* Compute queue: implementation also of buffer->image, image->image, and image clear.
*/
+/* GFX9 needs to use a 3D sampler to access 3D resources, so the shader has the options
+ * for that.
+ */
static nir_shader *
-build_nir_itob_compute_shader(struct radv_device *dev)
+build_nir_itob_compute_shader(struct radv_device *dev, bool is_3d)
{
nir_builder b;
- const struct glsl_type *sampler_type = glsl_sampler_type(GLSL_SAMPLER_DIM_2D,
+ enum glsl_sampler_dim dim = is_3d ? GLSL_SAMPLER_DIM_3D : GLSL_SAMPLER_DIM_2D;
+ const struct glsl_type *sampler_type = glsl_sampler_type(dim,
false,
false,
GLSL_TYPE_FLOAT);
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.name = ralloc_strdup(b.shader, is_3d ? "meta_itob_cs_3d" : "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_intrinsic_instr *offset = nir_intrinsic_instr_create(b.shader, nir_intrinsic_load_push_constant);
nir_intrinsic_set_base(offset, 0);
- nir_intrinsic_set_range(offset, 12);
+ nir_intrinsic_set_range(offset, 16);
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");
+ offset->num_components = is_3d ? 3 : 2;
+ nir_ssa_dest_init(&offset->instr, &offset->dest, is_3d ? 3 : 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);
nir_intrinsic_set_base(stride, 0);
- nir_intrinsic_set_range(stride, 12);
- stride->src[0] = nir_src_for_ssa(nir_imm_int(&b, 8));
+ nir_intrinsic_set_range(stride, 16);
+ stride->src[0] = nir_src_for_ssa(nir_imm_int(&b, 12));
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->sampler_dim = dim;
tex->op = nir_texop_txf;
tex->src[0].src_type = nir_tex_src_coord;
- tex->src[0].src = nir_src_for_ssa(nir_channels(&b, img_coord, 0x3));
+ tex->src[0].src = nir_src_for_ssa(nir_channels(&b, img_coord, is_3d ? 0x7 : 0x3));
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->coord_components = is_3d ? 3 : 2;
tex->texture = nir_deref_var_create(tex, input_img);
tex->sampler = NULL;
{
VkResult result;
struct radv_shader_module cs = { .nir = NULL };
+ struct radv_shader_module cs_3d = { .nir = NULL };
- cs.nir = build_nir_itob_compute_shader(device);
+ cs.nir = build_nir_itob_compute_shader(device, false);
+ if (device->physical_device->rad_info.chip_class >= GFX9)
+ cs_3d.nir = build_nir_itob_compute_shader(device, true);
/*
* two descriptors one for the image being sampled
.setLayoutCount = 1,
.pSetLayouts = &device->meta_state.itob.img_ds_layout,
.pushConstantRangeCount = 1,
- .pPushConstantRanges = &(VkPushConstantRange){VK_SHADER_STAGE_COMPUTE_BIT, 0, 12},
+ .pPushConstantRanges = &(VkPushConstantRange){VK_SHADER_STAGE_COMPUTE_BIT, 0, 16},
};
result = radv_CreatePipelineLayout(radv_device_to_handle(device),
if (result != VK_SUCCESS)
goto fail;
+ if (device->physical_device->rad_info.chip_class >= GFX9) {
+ VkPipelineShaderStageCreateInfo pipeline_shader_stage_3d = {
+ .sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO,
+ .stage = VK_SHADER_STAGE_COMPUTE_BIT,
+ .module = radv_shader_module_to_handle(&cs_3d),
+ .pName = "main",
+ .pSpecializationInfo = NULL,
+ };
+
+ VkComputePipelineCreateInfo vk_pipeline_info_3d = {
+ .sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO,
+ .stage = pipeline_shader_stage_3d,
+ .flags = 0,
+ .layout = device->meta_state.itob.img_p_layout,
+ };
+
+ result = radv_CreateComputePipelines(radv_device_to_handle(device),
+ radv_pipeline_cache_to_handle(&device->meta_state.cache),
+ 1, &vk_pipeline_info_3d, NULL,
+ &device->meta_state.itob.pipeline_3d);
+ if (result != VK_SUCCESS)
+ goto fail;
+ ralloc_free(cs_3d.nir);
+ }
ralloc_free(cs.nir);
+
return VK_SUCCESS;
fail:
ralloc_free(cs.nir);
+ ralloc_free(cs_3d.nir);
return result;
}
&state->alloc);
radv_DestroyPipeline(radv_device_to_handle(device),
state->itob.pipeline, &state->alloc);
+ if (device->physical_device->rad_info.chip_class >= GFX9)
+ radv_DestroyPipeline(radv_device_to_handle(device),
+ state->itob.pipeline_3d, &state->alloc);
}
static nir_shader *
struct radv_meta_blit2d_surf *surf,
struct radv_image_view *iview)
{
-
+ VkImageViewType view_type = cmd_buffer->device->physical_device->rad_info.chip_class < GFX9 ? VK_IMAGE_VIEW_TYPE_2D :
+ radv_meta_get_view_type(surf->image);
radv_image_view_init(iview, cmd_buffer->device,
&(VkImageViewCreateInfo) {
.sType = VK_STRUCTURE_TYPE_IMAGE_VIEW_CREATE_INFO,
.image = radv_image_to_handle(surf->image),
- .viewType = VK_IMAGE_VIEW_TYPE_2D,
+ .viewType = view_type,
.format = surf->format,
.subresourceRange = {
.aspectMask = surf->aspect_mask,
create_bview(cmd_buffer, dst->buffer, dst->offset, dst->format, &dst_view);
itob_bind_descriptors(cmd_buffer, &src_view, &dst_view);
+ if (device->physical_device->rad_info.chip_class >= GFX9 &&
+ src->image->type == VK_IMAGE_TYPE_3D)
+ pipeline = cmd_buffer->device->meta_state.itob.pipeline_3d;
radv_CmdBindPipeline(radv_cmd_buffer_to_handle(cmd_buffer),
VK_PIPELINE_BIND_POINT_COMPUTE, pipeline);
for (unsigned r = 0; r < num_rects; ++r) {
- unsigned push_constants[3] = {
+ unsigned push_constants[4] = {
rects[r].src_x,
rects[r].src_y,
+ src->layer,
dst->pitch
};
radv_CmdPushConstants(radv_cmd_buffer_to_handle(cmd_buffer),
device->meta_state.itob.img_p_layout,
- VK_SHADER_STAGE_COMPUTE_BIT, 0, 12,
+ VK_SHADER_STAGE_COMPUTE_BIT, 0, 16,
push_constants);
radv_unaligned_dispatch(cmd_buffer, rects[r].width, rects[r].height, 1);