mirrors
/
vkd3d-proton
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

vkd3d: Add wave size workaround for GravityMark. 

Signed-off-by: Hans-Kristian Arntzen <post@arntzen-software.no>
This commit is contained in:
Hans-Kristian Arntzen 2021-07-02 13:36:03 +02:00
parent 92c4f861e7
commit 419790ac77
3 changed files with 38 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
include/vkd3d.h
View File

@ -69,6 +69,7 @@ enum vkd3d_config_flags
VKD3D_CONFIG_FLAG_DESCRIPTOR_QA_CHECKS = 0x00000040,
VKD3D_CONFIG_FLAG_FORCE_RTV_EXCLUSIVE_QUEUE = 0x00000080,
VKD3D_CONFIG_FLAG_FORCE_DSV_EXCLUSIVE_QUEUE = 0x00000100,
VKD3D_CONFIG_FLAG_FORCE_MINIMUM_SUBGROUP_SIZE = 0x00000200,
};
typedef HRESULT (*PFN_vkd3d_signal_event)(HANDLE event);

1
libs/vkd3d/device.c
View File

@ -417,6 +417,7 @@ struct vkd3d_instance_application_meta
};
static const struct vkd3d_instance_application_meta application_override[] = {
/* MSVC fails to compile empty array. */
{ "GravityMark.exe", VKD3D_CONFIG_FLAG_FORCE_MINIMUM_SUBGROUP_SIZE, 0 },
{ NULL, 0, 0 }
};

41
libs/vkd3d/state.c
View File

@ -2060,13 +2060,14 @@ struct d3d12_pipeline_state *unsafe_impl_from_ID3D12PipelineState(ID3D12Pipeline
}
static HRESULT create_shader_stage(struct d3d12_device *device,
struct VkPipelineShaderStageCreateInfo *stage_desc, VkShaderStageFlagBits stage,
VkPipelineShaderStageCreateInfo *stage_desc, VkShaderStageFlagBits stage,
VkPipelineShaderStageRequiredSubgroupSizeCreateInfoEXT *required_subgroup_size_info,
const D3D12_SHADER_BYTECODE *code, const struct vkd3d_shader_interface_info *shader_interface,
const struct vkd3d_shader_compile_arguments *compile_args, struct vkd3d_shader_meta *meta)
{
struct vkd3d_shader_code dxbc = {code->pShaderBytecode, code->BytecodeLength};
const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs;
struct VkShaderModuleCreateInfo shader_desc;
VkShaderModuleCreateInfo shader_desc;
struct vkd3d_shader_code spirv = {0};
char hash_str[16 + 1];
VkResult vr;
@ -2095,7 +2096,35 @@ static HRESULT create_shader_stage(struct d3d12_device *device,
*meta = spirv.meta;
if (spirv.meta.uses_subgroup_size && device->device_info.subgroup_size_control_features.subgroupSizeControl)
stage_desc->flags |= VK_PIPELINE_SHADER_STAGE_CREATE_ALLOW_VARYING_SUBGROUP_SIZE_BIT_EXT;
{
/* GravityMark checks minSubgroupSize and based on that uses a shader variant.
* This shader variant unfortunately expects that a subgroup 32 variant will actually use wave32 on AMD.
* amdgpu-pro and AMDVLK happens to emit wave32, but RADV will emit wave64 here unless we force it to be wave32.
* This is an application bug, since the shader is not guaranteed a specific size, but we can only workaround ...
* This path will also be relevant in SM 6.6 where we have to handle [WaveSize(N)] attribute. */
uint32_t subgroup_size_alignment = device->device_info.subgroup_size_control_properties.maxSubgroupSize;
if (required_subgroup_size_info &&
(vkd3d_config_flags & VKD3D_CONFIG_FLAG_FORCE_MINIMUM_SUBGROUP_SIZE) &&
(device->device_info.subgroup_size_control_properties.requiredSubgroupSizeStages & stage))
{
subgroup_size_alignment = device->device_info.subgroup_size_control_properties.minSubgroupSize;
required_subgroup_size_info->sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_REQUIRED_SUBGROUP_SIZE_CREATE_INFO_EXT;
required_subgroup_size_info->pNext = NULL;
required_subgroup_size_info->requiredSubgroupSize = subgroup_size_alignment;
stage_desc->pNext = required_subgroup_size_info;
}
else
stage_desc->flags |= VK_PIPELINE_SHADER_STAGE_CREATE_ALLOW_VARYING_SUBGROUP_SIZE_BIT_EXT;
/* If we can, we should be explicit and enable FULL_SUBGROUPS bit as well. This should be default
* behavior, but cannot hurt. */
if (stage == VK_SHADER_STAGE_COMPUTE_BIT &&
device->device_info.subgroup_size_control_features.computeFullSubgroups &&
!(spirv.meta.cs_workgroup_size[0] % subgroup_size_alignment))
{
stage_desc->flags |= VK_PIPELINE_SHADER_STAGE_CREATE_REQUIRE_FULL_SUBGROUPS_BIT_EXT;
}
}
vr = VK_CALL(vkCreateShaderModule(device->vk_device, &shader_desc, NULL, &stage_desc->module));
vkd3d_shader_free_shader_code(&spirv);
@ -2117,6 +2146,7 @@ static HRESULT vkd3d_create_compute_pipeline(struct d3d12_device *device,
VkPipelineLayout vk_pipeline_layout, VkPipelineCache vk_cache, VkPipeline *vk_pipeline,
struct vkd3d_shader_meta *meta)
{
VkPipelineShaderStageRequiredSubgroupSizeCreateInfoEXT required_subgroup_size_info;
const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs;
struct vkd3d_shader_debug_ring_spec_info spec_info;
struct vkd3d_shader_compile_arguments compile_args;
@ -2134,7 +2164,8 @@ static HRESULT vkd3d_create_compute_pipeline(struct d3d12_device *device,
pipeline_info.pNext = NULL;
pipeline_info.flags = 0;
if (FAILED(hr = create_shader_stage(device, &pipeline_info.stage,
VK_SHADER_STAGE_COMPUTE_BIT, code, shader_interface, &compile_args, meta)))
VK_SHADER_STAGE_COMPUTE_BIT, &required_subgroup_size_info,
code, shader_interface, &compile_args, meta)))
return hr;
pipeline_info.layout = vk_pipeline_layout;
pipeline_info.basePipelineHandle = VK_NULL_HANDLE;
@ -3159,7 +3190,7 @@ static HRESULT d3d12_pipeline_state_init_graphics(struct d3d12_pipeline_state *s
shader_interface.xfb_info = shader_stages[i].stage == xfb_stage ? &xfb_info : NULL;
shader_interface.stage = shader_stages[i].stage;
if (FAILED(hr = create_shader_stage(device, &graphics->stages[graphics->stage_count],
shader_stages[i].stage, b, &shader_interface,
shader_stages[i].stage, NULL, b, &shader_interface,
shader_stages[i].stage == VK_SHADER_STAGE_FRAGMENT_BIT ? &ps_compile_args : &compile_args,
&graphics->stage_meta[graphics->stage_count])))
goto fail;