From 28e9601d0e681411b60a7de8be9f401b0df77d29 Mon Sep 17 00:00:00 2001 From: Jason Ekstrand Date: Wed, 15 Apr 2015 18:00:05 -0700 Subject: i965: Add a devinfo field to backend_visitor and use it for gen checks Reviewed-by: Matt Turner --- src/mesa/drivers/dri/i965/brw_fs.cpp | 72 ++++++------- .../drivers/dri/i965/brw_fs_combine_constants.cpp | 10 +- .../drivers/dri/i965/brw_fs_copy_propagation.cpp | 10 +- src/mesa/drivers/dri/i965/brw_fs_fp.cpp | 4 +- src/mesa/drivers/dri/i965/brw_fs_nir.cpp | 42 ++++---- .../dri/i965/brw_fs_peephole_predicated_break.cpp | 2 +- src/mesa/drivers/dri/i965/brw_fs_reg_allocate.cpp | 10 +- src/mesa/drivers/dri/i965/brw_fs_sel_peephole.cpp | 4 +- src/mesa/drivers/dri/i965/brw_fs_visitor.cpp | 108 +++++++++---------- src/mesa/drivers/dri/i965/brw_ir_fs.h | 2 +- src/mesa/drivers/dri/i965/brw_ir_vec4.h | 2 +- src/mesa/drivers/dri/i965/brw_shader.cpp | 3 +- src/mesa/drivers/dri/i965/brw_shader.h | 1 + src/mesa/drivers/dri/i965/brw_vec4.cpp | 22 ++-- .../drivers/dri/i965/brw_vec4_copy_propagation.cpp | 20 ++-- .../dri/i965/brw_vec4_dead_code_eliminate.cpp | 6 +- src/mesa/drivers/dri/i965/brw_vec4_gs_visitor.cpp | 4 +- src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp | 120 ++++++++++----------- src/mesa/drivers/dri/i965/brw_vec4_vp.cpp | 4 +- 19 files changed, 225 insertions(+), 221 deletions(-) diff --git a/src/mesa/drivers/dri/i965/brw_fs.cpp b/src/mesa/drivers/dri/i965/brw_fs.cpp index cf1c385f098..5f83470ea92 100644 --- a/src/mesa/drivers/dri/i965/brw_fs.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs.cpp @@ -313,7 +313,7 @@ fs_inst * fs_visitor::IF(const fs_reg &src0, const fs_reg &src1, enum brw_conditional_mod condition) { - assert(brw->gen == 6); + assert(devinfo->gen == 6); fs_inst *inst = new(mem_ctx) fs_inst(BRW_OPCODE_IF, dispatch_width, reg_null_d, src0, src1); inst->conditional_mod = condition; @@ -405,7 +405,7 @@ fs_visitor::VARYING_PULL_CONSTANT_LOAD(const fs_reg &dst, varying_offset, fs_reg(const_offset & ~3))); int scale = 1; - if (brw->gen == 4 && dst.width == 8) { + if (devinfo->gen == 4 && dst.width == 8) { /* Pre-gen5, we can either use a SIMD8 message that requires (header, * u, v, r) as parameters, or we can just use the SIMD16 message * consisting of (header, u). We choose the second, at the cost of a @@ -415,7 +415,7 @@ fs_visitor::VARYING_PULL_CONSTANT_LOAD(const fs_reg &dst, } enum opcode op; - if (brw->gen >= 7) + if (devinfo->gen >= 7) op = FS_OPCODE_VARYING_PULL_CONSTANT_LOAD_GEN7; else op = FS_OPCODE_VARYING_PULL_CONSTANT_LOAD; @@ -428,10 +428,10 @@ fs_visitor::VARYING_PULL_CONSTANT_LOAD(const fs_reg &dst, inst->regs_written = regs_written; instructions.push_tail(inst); - if (brw->gen < 7) { + if (devinfo->gen < 7) { inst->base_mrf = 13; inst->header_present = true; - if (brw->gen == 4) + if (devinfo->gen == 4) inst->mlen = 3; else inst->mlen = 1 + dispatch_width / 8; @@ -517,9 +517,9 @@ fs_inst::is_send_from_grf() const } bool -fs_inst::can_do_source_mods(struct brw_context *brw) +fs_inst::can_do_source_mods(const struct brw_device_info *devinfo) { - if (brw->gen == 6 && is_math()) + if (devinfo->gen == 6 && is_math()) return false; if (is_send_from_grf()) @@ -689,7 +689,7 @@ fs_visitor::type_size(const struct glsl_type *type) fs_reg fs_visitor::get_timestamp(fs_inst **out_mov) { - assert(brw->gen >= 7); + assert(devinfo->gen >= 7); fs_reg ts = fs_reg(retype(brw_vec4_reg(BRW_ARCHITECTURE_REGISTER_FILE, BRW_ARF_TIMESTAMP, @@ -1261,7 +1261,7 @@ fs_visitor::emit_fragcoord_interpolation(bool pixel_center_integer, wpos = offset(wpos, 1); /* gl_FragCoord.z */ - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { emit(MOV(wpos, fs_reg(brw_vec8_grf(payload.source_depth_reg, 0)))); } else { emit(FS_OPCODE_LINTERP, wpos, @@ -1282,7 +1282,7 @@ fs_visitor::emit_linterp(const fs_reg &attr, const fs_reg &interp, bool is_centroid, bool is_sample) { brw_wm_barycentric_interp_mode barycoord_mode; - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { if (is_centroid) { if (interpolation_mode == INTERP_QUALIFIER_SMOOTH) barycoord_mode = BRW_WM_PERSPECTIVE_CENTROID_BARYCENTRIC; @@ -1372,7 +1372,7 @@ fs_visitor::emit_general_interpolation(fs_reg attr, const char *name, /* Smooth/noperspective interpolation case. */ for (unsigned int k = 0; k < type->vector_elements; k++) { struct brw_reg interp = interp_reg(location, k); - if (brw->needs_unlit_centroid_workaround && mod_centroid) { + if (devinfo->needs_unlit_centroid_workaround && mod_centroid) { /* Get the pixel/sample mask into f0 so that we know * which pixels are lit. Then, for each channel that is * unlit, replace the centroid data with non-centroid @@ -1385,7 +1385,7 @@ fs_visitor::emit_general_interpolation(fs_reg attr, const char *name, false, false); inst->predicate = BRW_PREDICATE_NORMAL; inst->predicate_inverse = true; - if (brw->has_pln) + if (devinfo->has_pln) inst->no_dd_clear = true; inst = emit_linterp(attr, fs_reg(interp), interpolation_mode, @@ -1393,7 +1393,7 @@ fs_visitor::emit_general_interpolation(fs_reg attr, const char *name, mod_sample || key->persample_shading); inst->predicate = BRW_PREDICATE_NORMAL; inst->predicate_inverse = false; - if (brw->has_pln) + if (devinfo->has_pln) inst->no_dd_check = true; } else { @@ -1401,7 +1401,7 @@ fs_visitor::emit_general_interpolation(fs_reg attr, const char *name, mod_centroid && !key->persample_shading, mod_sample || key->persample_shading); } - if (brw->gen < 6 && interpolation_mode == INTERP_QUALIFIER_SMOOTH) { + if (devinfo->gen < 6 && interpolation_mode == INTERP_QUALIFIER_SMOOTH) { emit(BRW_OPCODE_MUL, attr, attr, this->pixel_w); } attr = offset(attr, 1); @@ -1418,7 +1418,7 @@ fs_visitor::emit_frontfacing_interpolation() { fs_reg *reg = new(this->mem_ctx) fs_reg(vgrf(glsl_type::bool_type)); - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { /* Bit 15 of g0.0 is 0 if the polygon is front facing. We want to create * a boolean result from this (~0/true or 0/false). * @@ -1480,7 +1480,7 @@ fs_visitor::compute_sample_position(fs_reg dst, fs_reg int_sample_pos) fs_reg * fs_visitor::emit_samplepos_setup() { - assert(brw->gen >= 6); + assert(devinfo->gen >= 6); this->current_annotation = "compute sample position"; fs_reg *reg = new(this->mem_ctx) fs_reg(vgrf(glsl_type::vec2_type)); @@ -1531,7 +1531,7 @@ fs_visitor::emit_sampleid_setup() { assert(stage == MESA_SHADER_FRAGMENT); brw_wm_prog_key *key = (brw_wm_prog_key*) this->key; - assert(brw->gen >= 6); + assert(devinfo->gen >= 6); this->current_annotation = "compute sample id"; fs_reg *reg = new(this->mem_ctx) fs_reg(vgrf(glsl_type::int_type)); @@ -1607,14 +1607,14 @@ fs_visitor::fix_math_operand(fs_reg src) * The hardware ignores source modifiers (negate and abs) on math * instructions, so we also move to a temp to set those up. */ - if (brw->gen == 6 && src.file != UNIFORM && src.file != IMM && + if (devinfo->gen == 6 && src.file != UNIFORM && src.file != IMM && !src.abs && !src.negate) return src; /* Gen7 relaxes most of the above restrictions, but still can't use IMM * operands to math */ - if (brw->gen >= 7 && src.file != IMM) + if (devinfo->gen >= 7 && src.file != IMM) return src; fs_reg expanded = vgrf(glsl_type::float_type); @@ -1647,12 +1647,12 @@ fs_visitor::emit_math(enum opcode opcode, fs_reg dst, fs_reg src) * Gen 6 hardware ignores source modifiers (negate and abs) on math * instructions, so we also move to a temp to set those up. */ - if (brw->gen == 6 || brw->gen == 7) + if (devinfo->gen == 6 || devinfo->gen == 7) src = fix_math_operand(src); fs_inst *inst = emit(opcode, dst, src); - if (brw->gen < 6) { + if (devinfo->gen < 6) { inst->base_mrf = 2; inst->mlen = dispatch_width / 8; } @@ -1666,9 +1666,9 @@ fs_visitor::emit_math(enum opcode opcode, fs_reg dst, fs_reg src0, fs_reg src1) int base_mrf = 2; fs_inst *inst; - if (brw->gen >= 8) { + if (devinfo->gen >= 8) { inst = emit(opcode, dst, src0, src1); - } else if (brw->gen >= 6) { + } else if (devinfo->gen >= 6) { src0 = fix_math_operand(src0); src1 = fix_math_operand(src1); @@ -1768,7 +1768,7 @@ fs_visitor::calculate_urb_setup() int urb_next = 0; /* Figure out where each of the incoming setup attributes lands. */ - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { if (_mesa_bitcount_64(prog->InputsRead & BRW_FS_VARYING_INPUT_MASK) <= 16) { /* The SF/SBE pipeline stage can do arbitrary rearrangement of the @@ -2548,7 +2548,7 @@ fs_visitor::opt_sampler_eot() { brw_wm_prog_key *key = (brw_wm_prog_key*) this->key; - if (brw->gen < 9 && !brw->is_cherryview) + if (devinfo->gen < 9 && !devinfo->is_cherryview) return false; /* FINISHME: It should be possible to implement this optimization when there @@ -2732,7 +2732,7 @@ fs_visitor::compute_to_mrf() int next_ip = 0; /* No MRFs on Gen >= 7. */ - if (brw->gen >= 7) + if (devinfo->gen >= 7) return false; calculate_live_intervals(); @@ -2797,7 +2797,7 @@ fs_visitor::compute_to_mrf() if (scan_inst->mlen) break; - if (brw->gen == 6) { + if (devinfo->gen == 6) { /* gen6 math instructions must have the destination be * GRF, so no compute-to-MRF for them. */ @@ -3164,7 +3164,7 @@ fs_visitor::insert_gen4_post_send_dependency_workarounds(bblock_t *block, fs_ins void fs_visitor::insert_gen4_send_dependency_workarounds() { - if (brw->gen != 4 || brw->is_g4x) + if (devinfo->gen != 4 || devinfo->is_g4x) return; bool progress = false; @@ -3208,7 +3208,7 @@ fs_visitor::lower_uniform_pull_constant_loads() if (inst->opcode != FS_OPCODE_UNIFORM_PULL_CONSTANT_LOAD) continue; - if (brw->gen >= 7) { + if (devinfo->gen >= 7) { /* The offset arg before was a vec4-aligned byte offset. We need to * turn it into a dword offset. */ @@ -3221,7 +3221,7 @@ fs_visitor::lower_uniform_pull_constant_loads() /* We have to use a message header on Skylake to get SIMD4x2 mode. * Reserve space for the register. */ - if (brw->gen >= 9) { + if (devinfo->gen >= 9) { payload.reg_offset++; alloc.sizes[payload.reg] = 2; } @@ -3306,7 +3306,7 @@ fs_visitor::lower_load_payload() /* Do nothing but otherwise increment as normal */ } else if (dst.file == MRF && dst.width == 8 && - brw->has_compr4 && + devinfo->has_compr4 && i + 4 < inst->sources && inst->src[i + 4].equals(horiz_offset(inst->src[i], 8))) { fs_reg compr4_dst = dst; @@ -3427,7 +3427,7 @@ fs_visitor::dump_instruction(backend_instruction *be_inst, FILE *file) if (inst->conditional_mod) { fprintf(file, "%s", conditional_modifier[inst->conditional_mod]); if (!inst->predicate && - (brw->gen < 5 || (inst->opcode != BRW_OPCODE_SEL && + (devinfo->gen < 5 || (inst->opcode != BRW_OPCODE_SEL && inst->opcode != BRW_OPCODE_IF && inst->opcode != BRW_OPCODE_WHILE))) { fprintf(file, ".f0.%d", inst->flag_subreg); @@ -3646,7 +3646,7 @@ fs_visitor::setup_payload_gen6() (stage == MESA_SHADER_FRAGMENT) ? ((brw_wm_prog_data*) this->prog_data)->barycentric_interp_modes : 0; - assert(brw->gen >= 6); + assert(devinfo->gen >= 6); /* R0-1: masks, pixel X/Y coordinates. */ payload.num_regs = 2; @@ -3701,7 +3701,7 @@ fs_visitor::setup_payload_gen6() /* R32: MSAA input coverage mask */ if (prog->SystemValuesRead & SYSTEM_BIT_SAMPLE_MASK_IN) { - assert(brw->gen >= 7); + assert(devinfo->gen >= 7); payload.sample_mask_in_reg = payload.num_regs; payload.num_regs++; if (dispatch_width == 16) { @@ -3978,7 +3978,7 @@ fs_visitor::run_fs() assign_binding_table_offsets(); - if (brw->gen >= 6) + if (devinfo->gen >= 6) setup_payload_gen6(); else setup_payload_gen4(); @@ -3993,7 +3993,7 @@ fs_visitor::run_fs() calculate_urb_setup(); if (prog->InputsRead > 0) { - if (brw->gen < 6) + if (devinfo->gen < 6) emit_interpolation_setup_gen4(); else emit_interpolation_setup_gen6(); diff --git a/src/mesa/drivers/dri/i965/brw_fs_combine_constants.cpp b/src/mesa/drivers/dri/i965/brw_fs_combine_constants.cpp index ebde8df2ff1..a51b726d4a4 100644 --- a/src/mesa/drivers/dri/i965/brw_fs_combine_constants.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs_combine_constants.cpp @@ -42,9 +42,9 @@ * replaced with a GRF source. */ static bool -could_coissue(const struct brw_context *brw, const fs_inst *inst) +could_coissue(const struct brw_device_info *devinfo, const fs_inst *inst) { - if (brw->gen != 7) + if (devinfo->gen != 7) return false; switch (inst->opcode) { @@ -207,7 +207,7 @@ fs_visitor::opt_combine_constants() foreach_block_and_inst(block, fs_inst, inst, cfg) { ip++; - if (!could_coissue(brw, inst) && !must_promote_imm(inst)) + if (!could_coissue(devinfo, inst) && !must_promote_imm(inst)) continue; for (int i = 0; i < inst->sources; i++) { @@ -224,7 +224,7 @@ fs_visitor::opt_combine_constants() imm->inst = NULL; imm->block = intersection; imm->uses->push_tail(link(const_ctx, &inst->src[i])); - imm->uses_by_coissue += could_coissue(brw, inst); + imm->uses_by_coissue += could_coissue(devinfo, inst); imm->must_promote = imm->must_promote || must_promote_imm(inst); imm->last_use_ip = ip; } else { @@ -234,7 +234,7 @@ fs_visitor::opt_combine_constants() imm->uses = new(const_ctx) exec_list(); imm->uses->push_tail(link(const_ctx, &inst->src[i])); imm->val = val; - imm->uses_by_coissue = could_coissue(brw, inst); + imm->uses_by_coissue = could_coissue(devinfo, inst); imm->must_promote = must_promote_imm(inst); imm->first_use_ip = ip; imm->last_use_ip = ip; diff --git a/src/mesa/drivers/dri/i965/brw_fs_copy_propagation.cpp b/src/mesa/drivers/dri/i965/brw_fs_copy_propagation.cpp index 6a8e7bf5b5d..6b6565fec20 100644 --- a/src/mesa/drivers/dri/i965/brw_fs_copy_propagation.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs_copy_propagation.cpp @@ -325,7 +325,7 @@ fs_visitor::try_copy_propagate(fs_inst *inst, int arg, acp_entry *entry) if ((has_source_modifiers || entry->src.file == UNIFORM || !entry->src.is_contiguous()) && - !inst->can_do_source_mods(brw)) + !inst->can_do_source_mods(devinfo)) return false; if (has_source_modifiers && @@ -361,7 +361,7 @@ fs_visitor::try_copy_propagate(fs_inst *inst, int arg, acp_entry *entry) !can_change_source_types(inst)) return false; - if (brw->gen >= 8 && (entry->src.negate || entry->src.abs) && + if (devinfo->gen >= 8 && (entry->src.negate || entry->src.abs) && is_logic_op(inst->opcode)) { return false; } @@ -479,14 +479,14 @@ fs_visitor::try_constant_propagate(fs_inst *inst, acp_entry *entry) val.type = inst->src[i].type; if (inst->src[i].abs) { - if ((brw->gen >= 8 && is_logic_op(inst->opcode)) || + if ((devinfo->gen >= 8 && is_logic_op(inst->opcode)) || !brw_abs_immediate(val.type, &val.fixed_hw_reg)) { continue; } } if (inst->src[i].negate) { - if ((brw->gen >= 8 && is_logic_op(inst->opcode)) || + if ((devinfo->gen >= 8 && is_logic_op(inst->opcode)) || !brw_negate_immediate(val.type, &val.fixed_hw_reg)) { continue; } @@ -502,7 +502,7 @@ fs_visitor::try_constant_propagate(fs_inst *inst, acp_entry *entry) case SHADER_OPCODE_POW: case SHADER_OPCODE_INT_QUOTIENT: case SHADER_OPCODE_INT_REMAINDER: - if (brw->gen < 8) + if (devinfo->gen < 8) break; /* fallthrough */ case BRW_OPCODE_BFI1: diff --git a/src/mesa/drivers/dri/i965/brw_fs_fp.cpp b/src/mesa/drivers/dri/i965/brw_fs_fp.cpp index 21d677a45a1..6518ff60c3b 100644 --- a/src/mesa/drivers/dri/i965/brw_fs_fp.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs_fp.cpp @@ -251,7 +251,7 @@ fs_visitor::emit_fragment_program_code() cmp->predicate = BRW_PREDICATE_NORMAL; cmp->flag_subreg = 1; - if (brw->gen >= 6) + if (devinfo->gen >= 6) emit_discard_jump(); } break; @@ -319,7 +319,7 @@ fs_visitor::emit_fragment_program_code() case OPCODE_MAD: for (int i = 0; i < 4; i++) { if (fpi->DstReg.WriteMask & (1 << i)) { - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { emit(MAD(offset(dst, i), offset(src[2], i), offset(src[1], i), offset(src[0], i))); } else { diff --git a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp index f2f78be5e89..0dcbc3b4675 100644 --- a/src/mesa/drivers/dri/i965/brw_fs_nir.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs_nir.cpp @@ -324,7 +324,7 @@ emit_system_values_block(nir_block *block, void *void_visitor) case nir_intrinsic_load_sample_mask_in: assert(v->stage == MESA_SHADER_FRAGMENT); - assert(v->brw->gen >= 7); + assert(v->devinfo->gen >= 7); reg = &v->nir_system_values[SYSTEM_VALUE_SAMPLE_MASK_IN]; if (reg->file == BAD_FILE) *reg = fs_reg(retype(brw_vec8_grf(v->payload.sample_mask_in_reg, 0), @@ -408,7 +408,7 @@ fs_visitor::nir_emit_if(nir_if *if_stmt) emit(BRW_OPCODE_ENDIF); - if (!try_replace_with_sel() && brw->gen < 6) { + if (!try_replace_with_sel() && devinfo->gen < 6) { no16("Can't support (non-uniform) control flow on SIMD16\n"); } } @@ -416,7 +416,7 @@ fs_visitor::nir_emit_if(nir_if *if_stmt) void fs_visitor::nir_emit_loop(nir_loop *loop) { - if (brw->gen < 6) { + if (devinfo->gen < 6) { no16("Can't support (non-uniform) control flow on SIMD16\n"); } @@ -517,7 +517,7 @@ fs_visitor::optimize_frontfacing_ternary(nir_alu_instr *instr, fs_reg tmp = vgrf(glsl_type::int_type); - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { /* Bit 15 of g0.0 is 0 if the polygon is front facing. */ fs_reg g0 = fs_reg(retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_W)); @@ -785,7 +785,7 @@ fs_visitor::nir_emit_alu(nir_alu_instr *instr) break; case nir_op_imul: { - if (brw->gen >= 8) { + if (devinfo->gen >= 8) { emit(MUL(result, op[0], op[1])); break; } else { @@ -793,14 +793,14 @@ fs_visitor::nir_emit_alu(nir_alu_instr *instr) nir_const_value *value1 = nir_src_as_const_value(instr->src[1].src); if (value0 && value0->u[0] < (1 << 16)) { - if (brw->gen < 7) { + if (devinfo->gen < 7) { emit(MUL(result, op[0], op[1])); } else { emit(MUL(result, op[1], op[0])); } break; } else if (value1 && value1->u[0] < (1 << 16)) { - if (brw->gen < 7) { + if (devinfo->gen < 7) { emit(MUL(result, op[1], op[0])); } else { emit(MUL(result, op[0], op[1])); @@ -809,7 +809,7 @@ fs_visitor::nir_emit_alu(nir_alu_instr *instr) } } - if (brw->gen >= 7) + if (devinfo->gen >= 7) no16("SIMD16 explicit accumulator operands unsupported\n"); struct brw_reg acc = retype(brw_acc_reg(dispatch_width), result.type); @@ -822,7 +822,7 @@ fs_visitor::nir_emit_alu(nir_alu_instr *instr) case nir_op_imul_high: case nir_op_umul_high: { - if (brw->gen >= 7) + if (devinfo->gen >= 7) no16("SIMD16 explicit accumulator operands unsupported\n"); struct brw_reg acc = retype(brw_acc_reg(dispatch_width), result.type); @@ -838,7 +838,7 @@ fs_visitor::nir_emit_alu(nir_alu_instr *instr) break; case nir_op_uadd_carry: { - if (brw->gen >= 7) + if (devinfo->gen >= 7) no16("SIMD16 explicit accumulator operands unsupported\n"); struct brw_reg acc = retype(brw_acc_reg(dispatch_width), @@ -850,7 +850,7 @@ fs_visitor::nir_emit_alu(nir_alu_instr *instr) } case nir_op_usub_borrow: { - if (brw->gen >= 7) + if (devinfo->gen >= 7) no16("SIMD16 explicit accumulator operands unsupported\n"); struct brw_reg acc = retype(brw_acc_reg(dispatch_width), @@ -888,27 +888,27 @@ fs_visitor::nir_emit_alu(nir_alu_instr *instr) break; case nir_op_inot: - if (brw->gen >= 8) { + if (devinfo->gen >= 8) { resolve_source_modifiers(&op[0]); } emit(NOT(result, op[0])); break; case nir_op_ixor: - if (brw->gen >= 8) { + if (devinfo->gen >= 8) { resolve_source_modifiers(&op[0]); resolve_source_modifiers(&op[1]); } emit(XOR(result, op[0], op[1])); break; case nir_op_ior: - if (brw->gen >= 8) { + if (devinfo->gen >= 8) { resolve_source_modifiers(&op[0]); resolve_source_modifiers(&op[1]); } emit(OR(result, op[0], op[1])); break; case nir_op_iand: - if (brw->gen >= 8) { + if (devinfo->gen >= 8) { resolve_source_modifiers(&op[0]); resolve_source_modifiers(&op[1]); } @@ -1013,7 +1013,7 @@ fs_visitor::nir_emit_alu(nir_alu_instr *instr) case nir_op_fmin: case nir_op_imin: case nir_op_umin: - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { inst = emit(BRW_OPCODE_SEL, result, op[0], op[1]); inst->conditional_mod = BRW_CONDITIONAL_L; } else { @@ -1027,7 +1027,7 @@ fs_visitor::nir_emit_alu(nir_alu_instr *instr) case nir_op_fmax: case nir_op_imax: case nir_op_umax: - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { inst = emit(BRW_OPCODE_SEL, result, op[0], op[1]); inst->conditional_mod = BRW_CONDITIONAL_GE; } else { @@ -1148,7 +1148,7 @@ fs_visitor::nir_emit_alu(nir_alu_instr *instr) /* If we need to do a boolean resolve, replace the result with -(x & 1) * to sign extend the low bit to 0/~0 */ - if (brw->gen <= 5 && + if (devinfo->gen <= 5 && (instr->instr.pass_flags & BRW_NIR_BOOLEAN_MASK) == BRW_NIR_BOOLEAN_NEEDS_RESOLVE) { fs_reg masked = vgrf(glsl_type::int_type); emit(AND(masked, result, fs_reg(1))); @@ -1257,7 +1257,7 @@ fs_visitor::nir_emit_intrinsic(nir_intrinsic_instr *instr) cmp->predicate = BRW_PREDICATE_NORMAL; cmp->flag_subreg = 1; - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { emit_discard_jump(); } break; @@ -1676,7 +1676,7 @@ fs_visitor::nir_emit_texture(nir_tex_instr *instr) case nir_tex_src_sampler_offset: { /* Figure out the highest possible sampler index and mark it as used */ uint32_t max_used = sampler + instr->sampler_array_size - 1; - if (instr->op == nir_texop_tg4 && brw->gen < 8) { + if (instr->op == nir_texop_tg4 && devinfo->gen < 8) { max_used += stage_prog_data->binding_table.gather_texture_start; } else { max_used += stage_prog_data->binding_table.texture_start; @@ -1696,7 +1696,7 @@ fs_visitor::nir_emit_texture(nir_tex_instr *instr) } if (instr->op == nir_texop_txf_ms) { - if (brw->gen >= 7 && + if (devinfo->gen >= 7 && key_tex->compressed_multisample_layout_mask & (1 << sampler)) { mcs = emit_mcs_fetch(coordinate, instr->coord_components, sampler_reg); } else { diff --git a/src/mesa/drivers/dri/i965/brw_fs_peephole_predicated_break.cpp b/src/mesa/drivers/dri/i965/brw_fs_peephole_predicated_break.cpp index 047c2c0b1cf..cf3da7b1882 100644 --- a/src/mesa/drivers/dri/i965/brw_fs_peephole_predicated_break.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs_peephole_predicated_break.cpp @@ -84,7 +84,7 @@ fs_visitor::opt_peephole_predicated_break() /* For Sandybridge with IF with embedded comparison we need to emit an * instruction to set the flag register. */ - if (brw->gen == 6 && if_inst->conditional_mod) { + if (devinfo->gen == 6 && if_inst->conditional_mod) { fs_inst *cmp_inst = CMP(reg_null_d, if_inst->src[0], if_inst->src[1], if_inst->conditional_mod); if_inst->insert_before(if_block, cmp_inst); diff --git a/src/mesa/drivers/dri/i965/brw_fs_reg_allocate.cpp b/src/mesa/drivers/dri/i965/brw_fs_reg_allocate.cpp index 47f5a42b187..92d29a4d530 100644 --- a/src/mesa/drivers/dri/i965/brw_fs_reg_allocate.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs_reg_allocate.cpp @@ -390,7 +390,7 @@ fs_visitor::setup_payload_interference(struct ra_graph *g, * in the arguments. Pre-gen6, the deltas are computed in normal * VGRFs. */ - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { int delta_x_arg = 0; if (inst->src[delta_x_arg].file == HW_REG && inst->src[delta_x_arg].fixed_hw_reg.file == @@ -443,7 +443,7 @@ fs_visitor::setup_payload_interference(struct ra_graph *g, * The alternative would be to have per-physical-register classes, which * would just be silly. */ - if (brw->intelScreen->devinfo->gen <= 5 && dispatch_width == 16) { + if (devinfo->gen <= 5 && dispatch_width == 16) { /* We have to divide by 2 here because we only have even numbered * registers. Some of the payload registers will be odd, but * that's ok because their physical register numbers have already @@ -541,7 +541,7 @@ fs_visitor::assign_regs(bool allow_spilling) int first_payload_node = node_count; node_count += payload_node_count; int first_mrf_hack_node = node_count; - if (brw->gen >= 7) + if (devinfo->gen >= 7) node_count += BRW_MAX_GRF - GEN7_MRF_HACK_START; struct ra_graph *g = ra_alloc_interference_graph(screen->wm_reg_sets[rsi].regs, node_count); @@ -579,7 +579,7 @@ fs_visitor::assign_regs(bool allow_spilling) } setup_payload_interference(g, payload_node_count, first_payload_node); - if (brw->gen >= 7) { + if (devinfo->gen >= 7) { setup_mrf_hack_interference(g, first_mrf_hack_node); foreach_block_and_inst(block, fs_inst, inst, cfg) { @@ -694,7 +694,7 @@ fs_visitor::emit_unspill(bblock_t *block, fs_inst *inst, fs_reg dst, for (int i = 0; i < count / reg_size; i++) { /* The gen7 descriptor-based offset is 12 bits of HWORD units. */ - bool gen7_read = brw->gen >= 7 && spill_offset < (1 << 12) * REG_SIZE; + bool gen7_read = devinfo->gen >= 7 && spill_offset < (1 << 12) * REG_SIZE; fs_inst *unspill_inst = new(mem_ctx) fs_inst(gen7_read ? diff --git a/src/mesa/drivers/dri/i965/brw_fs_sel_peephole.cpp b/src/mesa/drivers/dri/i965/brw_fs_sel_peephole.cpp index 740ba67bb94..52aa5590c2e 100644 --- a/src/mesa/drivers/dri/i965/brw_fs_sel_peephole.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs_sel_peephole.cpp @@ -158,7 +158,7 @@ fs_visitor::opt_peephole_sel() enum brw_predicate predicate; bool predicate_inverse; - if (brw->gen == 6 && if_inst->conditional_mod) { + if (devinfo->gen == 6 && if_inst->conditional_mod) { /* For Sandybridge with IF with embedded comparison */ predicate = BRW_PREDICATE_NORMAL; predicate_inverse = false; @@ -213,7 +213,7 @@ fs_visitor::opt_peephole_sel() continue; /* Emit a CMP if our IF used the embedded comparison */ - if (brw->gen == 6 && if_inst->conditional_mod) { + if (devinfo->gen == 6 && if_inst->conditional_mod) { fs_inst *cmp_inst = CMP(reg_null_d, if_inst->src[0], if_inst->src[1], if_inst->conditional_mod); if_inst->insert_before(block, cmp_inst); diff --git a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp index 4d1afb14074..11f38c26005 100644 --- a/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp +++ b/src/mesa/drivers/dri/i965/brw_fs_visitor.cpp @@ -197,7 +197,7 @@ fs_visitor::visit(ir_variable *ir) reg = emit_sampleid_setup(); break; case SYSTEM_VALUE_SAMPLE_MASK_IN: - assert(brw->gen >= 7); + assert(devinfo->gen >= 7); reg = new(mem_ctx) fs_reg(retype(brw_vec8_grf(payload.sample_mask_in_reg, 0), BRW_REGISTER_TYPE_D)); @@ -292,7 +292,7 @@ fs_inst * fs_visitor::emit_lrp(const fs_reg &dst, const fs_reg &x, const fs_reg &y, const fs_reg &a) { - if (brw->gen < 6) { + if (devinfo->gen < 6) { /* We can't use the LRP instruction. Emit x*(1-a) + y*a. */ fs_reg y_times_a = vgrf(glsl_type::float_type); fs_reg one_minus_a = vgrf(glsl_type::float_type); @@ -323,7 +323,7 @@ fs_visitor::emit_minmax(enum brw_conditional_mod conditionalmod, const fs_reg &d fs_inst *inst; - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { inst = emit(BRW_OPCODE_SEL, dst, src0, src1); inst->conditional_mod = conditionalmod; } else { @@ -420,7 +420,7 @@ bool fs_visitor::try_emit_mad(ir_expression *ir) { /* 3-src instructions were introduced in gen6. */ - if (brw->gen < 6) + if (devinfo->gen < 6) return false; /* MAD can only handle floating-point data. */ @@ -699,7 +699,7 @@ fs_visitor::visit(ir_expression *ir) /* Deal with the real oddball stuff first */ switch (ir->operation) { case ir_binop_add: - if (brw->gen <= 5 && try_emit_line(ir)) + if (devinfo->gen <= 5 && try_emit_line(ir)) return; if (try_emit_mad(ir)) return; @@ -719,7 +719,7 @@ fs_visitor::visit(ir_expression *ir) return; case ir_unop_b2f: - if (brw->gen <= 5 && try_emit_b2f_of_comparison(ir)) + if (devinfo->gen <= 5 && try_emit_b2f_of_comparison(ir)) return; break; @@ -860,24 +860,24 @@ fs_visitor::visit(ir_expression *ir) unreachable("not reached: should be handled by ir_sub_to_add_neg"); case ir_binop_mul: - if (brw->gen < 8 && ir->type->is_integer()) { + if (devinfo->gen < 8 && ir->type->is_integer()) { /* For integer multiplication, the MUL uses the low 16 bits * of one of the operands (src0 on gen6, src1 on gen7). The * MACH accumulates in the contribution of the upper 16 bits * of that operand. */ if (ir->operands[0]->is_uint16_constant()) { - if (brw->gen < 7) + if (devinfo->gen < 7) emit(MUL(this->result, op[0], op[1])); else emit(MUL(this->result, op[1], op[0])); } else if (ir->operands[1]->is_uint16_constant()) { - if (brw->gen < 7) + if (devinfo->gen < 7) emit(MUL(this->result, op[1], op[0])); else emit(MUL(this->result, op[0], op[1])); } else { - if (brw->gen >= 7) + if (devinfo->gen >= 7) no16("SIMD16 explicit accumulator operands unsupported\n"); struct brw_reg acc = retype(brw_acc_reg(dispatch_width), @@ -892,7 +892,7 @@ fs_visitor::visit(ir_expression *ir) } break; case ir_binop_imul_high: { - if (brw->gen == 7) + if (devinfo->gen == 7) no16("SIMD16 explicit accumulator operands unsupported\n"); struct brw_reg acc = retype(brw_acc_reg(dispatch_width), @@ -911,7 +911,7 @@ fs_visitor::visit(ir_expression *ir) * * FINISHME: Don't use source modifiers on src1. */ - if (brw->gen >= 8) { + if (devinfo->gen >= 8) { assert(mul->src[1].type == BRW_REGISTER_TYPE_D || mul->src[1].type == BRW_REGISTER_TYPE_UD); if (mul->src[1].type == BRW_REGISTER_TYPE_D) { @@ -929,7 +929,7 @@ fs_visitor::visit(ir_expression *ir) emit_math(SHADER_OPCODE_INT_QUOTIENT, this->result, op[0], op[1]); break; case ir_binop_carry: { - if (brw->gen == 7) + if (devinfo->gen == 7) no16("SIMD16 explicit accumulator operands unsupported\n"); struct brw_reg acc = retype(brw_acc_reg(dispatch_width), @@ -940,7 +940,7 @@ fs_visitor::visit(ir_expression *ir) break; } case ir_binop_borrow: { - if (brw->gen == 7) + if (devinfo->gen == 7) no16("SIMD16 explicit accumulator operands unsupported\n"); struct brw_reg acc = retype(brw_acc_reg(dispatch_width), @@ -964,7 +964,7 @@ fs_visitor::visit(ir_expression *ir) case ir_binop_all_equal: case ir_binop_nequal: case ir_binop_any_nequal: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { resolve_bool_comparison(ir->operands[0], &op[0]); resolve_bool_comparison(ir->operands[1], &op[1]); } @@ -1038,7 +1038,7 @@ fs_visitor::visit(ir_expression *ir) emit(AND(this->result, op[0], fs_reg(1))); break; case ir_unop_b2f: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { resolve_bool_comparison(ir->operands[0], &op[0]); } op[0].type = BRW_REGISTER_TYPE_D; @@ -1786,9 +1786,9 @@ fs_visitor::emit_texture_gen5(ir_texture_opcode op, fs_reg dst, } static bool -is_high_sampler(struct brw_context *brw, fs_reg sampler) +is_high_sampler(const struct brw_device_info *devinfo, fs_reg sampler) { - if (brw->gen < 8 && !brw->is_haswell) + if (devinfo->gen < 8 && !devinfo->is_haswell) return false; return sampler.file != IMM || sampler.fixed_hw_reg.dw1.ud >= 16; @@ -1812,7 +1812,7 @@ fs_visitor::emit_texture_gen7(ir_texture_opcode op, fs_reg dst, int length = 0; if (op == ir_tg4 || offset_value.file != BAD_FILE || - is_high_sampler(brw, sampler)) { + is_high_sampler(devinfo, sampler)) { /* For general texture offsets (no txf workaround), we need a header to * put them in. Note that for SIMD16 we're making space for two actual * hardware registers here, so the emit will have to fix up for this. @@ -1904,7 +1904,7 @@ fs_visitor::emit_texture_gen7(ir_texture_opcode op, fs_reg dst, coordinate = offset(coordinate, 1); length++; - if (brw->gen >= 9) { + if (devinfo->gen >= 9) { if (coord_components >= 2) { emit(MOV(retype(sources[length], BRW_REGISTER_TYPE_D), coordinate)); coordinate = offset(coordinate, 1); @@ -1915,7 +1915,7 @@ fs_visitor::emit_texture_gen7(ir_texture_opcode op, fs_reg dst, emit(MOV(retype(sources[length], BRW_REGISTER_TYPE_D), lod)); length++; - for (int i = brw->gen >= 9 ? 2 : 1; i < coord_components; i++) { + for (int i = devinfo->gen >= 9 ? 2 : 1; i < coord_components; i++) { emit(MOV(retype(sources[length], BRW_REGISTER_TYPE_D), coordinate)); coordinate = offset(coordinate, 1); length++; @@ -2038,9 +2038,9 @@ fs_visitor::rescale_texcoord(fs_reg coordinate, int coord_components, * tracking to get the scaling factor. */ if (is_rect && - (brw->gen < 6 || - (brw->gen >= 6 && (key_tex->gl_clamp_mask[0] & (1 << sampler) || - key_tex->gl_clamp_mask[1] & (1 << sampler))))) { + (devinfo->gen < 6 || + (devinfo->gen >= 6 && (key_tex->gl_clamp_mask[0] & (1 << sampler) || + key_tex->gl_clamp_mask[1] & (1 << sampler))))) { struct gl_program_parameter_list *params = prog->Parameters; int tokens[STATE_LENGTH] = { STATE_INTERNAL, @@ -2083,7 +2083,7 @@ fs_visitor::rescale_texcoord(fs_reg coordinate, int coord_components, * texture coordinates. We use the program parameter state * tracking to get the scaling factor. */ - if (brw->gen < 6 && is_rect) { + if (devinfo->gen < 6 && is_rect) { fs_reg dst = fs_reg(GRF, alloc.allocate(coord_components)); fs_reg src = coordinate; coordinate = dst; @@ -2217,12 +2217,12 @@ fs_visitor::emit_texture(ir_texture_opcode op, */ fs_reg dst = vgrf(glsl_type::get_instance(dest_type->base_type, 4, 1)); - if (brw->gen >= 7) { + if (devinfo->gen >= 7) { inst = emit_texture_gen7(op, dst, coordinate, coord_components, shadow_c, lod, lod2, grad_components, sample_index, mcs, sampler_reg, offset_value); - } else if (brw->gen >= 5) { + } else if (devinfo->gen >= 5) { inst = emit_texture_gen5(op, dst, coordinate, coord_components, shadow_c, lod, lod2, grad_components, sample_index, sampler, @@ -2246,7 +2246,7 @@ fs_visitor::emit_texture(ir_texture_opcode op, inst->offset |= gather_channel(gather_component, sampler) << 16; /* M0.2:16-17 */ - if (brw->gen == 6) + if (devinfo->gen == 6) emit_gen6_gather_wa(key_tex->gen6_gather_wa[sampler], dst); } @@ -2291,7 +2291,7 @@ fs_visitor::visit(ir_texture *ir) ->array->type->array_size(); uint32_t max_used = sampler + array_size - 1; - if (ir->op == ir_tg4 && brw->gen < 8) { + if (ir->op == ir_tg4 && devinfo->gen < 8) { max_used += stage_prog_data->binding_table.gather_texture_start; } else { max_used += stage_prog_data->binding_table.texture_start; @@ -2390,7 +2390,7 @@ fs_visitor::visit(ir_texture *ir) ir->lod_info.sample_index->accept(this); sample_index = this->result; - if (brw->gen >= 7 && + if (devinfo->gen >= 7 && key_tex->compressed_multisample_layout_mask & (1 << sampler)) { mcs = emit_mcs_fetch(coordinate, ir->coordinate->type->vector_elements, sampler_reg); @@ -2578,7 +2578,7 @@ fs_visitor::visit(ir_discard *ir) cmp->predicate = BRW_PREDICATE_NORMAL; cmp->flag_subreg = 1; - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { emit_discard_jump(); } } @@ -2691,7 +2691,7 @@ fs_visitor::emit_bool_to_cond_code_of_reg(ir_expression *expr, fs_reg op[3]) break; case ir_binop_logic_xor: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { fs_reg temp = vgrf(expr->type); emit(XOR(temp, op[0], op[1])); inst = emit(AND(reg_null_d, temp, fs_reg(1))); @@ -2702,7 +2702,7 @@ fs_visitor::emit_bool_to_cond_code_of_reg(ir_expression *expr, fs_reg op[3]) break; case ir_binop_logic_or: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { fs_reg temp = vgrf(expr->type); emit(OR(temp, op[0], op[1])); inst = emit(AND(reg_null_d, temp, fs_reg(1))); @@ -2713,7 +2713,7 @@ fs_visitor::emit_bool_to_cond_code_of_reg(ir_expression *expr, fs_reg op[3]) break; case ir_binop_logic_and: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { fs_reg temp = vgrf(expr->type); emit(AND(temp, op[0], op[1])); inst = emit(AND(reg_null_d, temp, fs_reg(1))); @@ -2724,7 +2724,7 @@ fs_visitor::emit_bool_to_cond_code_of_reg(ir_expression *expr, fs_reg op[3]) break; case ir_unop_f2b: - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { emit(CMP(reg_null_d, op[0], fs_reg(0.0f), BRW_CONDITIONAL_NZ)); } else { inst = emit(MOV(reg_null_f, op[0])); @@ -2733,7 +2733,7 @@ fs_visitor::emit_bool_to_cond_code_of_reg(ir_expression *expr, fs_reg op[3]) break; case ir_unop_i2b: - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { emit(CMP(reg_null_d, op[0], fs_reg(0), BRW_CONDITIONAL_NZ)); } else { inst = emit(MOV(reg_null_d, op[0])); @@ -2749,7 +2749,7 @@ fs_visitor::emit_bool_to_cond_code_of_reg(ir_expression *expr, fs_reg op[3]) case ir_binop_all_equal: case ir_binop_nequal: case ir_binop_any_nequal: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { resolve_bool_comparison(expr->operands[0], &op[0]); resolve_bool_comparison(expr->operands[1], &op[1]); } @@ -2839,7 +2839,7 @@ fs_visitor::emit_if_gen6(ir_if *ir) case ir_binop_all_equal: case ir_binop_nequal: case ir_binop_any_nequal: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { resolve_bool_comparison(expr->operands[0], &op[0]); resolve_bool_comparison(expr->operands[1], &op[1]); } @@ -2909,7 +2909,7 @@ fs_visitor::try_opt_frontfacing_ternary(ir_if *ir) dst.type = BRW_REGISTER_TYPE_D; fs_reg tmp = vgrf(glsl_type::int_type); - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { /* Bit 15 of g0.0 is 0 if the polygon is front facing. */ fs_reg g0 = fs_reg(retype(brw_vec1_grf(0, 0), BRW_REGISTER_TYPE_W)); @@ -3062,7 +3062,7 @@ fs_visitor::visit(ir_if *ir) */ this->base_ir = ir->condition; - if (brw->gen == 6) { + if (devinfo->gen == 6) { emit_if_gen6(ir); } else { emit_bool_to_cond_code(ir->condition); @@ -3086,7 +3086,7 @@ fs_visitor::visit(ir_if *ir) emit(BRW_OPCODE_ENDIF); - if (!try_replace_with_sel() && brw->gen < 6) { + if (!try_replace_with_sel() && devinfo->gen < 6) { no16("Can't support (non-uniform) control flow on SIMD16\n"); } } @@ -3094,7 +3094,7 @@ fs_visitor::visit(ir_if *ir) void fs_visitor::visit(ir_loop *ir) { - if (brw->gen < 6) { + if (devinfo->gen < 6) { no16("Can't support (non-uniform) control flow on SIMD16\n"); } @@ -3375,7 +3375,7 @@ fs_visitor::emit_dummy_fs() fs_inst *write; write = emit(FS_OPCODE_FB_WRITE); write->eot = true; - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { write->base_mrf = 2; write->mlen = 4 * reg_width; } else { @@ -3388,7 +3388,7 @@ fs_visitor::emit_dummy_fs() * varying to avoid GPU hangs, so set that. */ brw_wm_prog_data *wm_prog_data = (brw_wm_prog_data *) this->prog_data; - wm_prog_data->num_varying_inputs = brw->gen < 6 ? 1 : 0; + wm_prog_data->num_varying_inputs = devinfo->gen < 6 ? 1 : 0; memset(wm_prog_data->urb_setup, -1, sizeof(wm_prog_data->urb_setup[0]) * VARYING_SLOT_MAX); @@ -3446,7 +3446,7 @@ fs_visitor::emit_interpolation_setup_gen4() const fs_reg xstart(negate(brw_vec1_grf(1, 0))); const fs_reg ystart(negate(brw_vec1_grf(1, 1))); - if (brw->has_pln && dispatch_width == 16) { + if (devinfo->has_pln && dispatch_width == 16) { emit(ADD(half(offset(delta_xy, 0), 0), half(this->pixel_x, 0), xstart)); emit(ADD(half(offset(delta_xy, 0), 1), half(this->pixel_y, 0), ystart)); emit(ADD(half(offset(delta_xy, 1), 0), half(this->pixel_x, 1), xstart)) @@ -3562,7 +3562,7 @@ fs_visitor::setup_color_payload(fs_reg *dst, fs_reg color, unsigned components, colors_enabled = (1 << components) - 1; } - if (dispatch_width == 8 || (brw->gen >= 6 && !do_dual_src)) { + if (dispatch_width == 8 || (devinfo->gen >= 6 && !do_dual_src)) { /* SIMD8 write looks like: * m + 0: r0 * m + 1: r1 @@ -3593,7 +3593,7 @@ fs_visitor::setup_color_payload(fs_reg *dst, fs_reg color, unsigned components, len++; } return len; - } else if (brw->gen >= 6 && do_dual_src) { + } else if (devinfo->gen >= 6 && do_dual_src) { /* SIMD16 dual source blending for gen6+. * * From the SNB PRM, volume 4, part 1, page 193: @@ -3727,8 +3727,8 @@ fs_visitor::emit_single_fb_write(fs_reg color0, fs_reg color1, * dispatched. This field is only required for the end-of- * thread message and on all dual-source messages." */ - if (brw->gen >= 6 && - (brw->is_haswell || brw->gen >= 8 || !prog_data->uses_kill) && + if (devinfo->gen >= 6 && + (devinfo->is_haswell || devinfo->gen >= 8 || !prog_data->uses_kill) && color1.file == BAD_FILE && key->nr_color_regions == 1) { header_present = false; @@ -3785,7 +3785,7 @@ fs_visitor::emit_single_fb_write(fs_reg color0, fs_reg color1, } if (source_depth_to_render_target) { - if (brw->gen == 6) { + if (devinfo->gen == 6) { /* For outputting oDepth on gen6, SIMD8 writes have to be * used. This would require SIMD8 moves of each half to * message regs, kind of like pre-gen5 SIMD16 FB writes. @@ -3816,7 +3816,7 @@ fs_visitor::emit_single_fb_write(fs_reg color0, fs_reg color1, fs_inst *load; fs_inst *write; - if (brw->gen >= 7) { + if (devinfo->gen >= 7) { /* Send from the GRF */ fs_reg payload = fs_reg(GRF, -1, BRW_REGISTER_TYPE_F); load = emit(LOAD_PAYLOAD(payload, sources, length)); @@ -3892,7 +3892,7 @@ fs_visitor::emit_fb_writes() "FB write target %d", target); fs_reg src0_alpha; - if (brw->gen >= 6 && key->replicate_alpha && target != 0) + if (devinfo->gen >= 6 && key->replicate_alpha && target != 0) src0_alpha = offset(outputs[0], 3); inst = emit_single_fb_write(this->outputs[target], reg_undef, @@ -4163,7 +4163,7 @@ fs_visitor::resolve_ud_negate(fs_reg *reg) void fs_visitor::resolve_bool_comparison(ir_rvalue *rvalue, fs_reg *reg) { - assert(brw->gen <= 5); + assert(devinfo->gen <= 5); if (rvalue->type != glsl_type::bool_type) return; @@ -4244,7 +4244,7 @@ fs_visitor::init() this->source_depth_to_render_target = false; this->runtime_check_aads_emit = false; this->first_non_payload_grf = 0; - this->max_grf = brw->gen >= 7 ? GEN7_MRF_HACK_START : BRW_MAX_GRF; + this->max_grf = devinfo->gen >= 7 ? GEN7_MRF_HACK_START : BRW_MAX_GRF; this->current_annotation = NULL; this->base_ir = NULL; diff --git a/src/mesa/drivers/dri/i965/brw_ir_fs.h b/src/mesa/drivers/dri/i965/brw_ir_fs.h index 30c19f41a31..1e54bb23301 100644 --- a/src/mesa/drivers/dri/i965/brw_ir_fs.h +++ b/src/mesa/drivers/dri/i965/brw_ir_fs.h @@ -224,7 +224,7 @@ public: bool is_send_from_grf() const; bool is_partial_write() const; int regs_read(int arg) const; - bool can_do_source_mods(struct brw_context *brw); + bool can_do_source_mods(const struct brw_device_info *devinfo); bool has_side_effects() const; bool reads_flag() const; diff --git a/src/mesa/drivers/dri/i965/brw_ir_vec4.h b/src/mesa/drivers/dri/i965/brw_ir_vec4.h index d3bd64dbb71..955a62d5f9e 100644 --- a/src/mesa/drivers/dri/i965/brw_ir_vec4.h +++ b/src/mesa/drivers/dri/i965/brw_ir_vec4.h @@ -168,7 +168,7 @@ public: unsigned regs_read(unsigned arg) const; bool can_reswizzle(int dst_writemask, int swizzle, int swizzle_mask); void reswizzle(int dst_writemask, int swizzle); - bool can_do_source_mods(struct brw_context *brw); + bool can_do_source_mods(const struct brw_device_info *devinfo); bool reads_flag() { diff --git a/src/mesa/drivers/dri/i965/brw_shader.cpp b/src/mesa/drivers/dri/i965/brw_shader.cpp index 91ffb532d09..d90eac120cf 100644 --- a/src/mesa/drivers/dri/i965/brw_shader.cpp +++ b/src/mesa/drivers/dri/i965/brw_shader.cpp @@ -721,6 +721,7 @@ backend_visitor::backend_visitor(struct brw_context *brw, struct brw_stage_prog_data *stage_prog_data, gl_shader_stage stage) : brw(brw), + devinfo(brw->intelScreen->devinfo), ctx(&brw->ctx), shader(shader_prog ? (struct brw_shader *)shader_prog->_LinkedShaders[stage] : NULL), @@ -1183,7 +1184,7 @@ backend_visitor::assign_common_binding_table_offsets(uint32_t next_binding_table } if (prog->UsesGather) { - if (brw->gen >= 8) { + if (devinfo->gen >= 8) { stage_prog_data->binding_table.gather_texture_start = stage_prog_data->binding_table.texture_start; } else { diff --git a/src/mesa/drivers/dri/i965/brw_shader.h b/src/mesa/drivers/dri/i965/brw_shader.h index e6246ed7fc3..b80e7405dbc 100644 --- a/src/mesa/drivers/dri/i965/brw_shader.h +++ b/src/mesa/drivers/dri/i965/brw_shader.h @@ -169,6 +169,7 @@ protected: public: struct brw_context * const brw; + const struct brw_device_info * const devinfo; struct gl_context * const ctx; struct brw_shader * const shader; struct gl_shader_program * const shader_prog; diff --git a/src/mesa/drivers/dri/i965/brw_vec4.cpp b/src/mesa/drivers/dri/i965/brw_vec4.cpp index c4c77b28760..9398adce277 100644 --- a/src/mesa/drivers/dri/i965/brw_vec4.cpp +++ b/src/mesa/drivers/dri/i965/brw_vec4.cpp @@ -239,9 +239,9 @@ vec4_instruction::regs_read(unsigned arg) const } bool -vec4_instruction::can_do_source_mods(struct brw_context *brw) +vec4_instruction::can_do_source_mods(const struct brw_device_info *devinfo) { - if (brw->gen == 6 && is_math()) + if (devinfo->gen == 6 && is_math()) return false; if (is_send_from_grf()) @@ -783,7 +783,7 @@ vec4_visitor::is_dep_ctrl_unsafe(const vec4_instruction *inst) * multiply, DepCtrl must not be used." * May apply to future SoCs as well. */ - if (brw->is_cherryview) { + if (devinfo->is_cherryview) { if (inst->opcode == BRW_OPCODE_MUL && IS_DWORD(inst->src[0]) && IS_DWORD(inst->src[1])) @@ -791,7 +791,7 @@ vec4_visitor::is_dep_ctrl_unsafe(const vec4_instruction *inst) } #undef IS_DWORD - if (brw->gen >= 8) { + if (devinfo->gen >= 8) { if (inst->opcode == BRW_OPCODE_F32TO16) return true; } @@ -1012,7 +1012,7 @@ vec4_visitor::opt_register_coalesce() if (scan_inst->mlen) break; - if (brw->gen == 6) { + if (devinfo->gen == 6) { /* gen6 math instructions must have the destination be * GRF, so no compute-to-MRF for them. */ @@ -1209,9 +1209,9 @@ vec4_visitor::dump_instruction(backend_instruction *be_inst, FILE *file) if (inst->conditional_mod) { fprintf(file, "%s", conditional_modifier[inst->conditional_mod]); if (!inst->predicate && - (brw->gen < 5 || (inst->opcode != BRW_OPCODE_SEL && - inst->opcode != BRW_OPCODE_IF && - inst->opcode != BRW_OPCODE_WHILE))) { + (devinfo->gen < 5 || (inst->opcode != BRW_OPCODE_SEL && + inst->opcode != BRW_OPCODE_IF && + inst->opcode != BRW_OPCODE_WHILE))) { fprintf(file, ".f0.%d", inst->flag_subreg); } } @@ -1488,7 +1488,7 @@ vec4_vs_visitor::setup_attributes(int payload_reg) unsigned vue_entries = MAX2(nr_attributes, prog_data->vue_map.num_slots); - if (brw->gen == 6) + if (devinfo->gen == 6) prog_data->urb_entry_size = ALIGN(vue_entries, 8) / 8; else prog_data->urb_entry_size = ALIGN(vue_entries, 4) / 4; @@ -1504,7 +1504,7 @@ vec4_visitor::setup_uniforms(int reg) /* The pre-gen6 VS requires that some push constants get loaded no * matter what, or the GPU would hang. */ - if (brw->gen < 6 && this->uniforms == 0) { + if (devinfo->gen < 6 && this->uniforms == 0) { assert(this->uniforms < this->uniform_array_size); this->uniform_vector_size[this->uniforms] = 1; @@ -1557,7 +1557,7 @@ vec4_visitor::assign_binding_table_offsets() src_reg vec4_visitor::get_timestamp() { - assert(brw->gen >= 7); + assert(devinfo->gen >= 7); src_reg ts = src_reg(brw_reg(BRW_ARCHITECTURE_REGISTER_FILE, BRW_ARF_TIMESTAMP, diff --git a/src/mesa/drivers/dri/i965/brw_vec4_copy_propagation.cpp b/src/mesa/drivers/dri/i965/brw_vec4_copy_propagation.cpp index e897be299ba..0fbf0ab50a4 100644 --- a/src/mesa/drivers/dri/i965/brw_vec4_copy_propagation.cpp +++ b/src/mesa/drivers/dri/i965/brw_vec4_copy_propagation.cpp @@ -104,7 +104,8 @@ is_logic_op(enum opcode opcode) } static bool -try_constant_propagate(struct brw_context *brw, vec4_instruction *inst, +try_constant_propagate(const struct brw_device_info *devinfo, + vec4_instruction *inst, int arg, struct copy_entry *entry) { /* For constant propagation, we only handle the same constant @@ -132,14 +133,14 @@ try_constant_propagate(struct brw_context *brw, vec4_instruction *inst, } if (inst->src[arg].abs) { - if ((brw->gen >= 8 && is_logic_op(inst->opcode)) || + if ((devinfo->gen >= 8 && is_logic_op(inst->opcode)) || !brw_abs_immediate(value.type, &value.fixed_hw_reg)) { return false; } } if (inst->src[arg].negate) { - if ((brw->gen >= 8 && is_logic_op(inst->opcode)) || + if ((devinfo->gen >= 8 && is_logic_op(inst->opcode)) || !brw_negate_immediate(value.type, &value.fixed_hw_reg)) { return false; } @@ -157,7 +158,7 @@ try_constant_propagate(struct brw_context *brw, vec4_instruction *inst, case SHADER_OPCODE_POW: case SHADER_OPCODE_INT_QUOTIENT: case SHADER_OPCODE_INT_REMAINDER: - if (brw->gen < 8) + if (devinfo->gen < 8) break; /* fallthrough */ case BRW_OPCODE_DP2: @@ -246,7 +247,8 @@ try_constant_propagate(struct brw_context *brw, vec4_instruction *inst, } static bool -try_copy_propagate(struct brw_context *brw, vec4_instruction *inst, +try_copy_propagate(const struct brw_device_info *devinfo, + vec4_instruction *inst, int arg, struct copy_entry *entry) { /* For constant propagation, we only handle the same constant @@ -283,7 +285,7 @@ try_copy_propagate(struct brw_context *brw, vec4_instruction *inst, value.file != ATTR) return false; - if (brw->gen >= 8 && (value.negate || value.abs) && + if (devinfo->gen >= 8 && (value.negate || value.abs) && is_logic_op(inst->opcode)) { return false; } @@ -301,7 +303,7 @@ try_copy_propagate(struct brw_context *brw, vec4_instruction *inst, * instructions. */ if ((has_source_modifiers || value.file == UNIFORM || - value.swizzle != BRW_SWIZZLE_XYZW) && !inst->can_do_source_mods(brw)) + value.swizzle != BRW_SWIZZLE_XYZW) && !inst->can_do_source_mods(devinfo)) return false; if (has_source_modifiers && value.type != inst->src[arg].type) @@ -433,10 +435,10 @@ vec4_visitor::opt_copy_propagation(bool do_constant_prop) if (c != 4) continue; - if (do_constant_prop && try_constant_propagate(brw, inst, i, &entry)) + if (do_constant_prop && try_constant_propagate(devinfo, inst, i, &entry)) progress = true; - if (try_copy_propagate(brw, inst, i, &entry)) + if (try_copy_propagate(devinfo, inst, i, &entry)) progress = true; } diff --git a/src/mesa/drivers/dri/i965/brw_vec4_dead_code_eliminate.cpp b/src/mesa/drivers/dri/i965/brw_vec4_dead_code_eliminate.cpp index 70d2af5e911..8fc7a365bfc 100644 --- a/src/mesa/drivers/dri/i965/brw_vec4_dead_code_eliminate.cpp +++ b/src/mesa/drivers/dri/i965/brw_vec4_dead_code_eliminate.cpp @@ -37,7 +37,7 @@ using namespace brw; static bool -can_do_writemask(const struct brw_context *brw, +can_do_writemask(const struct brw_device_info *devinfo, const vec4_instruction *inst) { switch (inst->opcode) { @@ -50,7 +50,7 @@ can_do_writemask(const struct brw_context *brw, /* The MATH instruction on Gen6 only executes in align1 mode, which does * not support writemasking. */ - if (brw->gen == 6 && inst->is_math()) + if (devinfo->gen == 6 && inst->is_math()) return false; if (inst->is_tex()) @@ -90,7 +90,7 @@ vec4_visitor::dead_code_eliminate() /* If the instruction can't do writemasking, then it's all or * nothing. */ - if (!can_do_writemask(brw, inst)) { + if (!can_do_writemask(devinfo, inst)) { bool result = result_live[0] | result_live[1] | result_live[2] | result_live[3]; result_live[0] = result; diff --git a/src/mesa/drivers/dri/i965/brw_vec4_gs_visitor.cpp b/src/mesa/drivers/dri/i965/brw_vec4_gs_visitor.cpp index 18a44be7370..363e30e34e4 100644 --- a/src/mesa/drivers/dri/i965/brw_vec4_gs_visitor.cpp +++ b/src/mesa/drivers/dri/i965/brw_vec4_gs_visitor.cpp @@ -281,7 +281,7 @@ vec4_gs_visitor::emit_urb_write_opcode(bool complete) /* We need to increment Global Offset by 1 to make room for Broadwell's * extra "Vertex Count" payload at the beginning of the URB entry. */ - if (brw->gen >= 8) + if (devinfo->gen >= 8) inst->offset++; inst->urb_write_flags = BRW_URB_WRITE_PER_SLOT_OFFSET; @@ -424,7 +424,7 @@ vec4_gs_visitor::emit_control_data_bits() * URB entry. Since this is an OWord message, Global Offset is counted * in 128-bit units, so we must set it to 2. */ - if (brw->gen >= 8) + if (devinfo->gen >= 8) inst->offset = 2; inst->base_mrf = base_mrf; inst->mlen = 2; diff --git a/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp b/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp index c9a496e6c74..ec6913b788f 100644 --- a/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp +++ b/src/mesa/drivers/dri/i965/brw_vec4_visitor.cpp @@ -146,7 +146,7 @@ vec4_visitor::emit(enum opcode opcode) vec4_visitor::op(const dst_reg &dst, const src_reg &src0, \ const src_reg &src1, const src_reg &src2) \ { \ - assert(brw->gen >= 6); \ + assert(devinfo->gen >= 6); \ return new(mem_ctx) vec4_instruction(BRW_OPCODE_##op, dst, \ src0, src1, src2); \ } @@ -201,7 +201,7 @@ vec4_instruction * vec4_visitor::IF(src_reg src0, src_reg src1, enum brw_conditional_mod condition) { - assert(brw->gen == 6); + assert(devinfo->gen == 6); vec4_instruction *inst; @@ -315,7 +315,7 @@ vec4_visitor::fix_3src_operand(src_reg src) src_reg vec4_visitor::fix_math_operand(src_reg src) { - if (brw->gen < 6 || brw->gen >= 8 || src.file == BAD_FILE) + if (devinfo->gen < 6 || devinfo->gen >= 8 || src.file == BAD_FILE) return src; /* The gen6 math instruction ignores the source modifiers -- @@ -329,7 +329,7 @@ vec4_visitor::fix_math_operand(src_reg src) * can't use. */ - if (brw->gen == 7 && src.file != IMM) + if (devinfo->gen == 7 && src.file != IMM) return src; dst_reg expanded = dst_reg(this, glsl_type::vec4_type); @@ -346,12 +346,12 @@ vec4_visitor::emit_math(enum opcode opcode, vec4_instruction *math = emit(opcode, dst, fix_math_operand(src0), fix_math_operand(src1)); - if (brw->gen == 6 && dst.writemask != WRITEMASK_XYZW) { + if (devinfo->gen == 6 && dst.writemask != WRITEMASK_XYZW) { /* MATH on Gen6 must be align1, so we can't do writemasks. */ math->dst = dst_reg(this, glsl_type::vec4_type); math->dst.type = dst.type; emit(MOV(dst, src_reg(math->dst))); - } else if (brw->gen < 6) { + } else if (devinfo->gen < 6) { math->base_mrf = 1; math->mlen = src1.file == BAD_FILE ? 1 : 2; } @@ -360,7 +360,7 @@ vec4_visitor::emit_math(enum opcode opcode, void vec4_visitor::emit_pack_half_2x16(dst_reg dst, src_reg src0) { - if (brw->gen < 7) { + if (devinfo->gen < 7) { unreachable("ir_unop_pack_half_2x16 should be lowered"); } @@ -437,7 +437,7 @@ vec4_visitor::emit_pack_half_2x16(dst_reg dst, src_reg src0) void vec4_visitor::emit_unpack_half_2x16(dst_reg dst, src_reg src0) { - if (brw->gen < 7) { + if (devinfo->gen < 7) { unreachable("ir_unop_unpack_half_2x16 should be lowered"); } @@ -802,7 +802,7 @@ vec4_visitor::emit_bool_to_cond_code(ir_rvalue *ir, break; case ir_binop_logic_xor: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { src_reg temp = src_reg(this, ir->type); emit(XOR(dst_reg(temp), op[0], op[1])); inst = emit(AND(dst_null_d(), temp, src_reg(1))); @@ -813,7 +813,7 @@ vec4_visitor::emit_bool_to_cond_code(ir_rvalue *ir, break; case ir_binop_logic_or: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { src_reg temp = src_reg(this, ir->type); emit(OR(dst_reg(temp), op[0], op[1])); inst = emit(AND(dst_null_d(), temp, src_reg(1))); @@ -824,7 +824,7 @@ vec4_visitor::emit_bool_to_cond_code(ir_rvalue *ir, break; case ir_binop_logic_and: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { src_reg temp = src_reg(this, ir->type); emit(AND(dst_reg(temp), op[0], op[1])); inst = emit(AND(dst_null_d(), temp, src_reg(1))); @@ -835,7 +835,7 @@ vec4_visitor::emit_bool_to_cond_code(ir_rvalue *ir, break; case ir_unop_f2b: - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { emit(CMP(dst_null_d(), op[0], src_reg(0.0f), BRW_CONDITIONAL_NZ)); } else { inst = emit(MOV(dst_null_f(), op[0])); @@ -844,7 +844,7 @@ vec4_visitor::emit_bool_to_cond_code(ir_rvalue *ir, break; case ir_unop_i2b: - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { emit(CMP(dst_null_d(), op[0], src_reg(0), BRW_CONDITIONAL_NZ)); } else { inst = emit(MOV(dst_null_d(), op[0])); @@ -853,7 +853,7 @@ vec4_visitor::emit_bool_to_cond_code(ir_rvalue *ir, break; case ir_binop_all_equal: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { resolve_bool_comparison(expr->operands[0], &op[0]); resolve_bool_comparison(expr->operands[1], &op[1]); } @@ -862,7 +862,7 @@ vec4_visitor::emit_bool_to_cond_code(ir_rvalue *ir, break; case ir_binop_any_nequal: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { resolve_bool_comparison(expr->operands[0], &op[0]); resolve_bool_comparison(expr->operands[1], &op[1]); } @@ -871,7 +871,7 @@ vec4_visitor::emit_bool_to_cond_code(ir_rvalue *ir, break; case ir_unop_any: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { resolve_bool_comparison(expr->operands[0], &op[0]); } inst = emit(CMP(dst_null_d(), op[0], src_reg(0), BRW_CONDITIONAL_NZ)); @@ -884,7 +884,7 @@ vec4_visitor::emit_bool_to_cond_code(ir_rvalue *ir, case ir_binop_lequal: case ir_binop_equal: case ir_binop_nequal: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { resolve_bool_comparison(expr->operands[0], &op[0]); resolve_bool_comparison(expr->operands[1], &op[1]); } @@ -1147,7 +1147,7 @@ bool vec4_visitor::try_emit_mad(ir_expression *ir) { /* 3-src instructions were introduced in gen6. */ - if (brw->gen < 6) + if (devinfo->gen < 6) return false; /* MAD can only handle floating-point data. */ @@ -1209,7 +1209,7 @@ vec4_visitor::try_emit_b2f_of_compare(ir_expression *ir) * false. Early hardware only sets the least significant bit, and * leaves the other bits undefined. So we can't use it. */ - if (brw->gen < 6) + if (devinfo->gen < 6) return false; ir_expression *const cmp = ir->operands[0]->as_expression(); @@ -1257,7 +1257,7 @@ vec4_visitor::emit_minmax(enum brw_conditional_mod conditionalmod, dst_reg dst, { vec4_instruction *inst; - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { inst = emit(BRW_OPCODE_SEL, dst, src0, src1); inst->conditional_mod = conditionalmod; } else { @@ -1272,7 +1272,7 @@ void vec4_visitor::emit_lrp(const dst_reg &dst, const src_reg &x, const src_reg &y, const src_reg &a) { - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { /* Note that the instruction's argument order is reversed from GLSL * and the IR. */ @@ -1313,7 +1313,7 @@ vec4_visitor::emit_pull_constant_load_reg(dst_reg dst, vec4_instruction *pull; - if (brw->gen >= 9) { + if (devinfo->gen >= 9) { /* Gen9+ needs a message header in order to use SIMD4x2 mode */ src_reg header(this, glsl_type::uvec4_type, 2); @@ -1341,7 +1341,7 @@ vec4_visitor::emit_pull_constant_load_reg(dst_reg dst, header); pull->mlen = 2; pull->header_present = true; - } else if (brw->gen >= 7) { + } else if (devinfo->gen >= 7) { dst_reg grf_offset = dst_reg(this, glsl_type::int_type); grf_offset.type = offset_reg.type; @@ -1549,7 +1549,7 @@ vec4_visitor::visit(ir_expression *ir) unreachable("not reached: should be handled by ir_sub_to_add_neg"); case ir_binop_mul: - if (brw->gen < 8 && ir->type->is_integer()) { + if (devinfo->gen < 8 && ir->type->is_integer()) { /* For integer multiplication, the MUL uses the low 16 bits of one of * the operands (src0 through SNB, src1 on IVB and later). The MACH * accumulates in the contribution of the upper 16 bits of that @@ -1557,12 +1557,12 @@ vec4_visitor::visit(ir_expression *ir) * 16 bits, though, we can just emit a single MUL. */ if (ir->operands[0]->is_uint16_constant()) { - if (brw->gen < 7) + if (devinfo->gen < 7) emit(MUL(result_dst, op[0], op[1])); else emit(MUL(result_dst, op[1], op[0])); } else if (ir->operands[1]->is_uint16_constant()) { - if (brw->gen < 7) + if (devinfo->gen < 7) emit(MUL(result_dst, op[1], op[0])); else emit(MUL(result_dst, op[0], op[1])); @@ -1615,7 +1615,7 @@ vec4_visitor::visit(ir_expression *ir) case ir_binop_gequal: case ir_binop_equal: case ir_binop_nequal: { - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { resolve_bool_comparison(ir->operands[0], &op[0]); resolve_bool_comparison(ir->operands[1], &op[1]); } @@ -1625,7 +1625,7 @@ vec4_visitor::visit(ir_expression *ir) } case ir_binop_all_equal: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { resolve_bool_comparison(ir->operands[0], &op[0]); resolve_bool_comparison(ir->operands[1], &op[1]); } @@ -1642,7 +1642,7 @@ vec4_visitor::visit(ir_expression *ir) } break; case ir_binop_any_nequal: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { resolve_bool_comparison(ir->operands[0], &op[0]); resolve_bool_comparison(ir->operands[1], &op[1]); } @@ -1661,7 +1661,7 @@ vec4_visitor::visit(ir_expression *ir) break; case ir_unop_any: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { resolve_bool_comparison(ir->operands[0], &op[0]); } emit(CMP(dst_null_d(), op[0], src_reg(0), BRW_CONDITIONAL_NZ)); @@ -1724,7 +1724,7 @@ vec4_visitor::visit(ir_expression *ir) emit(AND(result_dst, op[0], src_reg(1))); break; case ir_unop_b2f: - if (brw->gen <= 5) { + if (devinfo->gen <= 5) { resolve_bool_comparison(ir->operands[0], &op[0]); } op[0].type = BRW_REGISTER_TYPE_D; @@ -1836,7 +1836,7 @@ vec4_visitor::visit(ir_expression *ir) } if (const_offset_ir) { - if (brw->gen >= 8) { + if (devinfo->gen >= 8) { /* Store the offset in a GRF so we can send-from-GRF. */ offset = src_reg(this, glsl_type::int_type); emit(MOV(dst_reg(offset), src_reg(const_offset / 16))); @@ -2467,9 +2467,9 @@ vec4_visitor::emit_mcs_fetch(ir_texture *ir, src_reg coordinate, src_reg sampler } static bool -is_high_sampler(struct brw_context *brw, src_reg sampler) +is_high_sampler(const struct brw_device_info *devinfo, src_reg sampler) { - if (brw->gen < 8 && !brw->is_haswell) + if (devinfo->gen < 8 && !devinfo->is_haswell) return false; return sampler.file != IMM || sampler.fixed_hw_reg.dw1.ud >= 16; @@ -2495,7 +2495,7 @@ vec4_visitor::visit(ir_texture *ir) ->array->type->array_size(); uint32_t max_used = sampler + array_size - 1; - if (ir->op == ir_tg4 && brw->gen < 8) { + if (ir->op == ir_tg4 && devinfo->gen < 8) { max_used += prog_data->base.binding_table.gather_texture_start; } else { max_used += prog_data->base.binding_table.texture_start; @@ -2582,7 +2582,7 @@ vec4_visitor::visit(ir_texture *ir) sample_index = this->result; sample_index_type = ir->lod_info.sample_index->type; - if (brw->gen >= 7 && key->tex.compressed_multisample_layout_mask & (1<gen >= 7 && key->tex.compressed_multisample_layout_mask & (1<header_present = - brw->gen < 5 || brw->gen >= 9 || + devinfo->gen < 5 || devinfo->gen >= 9 || inst->offset != 0 || ir->op == ir_tg4 || - is_high_sampler(brw, sampler_reg); + is_high_sampler(devinfo, sampler_reg); inst->base_mrf = 2; inst->mlen = inst->header_present + 1; /* always at least one */ inst->dst.writemask = WRITEMASK_XYZW; @@ -2656,7 +2656,7 @@ vec4_visitor::visit(ir_texture *ir) int param_base = inst->base_mrf + inst->header_present; if (ir->op == ir_txs || ir->op == ir_query_levels) { - int writemask = brw->gen == 4 ? WRITEMASK_W : WRITEMASK_X; + int writemask = devinfo->gen == 4 ? WRITEMASK_W : WRITEMASK_X; emit(MOV(dst_reg(MRF, param_base, lod_type, writemask), lod)); } else { /* Load the coordinate */ @@ -2682,7 +2682,7 @@ vec4_visitor::visit(ir_texture *ir) /* Load the LOD info */ if (ir->op == ir_tex || ir->op == ir_txl) { int mrf, writemask; - if (brw->gen >= 5) { + if (devinfo->gen >= 5) { mrf = param_base + 1; if (ir->shadow_comparitor) { writemask = WRITEMASK_Y; @@ -2691,7 +2691,7 @@ vec4_visitor::visit(ir_texture *ir) writemask = WRITEMASK_X; inst->mlen++; } - } else /* brw->gen == 4 */ { + } else /* devinfo->gen == 4 */ { mrf = param_base; writemask = WRITEMASK_W; } @@ -2701,7 +2701,7 @@ vec4_visitor::visit(ir_texture *ir) } else if (ir->op == ir_txf_ms) { emit(MOV(dst_reg(MRF, param_base + 1, sample_index_type, WRITEMASK_X), sample_index)); - if (brw->gen >= 7) { + if (devinfo->gen >= 7) { /* MCS data is in the first channel of `mcs`, but we need to get it into * the .y channel of the second vec4 of params, so replicate .x across * the whole vec4 and then mask off everything except .y @@ -2714,7 +2714,7 @@ vec4_visitor::visit(ir_texture *ir) } else if (ir->op == ir_txd) { const glsl_type *type = lod_type; - if (brw->gen >= 5) { + if (devinfo->gen >= 5) { dPdx.swizzle = BRW_SWIZZLE4(SWIZZLE_X,SWIZZLE_X,SWIZZLE_Y,SWIZZLE_Y); dPdy.swizzle = BRW_SWIZZLE4(SWIZZLE_X,SWIZZLE_X,SWIZZLE_Y,SWIZZLE_Y); emit(MOV(dst_reg(MRF, param_base + 1, type, WRITEMASK_XZ), dPdx)); @@ -2734,7 +2734,7 @@ vec4_visitor::visit(ir_texture *ir) shadow_comparitor)); } } - } else /* brw->gen == 4 */ { + } else /* devinfo->gen == 4 */ { emit(MOV(dst_reg(MRF, param_base + 1, type, WRITEMASK_XYZ), dPdx)); emit(MOV(dst_reg(MRF, param_base + 2, type, WRITEMASK_XYZ), dPdy)); inst->mlen += 2; @@ -2766,7 +2766,7 @@ vec4_visitor::visit(ir_texture *ir) } } - if (brw->gen == 6 && ir->op == ir_tg4) { + if (devinfo->gen == 6 && ir->op == ir_tg4) { emit_gen6_gather_wa(key->tex.gen6_gather_wa[sampler], inst->dst); } @@ -2901,7 +2901,7 @@ vec4_visitor::visit(ir_if *ir) */ this->base_ir = ir->condition; - if (brw->gen == 6) { + if (devinfo->gen == 6) { emit_if_gen6(ir); } else { enum brw_predicate predicate; @@ -3009,9 +3009,9 @@ vec4_visitor::emit_ndc_computation() void vec4_visitor::emit_psiz_and_flags(dst_reg reg) { - if (brw->gen < 6 && + if (devinfo->gen < 6 && ((prog_data->vue_map.slots_valid & VARYING_BIT_PSIZ) || - key->userclip_active || brw->has_negative_rhw_bug)) { + key->userclip_active || devinfo->has_negative_rhw_bug)) { dst_reg header1 = dst_reg(this, glsl_type::uvec4_type); dst_reg header1_w = header1; header1_w.writemask = WRITEMASK_W; @@ -3050,7 +3050,7 @@ vec4_visitor::emit_psiz_and_flags(dst_reg reg) * Later, clipping will detect ucp[6] and ensure the primitive is * clipped against all fixed planes. */ - if (brw->has_negative_rhw_bug) { + if (devinfo->has_negative_rhw_bug) { src_reg ndc_w = src_reg(output_reg[BRW_VARYING_SLOT_NDC]); ndc_w.swizzle = BRW_SWIZZLE_WWWW; emit(CMP(dst_null_f(), ndc_w, src_reg(0.0f), BRW_CONDITIONAL_L)); @@ -3062,7 +3062,7 @@ vec4_visitor::emit_psiz_and_flags(dst_reg reg) } emit(MOV(retype(reg, BRW_REGISTER_TYPE_UD), src_reg(header1))); - } else if (brw->gen < 6) { + } else if (devinfo->gen < 6) { emit(MOV(retype(reg, BRW_REGISTER_TYPE_UD), 0u)); } else { emit(MOV(retype(reg, BRW_REGISTER_TYPE_D), src_reg(0))); @@ -3182,9 +3182,9 @@ vec4_visitor::emit_urb_slot(dst_reg reg, int varying) } static int -align_interleaved_urb_mlen(struct brw_context *brw, int mlen) +align_interleaved_urb_mlen(const struct brw_device_info *devinfo, int mlen) { - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { /* URB data written (does not include the message header reg) must * be a multiple of 256 bits, or 2 VS registers. See vol5c.5, * section 5.4.3.2.2: URB_INTERLEAVED. @@ -3232,7 +3232,7 @@ vec4_visitor::emit_vertex() */ emit_urb_write_header(mrf++); - if (brw->gen < 6) { + if (devinfo->gen < 6) { emit_ndc_computation(); } @@ -3276,7 +3276,7 @@ vec4_visitor::emit_vertex() current_annotation = "URB write"; vec4_instruction *inst = emit_urb_write_opcode(complete); inst->base_mrf = base_mrf; - inst->mlen = align_interleaved_urb_mlen(brw, mrf - base_mrf); + inst->mlen = align_interleaved_urb_mlen(devinfo, mrf - base_mrf); inst->offset += offset; } while(!complete); } @@ -3294,7 +3294,7 @@ vec4_visitor::get_scratch_offset(bblock_t *block, vec4_instruction *inst, /* Pre-gen6, the message header uses byte offsets instead of vec4 * (16-byte) offset units. */ - if (brw->gen < 6) + if (devinfo->gen < 6) message_header_scale *= 16; if (reladdr) { @@ -3324,18 +3324,18 @@ vec4_visitor::get_pull_constant_offset(bblock_t * block, vec4_instruction *inst, /* Pre-gen6, the message header uses byte offsets instead of vec4 * (16-byte) offset units. */ - if (brw->gen < 6) { + if (devinfo->gen < 6) { emit_before(block, inst, MUL(dst_reg(index), index, src_reg(16))); } return index; - } else if (brw->gen >= 8) { + } else if (devinfo->gen >= 8) { /* Store the offset in a GRF so we can send-from-GRF. */ src_reg offset = src_reg(this, glsl_type::int_type); emit_before(block, inst, MOV(dst_reg(offset), src_reg(reg_offset))); return offset; } else { - int message_header_scale = brw->gen < 6 ? 16 : 1; + int message_header_scale = devinfo->gen < 6 ? 16 : 1; return src_reg(reg_offset * message_header_scale); } } @@ -3629,7 +3629,7 @@ vec4_visitor::resolve_ud_negate(src_reg *reg) void vec4_visitor::resolve_bool_comparison(ir_rvalue *rvalue, src_reg *reg) { - assert(brw->gen <= 5); + assert(devinfo->gen <= 5); if (!rvalue->type->is_boolean()) return; @@ -3681,7 +3681,7 @@ vec4_visitor::vec4_visitor(struct brw_context *brw, this->virtual_grf_end = NULL; this->live_intervals = NULL; - this->max_grf = brw->gen >= 7 ? GEN7_MRF_HACK_START : BRW_MAX_GRF; + this->max_grf = devinfo->gen >= 7 ? GEN7_MRF_HACK_START : BRW_MAX_GRF; this->uniforms = 0; diff --git a/src/mesa/drivers/dri/i965/brw_vec4_vp.cpp b/src/mesa/drivers/dri/i965/brw_vec4_vp.cpp index 8756bef79ab..92d108598a2 100644 --- a/src/mesa/drivers/dri/i965/brw_vec4_vp.cpp +++ b/src/mesa/drivers/dri/i965/brw_vec4_vp.cpp @@ -96,7 +96,7 @@ vec4_vs_visitor::emit_program_code() break; case OPCODE_ARL: - if (brw->gen >= 6) { + if (devinfo->gen >= 6) { dst.writemask = WRITEMASK_X; dst_reg dst_f = dst; dst_f.type = BRW_REGISTER_TYPE_F; @@ -532,7 +532,7 @@ vec4_vs_visitor::get_vp_src_reg(const prog_src_register &src) dst_reladdr.writemask = WRITEMASK_X; emit(ADD(dst_reladdr, this->vp_addr_reg, src_reg(src.Index))); - if (brw->gen < 6) + if (devinfo->gen < 6) emit(MUL(dst_reladdr, reladdr, src_reg(16))); #if 0 -- cgit v1.2.3