static void build_color_shaders(struct nir_shader **out_vs, struct nir_shader **out_fs, uint32_t frag_output) { nir_builder vs_b; nir_builder fs_b; nir_builder_init_simple_shader(&vs_b, NULL, MESA_SHADER_VERTEX, NULL); nir_builder_init_simple_shader(&fs_b, NULL, MESA_SHADER_FRAGMENT, NULL); vs_b.shader->info.name = ralloc_strdup(vs_b.shader, "meta_clear_color_vs"); fs_b.shader->info.name = ralloc_strdup(fs_b.shader, "meta_clear_color_fs"); const struct glsl_type *position_type = glsl_vec4_type(); const struct glsl_type *color_type = glsl_vec4_type(); nir_variable *vs_out_pos = nir_variable_create(vs_b.shader, nir_var_shader_out, position_type, "gl_Position"); vs_out_pos->data.location = VARYING_SLOT_POS; nir_intrinsic_instr *in_color_load = nir_intrinsic_instr_create(fs_b.shader, nir_intrinsic_load_push_constant); nir_intrinsic_set_base(in_color_load, 0); nir_intrinsic_set_range(in_color_load, 16); in_color_load->src[0] = nir_src_for_ssa(nir_imm_int(&fs_b, 0)); in_color_load->num_components = 4; nir_ssa_dest_init(&in_color_load->instr, &in_color_load->dest, 4, 32, "clear color"); nir_builder_instr_insert(&fs_b, &in_color_load->instr); nir_variable *fs_out_color = nir_variable_create(fs_b.shader, nir_var_shader_out, color_type, "f_color"); fs_out_color->data.location = FRAG_RESULT_DATA0 + frag_output; nir_store_var(&fs_b, fs_out_color, &in_color_load->dest.ssa, 0xf); nir_ssa_def *outvec = radv_meta_gen_rect_vertices(&vs_b); nir_store_var(&vs_b, vs_out_pos, outvec, 0xf); const struct glsl_type *layer_type = glsl_int_type(); nir_variable *vs_out_layer = nir_variable_create(vs_b.shader, nir_var_shader_out, layer_type, "v_layer"); vs_out_layer->data.location = VARYING_SLOT_LAYER; vs_out_layer->data.interpolation = INTERP_MODE_FLAT; nir_ssa_def *inst_id = nir_load_system_value(&vs_b, nir_intrinsic_load_instance_id, 0); nir_ssa_def *base_instance = nir_load_system_value(&vs_b, nir_intrinsic_load_base_instance, 0); nir_ssa_def *layer_id = nir_iadd(&vs_b, inst_id, base_instance); nir_store_var(&vs_b, vs_out_layer, layer_id, 0x1); *out_vs = vs_b.shader; *out_fs = fs_b.shader; }
static nir_shader * build_resolve_fragment_shader(struct radv_device *dev, bool is_integer, int samples) { nir_builder b; char name[64]; const struct glsl_type *vec2 = glsl_vector_type(GLSL_TYPE_FLOAT, 2); const struct glsl_type *vec4 = glsl_vec4_type(); const struct glsl_type *sampler_type = glsl_sampler_type(GLSL_SAMPLER_DIM_MS, false, false, GLSL_TYPE_FLOAT); snprintf(name, 64, "meta_resolve_fs-%d-%s", samples, is_integer ? "int" : "float"); nir_builder_init_simple_shader(&b, NULL, MESA_SHADER_FRAGMENT, NULL); b.shader->info.name = ralloc_strdup(b.shader, name); nir_variable *input_img = nir_variable_create(b.shader, nir_var_uniform, sampler_type, "s_tex"); input_img->data.descriptor_set = 0; input_img->data.binding = 0; nir_variable *fs_pos_in = nir_variable_create(b.shader, nir_var_shader_in, vec2, "fs_pos_in"); fs_pos_in->data.location = VARYING_SLOT_POS; nir_variable *color_out = nir_variable_create(b.shader, nir_var_shader_out, vec4, "f_color"); color_out->data.location = FRAG_RESULT_DATA0; nir_ssa_def *pos_in = nir_load_var(&b, fs_pos_in); nir_intrinsic_instr *src_offset = nir_intrinsic_instr_create(b.shader, nir_intrinsic_load_push_constant); nir_intrinsic_set_base(src_offset, 0); nir_intrinsic_set_range(src_offset, 8); src_offset->src[0] = nir_src_for_ssa(nir_imm_int(&b, 0)); src_offset->num_components = 2; nir_ssa_dest_init(&src_offset->instr, &src_offset->dest, 2, 32, "src_offset"); nir_builder_instr_insert(&b, &src_offset->instr); nir_ssa_def *pos_int = nir_f2i32(&b, pos_in); nir_ssa_def *img_coord = nir_channels(&b, nir_iadd(&b, pos_int, &src_offset->dest.ssa), 0x3); nir_variable *color = nir_local_variable_create(b.impl, glsl_vec4_type(), "color"); radv_meta_build_resolve_shader_core(&b, is_integer, samples, input_img, color, img_coord); nir_ssa_def *outval = nir_load_var(&b, color); nir_store_var(&b, color_out, outval, 0xf); return b.shader; }
static void brw_blorp_params_get_clear_kernel(struct brw_context *brw, struct brw_blorp_params *params, bool use_replicated_data) { struct brw_blorp_const_color_prog_key blorp_key; memset(&blorp_key, 0, sizeof(blorp_key)); blorp_key.use_simd16_replicated_data = use_replicated_data; if (brw_search_cache(&brw->cache, BRW_CACHE_BLORP_PROG, &blorp_key, sizeof(blorp_key), ¶ms->wm_prog_kernel, ¶ms->wm_prog_data)) return; void *mem_ctx = ralloc_context(NULL); nir_builder b; nir_builder_init_simple_shader(&b, NULL, MESA_SHADER_FRAGMENT, NULL); b.shader->info.name = ralloc_strdup(b.shader, "BLORP-clear"); nir_variable *v_color = nir_variable_create(b.shader, nir_var_shader_in, glsl_vec4_type(), "v_color"); v_color->data.location = VARYING_SLOT_VAR0; v_color->data.interpolation = INTERP_MODE_FLAT; nir_variable *frag_color = nir_variable_create(b.shader, nir_var_shader_out, glsl_vec4_type(), "gl_FragColor"); frag_color->data.location = FRAG_RESULT_COLOR; nir_copy_var(&b, frag_color, v_color); struct brw_wm_prog_key wm_key; brw_blorp_init_wm_prog_key(&wm_key); struct brw_blorp_prog_data prog_data; unsigned program_size; const unsigned *program = brw_blorp_compile_nir_shader(brw, b.shader, &wm_key, use_replicated_data, &prog_data, &program_size); brw_upload_cache(&brw->cache, BRW_CACHE_BLORP_PROG, &blorp_key, sizeof(blorp_key), program, program_size, &prog_data, sizeof(prog_data), ¶ms->wm_prog_kernel, ¶ms->wm_prog_data); ralloc_free(mem_ctx); }
static void build_color_shaders(struct nir_shader **out_vs, struct nir_shader **out_fs, uint32_t frag_output) { nir_builder vs_b; nir_builder fs_b; nir_builder_init_simple_shader(&vs_b, NULL, MESA_SHADER_VERTEX, NULL); nir_builder_init_simple_shader(&fs_b, NULL, MESA_SHADER_FRAGMENT, NULL); vs_b.shader->info.name = ralloc_strdup(vs_b.shader, "meta_clear_color_vs"); fs_b.shader->info.name = ralloc_strdup(fs_b.shader, "meta_clear_color_fs"); const struct glsl_type *position_type = glsl_vec4_type(); const struct glsl_type *color_type = glsl_vec4_type(); nir_variable *vs_in_pos = nir_variable_create(vs_b.shader, nir_var_shader_in, position_type, "a_position"); vs_in_pos->data.location = VERT_ATTRIB_GENERIC0; nir_variable *vs_out_pos = nir_variable_create(vs_b.shader, nir_var_shader_out, position_type, "gl_Position"); vs_out_pos->data.location = VARYING_SLOT_POS; nir_variable *vs_in_color = nir_variable_create(vs_b.shader, nir_var_shader_in, color_type, "a_color"); vs_in_color->data.location = VERT_ATTRIB_GENERIC1; nir_variable *vs_out_color = nir_variable_create(vs_b.shader, nir_var_shader_out, color_type, "v_color"); vs_out_color->data.location = VARYING_SLOT_VAR0; vs_out_color->data.interpolation = INTERP_MODE_FLAT; nir_variable *fs_in_color = nir_variable_create(fs_b.shader, nir_var_shader_in, color_type, "v_color"); fs_in_color->data.location = vs_out_color->data.location; fs_in_color->data.interpolation = vs_out_color->data.interpolation; nir_variable *fs_out_color = nir_variable_create(fs_b.shader, nir_var_shader_out, color_type, "f_color"); fs_out_color->data.location = FRAG_RESULT_DATA0 + frag_output; nir_copy_var(&vs_b, vs_out_pos, vs_in_pos); nir_copy_var(&vs_b, vs_out_color, vs_in_color); nir_copy_var(&fs_b, fs_out_color, fs_in_color); *out_vs = vs_b.shader; *out_fs = fs_b.shader; }
/* passthrough vertex shader */ static nir_shader * build_nir_vs(void) { const struct glsl_type *vec4 = glsl_vec4_type(); nir_builder b; nir_variable *a_position; nir_variable *v_position; nir_builder_init_simple_shader(&b, NULL, MESA_SHADER_VERTEX, NULL); b.shader->info.name = ralloc_strdup(b.shader, "meta_depth_decomp_vs"); a_position = nir_variable_create(b.shader, nir_var_shader_in, vec4, "a_position"); a_position->data.location = VERT_ATTRIB_GENERIC0; v_position = nir_variable_create(b.shader, nir_var_shader_out, vec4, "gl_Position"); v_position->data.location = VARYING_SLOT_POS; nir_copy_var(&b, v_position, a_position); return b.shader; }
static nir_shader * build_nir_vertex_shader(void) { const struct glsl_type *vec4 = glsl_vec4_type(); nir_builder b; nir_builder_init_simple_shader(&b, NULL, MESA_SHADER_VERTEX, NULL); b.shader->info.name = ralloc_strdup(b.shader, "meta_resolve_vs"); nir_variable *pos_out = nir_variable_create(b.shader, nir_var_shader_out, vec4, "gl_Position"); pos_out->data.location = VARYING_SLOT_POS; nir_ssa_def *outvec = radv_meta_gen_rect_vertices(&b); nir_store_var(&b, pos_out, outvec, 0xf); return b.shader; }
/* emit 0, 0, 0, 1 */ static nir_shader * build_nir_fs(void) { const struct glsl_type *vec4 = glsl_vec4_type(); nir_builder b; nir_variable *f_color; /* vec4, fragment output color */ nir_builder_init_simple_shader(&b, NULL, MESA_SHADER_FRAGMENT, NULL); b.shader->info.name = ralloc_asprintf(b.shader, "meta_resolve_fs"); f_color = nir_variable_create(b.shader, nir_var_shader_out, vec4, "f_color"); f_color->data.location = FRAG_RESULT_DATA0; nir_store_var(&b, f_color, nir_imm_vec4(&b, 0.0, 0.0, 0.0, 1.0), 0xf); return b.shader; }
static nir_ssa_def * get_transform(lower_wpos_ytransform_state *state) { if (state->transform == NULL) { /* NOTE: name must be prefixed w/ "gl_" to trigger slot based * special handling in uniform setup: */ nir_variable *var = nir_variable_create(state->shader, nir_var_uniform, glsl_vec4_type(), "gl_FbWposYTransform"); var->num_state_slots = 1; var->state_slots = ralloc_array(var, nir_state_slot, 1); var->state_slots[0].swizzle = SWIZZLE_XYZW; memcpy(var->state_slots[0].tokens, state->options->state_tokens, sizeof(var->state_slots[0].tokens)); state->transform = var; } return nir_load_var(&state->b, state->transform); }
static nir_shader * build_resolve_compute_shader(struct radv_device *dev, bool is_integer, int samples) { nir_builder b; char name[64]; nir_if *outer_if = NULL; const struct glsl_type *sampler_type = glsl_sampler_type(GLSL_SAMPLER_DIM_MS, false, false, GLSL_TYPE_FLOAT); const struct glsl_type *img_type = glsl_sampler_type(GLSL_SAMPLER_DIM_2D, false, false, GLSL_TYPE_FLOAT); snprintf(name, 64, "meta_resolve_cs-%d-%s", samples, is_integer ? "int" : "float"); nir_builder_init_simple_shader(&b, NULL, MESA_SHADER_COMPUTE, NULL); b.shader->info->name = ralloc_strdup(b.shader, name); b.shader->info->cs.local_size[0] = 16; b.shader->info->cs.local_size[1] = 16; b.shader->info->cs.local_size[2] = 1; nir_variable *input_img = nir_variable_create(b.shader, nir_var_uniform, sampler_type, "s_tex"); input_img->data.descriptor_set = 0; input_img->data.binding = 0; nir_variable *output_img = nir_variable_create(b.shader, nir_var_uniform, img_type, "out_img"); output_img->data.descriptor_set = 0; output_img->data.binding = 1; nir_ssa_def *invoc_id = nir_load_system_value(&b, nir_intrinsic_load_local_invocation_id, 0); nir_ssa_def *wg_id = nir_load_system_value(&b, nir_intrinsic_load_work_group_id, 0); nir_ssa_def *block_size = nir_imm_ivec4(&b, b.shader->info->cs.local_size[0], b.shader->info->cs.local_size[1], b.shader->info->cs.local_size[2], 0); nir_ssa_def *global_id = nir_iadd(&b, nir_imul(&b, wg_id, block_size), invoc_id); nir_intrinsic_instr *src_offset = nir_intrinsic_instr_create(b.shader, nir_intrinsic_load_push_constant); src_offset->src[0] = nir_src_for_ssa(nir_imm_int(&b, 0)); src_offset->num_components = 2; nir_ssa_dest_init(&src_offset->instr, &src_offset->dest, 2, 32, "src_offset"); nir_builder_instr_insert(&b, &src_offset->instr); nir_intrinsic_instr *dst_offset = nir_intrinsic_instr_create(b.shader, nir_intrinsic_load_push_constant); dst_offset->src[0] = nir_src_for_ssa(nir_imm_int(&b, 8)); dst_offset->num_components = 2; nir_ssa_dest_init(&dst_offset->instr, &dst_offset->dest, 2, 32, "dst_offset"); nir_builder_instr_insert(&b, &dst_offset->instr); nir_ssa_def *img_coord = nir_iadd(&b, global_id, &src_offset->dest.ssa); /* do a txf_ms on each sample */ nir_ssa_def *tmp; nir_tex_instr *tex = nir_tex_instr_create(b.shader, 2); tex->sampler_dim = GLSL_SAMPLER_DIM_MS; tex->op = nir_texop_txf_ms; tex->src[0].src_type = nir_tex_src_coord; tex->src[0].src = nir_src_for_ssa(img_coord); tex->src[1].src_type = nir_tex_src_ms_index; tex->src[1].src = nir_src_for_ssa(nir_imm_int(&b, 0)); tex->dest_type = nir_type_float; tex->is_array = false; tex->coord_components = 2; tex->texture = nir_deref_var_create(tex, input_img); tex->sampler = NULL; nir_ssa_dest_init(&tex->instr, &tex->dest, 4, 32, "tex"); nir_builder_instr_insert(&b, &tex->instr); tmp = &tex->dest.ssa; nir_variable *color = nir_local_variable_create(b.impl, glsl_vec4_type(), "color"); if (!is_integer && samples > 1) { nir_tex_instr *tex_all_same = nir_tex_instr_create(b.shader, 1); tex_all_same->sampler_dim = GLSL_SAMPLER_DIM_MS; tex_all_same->op = nir_texop_samples_identical; tex_all_same->src[0].src_type = nir_tex_src_coord; tex_all_same->src[0].src = nir_src_for_ssa(img_coord); tex_all_same->dest_type = nir_type_float; tex_all_same->is_array = false; tex_all_same->coord_components = 2; tex_all_same->texture = nir_deref_var_create(tex_all_same, input_img); tex_all_same->sampler = NULL; nir_ssa_dest_init(&tex_all_same->instr, &tex_all_same->dest, 1, 32, "tex"); nir_builder_instr_insert(&b, &tex_all_same->instr); nir_ssa_def *all_same = nir_ine(&b, &tex_all_same->dest.ssa, nir_imm_int(&b, 0)); nir_if *if_stmt = nir_if_create(b.shader); if_stmt->condition = nir_src_for_ssa(all_same); nir_cf_node_insert(b.cursor, &if_stmt->cf_node); b.cursor = nir_after_cf_list(&if_stmt->then_list); for (int i = 1; i < samples; i++) { nir_tex_instr *tex_add = nir_tex_instr_create(b.shader, 2); tex_add->sampler_dim = GLSL_SAMPLER_DIM_MS; tex_add->op = nir_texop_txf_ms; tex_add->src[0].src_type = nir_tex_src_coord; tex_add->src[0].src = nir_src_for_ssa(img_coord); tex_add->src[1].src_type = nir_tex_src_ms_index; tex_add->src[1].src = nir_src_for_ssa(nir_imm_int(&b, i)); tex_add->dest_type = nir_type_float; tex_add->is_array = false; tex_add->coord_components = 2; tex_add->texture = nir_deref_var_create(tex_add, input_img); tex_add->sampler = NULL; nir_ssa_dest_init(&tex_add->instr, &tex_add->dest, 4, 32, "tex"); nir_builder_instr_insert(&b, &tex_add->instr); tmp = nir_fadd(&b, tmp, &tex_add->dest.ssa); } tmp = nir_fdiv(&b, tmp, nir_imm_float(&b, samples)); nir_store_var(&b, color, tmp, 0xf); b.cursor = nir_after_cf_list(&if_stmt->else_list); outer_if = if_stmt; } nir_store_var(&b, color, &tex->dest.ssa, 0xf); if (outer_if) b.cursor = nir_after_cf_node(&outer_if->cf_node); nir_ssa_def *newv = nir_load_var(&b, color); nir_ssa_def *coord = nir_iadd(&b, global_id, &dst_offset->dest.ssa); nir_intrinsic_instr *store = nir_intrinsic_instr_create(b.shader, nir_intrinsic_image_store); store->src[0] = nir_src_for_ssa(coord); store->src[1] = nir_src_for_ssa(nir_ssa_undef(&b, 1, 32)); store->src[2] = nir_src_for_ssa(newv); store->variables[0] = nir_deref_var_create(store, output_img); nir_builder_instr_insert(&b, &store->instr); return b.shader; }
static nir_shader * create_passthrough_tcs(void *mem_ctx, const struct brw_compiler *compiler, const nir_shader_compiler_options *options, const struct brw_tcs_prog_key *key) { nir_builder b; nir_builder_init_simple_shader(&b, mem_ctx, MESA_SHADER_TESS_CTRL, options); nir_shader *nir = b.shader; nir_variable *var; nir_intrinsic_instr *load; nir_intrinsic_instr *store; nir_ssa_def *zero = nir_imm_int(&b, 0); nir_ssa_def *invoc_id = nir_load_system_value(&b, nir_intrinsic_load_invocation_id, 0); nir->info->inputs_read = key->outputs_written; nir->info->outputs_written = key->outputs_written; nir->info->tcs.vertices_out = key->input_vertices; nir->info->name = ralloc_strdup(nir, "passthrough"); nir->num_uniforms = 8 * sizeof(uint32_t); var = nir_variable_create(nir, nir_var_uniform, glsl_vec4_type(), "hdr_0"); var->data.location = 0; var = nir_variable_create(nir, nir_var_uniform, glsl_vec4_type(), "hdr_1"); var->data.location = 1; /* Write the patch URB header. */ for (int i = 0; i <= 1; i++) { load = nir_intrinsic_instr_create(nir, nir_intrinsic_load_uniform); load->num_components = 4; load->src[0] = nir_src_for_ssa(zero); nir_ssa_dest_init(&load->instr, &load->dest, 4, 32, NULL); nir_intrinsic_set_base(load, i * 4 * sizeof(uint32_t)); nir_builder_instr_insert(&b, &load->instr); store = nir_intrinsic_instr_create(nir, nir_intrinsic_store_output); store->num_components = 4; store->src[0] = nir_src_for_ssa(&load->dest.ssa); store->src[1] = nir_src_for_ssa(zero); nir_intrinsic_set_base(store, VARYING_SLOT_TESS_LEVEL_INNER - i); nir_intrinsic_set_write_mask(store, WRITEMASK_XYZW); nir_builder_instr_insert(&b, &store->instr); } /* Copy inputs to outputs. */ uint64_t varyings = key->outputs_written; while (varyings != 0) { const int varying = ffsll(varyings) - 1; load = nir_intrinsic_instr_create(nir, nir_intrinsic_load_per_vertex_input); load->num_components = 4; load->src[0] = nir_src_for_ssa(invoc_id); load->src[1] = nir_src_for_ssa(zero); nir_ssa_dest_init(&load->instr, &load->dest, 4, 32, NULL); nir_intrinsic_set_base(load, varying); nir_builder_instr_insert(&b, &load->instr); store = nir_intrinsic_instr_create(nir, nir_intrinsic_store_per_vertex_output); store->num_components = 4; store->src[0] = nir_src_for_ssa(&load->dest.ssa); store->src[1] = nir_src_for_ssa(invoc_id); store->src[2] = nir_src_for_ssa(zero); nir_intrinsic_set_base(store, varying); nir_intrinsic_set_write_mask(store, WRITEMASK_XYZW); nir_builder_instr_insert(&b, &store->instr); varyings &= ~BITFIELD64_BIT(varying); } nir_validate_shader(nir); nir = brw_preprocess_nir(compiler, nir); return nir; }
static nir_shader * build_nir_itob_compute_shader(struct radv_device *dev) { nir_builder b; const struct glsl_type *sampler_type = glsl_sampler_type(GLSL_SAMPLER_DIM_2D, false, false, GLSL_TYPE_FLOAT); const struct glsl_type *img_type = glsl_sampler_type(GLSL_SAMPLER_DIM_BUF, false, false, GLSL_TYPE_FLOAT); nir_builder_init_simple_shader(&b, NULL, MESA_SHADER_COMPUTE, NULL); b.shader->info->name = ralloc_strdup(b.shader, "meta_itob_cs"); b.shader->info->cs.local_size[0] = 16; b.shader->info->cs.local_size[1] = 16; b.shader->info->cs.local_size[2] = 1; nir_variable *input_img = nir_variable_create(b.shader, nir_var_uniform, sampler_type, "s_tex"); input_img->data.descriptor_set = 0; input_img->data.binding = 0; nir_variable *output_img = nir_variable_create(b.shader, nir_var_uniform, img_type, "out_img"); output_img->data.descriptor_set = 0; output_img->data.binding = 1; nir_ssa_def *invoc_id = nir_load_system_value(&b, nir_intrinsic_load_local_invocation_id, 0); nir_ssa_def *wg_id = nir_load_system_value(&b, nir_intrinsic_load_work_group_id, 0); nir_ssa_def *block_size = nir_imm_ivec4(&b, b.shader->info->cs.local_size[0], b.shader->info->cs.local_size[1], b.shader->info->cs.local_size[2], 0); nir_ssa_def *global_id = nir_iadd(&b, nir_imul(&b, wg_id, block_size), invoc_id); nir_intrinsic_instr *offset = nir_intrinsic_instr_create(b.shader, nir_intrinsic_load_push_constant); offset->src[0] = nir_src_for_ssa(nir_imm_int(&b, 0)); offset->num_components = 2; nir_ssa_dest_init(&offset->instr, &offset->dest, 2, 32, "offset"); nir_builder_instr_insert(&b, &offset->instr); nir_intrinsic_instr *stride = nir_intrinsic_instr_create(b.shader, nir_intrinsic_load_push_constant); stride->src[0] = nir_src_for_ssa(nir_imm_int(&b, 8)); stride->num_components = 1; nir_ssa_dest_init(&stride->instr, &stride->dest, 1, 32, "stride"); nir_builder_instr_insert(&b, &stride->instr); nir_ssa_def *img_coord = nir_iadd(&b, global_id, &offset->dest.ssa); nir_tex_instr *tex = nir_tex_instr_create(b.shader, 2); tex->sampler_dim = GLSL_SAMPLER_DIM_2D; tex->op = nir_texop_txf; tex->src[0].src_type = nir_tex_src_coord; tex->src[0].src = nir_src_for_ssa(img_coord); tex->src[1].src_type = nir_tex_src_lod; tex->src[1].src = nir_src_for_ssa(nir_imm_int(&b, 0)); tex->dest_type = nir_type_float; tex->is_array = false; tex->coord_components = 2; tex->texture = nir_deref_var_create(tex, input_img); tex->sampler = NULL; nir_ssa_dest_init(&tex->instr, &tex->dest, 4, 32, "tex"); nir_builder_instr_insert(&b, &tex->instr); nir_ssa_def *pos_x = nir_channel(&b, global_id, 0); nir_ssa_def *pos_y = nir_channel(&b, global_id, 1); nir_ssa_def *tmp = nir_imul(&b, pos_y, &stride->dest.ssa); tmp = nir_iadd(&b, tmp, pos_x); nir_ssa_def *coord = nir_vec4(&b, tmp, tmp, tmp, tmp); nir_ssa_def *outval = &tex->dest.ssa; nir_intrinsic_instr *store = nir_intrinsic_instr_create(b.shader, nir_intrinsic_image_store); store->src[0] = nir_src_for_ssa(coord); store->src[1] = nir_src_for_ssa(nir_ssa_undef(&b, 1, 32)); store->src[2] = nir_src_for_ssa(outval); store->variables[0] = nir_deref_var_create(store, output_img); nir_builder_instr_insert(&b, &store->instr); return b.shader; }
static void *vc4_get_yuv_vs(struct pipe_context *pctx) { struct vc4_context *vc4 = vc4_context(pctx); struct pipe_screen *pscreen = pctx->screen; if (vc4->yuv_linear_blit_vs) return vc4->yuv_linear_blit_vs; const struct nir_shader_compiler_options *options = pscreen->get_compiler_options(pscreen, PIPE_SHADER_IR_NIR, PIPE_SHADER_VERTEX); nir_builder b; nir_builder_init_simple_shader(&b, NULL, MESA_SHADER_VERTEX, options); b.shader->info.name = ralloc_strdup(b.shader, "linear_blit_vs"); const struct glsl_type *vec4 = glsl_vec4_type(); nir_variable *pos_in = nir_variable_create(b.shader, nir_var_shader_in, vec4, "pos"); nir_variable *pos_out = nir_variable_create(b.shader, nir_var_shader_out, vec4, "gl_Position"); pos_out->data.location = VARYING_SLOT_POS; nir_store_var(&b, pos_out, nir_load_var(&b, pos_in), 0xf); struct pipe_shader_state shader_tmpl = { .type = PIPE_SHADER_IR_NIR, .ir.nir = b.shader, }; vc4->yuv_linear_blit_vs = pctx->create_vs_state(pctx, &shader_tmpl); return vc4->yuv_linear_blit_vs; } static void *vc4_get_yuv_fs(struct pipe_context *pctx, int cpp) { struct vc4_context *vc4 = vc4_context(pctx); struct pipe_screen *pscreen = pctx->screen; struct pipe_shader_state **cached_shader; const char *name; if (cpp == 1) { cached_shader = &vc4->yuv_linear_blit_fs_8bit; name = "linear_blit_8bit_fs"; } else { cached_shader = &vc4->yuv_linear_blit_fs_16bit; name = "linear_blit_16bit_fs"; } if (*cached_shader) return *cached_shader; const struct nir_shader_compiler_options *options = pscreen->get_compiler_options(pscreen, PIPE_SHADER_IR_NIR, PIPE_SHADER_FRAGMENT); nir_builder b; nir_builder_init_simple_shader(&b, NULL, MESA_SHADER_FRAGMENT, options); b.shader->info.name = ralloc_strdup(b.shader, name); const struct glsl_type *vec4 = glsl_vec4_type(); const struct glsl_type *glsl_int = glsl_int_type(); nir_variable *color_out = nir_variable_create(b.shader, nir_var_shader_out, vec4, "f_color"); color_out->data.location = FRAG_RESULT_COLOR; nir_variable *pos_in = nir_variable_create(b.shader, nir_var_shader_in, vec4, "pos"); pos_in->data.location = VARYING_SLOT_POS; nir_ssa_def *pos = nir_load_var(&b, pos_in); nir_ssa_def *one = nir_imm_int(&b, 1); nir_ssa_def *two = nir_imm_int(&b, 2); nir_ssa_def *x = nir_f2i32(&b, nir_channel(&b, pos, 0)); nir_ssa_def *y = nir_f2i32(&b, nir_channel(&b, pos, 1)); nir_variable *stride_in = nir_variable_create(b.shader, nir_var_uniform, glsl_int, "stride"); nir_ssa_def *stride = nir_load_var(&b, stride_in); nir_ssa_def *x_offset; nir_ssa_def *y_offset; if (cpp == 1) { nir_ssa_def *intra_utile_x_offset = nir_ishl(&b, nir_iand(&b, x, one), two); nir_ssa_def *inter_utile_x_offset = nir_ishl(&b, nir_iand(&b, x, nir_imm_int(&b, ~3)), one); x_offset = nir_iadd(&b, intra_utile_x_offset, inter_utile_x_offset); y_offset = nir_imul(&b, nir_iadd(&b, nir_ishl(&b, y, one), nir_ushr(&b, nir_iand(&b, x, two), one)), stride); } else { x_offset = nir_ishl(&b, x, two); y_offset = nir_imul(&b, y, stride); } nir_intrinsic_instr *load = nir_intrinsic_instr_create(b.shader, nir_intrinsic_load_ubo); load->num_components = 1; nir_ssa_dest_init(&load->instr, &load->dest, load->num_components, 32, NULL); load->src[0] = nir_src_for_ssa(one); load->src[1] = nir_src_for_ssa(nir_iadd(&b, x_offset, y_offset)); nir_builder_instr_insert(&b, &load->instr); nir_store_var(&b, color_out, nir_unpack_unorm_4x8(&b, &load->dest.ssa), 0xf); struct pipe_shader_state shader_tmpl = { .type = PIPE_SHADER_IR_NIR, .ir.nir = b.shader, }; *cached_shader = pctx->create_fs_state(pctx, &shader_tmpl); return *cached_shader; } static bool vc4_yuv_blit(struct pipe_context *pctx, const struct pipe_blit_info *info) { struct vc4_context *vc4 = vc4_context(pctx); struct vc4_resource *src = vc4_resource(info->src.resource); struct vc4_resource *dst = vc4_resource(info->dst.resource); bool ok; if (src->tiled) return false; if (src->base.format != PIPE_FORMAT_R8_UNORM && src->base.format != PIPE_FORMAT_R8G8_UNORM) return false; /* YUV blits always turn raster-order to tiled */ assert(dst->base.format == src->base.format); assert(dst->tiled); /* Always 1:1 and at the origin */ assert(info->src.box.x == 0 && info->dst.box.x == 0); assert(info->src.box.y == 0 && info->dst.box.y == 0); assert(info->src.box.width == info->dst.box.width); assert(info->src.box.height == info->dst.box.height); if ((src->slices[info->src.level].offset & 3) || (src->slices[info->src.level].stride & 3)) { perf_debug("YUV-blit src texture offset/stride misaligned: 0x%08x/%d\n", src->slices[info->src.level].offset, src->slices[info->src.level].stride); goto fallback; } vc4_blitter_save(vc4); /* Create a renderable surface mapping the T-tiled shadow buffer. */ struct pipe_surface dst_tmpl; util_blitter_default_dst_texture(&dst_tmpl, info->dst.resource, info->dst.level, info->dst.box.z); dst_tmpl.format = PIPE_FORMAT_RGBA8888_UNORM; struct pipe_surface *dst_surf = pctx->create_surface(pctx, info->dst.resource, &dst_tmpl); if (!dst_surf) { fprintf(stderr, "Failed to create YUV dst surface\n"); util_blitter_unset_running_flag(vc4->blitter); return false; } dst_surf->width /= 2; if (dst->cpp == 1) dst_surf->height /= 2; /* Set the constant buffer. */ uint32_t stride = src->slices[info->src.level].stride; struct pipe_constant_buffer cb_uniforms = { .user_buffer = &stride, .buffer_size = sizeof(stride), }; pctx->set_constant_buffer(pctx, PIPE_SHADER_FRAGMENT, 0, &cb_uniforms); struct pipe_constant_buffer cb_src = { .buffer = info->src.resource, .buffer_offset = src->slices[info->src.level].offset, .buffer_size = (src->bo->size - src->slices[info->src.level].offset), }; pctx->set_constant_buffer(pctx, PIPE_SHADER_FRAGMENT, 1, &cb_src); /* Unbind the textures, to make sure we don't try to recurse into the * shadow blit. */ pctx->set_sampler_views(pctx, PIPE_SHADER_FRAGMENT, 0, 0, NULL); pctx->bind_sampler_states(pctx, PIPE_SHADER_FRAGMENT, 0, 0, NULL); util_blitter_custom_shader(vc4->blitter, dst_surf, vc4_get_yuv_vs(pctx), vc4_get_yuv_fs(pctx, src->cpp)); util_blitter_restore_textures(vc4->blitter); util_blitter_restore_constant_buffer_state(vc4->blitter); /* Restore cb1 (util_blitter doesn't handle this one). */ struct pipe_constant_buffer cb_disabled = { 0 }; pctx->set_constant_buffer(pctx, PIPE_SHADER_FRAGMENT, 1, &cb_disabled); pipe_surface_reference(&dst_surf, NULL); return true; fallback: /* Do an immediate SW fallback, since the render blit path * would just recurse. */ ok = util_try_blit_via_copy_region(pctx, info); assert(ok); (void)ok; return true; } static bool vc4_render_blit(struct pipe_context *ctx, struct pipe_blit_info *info) { struct vc4_context *vc4 = vc4_context(ctx); if (!util_blitter_is_blit_supported(vc4->blitter, info)) { fprintf(stderr, "blit unsupported %s -> %s\n", util_format_short_name(info->src.resource->format), util_format_short_name(info->dst.resource->format)); return false; } /* Enable the scissor, so we get a minimal set of tiles rendered. */ if (!info->scissor_enable) { info->scissor_enable = true; info->scissor.minx = info->dst.box.x; info->scissor.miny = info->dst.box.y; info->scissor.maxx = info->dst.box.x + info->dst.box.width; info->scissor.maxy = info->dst.box.y + info->dst.box.height; } vc4_blitter_save(vc4); util_blitter_blit(vc4->blitter, info); return true; } /* Optimal hardware path for blitting pixels. * Scaling, format conversion, up- and downsampling (resolve) are allowed. */ void vc4_blit(struct pipe_context *pctx, const struct pipe_blit_info *blit_info) { struct pipe_blit_info info = *blit_info; if (vc4_yuv_blit(pctx, blit_info)) return; if (vc4_tile_blit(pctx, blit_info)) return; if (info.mask & PIPE_MASK_S) { if (util_try_blit_via_copy_region(pctx, &info)) return; info.mask &= ~PIPE_MASK_S; fprintf(stderr, "cannot blit stencil, skipping\n"); } if (vc4_render_blit(pctx, &info)) return; fprintf(stderr, "Unsupported blit\n"); }