @@ -186,9 +194,11 @@ void brw_gs_lines( struct brw_gs_compile *c ) void brw_gs_points( struct brw_gs_compile *c ) { + struct intel_context *intel = &c->func.brw->intel; + brw_gs_alloc_regs(c, 1); - if (c->need_ff_sync) + if (intel->needs_ff_sync) brw_gs_ff_sync(c, 1); brw_gs_emit_vue(c, c->reg.vertex[0], 1, ((_3DPRIM_POINTLIST << 2) | R02_PRIM_START | R02_PRIM_END)); } diff --git a/src/mesa/drivers/dri/i965/brw_gs_state.c b/src/mesa/drivers/dri/i965/brw_gs_state.c index ed9d2ffe60..1af5790a67 100644 --- a/src/mesa/drivers/dri/i965/brw_gs_state.c +++ b/src/mesa/drivers/dri/i965/brw_gs_state.c @@ -72,6 +72,7 @@ gs_unit_populate_key(struct brw_context *brw, struct brw_gs_unit_key *key) static dri_bo * gs_unit_create_from_key(struct brw_context *brw, struct brw_gs_unit_key *key) { + struct intel_context *intel = &brw->intel; struct brw_gs_unit_state gs; dri_bo *bo; @@ -98,7 +99,7 @@ gs_unit_create_from_key(struct brw_context *brw, struct brw_gs_unit_key *key) else gs.thread4.max_threads = 0; - if (BRW_IS_IGDNG(brw)) + if (intel->is_ironlake) gs.thread4.rendering_enable = 1; if (INTEL_DEBUG & DEBUG_STATS) diff --git a/src/mesa/drivers/dri/i965/brw_misc_state.c b/src/mesa/drivers/dri/i965/brw_misc_state.c index 4b0d598336..6cb2da28e6 100644 --- a/src/mesa/drivers/dri/i965/brw_misc_state.c +++ b/src/mesa/drivers/dri/i965/brw_misc_state.c @@ -212,7 +212,7 @@ static void emit_depthbuffer(struct brw_context *brw) { struct intel_context *intel = &brw->intel; struct intel_region *region = brw->state.depth_region; - unsigned int len = (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) ? 6 : 5; + unsigned int len = (BRW_IS_G4X(brw) || intel->is_ironlake) ? 6 : 5; if (region == NULL) { BEGIN_BATCH(len, IGNORE_CLIPRECTS); @@ -223,7 +223,7 @@ static void emit_depthbuffer(struct brw_context *brw) OUT_BATCH(0); OUT_BATCH(0); - if (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) + if (BRW_IS_G4X(brw) || intel->is_ironlake) OUT_BATCH(0); ADVANCE_BATCH(); @@ -262,7 +262,7 @@ static void emit_depthbuffer(struct brw_context *brw) ((region->height - 1) << 19)); OUT_BATCH(0); - if (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw)) + if (BRW_IS_G4X(brw) || intel->is_ironlake) OUT_BATCH(0); ADVANCE_BATCH(); @@ -512,7 +512,7 @@ static void upload_state_base_address( struct brw_context *brw ) /* Output the structure (brw_state_base_address) directly to the * batchbuffer, so we can emit relocations inline. */ - if (BRW_IS_IGDNG(brw)) { + if (intel->is_ironlake) { BEGIN_BATCH(8, IGNORE_CLIPRECTS); OUT_BATCH(CMD_STATE_BASE_ADDRESS << 16 | (8 - 2)); OUT_BATCH(1); /* General state base address */ diff --git a/src/mesa/drivers/dri/i965/brw_sf_emit.c b/src/mesa/drivers/dri/i965/brw_sf_emit.c index 3eae41ee74..bb08055e3b 100644 --- a/src/mesa/drivers/dri/i965/brw_sf_emit.c +++ b/src/mesa/drivers/dri/i965/brw_sf_emit.c @@ -149,6 +149,7 @@ static void copy_colors( struct brw_sf_compile *c, static void do_flatshade_triangle( struct brw_sf_compile *c ) { struct brw_compile *p = &c->func; + struct intel_context *intel = &p->brw->intel; struct brw_reg ip = brw_ip_reg(); GLuint nr = brw_count_bits(c->key.attrs & VERT_RESULT_COLOR_BITS); GLuint jmpi = 1; @@ -161,7 +162,7 @@ static void do_flatshade_triangle( struct brw_sf_compile *c ) if (c->key.primitive == SF_UNFILLED_TRIS) return; - if (BRW_IS_IGDNG(p->brw)) + if (intel->is_ironlake) jmpi = 2; brw_push_insn_state(p); @@ -187,6 +188,7 @@ static void do_flatshade_triangle( struct brw_sf_compile *c ) static void do_flatshade_line( struct brw_sf_compile *c ) { struct brw_compile *p = &c->func; + struct intel_context *intel = &p->brw->intel; struct brw_reg ip = brw_ip_reg(); GLuint nr = brw_count_bits(c->key.attrs & VERT_RESULT_COLOR_BITS); GLuint jmpi = 1; @@ -199,7 +201,7 @@ static void do_flatshade_line( struct brw_sf_compile *c ) if (c->key.primitive == SF_UNFILLED_TRIS) return; - if (BRW_IS_IGDNG(p->brw)) + if (intel->is_ironlake) jmpi = 2; brw_push_insn_state(p); diff --git a/src/mesa/drivers/dri/i965/brw_sf_state.c b/src/mesa/drivers/dri/i965/brw_sf_state.c index bb69435ec0..09223b7cfb 100644 --- a/src/mesa/drivers/dri/i965/brw_sf_state.c +++ b/src/mesa/drivers/dri/i965/brw_sf_state.c @@ -165,6 +165,7 @@ static dri_bo * sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key, dri_bo **reloc_bufs) { + struct intel_context *intel = &brw->intel; struct brw_sf_unit_state sf; dri_bo *bo; int chipset_max_threads; @@ -177,7 +178,7 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key, sf.thread3.dispatch_grf_start_reg = 3; - if (BRW_IS_IGDNG(brw)) + if (intel->is_ironlake) sf.thread3.urb_entry_read_offset = 3; else sf.thread3.urb_entry_read_offset = 1; @@ -187,10 +188,10 @@ sf_unit_create_from_key(struct brw_context *brw, struct brw_sf_unit_key *key, sf.thread4.nr_urb_entries = key->nr_urb_entries; sf.thread4.urb_entry_allocation_size = key->sfsize - 1; - /* Each SF thread produces 1 PUE, and there can be up to 24(Pre-IGDNG) or - * 48(IGDNG) threads + /* Each SF thread produces 1 PUE, and there can be up to 24 (Pre-Ironlake) or + * 48 (Ironlake) threads. */ - if (BRW_IS_IGDNG(brw)) + if (intel->is_ironlake) chipset_max_threads = 48; else chipset_max_threads = 24; diff --git a/src/mesa/drivers/dri/i965/brw_tex_layout.c b/src/mesa/drivers/dri/i965/brw_tex_layout.c index e59e52ed86..64a9535282 100644 --- a/src/mesa/drivers/dri/i965/brw_tex_layout.c +++ b/src/mesa/drivers/dri/i965/brw_tex_layout.c @@ -49,7 +49,7 @@ GLboolean brw_miptree_layout(struct intel_context *intel, switch (mt->target) { case GL_TEXTURE_CUBE_MAP: - if (IS_IGDNG(intel->intelScreen->deviceID)) { + if (intel->is_ironlake) { GLuint align_h = 2, align_w = 4; GLuint level; GLuint x = 0; diff --git a/src/mesa/drivers/dri/i965/brw_urb.c b/src/mesa/drivers/dri/i965/brw_urb.c index 8c6f4355a6..e1e8a90ca7 100644 --- a/src/mesa/drivers/dri/i965/brw_urb.c +++ b/src/mesa/drivers/dri/i965/brw_urb.c @@ -113,6 +113,7 @@ static GLboolean check_urb_layout( struct brw_context *brw ) */ static void recalculate_urb_fence( struct brw_context *brw ) { + struct intel_context *intel = &brw->intel; GLuint csize = brw->curbe.total_size; GLuint vsize = brw->vs.prog_data->urb_entry_size; GLuint sfsize = brw->sf.prog_data->urb_entry_size; @@ -146,7 +147,7 @@ static void recalculate_urb_fence( struct brw_context *brw ) brw->urb.constrained = 0; - if (BRW_IS_IGDNG(brw)) { + if (intel->is_ironlake) { brw->urb.nr_vs_entries = 128; brw->urb.nr_sf_entries = 48; if (check_urb_layout(brw)) { diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c index 27aac8b54a..21d2fec707 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_emit.c +++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c @@ -67,6 +67,7 @@ static void release_tmps( struct brw_vs_compile *c ) */ static void brw_vs_alloc_regs( struct brw_vs_compile *c ) { + struct intel_context *intel = &c->func.brw->intel; GLuint i, reg = 0, mrf; int attributes_in_vue; @@ -141,7 +142,7 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c ) c->first_output = reg; c->first_overflow_output = 0; - if (BRW_IS_IGDNG(c->func.brw)) + if (intel->is_ironlake) mrf = 8; else mrf = 4; @@ -238,7 +239,7 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c ) */ attributes_in_vue = MAX2(c->nr_outputs, c->nr_inputs); - if (BRW_IS_IGDNG(c->func.brw)) + if (intel->is_ironlake) c->prog_data.urb_entry_size = (attributes_in_vue + 6 + 3) / 4; else c->prog_data.urb_entry_size = (attributes_in_vue + 2 + 3) / 4; @@ -1113,6 +1114,7 @@ static void emit_swz( struct brw_vs_compile *c, static void emit_vertex_write( struct brw_vs_compile *c) { struct brw_compile *p = &c->func; + struct intel_context *intel = &p->brw->intel; struct brw_reg m0 = brw_message_reg(0); struct brw_reg pos = c->regs[PROGRAM_OUTPUT][VERT_RESULT_HPOS]; struct brw_reg ndc; @@ -1195,8 +1197,8 @@ static void emit_vertex_write( struct brw_vs_compile *c) brw_set_access_mode(p, BRW_ALIGN_1); brw_MOV(p, offset(m0, 2), ndc); - if (BRW_IS_IGDNG(p->brw)) { - /* There are 20 DWs (D0-D19) in VUE vertex header on IGDNG */ + if (intel->is_ironlake) { + /* There are 20 DWs (D0-D19) in VUE vertex header on Ironlake */ brw_MOV(p, offset(m0, 3), pos); /* a portion of vertex header */ /* m4, m5 contain the distances from vertex to the user clip planeXXX. * Seems it is useless for us. @@ -1359,6 +1361,7 @@ void brw_vs_emit(struct brw_vs_compile *c ) #define MAX_LOOP_DEPTH 32 struct brw_compile *p = &c->func; struct brw_context *brw = p->brw; + struct intel_context *intel = &brw->intel; const GLuint nr_insns = c->vp->program.Base.NumInstructions; GLuint insn, if_depth = 0, loop_depth = 0; GLuint end_offset = 0; @@ -1592,7 +1595,7 @@ void brw_vs_emit(struct brw_vs_compile *c ) loop_depth--; - if (BRW_IS_IGDNG(brw)) + if (intel->is_ironlake) br = 2; inst0 = inst1 = brw_WHILE(p, loop_inst[loop_depth]); diff --git a/src/mesa/drivers/dri/i965/brw_vs_state.c b/src/mesa/drivers/dri/i965/brw_vs_state.c index 7285466645..fdc55aa1b2 100644 --- a/src/mesa/drivers/dri/i965/brw_vs_state.c +++ b/src/mesa/drivers/dri/i965/brw_vs_state.c @@ -82,6 +82,7 @@ vs_unit_populate_key(struct brw_context *brw, struct brw_vs_unit_key *key) static dri_bo * vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key) { + struct intel_context *intel = &brw->intel; struct brw_vs_unit_state vs; dri_bo *bo; int chipset_max_threads; @@ -98,7 +99,7 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key) */ vs.thread1.single_program_flow = 0; - if (BRW_IS_IGDNG(brw)) + if (intel->is_ironlake) vs.thread1.binding_table_entry_count = 0; /* hardware requirement */ else vs.thread1.binding_table_entry_count = key->nr_surfaces; @@ -109,7 +110,7 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key) vs.thread3.urb_entry_read_offset = 0; vs.thread3.const_urb_entry_read_offset = key->curbe_offset * 2; - if (BRW_IS_IGDNG(brw)) { + if (intel->is_ironlake) { switch (key->nr_urb_entries) { case 8: case 12: @@ -145,7 +146,7 @@ vs_unit_create_from_key(struct brw_context *brw, struct brw_vs_unit_key *key) vs.thread4.urb_entry_allocation_size = key->urb_size - 1; - if (BRW_IS_IGDNG(brw)) + if (intel->is_ironlake) chipset_max_threads = 72; else if (BRW_IS_G4X(brw)) chipset_max_threads = 32; diff --git a/src/mesa/drivers/dri/i965/brw_wm_emit.c b/src/mesa/drivers/dri/i965/brw_wm_emit.c index 5390fd2584..bcaa793de1 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_emit.c +++ b/src/mesa/drivers/dri/i965/brw_wm_emit.c @@ -830,6 +830,7 @@ void emit_tex(struct brw_wm_compile *c, GLboolean shadow) { struct brw_compile *p = &c->func; + struct intel_context *intel = &p->brw->intel; struct brw_reg dst_retyped; GLuint cur_mrf = 2, response_length; GLuint i, nr_texcoords; @@ -873,7 +874,7 @@ void emit_tex(struct brw_wm_compile *c, } /* Pre-Ironlake, the 8-wide sampler always took u,v,r. */ - if (!BRW_IS_IGDNG(p->brw) && c->dispatch_width == 8) + if (!intel->is_ironlake && c->dispatch_width == 8) nr_texcoords = 3; /* For shadow comparisons, we have to supply u,v,r. */ @@ -891,7 +892,7 @@ void emit_tex(struct brw_wm_compile *c, /* Fill in the shadow comparison reference value. */ if (shadow) { - if (BRW_IS_IGDNG(p->brw)) { + if (intel->is_ironlake) { /* Fill in the cube map array index value. */ brw_MOV(p, brw_message_reg(cur_mrf), brw_imm_f(0)); cur_mrf += mrf_per_channel; @@ -904,7 +905,7 @@ void emit_tex(struct brw_wm_compile *c, cur_mrf += mrf_per_channel; } - if (BRW_IS_IGDNG(p->brw)) { + if (intel->is_ironlake) { if (shadow) msg_type = BRW_SAMPLER_MESSAGE_SAMPLE_COMPARE_IGDNG; else @@ -944,6 +945,7 @@ void emit_txb(struct brw_wm_compile *c, GLuint sampler) { struct brw_compile *p = &c->func; + struct intel_context *intel = &p->brw->intel; GLuint msgLength; GLuint msg_type; GLuint mrf_per_channel; @@ -955,8 +957,8 @@ void emit_txb(struct brw_wm_compile *c, * undefined, and trust the execution mask to keep the undefined pixels * from mattering. */ - if (c->dispatch_width == 16 || !BRW_IS_IGDNG(p->brw)) { - if (BRW_IS_IGDNG(p->brw)) + if (c->dispatch_width == 16 || !intel->is_ironlake) { + if (intel->is_ironlake) msg_type = BRW_SAMPLER_MESSAGE_SAMPLE_BIAS_IGDNG; else msg_type = BRW_SAMPLER_MESSAGE_SIMD16_SAMPLE_BIAS; @@ -1160,6 +1162,7 @@ void emit_fb_write(struct brw_wm_compile *c, { struct brw_compile *p = &c->func; struct brw_context *brw = p->brw; + struct intel_context *intel = &brw->intel; GLuint nr = 2; GLuint channel; @@ -1174,7 +1177,7 @@ void emit_fb_write(struct brw_wm_compile *c, brw_push_insn_state(p); for (channel = 0; channel < 4; channel++) { - if (c->dispatch_width == 16 && (BRW_IS_G4X(brw) || BRW_IS_IGDNG(brw))) { + if (c->dispatch_width == 16 && (BRW_IS_G4X(brw) || intel->is_ironlake)) { /* By setting the high bit of the MRF register number, we indicate * that we want COMPR4 mode - instead of doing the usual destination * + 1 for the second half we get destination + 4. diff --git a/src/mesa/drivers/dri/i965/brw_wm_glsl.c b/src/mesa/drivers/dri/i965/brw_wm_glsl.c index e8c2cb66ec..f294387c80 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_glsl.c +++ b/src/mesa/drivers/dri/i965/brw_wm_glsl.c @@ -1826,6 +1826,7 @@ get_argument_regs(struct brw_wm_compile *c, static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c) { + struct intel_context *intel = &brw->intel; #define MAX_IF_DEPTH 32 #define MAX_LOOP_DEPTH 32 struct brw_instruction *if_inst[MAX_IF_DEPTH], *loop_inst[MAX_LOOP_DEPTH]; @@ -2096,7 +2097,7 @@ static void brw_wm_emit_glsl(struct brw_context *brw, struct brw_wm_compile *c) struct brw_instruction *inst0, *inst1; GLuint br = 1; - if (BRW_IS_IGDNG(brw)) + if (intel->is_ironlake) br = 2; loop_depth--; diff --git a/src/mesa/drivers/dri/i965/brw_wm_state.c b/src/mesa/drivers/dri/i965/brw_wm_state.c index f89ed9bce7..fe795fadf8 100644 --- a/src/mesa/drivers/dri/i965/brw_wm_state.c +++ b/src/mesa/drivers/dri/i965/brw_wm_state.c @@ -71,7 +71,7 @@ wm_unit_populate_key(struct brw_context *brw, struct brw_wm_unit_key *key) key->max_threads = 1; else { /* WM maximum threads is number of EUs times number of threads per EU. */ - if (BRW_IS_IGDNG(brw)) + if (intel->is_ironlake) key->max_threads = 12 * 6; else if (BRW_IS_G4X(brw)) key->max_threads = 10 * 5; @@ -140,6 +140,7 @@ static dri_bo * wm_unit_create_from_key(struct brw_context *brw, struct brw_wm_unit_key *key, dri_bo **reloc_bufs) { + struct intel_context *intel = &brw->intel; struct brw_wm_unit_state wm; dri_bo *bo; @@ -150,7 +151,7 @@ wm_unit_create_from_key(struct brw_context *brw, struct brw_wm_unit_key *key, wm.thread1.depth_coef_urb_read_offset = 1; wm.thread1.floating_point_mode = BRW_FLOATING_POINT_NON_IEEE_754; - if (BRW_IS_IGDNG(brw)) + if (intel->is_ironlake) wm.thread1.binding_table_entry_count = 0; /* hardware requirement */ else wm.thread1.binding_table_entry_count = key->nr_surfaces; @@ -170,7 +171,7 @@ wm_unit_create_from_key(struct brw_context *brw, struct brw_wm_unit_key *key, wm.thread3.const_urb_entry_read_length = key->curb_entry_read_length; wm.thread3.const_urb_entry_read_offset = key->curbe_offset * 2; - if (BRW_IS_IGDNG(brw)) + if (intel->is_ironlake) wm.wm4.sampler_count = 0; /* hardware requirement */ else wm.wm4.sampler_count = (key->sampler_count + 1) / 4; diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c index c616d32ebb..234f99a00e 100644 --- a/src/mesa/drivers/dri/intel/intel_context.c +++ b/src/mesa/drivers/dri/intel/intel_context.c @@ -620,6 +620,11 @@ intelInitContext(struct intel_context *intel, else intel->gen = 2; + if (IS_IGDNG(intel->intelScreen->deviceID)) { + intel->is_ironlake = GL_TRUE; + intel->needs_ff_sync = GL_TRUE; + } + /* Dri stuff */ intel->hHWContext = driContextPriv->hHWContext; intel->driFd = sPriv->fd; diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h index 481202c971..f7d9016d71 100644 --- a/src/mesa/drivers/dri/intel/intel_context.h +++ b/src/mesa/drivers/dri/intel/intel_context.h @@ -176,6 +176,8 @@ struct intel_context * Generation number of the hardware: 2 is 8xx, 3 is 9xx pre-965, 4 is 965. */ int gen; + GLboolean needs_ff_sync; + GLboolean is_ironlake; struct intel_region *front_region; struct intel_region *back_region; -- cgit v1.2.3