/* * Copyright (c) 2022-2025 Intel Corporation * SPDX-License-Identifier: MIT */ #include "brw_nir.h" /* * Wa_18019110168 for gfx 12.5. * * This file implements workaround for HW bug, which leads to fragment shader * reading incorrect per-primitive data if mesh shader, in addition to writing * per-primitive data, also writes to gl_ClipDistance. * * The suggested solution to that bug is to not use per-primitive data by: * - creating new vertices for provoking vertices shared by multiple primitives * - converting per-primitive attributes read by fragment shader to flat * per-vertex attributes for the provoking vertex * - modifying fragment shader to read those per-vertex attributes * * There are at least 2 type of failures not handled very well: * - if the number of varying slots overflows, than only some attributes will * be converted, leading to corruption of those unconverted attributes * - if the overall MUE size is so large it doesn't fit in URB, then URB * allocation will fail in some way; unfortunately there's no good way to * say how big MUE will be at this moment and back out */ static bool copy_primitive_count_write(nir_builder *b, nir_intrinsic_instr *intrin, void *data) { if (intrin->intrinsic != nir_intrinsic_set_vertex_and_primitive_count) return false; b->cursor = nir_after_instr(&intrin->instr); nir_variable *primitive_count = (nir_variable *)data; nir_store_var(b, primitive_count, intrin->src[1].ssa, 0x1); return true; } static nir_variable * copy_primitive_count_writes(nir_shader *nir) { nir_variable *primitive_count = nir_local_variable_create(nir_shader_get_entrypoint(nir), glsl_uint_type(), "Wa_18019110168_primitive_count"); nir_shader_intrinsics_pass(nir, copy_primitive_count_write, nir_metadata_control_flow, primitive_count); return primitive_count; } struct mapping { nir_variable *temp_var; nir_deref_instr *per_prim_deref; nir_deref_instr *per_vert_deref; }; static bool rewrite_derefs_to_per_prim_vars(nir_builder *b, nir_intrinsic_instr *intrin, void *data) { if (intrin->intrinsic != nir_intrinsic_load_deref && intrin->intrinsic != nir_intrinsic_store_deref) return false; nir_deref_instr *old_deref = nir_def_as_deref(intrin->src[0].ssa); nir_variable *var = nir_deref_instr_get_variable(old_deref); if (var == NULL) return false; struct mapping *mapping = data; if (mapping[var->data.location].temp_var == NULL) return false; b->cursor = nir_before_instr(&intrin->instr); nir_deref_instr *new_deref = nir_clone_deref_instr( b, mapping[var->data.location].temp_var, old_deref); nir_src_rewrite(&intrin->src[0], &new_deref->def); return true; } static bool mesh_convert_attrs_prim_to_vert(struct nir_shader *nir, struct brw_compile_mesh_params *params, int *wa_mapping) { const uint64_t outputs_written = nir->info.outputs_written; const uint64_t per_primitive_outputs = nir->info.per_primitive_outputs & ~VARYING_BIT_PRIMITIVE_INDICES; const uint64_t other_outputs = outputs_written & ~per_primitive_outputs; uint64_t all_outputs = outputs_written; const uint64_t remapped_outputs = outputs_written & nir->info.per_primitive_outputs & ~(VARYING_BIT_CULL_PRIMITIVE | VARYING_BIT_PRIMITIVE_INDICES | VARYING_BIT_PRIMITIVE_COUNT | VARYING_BIT_LAYER | VARYING_BIT_VIEWPORT | VARYING_BIT_PRIMITIVE_SHADING_RATE); /* indexed by slot of per-prim attribute */ struct mapping mapping[VARYING_SLOT_MAX] = { {NULL, NULL, NULL}, }; /* Figure out the mapping between per-primitive and new per-vertex outputs. */ nir_function_impl *impl = nir_shader_get_entrypoint(nir); nir_foreach_shader_out_variable(var, nir) { int location = var->data.location; if (!(BITFIELD64_BIT(location) & remapped_outputs)) continue; assert(location == VARYING_SLOT_PRIMITIVE_ID || location >= VARYING_SLOT_VAR0); const struct glsl_type *type = var->type; if (nir_is_arrayed_io(var, MESA_SHADER_MESH)) { assert(glsl_type_is_array(type)); type = glsl_get_array_element(type); } unsigned num_slots = glsl_count_attribute_slots(type, false); for (gl_varying_slot slot = VARYING_SLOT_VAR0; slot <= VARYING_SLOT_VAR31; slot++) { uint64_t mask = BITFIELD64_MASK(num_slots) << slot; if ((all_outputs & mask) == 0) { wa_mapping[location] = slot; all_outputs |= mask; break; } } if (wa_mapping[location] == 0) { fprintf(stderr, "Not enough space for hardware per-primitive data corruption work around.\n"); return false; } mapping[location].temp_var = nir_local_variable_create(impl, glsl_array_type(type, nir->info.mesh.max_primitives_out, glsl_get_std140_size(type, false)), var->name); } /* Rewrite all the per-primitive variable reads/writes to the temporary * variables. */ NIR_PASS(_, nir, nir_shader_intrinsics_pass, rewrite_derefs_to_per_prim_vars, nir_metadata_control_flow, mapping); void *mem_ctx = ralloc_context(NULL); unsigned vertices_per_primitive = mesa_vertices_per_prim(nir->info.mesh.primitive_type); nir_variable *primitive_count_var = copy_primitive_count_writes(nir); nir_builder _b = nir_builder_at(nir_after_impl(impl)), *b = &_b; /* wait for all subgroups to finish */ nir_barrier(b, SCOPE_WORKGROUP); /* Build a list of per-vertex variables we might need to copy */ unsigned num_other_variables = 0; nir_foreach_shader_out_variable(var, nir) { if ((BITFIELD64_BIT(var->data.location) & other_outputs) == 0) continue; num_other_variables++; } nir_variable *primitive_indices_var = NULL; nir_deref_instr **per_vertex_derefs = ralloc_array(mem_ctx, nir_deref_instr *, num_other_variables); unsigned num_per_vertex_variables = 0; unsigned processed = 0; nir_foreach_shader_out_variable(var, nir) { if ((BITFIELD64_BIT(var->data.location) & other_outputs) == 0) continue; switch (var->data.location) { case VARYING_SLOT_PRIMITIVE_COUNT: break; case VARYING_SLOT_PRIMITIVE_INDICES: primitive_indices_var = var; break; default: { const struct glsl_type *type = var->type; assert(glsl_type_is_array(type)); const struct glsl_type *array_element_type = glsl_get_array_element(type); /* Resize type of array output to make space for one extra vertex * attribute for each primitive, so we ensure that the provoking * vertex is not shared between primitives. */ const struct glsl_type *new_type = glsl_array_type(array_element_type, glsl_get_length(type) + nir->info.mesh.max_primitives_out, 0); var->type = new_type; per_vertex_derefs[num_per_vertex_variables++] = nir_build_deref_var(b, var); break; } } ++processed; } assert(processed == num_other_variables); nir_def *zero = nir_imm_int(b, 0); nir_def *provoking_vertex = params->load_provoking_vertex(b, params->load_provoking_vertex_data); nir_def *local_invocation_index = nir_load_local_invocation_index(b); nir_def *cmp = nir_ieq(b, local_invocation_index, zero); nir_if *if_stmt = nir_push_if(b, cmp); { assert(primitive_count_var != NULL); assert(primitive_indices_var != NULL); /* Update types of derefs to match type of variables they (de)reference. */ nir_foreach_function_impl(impl, nir) { nir_foreach_block(block, impl) { nir_foreach_instr(instr, block) { if (instr->type != nir_instr_type_deref) continue; nir_deref_instr *deref = nir_instr_as_deref(instr); if (deref->deref_type != nir_deref_type_var) continue; if (deref->var->type != deref->type) deref->type = deref->var->type; } } } /* Create new per-vertex output variables mirroring per-primitive variables * and create derefs for both old and new variables. */ nir_foreach_shader_out_variable(var, nir) { gl_varying_slot location = var->data.location; if ((BITFIELD64_BIT(location) & remapped_outputs) == 0) continue; const struct glsl_type *type = var->type; assert(glsl_type_is_array(type)); const struct glsl_type *array_element_type = glsl_get_array_element(type); const struct glsl_type *new_type = glsl_array_type(array_element_type, nir->info.mesh.max_vertices_out + nir->info.mesh.max_primitives_out, 0); nir_variable *new_var = nir_variable_create(nir, nir_var_shader_out, new_type, var->name); assert(wa_mapping[location] >= VARYING_SLOT_VAR0); assert(wa_mapping[location] <= VARYING_SLOT_VAR31); new_var->data.location = wa_mapping[location]; new_var->data.interpolation = INTERP_MODE_FLAT; mapping[location].per_vert_deref = nir_build_deref_var(b, new_var); mapping[location].per_prim_deref = nir_build_deref_var(b, mapping[location].temp_var); } nir_def *trueconst = nir_imm_true(b); /* * for each Primitive (0 : primitiveCount) * if VertexUsed[PrimitiveIndices[Primitive][provoking vertex]] * create 1 new vertex at offset "Vertex" * copy per vert attributes of provoking vertex to the new one * update PrimitiveIndices[Primitive][provoking vertex] * Vertex++ * else * VertexUsed[PrimitiveIndices[Primitive][provoking vertex]] := true * * for each attribute : mapping * copy per_prim_attr(Primitive) to per_vert_attr[Primitive][provoking vertex] */ /* primitive count */ nir_def *primitive_count = nir_load_var(b, primitive_count_var); /* primitive index */ nir_variable *primitive_var = nir_local_variable_create(impl, glsl_uint_type(), "Primitive"); nir_deref_instr *primitive_deref = nir_build_deref_var(b, primitive_var); nir_store_deref(b, primitive_deref, zero, 1); /* vertex index */ nir_variable *vertex_var = nir_local_variable_create(impl, glsl_uint_type(), "Vertex"); nir_deref_instr *vertex_deref = nir_build_deref_var(b, vertex_var); nir_store_deref(b, vertex_deref, nir_imm_int(b, nir->info.mesh.max_vertices_out), 1); /* used vertices bitvector */ const struct glsl_type *used_vertex_type = glsl_array_type(glsl_bool_type(), nir->info.mesh.max_vertices_out, 0); nir_variable *used_vertex_var = nir_local_variable_create(impl, used_vertex_type, "VertexUsed"); nir_deref_instr *used_vertex_deref = nir_build_deref_var(b, used_vertex_var); /* Initialize it as "not used" */ for (unsigned i = 0; i < nir->info.mesh.max_vertices_out; ++i) { nir_deref_instr *indexed_used_vertex_deref = nir_build_deref_array(b, used_vertex_deref, nir_imm_int(b, i)); nir_store_deref(b, indexed_used_vertex_deref, nir_imm_false(b), 1); } nir_loop *loop = nir_push_loop(b); { nir_def *primitive_id = nir_load_deref(b, primitive_deref); nir_def *cmp = nir_ige(b, primitive_id, primitive_count); nir_push_if(b, cmp); { nir_jump(b, nir_jump_break); } nir_pop_if(b, NULL); nir_deref_instr *primitive_indices_deref = nir_build_deref_var(b, primitive_indices_var); nir_deref_instr *indexed_primitive_indices_deref; nir_def *src_vertex; nir_def *prim_indices; /* array of vectors, we have to extract index out of array deref */ indexed_primitive_indices_deref = nir_build_deref_array(b, primitive_indices_deref, primitive_id); prim_indices = nir_load_deref(b, indexed_primitive_indices_deref); src_vertex = nir_vector_extract(b, prim_indices, provoking_vertex); nir_def *dst_vertex = nir_load_deref(b, vertex_deref); nir_deref_instr *indexed_used_vertex_deref = nir_build_deref_array(b, used_vertex_deref, src_vertex); nir_def *used_vertex = nir_load_deref(b, indexed_used_vertex_deref); nir_push_if(b, used_vertex); { /* If the vertex is used by another primitive, we need to make an * entire copy of the per-vertex variables. */ for (unsigned a = 0; a < num_per_vertex_variables; ++a) { nir_deref_instr *attr_arr = per_vertex_derefs[a]; nir_deref_instr *src = nir_build_deref_array(b, attr_arr, src_vertex); nir_deref_instr *dst = nir_build_deref_array(b, attr_arr, dst_vertex); assert(per_vertex_derefs[a]->instr.type == nir_instr_type_deref); nir_copy_deref(b, dst, src); } /* Rebuild the vertices indices for the primitive by pointing to * the new provoking vertex */ nir_def *new_val = nir_vector_insert(b, prim_indices, dst_vertex, provoking_vertex); nir_store_deref(b, indexed_primitive_indices_deref, new_val, BITFIELD_MASK(vertices_per_primitive)); nir_store_deref(b, vertex_deref, nir_iadd_imm(b, dst_vertex, 1), 1); /* Finally write the per-primitive values into the per-vertex * block at remapped locations. */ for (unsigned i = 0; i < ARRAY_SIZE(mapping); ++i) { if (!mapping[i].per_vert_deref) continue; assert(mapping[i].per_prim_deref->instr.type == nir_instr_type_deref); nir_deref_instr *src = nir_build_deref_array(b, mapping[i].per_prim_deref, primitive_id); nir_deref_instr *dst = nir_build_deref_array(b, mapping[i].per_vert_deref, dst_vertex); nir_copy_deref(b, dst, src); } } nir_push_else(b, NULL); { /* If the vertex is not used yet by any primitive, we just have to * write the per-primitive values into the per-vertex block at * remapped locations. */ nir_store_deref(b, indexed_used_vertex_deref, trueconst, 1); for (unsigned i = 0; i < ARRAY_SIZE(mapping); ++i) { if (!mapping[i].per_vert_deref) continue; assert(mapping[i].per_prim_deref->instr.type == nir_instr_type_deref); nir_deref_instr *src = nir_build_deref_array(b, mapping[i].per_prim_deref, primitive_id); nir_deref_instr *dst = nir_build_deref_array(b, mapping[i].per_vert_deref, src_vertex); nir_copy_deref(b, dst, src); } } nir_pop_if(b, NULL); nir_store_deref(b, primitive_deref, nir_iadd_imm(b, primitive_id, 1), 1); } nir_pop_loop(b, loop); } nir_pop_if(b, if_stmt); /* local_invocation_index == 0 */ nir->info.mesh.max_vertices_out += nir->info.mesh.max_primitives_out; ralloc_free(mem_ctx); return true; } void brw_nir_mesh_convert_attrs_prim_to_vert(struct nir_shader *nir, struct brw_compile_mesh_params *params, int *wa_mapping) { NIR_PASS(_, nir, mesh_convert_attrs_prim_to_vert, params, wa_mapping); /* Remove per-primitive references */ NIR_PASS(_, nir, nir_opt_dce); NIR_PASS(_, nir, nir_remove_dead_variables, nir_var_shader_out, NULL); nir_shader_gather_info(nir, nir_shader_get_entrypoint(nir)); /* deal with copy_derefs */ NIR_PASS(_, nir, nir_split_var_copies); NIR_PASS(_, nir, nir_lower_var_copies); } static bool frag_update_derefs_instr(struct nir_builder *b, nir_instr *instr, void *data) { if (instr->type != nir_instr_type_deref) return false; nir_deref_instr *deref = nir_instr_as_deref(instr); if (deref->deref_type != nir_deref_type_var) return false; nir_variable *var = deref->var; if (!(var->data.mode & nir_var_shader_in)) return false; int location = var->data.location; nir_deref_instr **new_derefs = (nir_deref_instr **)data; if (new_derefs[location] == NULL) return false; nir_def_replace(&deref->def, &new_derefs[location]->def); return true; } static bool frag_update_derefs(nir_shader *shader, nir_deref_instr **mapping) { return nir_shader_instructions_pass(shader, frag_update_derefs_instr, nir_metadata_none, (void *)mapping); } bool brw_nir_frag_convert_attrs_prim_to_vert(struct nir_shader *nir, const int *wa_mapping) { /* indexed by slot of per-prim attribute */ nir_deref_instr *new_derefs[VARYING_SLOT_MAX] = {NULL, }; nir_function_impl *impl = nir_shader_get_entrypoint(nir); nir_builder _b = nir_builder_at(nir_before_impl(impl)), *b = &_b; uint64_t remapped_inputs = 0; nir_foreach_shader_in_variable_safe(var, nir) { gl_varying_slot location = var->data.location; if (location == VARYING_SLOT_PRIMITIVE_COUNT || location == VARYING_SLOT_PRIMITIVE_INDICES || location == VARYING_SLOT_PRIMITIVE_SHADING_RATE || location == VARYING_SLOT_LAYER || location == VARYING_SLOT_VIEWPORT || location == VARYING_SLOT_CULL_PRIMITIVE) continue; gl_varying_slot new_location = wa_mapping[location]; if (new_location == -1) continue; assert(wa_mapping[new_location] == -1); nir_variable *new_var = nir_variable_create(nir, nir_var_shader_in, var->type, var->name); new_var->data.location = new_location; new_var->data.location_frac = var->data.location_frac; new_var->data.interpolation = INTERP_MODE_FLAT; new_derefs[location] = nir_build_deref_var(b, new_var); } nir->info.inputs_read |= remapped_inputs; nir->info.per_primitive_inputs &= ~remapped_inputs; NIR_PASS(_, nir, frag_update_derefs, new_derefs); nir_shader_gather_info(nir, impl); return true; } bool brw_nir_frag_convert_attrs_prim_to_vert_indirect(struct nir_shader *nir, const struct intel_device_info *devinfo, struct brw_compile_fs_params *params) { nir_function_impl *impl = nir_shader_get_entrypoint(nir); nir_builder _b = nir_builder_at(nir_before_impl(impl)), *b = &_b; const uint64_t per_primitive_inputs = nir->info.inputs_read & (nir->info.per_primitive_inputs | VARYING_BIT_PRIMITIVE_ID); int per_primitive_offsets[VARYING_SLOT_MAX]; uint32_t first_read_offset = 0, per_primitive_stride = 0; brw_compute_per_primitive_map(per_primitive_offsets, &per_primitive_stride, &first_read_offset, 0, nir, nir_var_shader_in, per_primitive_inputs, true /* separate_shader */); per_primitive_stride = align(per_primitive_stride, devinfo->grf_size); nir_def *msaa_flags = nir_load_fs_msaa_intel(b); nir_def *needs_remapping = nir_test_mask( b, msaa_flags, INTEL_MSAA_FLAG_PER_PRIMITIVE_REMAPPING); nir_push_if(b, needs_remapping); { nir_def *first_slot = nir_ubitfield_extract_imm( b, msaa_flags, INTEL_MSAA_FLAG_FIRST_VUE_SLOT_OFFSET, INTEL_MSAA_FLAG_FIRST_VUE_SLOT_SIZE); nir_def *remap_table_addr = nir_pack_64_2x32_split( b, nir_load_per_primitive_remap_intel(b), nir_load_reloc_const_intel( b, BRW_SHADER_RELOC_INSTRUCTION_BASE_ADDR_HIGH)); u_foreach_bit64(location, per_primitive_inputs) { if (location < VARYING_SLOT_VAR0 && location != VARYING_SLOT_PRIMITIVE_ID) continue; /* Read the varying_to_slot[] array from the mesh shader constants * space in the instruction heap. */ nir_def *data = nir_load_global_constant(b, 1, 32, nir_iadd_imm(b, remap_table_addr, ROUND_DOWN_TO(location, 4))); const unsigned bit_offset = (8 * location) % 32; nir_def *absolute_attr_idx = nir_ubitfield_extract_imm(b, data, bit_offset, 4); /* Now remove the first slot visible in the FS payload */ nir_def *payload_attr_idx = nir_iadd(b, absolute_attr_idx, nir_ineg(b, first_slot)); for (unsigned c = 0; c < 4; c++) { /* brw_nir_vertex_attribute_offset works in scalar */ nir_def *attr_idx = nir_iadd_imm( b, nir_imul_imm(b, payload_attr_idx, 4), c); /* Turn the scalar attribute index into register byte offset */ nir_def *per_vertex_offset = nir_iadd_imm( b, brw_nir_vertex_attribute_offset(b, attr_idx, devinfo), per_primitive_stride); nir_def *value = nir_load_attribute_payload_intel(b, 1, 32, per_vertex_offset); /* Write back the values into the per-primitive location */ nir_store_per_primitive_payload_intel( b, value, .base = location, .component = c); } } } nir_pop_if(b, NULL); return nir_progress(true, impl, nir_metadata_none); }