Edit
IABSD.fr/xenocara/lib/mesa/src/intel/compiler/brw_opt.cpp
jsg
- lib/mesa/src/intel/compiler/brw_opt.cpp
-
/* * Copyright © 2010 Intel Corporation * SPDX-License-Identifier: MIT */ #include "brw_eu.h" #include "brw_fs.h" #include "brw_builder.h" #include "dev/intel_debug.h" using namespace brw; void brw_optimize(fs_visitor &s) { const nir_shader *nir = s.nir; s.debug_optimizer(nir, "start", 0, 0); /* Start by validating the shader we currently have. */ brw_validate(s); /* Track how much non-SSA at this point. */ { const brw::def_analysis &defs = s.def_analysis.require(); s.shader_stats.non_ssa_registers_after_nir = defs.count() - defs.ssa_count(); } bool progress = false; int iteration = 0; int pass_num = 0; #define OPT(pass, ...) ({ \ pass_num++; \ bool this_progress = pass(s, ##__VA_ARGS__); \ \ if (this_progress) \ s.debug_optimizer(nir, #pass, iteration, pass_num); \ \ brw_validate(s); \ \ progress = progress || this_progress; \ this_progress; \ }) if (s.compiler->lower_dpas) OPT(brw_lower_dpas); OPT(brw_opt_split_virtual_grfs); /* Before anything else, eliminate dead code. The results of some NIR * instructions may effectively be calculated twice. Once when the * instruction is encountered, and again when the user of that result is * encountered. Wipe those away before algebraic optimizations and * especially copy propagation can mix things up. */ OPT(brw_opt_dead_code_eliminate); OPT(brw_opt_remove_extra_rounding_modes); OPT(brw_opt_eliminate_find_live_channel); do { progress = false; pass_num = 0; iteration++; OPT(brw_opt_algebraic); OPT(brw_opt_cse_defs); if (!OPT(brw_opt_copy_propagation_defs)) OPT(brw_opt_copy_propagation); OPT(brw_opt_cmod_propagation); OPT(brw_opt_dead_code_eliminate); OPT(brw_opt_saturate_propagation); OPT(brw_opt_register_coalesce); OPT(brw_opt_compact_virtual_grfs); } while (progress); brw_shader_phase_update(s, BRW_SHADER_PHASE_AFTER_OPT_LOOP); progress = false; pass_num = 0; if (OPT(brw_opt_combine_convergent_txf)) OPT(brw_opt_copy_propagation_defs); if (OPT(brw_lower_pack)) { OPT(brw_opt_register_coalesce); OPT(brw_opt_dead_code_eliminate); } OPT(brw_lower_subgroup_ops); OPT(brw_lower_csel); OPT(brw_lower_simd_width); OPT(brw_lower_scalar_fp64_MAD); OPT(brw_lower_barycentrics); OPT(brw_lower_logical_sends); brw_shader_phase_update(s, BRW_SHADER_PHASE_AFTER_EARLY_LOWERING); /* After logical SEND lowering. */ if (!OPT(brw_opt_copy_propagation_defs)) OPT(brw_opt_copy_propagation); /* Identify trailing zeros LOAD_PAYLOAD of sampler messages. * Do this before splitting SENDs. */ if (OPT(brw_opt_zero_samples)) { if (!OPT(brw_opt_copy_propagation_defs)) { OPT(brw_opt_copy_propagation); } } if (s.devinfo->ver >= 30) OPT(brw_opt_send_to_send_gather); OPT(brw_opt_split_sends); OPT(brw_workaround_nomask_control_flow); if (progress) { /* Do both forms of copy propagation because it is important to * eliminate as many cases of load_payload-of-load_payload as possible. */ OPT(brw_opt_copy_propagation_defs); OPT(brw_opt_copy_propagation); /* Run after logical send lowering to give it a chance to CSE the * LOAD_PAYLOAD instructions created to construct the payloads of * e.g. texturing messages in cases where it wasn't possible to CSE the * whole logical instruction. */ OPT(brw_opt_cse_defs); OPT(brw_opt_register_coalesce); OPT(brw_opt_dead_code_eliminate); } OPT(brw_opt_remove_redundant_halts); if (OPT(brw_lower_load_payload)) { OPT(brw_opt_split_virtual_grfs); OPT(brw_opt_register_coalesce); OPT(brw_lower_simd_width); OPT(brw_opt_dead_code_eliminate); } brw_shader_phase_update(s, BRW_SHADER_PHASE_AFTER_MIDDLE_LOWERING); OPT(brw_lower_alu_restrictions); OPT(brw_opt_combine_constants); if (OPT(brw_lower_integer_multiplication)) { /* If lower_integer_multiplication made progress, it may have produced * some 32x32-bit MULs in the process of lowering 64-bit MULs. Run it * one more time to clean those up if they exist. */ OPT(brw_lower_integer_multiplication); } OPT(brw_lower_sub_sat); progress = false; OPT(brw_lower_derivatives); OPT(brw_lower_regioning); /* Try both copy propagation passes. The defs one will likely not be * able to handle everything at this point. */ const bool cp1 = OPT(brw_opt_copy_propagation_defs); const bool cp2 = OPT(brw_opt_copy_propagation); if (cp1 || cp2) OPT(brw_opt_combine_constants); OPT(brw_opt_dead_code_eliminate); OPT(brw_opt_register_coalesce); if (progress) OPT(brw_lower_simd_width); if (s.devinfo->ver >= 30) OPT(brw_opt_send_gather_to_send); OPT(brw_lower_uniform_pull_constant_loads); /* Do this before brw_lower_send_descriptors. */ OPT(brw_workaround_memory_fence_before_eot); if (OPT(brw_lower_send_descriptors)) { /* No need for standard copy_propagation since * brw_fs_opt_address_reg_load will only optimize defs. */ if (OPT(brw_opt_copy_propagation_defs)) OPT(brw_opt_algebraic); OPT(brw_opt_address_reg_load); OPT(brw_opt_dead_code_eliminate); } OPT(brw_lower_sends_overlapping_payload); OPT(brw_lower_indirect_mov); OPT(brw_lower_find_live_channel); OPT(brw_lower_load_subgroup_invocation); brw_shader_phase_update(s, BRW_SHADER_PHASE_AFTER_LATE_LOWERING); } static unsigned load_payload_sources_read_for_size(fs_inst *lp, unsigned size_read) { assert(lp->opcode == SHADER_OPCODE_LOAD_PAYLOAD); assert(size_read >= lp->header_size * REG_SIZE); unsigned i; unsigned size = lp->header_size * REG_SIZE; for (i = lp->header_size; size < size_read && i < lp->sources; i++) size += lp->exec_size * brw_type_size_bytes(lp->src[i].type); /* Size read must cover exactly a subset of sources. */ assert(size == size_read); return i; } /** * Optimize sample messages that have constant zero values for the trailing * parameters. We can just reduce the message length for these * instructions instead of reserving a register for it. Trailing parameters * that aren't sent default to zero anyway. This will cause the dead code * eliminator to remove the MOV instruction that would otherwise be emitted to * set up the zero value. */ bool brw_opt_zero_samples(fs_visitor &s) { bool progress = false; foreach_block_and_inst(block, fs_inst, send, s.cfg) { if (send->opcode != SHADER_OPCODE_SEND || send->sfid != BRW_SFID_SAMPLER) continue; /* Wa_14012688258: * * Don't trim zeros at the end of payload for sample operations * in cube and cube arrays. */ if (send->keep_payload_trailing_zeros) continue; /* This pass works on SENDs before splitting. */ if (send->ex_mlen > 0) continue; fs_inst *lp = (fs_inst *) send->prev; if (lp->is_head_sentinel() || lp->opcode != SHADER_OPCODE_LOAD_PAYLOAD) continue; /* How much of the payload are actually read by this SEND. */ const unsigned params = load_payload_sources_read_for_size(lp, send->mlen * REG_SIZE); /* We don't want to remove the message header or the first parameter. * Removing the first parameter is not allowed, see the Haswell PRM * volume 7, page 149: * * "Parameter 0 is required except for the sampleinfo message, which * has no parameter 0" */ const unsigned first_param_idx = lp->header_size; unsigned zero_size = 0; for (unsigned i = params - 1; i > first_param_idx; i--) { if (lp->src[i].file != BAD_FILE && !lp->src[i].is_zero()) break; zero_size += lp->exec_size * brw_type_size_bytes(lp->src[i].type) * lp->dst.stride; } /* Round down to ensure to only consider full registers. */ const unsigned zero_len = ROUND_DOWN_TO(zero_size / REG_SIZE, reg_unit(s.devinfo)); if (zero_len > 0) { /* Note mlen is in REG_SIZE units. */ send->mlen -= zero_len; progress = true; } } if (progress) s.invalidate_analysis(DEPENDENCY_INSTRUCTION_DETAIL); return progress; } /** * Opportunistically split SEND message payloads. * * Gfx9+ supports "split" SEND messages, which take two payloads that are * implicitly concatenated. If we find a SEND message with a single payload, * we can split that payload in two. This results in smaller contiguous * register blocks for us to allocate. But it can help beyond that, too. * * We try and split a LOAD_PAYLOAD between sources which change registers. * For example, a sampler message often contains a x/y/z coordinate that may * already be in a contiguous VGRF, combined with an LOD, shadow comparitor, * or array index, which comes from elsewhere. In this case, the first few * sources will be different offsets of the same VGRF, then a later source * will be a different VGRF. So we split there, possibly eliminating the * payload concatenation altogether. */ bool brw_opt_split_sends(fs_visitor &s) { bool progress = false; foreach_block_and_inst(block, fs_inst, send, s.cfg) { if (send->opcode != SHADER_OPCODE_SEND || send->mlen <= reg_unit(s.devinfo) || send->ex_mlen > 0 || send->src[2].file != VGRF) continue; /* Currently don't split sends that reuse a previously used payload. */ fs_inst *lp = (fs_inst *) send->prev; if (lp->is_head_sentinel() || lp->opcode != SHADER_OPCODE_LOAD_PAYLOAD) continue; if (lp->dst.file != send->src[2].file || lp->dst.nr != send->src[2].nr) continue; /* Split either after the header (if present), or when consecutive * sources switch from one VGRF to a different one. */ unsigned mid = lp->header_size; if (mid == 0) { for (mid = 1; mid < lp->sources; mid++) { if (lp->src[mid].file == BAD_FILE) continue; if (lp->src[0].file != lp->src[mid].file || lp->src[0].nr != lp->src[mid].nr) break; } } /* SEND mlen might be smaller than what LOAD_PAYLOAD provides, so * find out how many sources from the payload does it really need. */ const unsigned end = load_payload_sources_read_for_size(lp, send->mlen * REG_SIZE); /* Nothing to split. */ if (end <= mid) continue; const brw_builder ibld(&s, block, lp); fs_inst *lp1 = ibld.LOAD_PAYLOAD(lp->dst, &lp->src[0], mid, lp->header_size); fs_inst *lp2 = ibld.LOAD_PAYLOAD(lp->dst, &lp->src[mid], end - mid, 0); assert(lp1->size_written % REG_SIZE == 0); assert(lp2->size_written % REG_SIZE == 0); assert((lp1->size_written + lp2->size_written) / REG_SIZE == send->mlen); lp1->dst = brw_vgrf(s.alloc.allocate(lp1->size_written / REG_SIZE), lp1->dst.type); lp2->dst = brw_vgrf(s.alloc.allocate(lp2->size_written / REG_SIZE), lp2->dst.type); send->resize_sources(4); send->src[2] = lp1->dst; send->src[3] = lp2->dst; send->ex_mlen = lp2->size_written / REG_SIZE; send->mlen -= send->ex_mlen; progress = true; } if (progress) s.invalidate_analysis(DEPENDENCY_INSTRUCTIONS | DEPENDENCY_VARIABLES); return progress; } /** * Remove redundant or useless halts. * * For example, we can eliminate halts in the following sequence: * * halt (redundant with the next halt) * halt (useless; jumps to the next instruction) * halt-target */ bool brw_opt_remove_redundant_halts(fs_visitor &s) { bool progress = false; unsigned halt_count = 0; fs_inst *halt_target = NULL; bblock_t *halt_target_block = NULL; foreach_block_and_inst(block, fs_inst, inst, s.cfg) { if (inst->opcode == BRW_OPCODE_HALT) halt_count++; if (inst->opcode == SHADER_OPCODE_HALT_TARGET) { halt_target = inst; halt_target_block = block; break; } } if (!halt_target) { assert(halt_count == 0); return false; } /* Delete any HALTs immediately before the halt target. */ for (fs_inst *prev = (fs_inst *) halt_target->prev; !prev->is_head_sentinel() && prev->opcode == BRW_OPCODE_HALT; prev = (fs_inst *) halt_target->prev) { prev->remove(halt_target_block); halt_count--; progress = true; } if (halt_count == 0) { halt_target->remove(halt_target_block); progress = true; } if (progress) s.invalidate_analysis(DEPENDENCY_INSTRUCTIONS); return progress; } /** * Eliminate FIND_LIVE_CHANNEL instructions occurring outside any control * flow. We could probably do better here with some form of divergence * analysis. */ bool brw_opt_eliminate_find_live_channel(fs_visitor &s) { bool progress = false; unsigned depth = 0; if (!brw_stage_has_packed_dispatch(s.devinfo, s.stage, s.max_polygons, s.prog_data)) { /* The optimization below assumes that channel zero is live on thread * dispatch, which may not be the case if the fixed function dispatches * threads sparsely. */ return false; } foreach_block_and_inst_safe(block, fs_inst, inst, s.cfg) { switch (inst->opcode) { case BRW_OPCODE_IF: case BRW_OPCODE_DO: depth++; break; case BRW_OPCODE_ENDIF: case BRW_OPCODE_WHILE: depth--; break; case BRW_OPCODE_HALT: /* This can potentially make control flow non-uniform until the end * of the program. */ goto out; case SHADER_OPCODE_FIND_LIVE_CHANNEL: if (depth == 0) { inst->opcode = BRW_OPCODE_MOV; inst->src[0] = brw_imm_ud(0u); inst->force_writemask_all = true; /* FIND_LIVE_CHANNEL emitted by emit_uniformize will have * size_written set by hand to a smaller value. In this case, * munge the exec_size to match. */ if (inst->size_written == inst->dst.component_size(8 * reg_unit(s.devinfo))) inst->exec_size = 8 * reg_unit(s.devinfo); inst->resize_sources(1); progress = true; /* emit_uniformize() frequently emits FIND_LIVE_CHANNEL paired * with a BROADCAST. Save some work for opt_copy_propagation * and opt_algebraic by trivially cleaning up both together. */ assert(!inst->next->is_tail_sentinel()); fs_inst *bcast = (fs_inst *) inst->next; /* Ignore stride when comparing */ if (bcast->opcode == SHADER_OPCODE_BROADCAST && inst->dst.file == VGRF && inst->dst.file == bcast->src[1].file && inst->dst.nr == bcast->src[1].nr && inst->dst.offset == bcast->src[1].offset) { bcast->opcode = BRW_OPCODE_MOV; if (!is_uniform(bcast->src[0])) bcast->src[0] = component(bcast->src[0], 0); bcast->force_writemask_all = true; bcast->exec_size = 8 * reg_unit(s.devinfo); assert(bcast->size_written == bcast->dst.component_size(bcast->exec_size)); bcast->resize_sources(1); } } break; default: break; } } out: if (progress) s.invalidate_analysis(DEPENDENCY_INSTRUCTION_DETAIL); return progress; } /** * Rounding modes for conversion instructions are included for each * conversion, but right now it is a state. So once it is set, * we don't need to call it again for subsequent calls. * * This is useful for vector/matrices conversions, as setting the * mode once is enough for the full vector/matrix */ bool brw_opt_remove_extra_rounding_modes(fs_visitor &s) { bool progress = false; unsigned execution_mode = s.nir->info.float_controls_execution_mode; brw_rnd_mode base_mode = BRW_RND_MODE_UNSPECIFIED; if ((FLOAT_CONTROLS_ROUNDING_MODE_RTE_FP16 | FLOAT_CONTROLS_ROUNDING_MODE_RTE_FP32 | FLOAT_CONTROLS_ROUNDING_MODE_RTE_FP64) & execution_mode) base_mode = BRW_RND_MODE_RTNE; if ((FLOAT_CONTROLS_ROUNDING_MODE_RTZ_FP16 | FLOAT_CONTROLS_ROUNDING_MODE_RTZ_FP32 | FLOAT_CONTROLS_ROUNDING_MODE_RTZ_FP64) & execution_mode) base_mode = BRW_RND_MODE_RTZ; foreach_block (block, s.cfg) { brw_rnd_mode prev_mode = base_mode; foreach_inst_in_block_safe (fs_inst, inst, block) { if (inst->opcode == SHADER_OPCODE_RND_MODE) { assert(inst->src[0].file == IMM); const brw_rnd_mode mode = (brw_rnd_mode) inst->src[0].d; if (mode == prev_mode) { inst->remove(block); progress = true; } else { prev_mode = mode; } } } } if (progress) s.invalidate_analysis(DEPENDENCY_INSTRUCTIONS); return progress; } bool brw_opt_send_to_send_gather(fs_visitor &s) { const intel_device_info *devinfo = s.devinfo; bool progress = false; assert(devinfo->ver >= 30); const unsigned unit = reg_unit(devinfo); assert(unit == 2); unsigned count = 0; foreach_block_and_inst_safe(block, fs_inst, inst, s.cfg) { if (inst->opcode != SHADER_OPCODE_SEND) continue; /* For 1-2 registers, send-gather offers no benefits over split-send. */ if (inst->mlen + inst->ex_mlen <= 2 * unit) continue; assert(inst->mlen % unit == 0); assert(inst->ex_mlen % unit == 0); struct { brw_reg src; unsigned phys_len; } payload[2] = { { inst->src[2], inst->mlen / unit }, { inst->src[3], inst->ex_mlen / unit }, }; const unsigned num_payload_sources = payload[0].phys_len + payload[1].phys_len; /* Limited by Src0.Length in the SEND instruction. */ if (num_payload_sources > 15) continue; if (INTEL_DEBUG(DEBUG_NO_SEND_GATHER)) { count++; continue; } inst->resize_sources(3 + num_payload_sources); /* Sources 0 and 1 remain the same. Source 2 will be filled * after register allocation. */ inst->src[2] = {}; int idx = 3; for (unsigned p = 0; p < ARRAY_SIZE(payload); p++) { for (unsigned i = 0; i < payload[p].phys_len; i++) { inst->src[idx++] = byte_offset(payload[p].src, i * reg_unit(devinfo) * REG_SIZE); } } assert(idx == inst->sources); inst->opcode = SHADER_OPCODE_SEND_GATHER; inst->mlen = 0; inst->ex_mlen = 0; progress = true; } if (INTEL_DEBUG(DEBUG_NO_SEND_GATHER)) { fprintf(stderr, "Ignored %u opportunities to try SEND_GATHER in %s shader.\n", count, _mesa_shader_stage_to_string(s.stage)); } if (progress) s.invalidate_analysis(DEPENDENCY_INSTRUCTION_DETAIL | DEPENDENCY_INSTRUCTION_DATA_FLOW); return progress; } /* If after optimizations, the sources are *still* contiguous in a * SEND_GATHER, prefer to use the regular SEND, which would save * having to write the ARF scalar register. */ bool brw_opt_send_gather_to_send(fs_visitor &s) { const intel_device_info *devinfo = s.devinfo; bool progress = false; assert(devinfo->ver >= 30); const unsigned unit = reg_unit(devinfo); assert(unit == 2); foreach_block_and_inst_safe(block, fs_inst, inst, s.cfg) { if (inst->opcode != SHADER_OPCODE_SEND_GATHER) continue; assert(inst->sources > 2); assert(inst->src[2].file == BAD_FILE); const int num_payload_sources = inst->sources - 3; assert(num_payload_sources > 0); /* Limited by Src0.Length in the SEND instruction. */ assert(num_payload_sources < 16); /* Determine whether the sources are still spread in either one or two * spans. In those cases the regular SEND instruction can be used * and there's no need to use SEND_GATHER (which would set ARF scalar register * adding an extra instruction). */ const brw_reg *payload = &inst->src[3]; brw_reg payload1 = payload[0]; brw_reg payload2 = {}; int payload1_len = 0; int payload2_len = 0; for (int i = 0; i < num_payload_sources; i++) { if (payload[i].file == VGRF && payload[i].nr == payload1.nr && payload[i].offset == payload1_len * REG_SIZE * unit) payload1_len++; else { payload2 = payload[i]; break; } } if (payload2.file == VGRF) { for (int i = payload1_len; i < num_payload_sources; i++) { if (payload[i].file == VGRF && payload[i].nr == payload2.nr && payload[i].offset == payload2_len * REG_SIZE * unit) payload2_len++; else break; } } else { payload2 = brw_null_reg(); } if (payload1_len + payload2_len != num_payload_sources) continue; /* Bspec 57058 (r64705) says * * When a source data payload is used in dataport message, that payload * must be specified as Source 1 portion of a Split Send message. * * But at this point the split point is not guaranteed to respect that. * * TODO: Pass LSC address length or infer it so valid splits can work. */ if (payload2_len && (inst->sfid == GFX12_SFID_UGM || inst->sfid == GFX12_SFID_TGM || inst->sfid == GFX12_SFID_SLM || inst->sfid == BRW_SFID_URB)) { enum lsc_opcode lsc_op = lsc_msg_desc_opcode(devinfo, inst->desc); if (lsc_op_num_data_values(lsc_op) > 0) continue; } inst->resize_sources(4); inst->opcode = SHADER_OPCODE_SEND; inst->src[2] = payload1; inst->src[3] = payload2; inst->mlen = payload1_len * unit; inst->ex_mlen = payload2_len * unit; progress = true; } if (progress) { s.invalidate_analysis(DEPENDENCY_INSTRUCTION_DETAIL | DEPENDENCY_INSTRUCTION_DATA_FLOW); } return progress; }