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

vkd3d: Rewrite the pipeline library implementation. 

This became basically a rewrite in the end, and it got too awkward to
split these commits in any meaningful way.

The goals here were primarily to:

- Support serializing SPIR-V and load SPIR-V.
  To do this robustly requires a lot more validation and checks to make
  sure end up compiling the same SPIR-V that we load from cache.
  This is critical for performance when games have primed their pipeline
  libraries and expect that loading a PSO should be fast. Without this,
  we will hit vkd3d-shader for every PSO, causing very long load times.
- Implement the required validation for mismatched PSO descriptions.
- Rewrite the binary layout of the pipeline library for flexibility
  concerns and performance.
  If the pipeline library is mmap-ed from disk - which appears to be
  the intended use - we only need to scan through the TOC to fully parse
  the library contents.
  From a flexibility concern, a blob needs to support inlined data,
  but a library can use referential links. We introduce separate
  hashmaps which store deduplicated SPIR-V and pipeline cache blobs,
  which significantly drop memory and storage requirements.
  For future improvements, it should be fairly easy to add information
  which lets us avoid SPIR-V or pipeline cache data altogether if
  relevant changes to Vulkan/drivers are made.

Signed-off-by: Hans-Kristian Arntzen <post@arntzen-software.no>
This commit is contained in:
Hans-Kristian Arntzen 2022-01-20 14:20:19 +01:00
parent a078197e16
commit 5d345f47cc
5 changed files with 1549 additions and 237 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1440
libs/vkd3d/cache.c
View File

File diff suppressed because it is too large Load Diff

6
libs/vkd3d/command.c
View File

@ -6934,15 +6934,15 @@ static void STDMETHODCALLTYPE d3d12_command_list_SetPipelineState(d3d12_command_
if (state->vk_bind_point == VK_PIPELINE_BIND_POINT_COMPUTE)
{
TRACE("Binding compute module with hash: %016"PRIx64".\n", state->compute.meta.hash);
TRACE("Binding compute module with hash: %016"PRIx64".\n", state->compute.code.meta.hash);
}
else if (state->vk_bind_point == VK_PIPELINE_BIND_POINT_GRAPHICS)
{
for (i = 0; i < state->graphics.stage_count; i++)
{
TRACE("Binding graphics module with hash: %016"PRIx64" (replaced: %s).\n",
state->graphics.stage_meta[i].hash,
(state->graphics.stage_meta[i].flags & VKD3D_SHADER_META_FLAG_REPLACED) ? "yes" : "no");
state->graphics.code[i].meta.hash,
(state->graphics.code[i].meta.flags & VKD3D_SHADER_META_FLAG_REPLACED) ? "yes" : "no");
}
}
}

4
libs/vkd3d/renderdoc.c
View File

@ -238,7 +238,7 @@ void vkd3d_renderdoc_command_list_check_capture(struct d3d12_command_list *list,
{
if (state->vk_bind_point == VK_PIPELINE_BIND_POINT_COMPUTE)
{
if (vkd3d_renderdoc_should_capture_shader_hash(state->compute.meta.hash))
if (vkd3d_renderdoc_should_capture_shader_hash(state->compute.code.meta.hash))
{
WARN("Triggering RenderDoc capture for this command list.\n");
list->debug_capture = true;
@ -248,7 +248,7 @@ void vkd3d_renderdoc_command_list_check_capture(struct d3d12_command_list *list,
{
for (i = 0; i < state->graphics.stage_count; i++)
{
if (vkd3d_renderdoc_should_capture_shader_hash(state->graphics.stage_meta[i].hash))
if (vkd3d_renderdoc_should_capture_shader_hash(state->graphics.code[i].meta.hash))
{
WARN("Triggering RenderDoc capture for this command list.\n");
list->debug_capture = true;

302
libs/vkd3d/state.c
View File

@ -2051,6 +2051,74 @@ static ULONG STDMETHODCALLTYPE d3d12_pipeline_state_AddRef(ID3D12PipelineState *
return refcount;
}
static HRESULT d3d12_pipeline_state_create_shader_module(struct d3d12_device *device,
VkPipelineShaderStageCreateInfo *stage_desc, const struct vkd3d_shader_code *code)
{
const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs;
VkShaderModuleCreateInfo shader_desc;
char hash_str[16 + 1];
VkResult vr;
/* If we kept the module around, no need to create it again. */
if (stage_desc->module != VK_NULL_HANDLE)
return S_OK;
shader_desc.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO;
shader_desc.pNext = NULL;
shader_desc.flags = 0;
shader_desc.codeSize = code->size;
shader_desc.pCode = code->code;
vr = VK_CALL(vkCreateShaderModule(device->vk_device, &shader_desc, NULL, &stage_desc->module));
if (vr < 0)
{
WARN("Failed to create Vulkan shader module, vr %d.\n", vr);
return hresult_from_vk_result(vr);
}
/* Helpful for tooling like RenderDoc. */
sprintf(hash_str, "%016"PRIx64, code->meta.hash);
vkd3d_set_vk_object_name(device, (uint64_t)stage_desc->module, VK_OBJECT_TYPE_SHADER_MODULE, hash_str);
return S_OK;
}
static void d3d12_pipeline_state_free_spirv_code(struct d3d12_pipeline_state *state)
{
unsigned int i;
if (d3d12_pipeline_state_is_graphics(state))
{
for (i = 0; i < state->graphics.stage_count; i++)
{
vkd3d_shader_free_shader_code(&state->graphics.code[i]);
/* Keep meta. */
state->graphics.code[i].code = NULL;
state->graphics.code[i].size = 0;
}
}
else if (d3d12_pipeline_state_is_compute(state))
{
vkd3d_shader_free_shader_code(&state->compute.code);
/* Keep meta. */
state->compute.code.code = NULL;
state->compute.code.size = 0;
}
}
static void d3d12_pipeline_state_destroy_shader_modules(struct d3d12_pipeline_state *state, struct d3d12_device *device)
{
const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs;
unsigned int i;
if (d3d12_pipeline_state_is_graphics(state))
{
for (i = 0; i < state->graphics.stage_count; i++)
{
VK_CALL(vkDestroyShaderModule(device->vk_device, state->graphics.stages[i].module, NULL));
state->graphics.stages[i].module = VK_NULL_HANDLE;
}
}
}
static void d3d12_pipeline_state_destroy_graphics(struct d3d12_pipeline_state *state,
struct d3d12_device *device)
{
@ -2059,10 +2127,7 @@ static void d3d12_pipeline_state_destroy_graphics(struct d3d12_pipeline_state *s
struct vkd3d_compiled_pipeline *current, *e;
unsigned int i;
for (i = 0; i < graphics->stage_count; ++i)
{
VK_CALL(vkDestroyShaderModule(device->vk_device, graphics->stages[i].module, NULL));
}
d3d12_pipeline_state_destroy_shader_modules(state, device);
LIST_FOR_EACH_ENTRY_SAFE(current, e, &graphics->compiled_fallback_pipelines, struct vkd3d_compiled_pipeline, entry)
{
@ -2097,6 +2162,7 @@ static ULONG STDMETHODCALLTYPE d3d12_pipeline_state_Release(ID3D12PipelineState
vkd3d_private_store_destroy(&state->private_store);
d3d12_pipeline_state_free_spirv_code(state);
if (d3d12_pipeline_state_is_graphics(state))
d3d12_pipeline_state_destroy_graphics(state, device);
else if (d3d12_pipeline_state_is_compute(state))
@ -2169,13 +2235,13 @@ static HRESULT STDMETHODCALLTYPE d3d12_pipeline_state_GetCachedBlob(ID3D12Pipeli
TRACE("iface %p, blob %p.\n", iface, blob);
if ((vr = vkd3d_serialize_pipeline_state(state, &cache_size, NULL)))
if ((vr = vkd3d_serialize_pipeline_state(NULL, state, &cache_size, NULL)))
return hresult_from_vk_result(vr);
if (!(cache_data = malloc(cache_size)))
return E_OUTOFMEMORY;
if ((vr = vkd3d_serialize_pipeline_state(state, &cache_size, cache_data)))
if ((vr = vkd3d_serialize_pipeline_state(NULL, state, &cache_size, cache_data)))
{
vkd3d_free(cache_data);
return hresult_from_vk_result(vr);
@ -2212,15 +2278,14 @@ CONST_VTBL struct ID3D12PipelineStateVtbl d3d12_pipeline_state_vtbl =
static HRESULT create_shader_stage(struct d3d12_device *device,
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)
const D3D12_SHADER_BYTECODE *code, const struct d3d12_cached_pipeline_state *cached_state,
const struct vkd3d_shader_interface_info *shader_interface,
const struct vkd3d_shader_compile_arguments *compile_args, struct vkd3d_shader_code *spirv_code)
{
struct vkd3d_shader_code dxbc = {code->pShaderBytecode, code->BytecodeLength};
const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs;
VkShaderModuleCreateInfo shader_desc;
struct vkd3d_shader_code spirv = {0};
char hash_str[16 + 1];
VkResult vr;
vkd3d_shader_hash_t recovered_hash = 0;
vkd3d_shader_hash_t compiled_hash = 0;
HRESULT hr;
int ret;
stage_desc->sType = VK_STRUCTURE_TYPE_PIPELINE_SHADER_STAGE_CREATE_INFO;
@ -2230,37 +2295,81 @@ static HRESULT create_shader_stage(struct d3d12_device *device,
stage_desc->pName = "main";
stage_desc->pSpecializationInfo = NULL;
shader_desc.sType = VK_STRUCTURE_TYPE_SHADER_MODULE_CREATE_INFO;
shader_desc.pNext = NULL;
shader_desc.flags = 0;
TRACE("Calling vkd3d_shader_compile_dxbc.\n");
if ((ret = vkd3d_shader_compile_dxbc(&dxbc, &spirv, 0, shader_interface, compile_args)) < 0)
if (!(vkd3d_config_flags & VKD3D_CONFIG_FLAG_PIPELINE_LIBRARY_IGNORE_SPIRV))
{
WARN("Failed to compile shader, vkd3d result %d.\n", ret);
return hresult_from_vkd3d_result(ret);
}
TRACE("Called vkd3d_shader_compile_dxbc.\n");
shader_desc.codeSize = spirv.size;
shader_desc.pCode = spirv.code;
*meta = spirv.meta;
hr = vkd3d_get_cached_spirv_code_from_d3d12_desc(cached_state, stage, spirv_code);
if (!d3d12_device_validate_shader_meta(device, &spirv.meta))
if (vkd3d_config_flags & VKD3D_CONFIG_FLAG_PIPELINE_LIBRARY_LOG)
{
if (SUCCEEDED(hr))
{
INFO("SPIR-V (stage: %x) for blob hash %016"PRIx64" received from cached pipeline state.\n",
stage, spirv_code->meta.hash);
}
else if (hr == E_FAIL)
{
if (cached_state->blob.CachedBlobSizeInBytes)
INFO("SPIR-V chunk was not found in cached PSO state.\n");
else
INFO("SPIR-V chunk was not found due to no Cached PSO state being provided.\n");
}
else if (hr == E_INVALIDARG)
INFO("Pipeline could not be created to mismatch in either root signature or DXBC blobs.\n");
else
INFO("Unexpected error when unserializing SPIR-V (hr %x).\n", hr);
}
/* For debug/dev purposes. */
if (vkd3d_config_flags & VKD3D_CONFIG_FLAG_PIPELINE_LIBRARY_SANITIZE_SPIRV)
{
if (SUCCEEDED(hr))
{
recovered_hash = vkd3d_shader_hash(spirv_code);
vkd3d_shader_free_shader_code(spirv_code);
}
hr = E_FAIL;
}
}
else
hr = E_FAIL;
if (FAILED(hr))
{
TRACE("Calling vkd3d_shader_compile_dxbc.\n");
if ((ret = vkd3d_shader_compile_dxbc(&dxbc, spirv_code, 0, shader_interface, compile_args)) < 0)
{
WARN("Failed to compile shader, vkd3d result %d.\n", ret);
return hresult_from_vkd3d_result(ret);
}
TRACE("Called vkd3d_shader_compile_dxbc.\n");
}
/* Debug compare SPIR-V we got from cache, and SPIR-V we got from compilation. */
if (recovered_hash)
{
compiled_hash = vkd3d_shader_hash(spirv_code);
if (compiled_hash == recovered_hash)
INFO("SPIR-V match for cache reference OK!\n");
else
INFO("SPIR-V mismatch for cache reference!\n");
}
if (!d3d12_device_validate_shader_meta(device, &spirv_code->meta))
return E_INVALIDARG;
if (((spirv.meta.flags & VKD3D_SHADER_META_FLAG_USES_SUBGROUP_SIZE) &&
if (((spirv_code->meta.flags & VKD3D_SHADER_META_FLAG_USES_SUBGROUP_SIZE) &&
device->device_info.subgroup_size_control_features.subgroupSizeControl) ||
spirv.meta.cs_required_wave_size)
spirv_code->meta.cs_required_wave_size)
{
uint32_t subgroup_size_alignment = device->device_info.subgroup_size_control_properties.maxSubgroupSize;
stage_desc->flags |= VK_PIPELINE_SHADER_STAGE_CREATE_ALLOW_VARYING_SUBGROUP_SIZE_BIT_EXT;
if (required_subgroup_size_info)
{
if (spirv.meta.cs_required_wave_size)
if (spirv_code->meta.cs_required_wave_size)
{
/* [WaveSize(N)] attribute in SM 6.6. */
subgroup_size_alignment = spirv.meta.cs_required_wave_size;
subgroup_size_alignment = spirv_code->meta.cs_required_wave_size;
stage_desc->pNext = required_subgroup_size_info;
stage_desc->flags &= ~VK_PIPELINE_SHADER_STAGE_CREATE_ALLOW_VARYING_SUBGROUP_SIZE_BIT_EXT;
}
@ -2285,25 +2394,14 @@ static HRESULT create_shader_stage(struct d3d12_device *device,
* 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))
!(spirv_code->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);
if (vr < 0)
{
WARN("Failed to create Vulkan shader module, vr %d.\n", vr);
return hresult_from_vk_result(vr);
}
/* Helpful for tooling like RenderDoc. */
sprintf(hash_str, "%016"PRIx64, spirv.meta.hash);
vkd3d_set_vk_object_name(device, (uint64_t)stage_desc->module, VK_OBJECT_TYPE_SHADER_MODULE, hash_str);
return S_OK;
stage_desc->module = VK_NULL_HANDLE;
return d3d12_pipeline_state_create_shader_module(device, stage_desc, spirv_code);
}
static void vkd3d_report_pipeline_creation_feedback_results(const VkPipelineCreationFeedbackCreateInfoEXT *feedback)
@ -2348,9 +2446,10 @@ static void vkd3d_report_pipeline_creation_feedback_results(const VkPipelineCrea
}
static HRESULT vkd3d_create_compute_pipeline(struct d3d12_device *device,
const D3D12_SHADER_BYTECODE *code, const struct vkd3d_shader_interface_info *shader_interface,
const D3D12_SHADER_BYTECODE *code, const struct d3d12_cached_pipeline_state *cached_state,
const struct vkd3d_shader_interface_info *shader_interface,
VkPipelineLayout vk_pipeline_layout, VkPipelineCache vk_cache, VkPipeline *vk_pipeline,
struct vkd3d_shader_meta *meta)
struct vkd3d_shader_code *spirv_code)
{
VkPipelineShaderStageRequiredSubgroupSizeCreateInfoEXT required_subgroup_size_info;
const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs;
@ -2372,17 +2471,18 @@ static HRESULT vkd3d_create_compute_pipeline(struct d3d12_device *device,
pipeline_info.sType = VK_STRUCTURE_TYPE_COMPUTE_PIPELINE_CREATE_INFO;
pipeline_info.pNext = NULL;
pipeline_info.flags = 0;
if (FAILED(hr = create_shader_stage(device, &pipeline_info.stage,
if (FAILED(hr = create_shader_stage(device,
&pipeline_info.stage,
VK_SHADER_STAGE_COMPUTE_BIT, &required_subgroup_size_info,
code, shader_interface, &compile_args, meta)))
code, cached_state, shader_interface, &compile_args, spirv_code)))
return hr;
pipeline_info.layout = vk_pipeline_layout;
pipeline_info.basePipelineHandle = VK_NULL_HANDLE;
pipeline_info.basePipelineIndex = -1;
if ((meta->flags & VKD3D_SHADER_META_FLAG_REPLACED) && device->debug_ring.active)
if ((spirv_code->meta.flags & VKD3D_SHADER_META_FLAG_REPLACED) && device->debug_ring.active)
{
vkd3d_shader_debug_ring_init_spec_constant(device, &spec_info, meta->hash);
vkd3d_shader_debug_ring_init_spec_constant(device, &spec_info, spirv_code->meta.hash);
pipeline_info.stage.pSpecializationInfo = &spec_info.spec_info;
}
@ -2426,6 +2526,7 @@ static HRESULT d3d12_pipeline_state_init_compute(struct d3d12_pipeline_state *st
state->ID3D12PipelineState_iface.lpVtbl = &d3d12_pipeline_state_vtbl;
state->refcount = 1;
state->vk_bind_point = VK_PIPELINE_BIND_POINT_COMPUTE;
if (desc->root_signature)
root_signature = impl_from_ID3D12RootSignature(desc->root_signature);
@ -2458,11 +2559,12 @@ static HRESULT d3d12_pipeline_state_init_compute(struct d3d12_pipeline_state *st
}
}
hr = vkd3d_create_compute_pipeline(device, &desc->cs, &shader_interface,
hr = vkd3d_create_compute_pipeline(device,
&desc->cs, &desc->cached_pso, &shader_interface,
root_signature->compute.vk_pipeline_layout,
state->vk_pso_cache ? state->vk_pso_cache : device->global_pipeline_cache,
&state->compute.vk_pipeline,
&state->compute.meta);
&state->compute.code);
if (FAILED(hr))
{
@ -2476,7 +2578,6 @@ static HRESULT d3d12_pipeline_state_init_compute(struct d3d12_pipeline_state *st
return hr;
}
state->vk_bind_point = VK_PIPELINE_BIND_POINT_COMPUTE;
d3d12_device_add_ref(state->device = device);
return S_OK;
@ -3205,7 +3306,6 @@ static HRESULT d3d12_pipeline_state_init_graphics(struct d3d12_pipeline_state *s
unsigned int ps_output_swizzle[D3D12_SIMULTANEOUS_RENDER_TARGET_COUNT];
struct vkd3d_shader_compile_arguments compile_args, ps_compile_args;
struct d3d12_graphics_pipeline_state *graphics = &state->graphics;
const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs;
const D3D12_STREAM_OUTPUT_DESC *so_desc = &desc->stream_output;
VkVertexInputBindingDivisorDescriptionEXT *binding_divisor;
const struct vkd3d_vulkan_info *vk_info = &device->vk_info;
@ -3244,6 +3344,7 @@ static HRESULT d3d12_pipeline_state_init_graphics(struct d3d12_pipeline_state *s
state->ID3D12PipelineState_iface.lpVtbl = &d3d12_pipeline_state_vtbl;
state->refcount = 1;
state->vk_bind_point = VK_PIPELINE_BIND_POINT_GRAPHICS;
graphics->stage_count = 0;
graphics->primitive_topology_type = desc->primitive_topology_type;
@ -3539,21 +3640,22 @@ 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, NULL, b, &shader_interface,
if (FAILED(hr = create_shader_stage(device,
&graphics->stages[graphics->stage_count],
shader_stages[i].stage, NULL, b, &desc->cached_pso, &shader_interface,
shader_stages[i].stage == VK_SHADER_STAGE_FRAGMENT_BIT ? &ps_compile_args : &compile_args,
&graphics->stage_meta[graphics->stage_count])))
&graphics->code[graphics->stage_count])))
goto fail;
if (shader_stages[i].stage == VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT)
graphics->patch_vertex_count = graphics->stage_meta[graphics->stage_count].patch_vertex_count;
graphics->patch_vertex_count = graphics->code[graphics->stage_count].meta.patch_vertex_count;
if ((graphics->stage_meta[graphics->stage_count].flags & VKD3D_SHADER_META_FLAG_REPLACED) &&
if ((graphics->code[graphics->stage_count].meta.flags & VKD3D_SHADER_META_FLAG_REPLACED) &&
device->debug_ring.active)
{
vkd3d_shader_debug_ring_init_spec_constant(device,
&graphics->spec_info[graphics->stage_count],
graphics->stage_meta[graphics->stage_count].hash);
graphics->code[graphics->stage_count].meta.hash);
graphics->stages[graphics->stage_count].pSpecializationInfo = &graphics->spec_info[graphics->stage_count].spec_info;
}
@ -3810,16 +3912,11 @@ static HRESULT d3d12_pipeline_state_init_graphics(struct d3d12_pipeline_state *s
if (FAILED(hr = vkd3d_private_store_init(&state->private_store)))
goto fail;
state->vk_bind_point = VK_PIPELINE_BIND_POINT_GRAPHICS;
d3d12_device_add_ref(state->device);
return S_OK;
fail:
for (i = 0; i < graphics->stage_count; ++i)
{
VK_CALL(vkDestroyShaderModule(device->vk_device, state->graphics.stages[i].module, NULL));
}
vkd3d_shader_free_shader_signature(&input_signature);
vkd3d_shader_free_shader_signature(&output_signature);
@ -3830,11 +3927,11 @@ bool d3d12_pipeline_state_has_replaced_shaders(struct d3d12_pipeline_state *stat
{
unsigned int i;
if (state->vk_bind_point == VK_PIPELINE_BIND_POINT_COMPUTE)
return !!(state->compute.meta.flags & VKD3D_SHADER_META_FLAG_REPLACED);
return !!(state->compute.code.meta.flags & VKD3D_SHADER_META_FLAG_REPLACED);
else if (state->vk_bind_point == VK_PIPELINE_BIND_POINT_GRAPHICS)
{
for (i = 0; i < state->graphics.stage_count; i++)
if (state->graphics.stage_meta[i].flags & VKD3D_SHADER_META_FLAG_REPLACED)
if (state->graphics.code[i].meta.flags & VKD3D_SHADER_META_FLAG_REPLACED)
return true;
return false;
}
@ -3882,8 +3979,21 @@ HRESULT d3d12_pipeline_state_create(struct d3d12_device *device, VkPipelineBindP
else
root_signature = impl_from_ID3D12RootSignature(desc->root_signature);
vkd3d_pipeline_cache_compat_from_state_desc(&object->pipeline_cache_compat, desc);
if (root_signature)
object->root_signature_compat_hash = root_signature->compatibility_hash;
object->pipeline_cache_compat.root_signature_compat_hash = root_signature->compatibility_hash;
if (desc->cached_pso.blob.CachedBlobSizeInBytes)
{
if (FAILED(hr = d3d12_cached_pipeline_state_validate(device, &desc->cached_pso,
&object->pipeline_cache_compat)))
{
if (object->private_root_signature)
ID3D12RootSignature_Release(object->private_root_signature);
vkd3d_free(object);
return hr;
}
}
switch (bind_point)
{
@ -3904,12 +4014,37 @@ HRESULT d3d12_pipeline_state_create(struct d3d12_device *device, VkPipelineBindP
{
if (object->private_root_signature)
ID3D12RootSignature_Release(object->private_root_signature);
d3d12_pipeline_state_free_spirv_code(object);
d3d12_pipeline_state_destroy_shader_modules(object, device);
VK_CALL(vkDestroyPipelineCache(device->vk_device, object->vk_pso_cache, NULL));
vkd3d_free(object);
return hr;
}
/* The strategy here is that we need to keep the SPIR-V alive somehow.
* If we don't need to serialize SPIR-V from the PSO, then we don't need to keep the code alive as pointer/size pairs.
* The scenarios for this case is when we choose to not serialize SPIR-V at all with VKD3D_CONFIG,
* or the PSO was loaded from a cached blob. It's extremely unlikely that anyone is going to try
* serializing that PSO again, so there should be no need to keep it alive.
* The worst that would happen is a performance loss should that entry be reloaded later.
* For graphics pipelines, we have to keep VkShaderModules around in case we need fallback pipelines.
* If we keep the SPIR-V around in memory, we can always create shader modules on-demand in case we
* need to actually create fallback pipelines. This avoids unnecessary memory bloat. */
if (desc->cached_pso.blob.CachedBlobSizeInBytes ||
(vkd3d_config_flags & VKD3D_CONFIG_FLAG_PIPELINE_LIBRARY_NO_SERIALIZE_SPIRV))
d3d12_pipeline_state_free_spirv_code(object);
else
d3d12_pipeline_state_destroy_shader_modules(object, device);
/* We don't expect to serialize the PSO blob if we loaded it from cache.
* Free the cache now to save on memory. */
if (desc->cached_pso.blob.CachedBlobSizeInBytes)
{
VK_CALL(vkDestroyPipelineCache(device->vk_device, object->vk_pso_cache, NULL));
object->vk_pso_cache = VK_NULL_HANDLE;
}
TRACE("Created pipeline state %p.\n", object);
*state = object;
@ -4087,6 +4222,7 @@ VkPipeline d3d12_pipeline_state_create_pipeline_variant(struct d3d12_pipeline_st
struct d3d12_graphics_pipeline_state *graphics = &state->graphics;
VkPipelineVertexInputDivisorStateCreateInfoEXT input_divisor_info;
VkPipelineCreationFeedbackEXT feedbacks[VKD3D_MAX_SHADER_STAGES];
VkPipelineShaderStageCreateInfo stages[VKD3D_MAX_SHADER_STAGES];
VkPipelineTessellationStateCreateInfo tessellation_info;
VkPipelineDepthStencilStateCreateInfo fallback_ds_desc;
VkPipelineCreationFeedbackCreateInfoEXT feedback_info;
@ -4209,6 +4345,28 @@ VkPipeline d3d12_pipeline_state_create_pipeline_variant(struct d3d12_pipeline_st
/* Any of these is fine from a compatibility PoV. */
pipeline_desc.renderPass = render_pass_compat->dsv_layouts[0];
if (key)
{
/* In a fallback pipeline, we might have to re-create shader modules.
* This can happen from multiple threads, so need temporary pStages array. */
memcpy(stages, graphics->stages, graphics->stage_count * sizeof(stages[0]));
for (i = 0; i < graphics->stage_count; i++)
{
if (stages[i].module == VK_NULL_HANDLE && graphics->code[i].code)
{
if (FAILED(hr = d3d12_pipeline_state_create_shader_module(device, &stages[i], &graphics->code[i])))
{
/* This is kind of fatal and should only happen for out-of-memory. */
ERR("Unexpected failure (hr %x) in creating fallback SPIR-V module.\n", hr);
return VK_NULL_HANDLE;
}
pipeline_desc.pStages = stages;
}
}
}
TRACE("Calling vkCreateGraphicsPipelines.\n");
if ((vkd3d_config_flags & VKD3D_CONFIG_FLAG_PIPELINE_LIBRARY_LOG) &&
@ -4230,6 +4388,12 @@ VkPipeline d3d12_pipeline_state_create_pipeline_variant(struct d3d12_pipeline_st
}
TRACE("Completed vkCreateGraphicsPipelines.\n");
/* Clean up any temporary SPIR-V modules we created. */
if (pipeline_desc.pStages == stages)
for (i = 0; i < graphics->stage_count; i++)
if (stages[i].module != graphics->stages[i].module)
VK_CALL(vkDestroyShaderModule(device->vk_device, stages[i].module, NULL));
if (pipeline_desc.pNext)
vkd3d_report_pipeline_creation_feedback_results(&feedback_info);

34
libs/vkd3d/vkd3d_private.h
View File

@ -1453,7 +1453,7 @@ struct d3d12_graphics_pipeline_state
{
struct vkd3d_shader_debug_ring_spec_info spec_info[VKD3D_MAX_SHADER_STAGES];
VkPipelineShaderStageCreateInfo stages[VKD3D_MAX_SHADER_STAGES];
struct vkd3d_shader_meta stage_meta[VKD3D_MAX_SHADER_STAGES];
struct vkd3d_shader_code code[VKD3D_MAX_SHADER_STAGES];
size_t stage_count;
VkVertexInputAttributeDescription attributes[D3D12_VS_INPUT_REGISTER_COUNT];
@ -1506,7 +1506,16 @@ static inline unsigned int dsv_attachment_mask(const struct d3d12_graphics_pipel
struct d3d12_compute_pipeline_state
{
VkPipeline vk_pipeline;
struct vkd3d_shader_meta meta;
struct vkd3d_shader_code code;
};
/* To be able to load a pipeline from cache, this information must match exactly,
* otherwise, we must regard the PSO as incompatible (which is invalid usage and must be validated). */
struct vkd3d_pipeline_cache_compatibility
{
uint64_t state_desc_compat_hash;
uint64_t root_signature_compat_hash;
uint64_t dxbc_blob_hashes[VKD3D_MAX_SHADER_STAGES];
};
/* ID3D12PipelineState */
@ -1524,7 +1533,7 @@ struct d3d12_pipeline_state
VkPipelineCache vk_pso_cache;
spinlock_t lock;
vkd3d_shader_hash_t root_signature_compat_hash;
struct vkd3d_pipeline_cache_compatibility pipeline_cache_compat;
ID3D12RootSignature *private_root_signature;
struct d3d12_device *device;
@ -1640,7 +1649,12 @@ struct d3d12_pipeline_library
struct d3d12_device *device;
rwlock_t mutex;
struct hash_map map;
struct hash_map pso_map;
struct hash_map driver_cache_map;
struct hash_map spirv_cache_map;
size_t total_name_table_size;
size_t total_blob_size;
struct vkd3d_private_store private_store;
};
@ -1652,7 +1666,17 @@ VkResult vkd3d_create_pipeline_cache(struct d3d12_device *device,
size_t size, const void *data, VkPipelineCache *cache);
HRESULT vkd3d_create_pipeline_cache_from_d3d12_desc(struct d3d12_device *device,
const struct d3d12_cached_pipeline_state *state, VkPipelineCache *cache);
VkResult vkd3d_serialize_pipeline_state(const struct d3d12_pipeline_state *state, size_t *size, void *data);
HRESULT vkd3d_get_cached_spirv_code_from_d3d12_desc(
const struct d3d12_cached_pipeline_state *state,
VkShaderStageFlagBits stage,
struct vkd3d_shader_code *spirv_code);
VkResult vkd3d_serialize_pipeline_state(struct d3d12_pipeline_library *pipeline_library,
const struct d3d12_pipeline_state *state, size_t *size, void *data);
HRESULT d3d12_cached_pipeline_state_validate(struct d3d12_device *device,
const struct d3d12_cached_pipeline_state *state,
const struct vkd3d_pipeline_cache_compatibility *compat);
void vkd3d_pipeline_cache_compat_from_state_desc(struct vkd3d_pipeline_cache_compatibility *compat,
const struct d3d12_pipeline_state_desc *desc);
struct vkd3d_buffer
{