.lower_workgroup_access_to_offsets = true,
.caps = {
.derivative_group = true,
+ .descriptor_array_dynamic_indexing = true,
+ .descriptor_array_non_uniform_indexing = true,
+ .descriptor_indexing = true,
.device_group = true,
.draw_parameters = true,
.float16 = pdevice->info.gen >= 8,
.multiview = true,
.physical_storage_buffer_address = pdevice->has_a64_buffer_access,
.post_depth_coverage = pdevice->info.gen >= 9,
+ .runtime_descriptor_array = true,
.shader_viewport_index_layer = true,
.stencil_export = pdevice->info.gen >= 9,
.storage_8bit = pdevice->info.gen >= 8,
.transform_feedback = pdevice->info.gen >= 8,
.variable_pointers = true,
},
- .ubo_ptr_type = glsl_vector_type(GLSL_TYPE_UINT, 2),
- .phys_ssbo_ptr_type = glsl_vector_type(GLSL_TYPE_UINT64, 1),
- .push_const_ptr_type = glsl_uint_type(),
- .shared_ptr_type = glsl_uint_type(),
+ .ubo_addr_format = nir_address_format_32bit_index_offset,
+ .ssbo_addr_format =
+ anv_nir_ssbo_addr_format(pdevice, device->robust_buffer_access),
+ .phys_ssbo_addr_format = nir_address_format_64bit_global,
+ .push_const_addr_format = nir_address_format_logical,
+
+ /* TODO: Consider changing this to an address format that has the NULL
+ * pointer equals to 0. That might be a better format to play nice
+ * with certain code / code generators.
+ */
+ .shared_addr_format = nir_address_format_32bit_offset,
};
- if (pdevice->has_a64_buffer_access) {
- if (device->robust_buffer_access)
- spirv_options.ssbo_ptr_type = glsl_vector_type(GLSL_TYPE_UINT, 4);
- else
- spirv_options.ssbo_ptr_type = glsl_vector_type(GLSL_TYPE_UINT64, 1);
- } else {
- spirv_options.ssbo_ptr_type = glsl_vector_type(GLSL_TYPE_UINT, 2);
- }
- nir_function *entry_point =
+ nir_shader *nir =
spirv_to_nir(spirv, module->size / 4,
spec_entries, num_spec_entries,
stage, entrypoint_name, &spirv_options, nir_options);
- nir_shader *nir = entry_point->shader;
assert(nir->info.stage == stage);
nir_validate_shader(nir, "after spirv_to_nir");
ralloc_steal(mem_ctx, nir);
/* Pick off the single entrypoint that we want */
foreach_list_typed_safe(nir_function, func, node, &nir->functions) {
- if (func != entry_point)
+ if (!func->is_entrypoint)
exec_node_remove(&func->node);
}
assert(exec_list_length(&nir->functions) == 1);
NIR_PASS_V(nir, nir_propagate_invariant);
NIR_PASS_V(nir, nir_lower_io_to_temporaries,
- entry_point->impl, true, false);
+ nir_shader_get_entrypoint(nir), true, false);
NIR_PASS_V(nir, nir_lower_frexp);
* harmless to compute it and then let dead-code take care of it.
*/
if (ms_info->rasterizationSamples > 1) {
- key->persample_interp =
+ key->persample_interp = ms_info->sampleShadingEnable &&
(ms_info->minSampleShading * ms_info->rasterizationSamples) > 1;
key->multisample_fbo = true;
}
- key->frag_coord_adds_sample_pos = ms_info->sampleShadingEnable;
+ key->frag_coord_adds_sample_pos = key->persample_interp;
}
}
NIR_PASS_V(nir, nir_lower_explicit_io, nir_var_mem_ubo,
nir_address_format_32bit_index_offset);
-
- nir_address_format ssbo_address_format;
- if (pdevice->has_a64_buffer_access) {
- if (pipeline->device->robust_buffer_access)
- ssbo_address_format = nir_address_format_64bit_bounded_global;
- else
- ssbo_address_format = nir_address_format_64bit_global;
- } else {
- ssbo_address_format = nir_address_format_32bit_index_offset;
- }
NIR_PASS_V(nir, nir_lower_explicit_io, nir_var_mem_ssbo,
- ssbo_address_format);
+ anv_nir_ssbo_addr_format(pdevice,
+ pipeline->device->robust_buffer_access));
NIR_PASS_V(nir, nir_opt_constant_folding);
+
+ /* We don't support non-uniform UBOs and non-uniform SSBO access is
+ * handled naturally by falling back to A64 messages.
+ */
+ NIR_PASS_V(nir, nir_lower_non_uniform_access,
+ nir_lower_non_uniform_texture_access |
+ nir_lower_non_uniform_image_access);
}
if (nir->info.stage != MESA_SHADER_COMPUTE)
continue;
const unsigned rt = var->data.location - FRAG_RESULT_DATA0;
- /* Unused or out-of-bounds */
- if (rt >= MAX_RTS || !(stage->key.wm.color_outputs_valid & (1 << rt)))
+ /* Out-of-bounds */
+ if (rt >= MAX_RTS)
continue;
const unsigned array_len =
glsl_type_is_array(var->type) ? glsl_get_length(var->type) : 1;
assert(rt + array_len <= max_rt);
+ /* Unused */
+ if (!(stage->key.wm.color_outputs_valid & BITFIELD_RANGE(rt, array_len))) {
+ /* If this is the RT at location 0 and we have alpha to coverage
+ * enabled we will have to create a null RT for it, so mark it as
+ * used.
+ */
+ if (rt > 0 || !stage->key.wm.alpha_to_coverage)
+ continue;
+ }
+
for (unsigned i = 0; i < array_len; i++)
rt_used[rt + i] = true;
}
continue;
rt_to_bindings[i] = num_rts;
- rt_bindings[rt_to_bindings[i]] = (struct anv_pipeline_binding) {
- .set = ANV_DESCRIPTOR_SET_COLOR_ATTACHMENTS,
- .binding = 0,
- .index = i,
- };
+
+ if (stage->key.wm.color_outputs_valid & (1 << i)) {
+ rt_bindings[rt_to_bindings[i]] = (struct anv_pipeline_binding) {
+ .set = ANV_DESCRIPTOR_SET_COLOR_ATTACHMENTS,
+ .binding = 0,
+ .index = i,
+ };
+ } else {
+ /* Setup a null render target */
+ rt_bindings[rt_to_bindings[i]] = (struct anv_pipeline_binding) {
+ .set = ANV_DESCRIPTOR_SET_COLOR_ATTACHMENTS,
+ .binding = 0,
+ .index = UINT32_MAX,
+ };
+ }
+
num_rts++;
}
continue;
const unsigned rt = var->data.location - FRAG_RESULT_DATA0;
- if (rt >= MAX_RTS ||
- !(stage->key.wm.color_outputs_valid & (1 << rt))) {
- /* Unused or out-of-bounds, throw it away */
+
+ if (rt >= MAX_RTS || !rt_used[rt]) {
+ /* Unused or out-of-bounds, throw it away, unless it is the first
+ * RT and we have alpha to coverage enabled.
+ */
deleted_output = true;
var->data.mode = nir_var_function_temp;
exec_node_remove(&var->node);