April 2019 - Wine-Devel - list.winehq.org

[PATCH vkd3d 4/9] vkd3d: Share Vulkan render pass objects between D3D12 pipeline states.
by Józef Kucia April 29, 2019

April 29, 2019

From: Józef Kucia <jkucia(a)codeweavers.com> For example, World of Warcraft creates 28 render passes for pipeline state objects instead of 600+ render passes. Other games appears to use similarly low number of render passes, e.g. 8, 17, 26, 60. Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- libs/vkd3d/device.c | 6 +- libs/vkd3d/state.c | 260 +++++++++++++++++++++++++++++++------ libs/vkd3d/vkd3d_private.h | 30 ++++- 3 files changed, 254 insertions(+), 42 deletions(-) diff --git a/libs/vkd3d/device.c b/libs/vkd3d/device.c index 5c0c880d9d25..5364df28bbc9 100644 --- a/libs/vkd3d/device.c +++ b/libs/vkd3d/device.c @@ -1665,7 +1665,7 @@ static HRESULT d3d12_device_init_pipeline_cache(struct d3d12_device *device) VkResult vr; int rc; - if ((rc = pthread_mutex_init(&device->pipeline_cache_mutex, NULL))) + if ((rc = pthread_mutex_init(&device->mutex, NULL))) { ERR("Failed to initialize mutex, error %d.\n", rc); return hresult_from_errno(rc); @@ -1693,7 +1693,7 @@ static void d3d12_device_destroy_pipeline_cache(struct d3d12_device *device) if (device->vk_pipeline_cache) VK_CALL(vkDestroyPipelineCache(device->vk_device, device->vk_pipeline_cache, NULL)); - pthread_mutex_destroy(&device->pipeline_cache_mutex); + pthread_mutex_destroy(&device->mutex); } D3D12_GPU_VIRTUAL_ADDRESS vkd3d_gpu_va_allocator_allocate(struct vkd3d_gpu_va_allocator *allocator, @@ -1866,6 +1866,7 @@ static ULONG STDMETHODCALLTYPE d3d12_device_Release(ID3D12Device *iface) vkd3d_destroy_null_resources(&device->null_resources, device); vkd3d_gpu_va_allocator_cleanup(&device->gpu_va_allocator); + vkd3d_render_pass_cache_cleanup(&device->render_pass_cache, device); vkd3d_fence_worker_stop(&device->fence_worker, device); d3d12_device_destroy_pipeline_cache(device); d3d12_device_destroy_vkd3d_queues(device); @@ -3030,6 +3031,7 @@ static HRESULT d3d12_device_init(struct d3d12_device *device, if (FAILED(hr = vkd3d_init_null_resources(&device->null_resources, device))) goto out_stop_fence_worker; + vkd3d_render_pass_cache_init(&device->render_pass_cache); vkd3d_gpu_va_allocator_init(&device->gpu_va_allocator); if ((device->parent = create_info->parent)) diff --git a/libs/vkd3d/state.c b/libs/vkd3d/state.c index edf3399b1aa8..b23101e868a6 100644 --- a/libs/vkd3d/state.c +++ b/libs/vkd3d/state.c @@ -1033,6 +1033,196 @@ HRESULT d3d12_root_signature_create(struct d3d12_device *device, return S_OK; } +/* vkd3d_render_pass_cache */ +struct vkd3d_render_pass_entry +{ + struct vkd3d_render_pass_key key; + VkRenderPass vk_render_pass; +}; + +STATIC_ASSERT(sizeof(struct vkd3d_render_pass_key) == 48); + +static HRESULT vkd3d_render_pass_cache_create_pass_locked(struct vkd3d_render_pass_cache *cache, + struct d3d12_device *device, const struct vkd3d_render_pass_key *key, VkRenderPass *vk_render_pass) +{ + VkAttachmentReference attachment_references[D3D12_SIMULTANEOUS_RENDER_TARGET_COUNT + 1]; + VkAttachmentDescription attachments[D3D12_SIMULTANEOUS_RENDER_TARGET_COUNT + 1]; + const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs; + struct vkd3d_render_pass_entry *entry; + unsigned int color_attachment_index; + VkSubpassDescription sub_pass_desc; + VkRenderPassCreateInfo pass_info; + bool have_depth_stencil; + unsigned int index; + VkResult vr; + + if (!vkd3d_array_reserve((void **)&cache->render_passes, &cache->render_passes_size, + cache->render_pass_count + 1, sizeof(*cache->render_passes))) + { + *vk_render_pass = VK_NULL_HANDLE; + return E_OUTOFMEMORY; + } + + entry = &cache->render_passes[cache->render_pass_count]; + + entry->key = *key; + + have_depth_stencil = key->depth_enable || key->stencil_enable; + color_attachment_index = have_depth_stencil ? 1 : 0; + + index = 0; + if (have_depth_stencil) + { + VkImageLayout depth_layout = key->depth_stencil_write + ? VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL + : VK_IMAGE_LAYOUT_DEPTH_STENCIL_READ_ONLY_OPTIMAL; + + attachments[index].flags = 0; + attachments[index].format = key->vk_formats[index]; + attachments[index].samples = key->sample_count; + + if (key->depth_enable) + { + attachments[index].loadOp = VK_ATTACHMENT_LOAD_OP_LOAD; + attachments[index].storeOp = VK_ATTACHMENT_STORE_OP_STORE; + } + else + { + attachments[index].loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; + attachments[index].storeOp = VK_ATTACHMENT_STORE_OP_DONT_CARE; + } + if (key->stencil_enable) + { + attachments[index].stencilLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD; + attachments[index].stencilStoreOp = VK_ATTACHMENT_STORE_OP_STORE; + } + else + { + attachments[index].stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; + attachments[index].stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE; + } + attachments[index].initialLayout = depth_layout; + attachments[index].finalLayout = depth_layout; + + attachment_references[index].attachment = 0; + attachment_references[index].layout = depth_layout; + + ++index; + } + + assert(index == color_attachment_index); + for (; index < key->attachment_count; ++index) + { + + attachments[index].flags = 0; + attachments[index].format = key->vk_formats[index]; + attachments[index].samples = key->sample_count; + attachments[index].loadOp = VK_ATTACHMENT_LOAD_OP_LOAD; + attachments[index].storeOp = VK_ATTACHMENT_STORE_OP_STORE; + attachments[index].stencilLoadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; + attachments[index].stencilStoreOp = VK_ATTACHMENT_STORE_OP_DONT_CARE; + attachments[index].initialLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; + attachments[index].finalLayout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; + + attachment_references[index].attachment = index; + attachment_references[index].layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; + } + + sub_pass_desc.flags = 0; + sub_pass_desc.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS; + sub_pass_desc.inputAttachmentCount = 0; + sub_pass_desc.pInputAttachments = NULL; + sub_pass_desc.colorAttachmentCount = key->attachment_count - color_attachment_index; + sub_pass_desc.pColorAttachments = &attachment_references[color_attachment_index]; + sub_pass_desc.pResolveAttachments = NULL; + if (have_depth_stencil) + sub_pass_desc.pDepthStencilAttachment = &attachment_references[0]; + else + sub_pass_desc.pDepthStencilAttachment = NULL; + sub_pass_desc.preserveAttachmentCount = 0; + sub_pass_desc.pPreserveAttachments = NULL; + + pass_info.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO; + pass_info.pNext = NULL; + pass_info.flags = 0; + pass_info.attachmentCount = key->attachment_count; + pass_info.pAttachments = attachments; + pass_info.subpassCount = 1; + pass_info.pSubpasses = &sub_pass_desc; + pass_info.dependencyCount = 0; + pass_info.pDependencies = NULL; + if ((vr = VK_CALL(vkCreateRenderPass(device->vk_device, &pass_info, NULL, vk_render_pass))) >= 0) + { + entry->vk_render_pass = *vk_render_pass; + ++cache->render_pass_count; + } + else + { + WARN("Failed to create Vulkan render pass, vr %d.\n", vr); + *vk_render_pass = VK_NULL_HANDLE; + } + + return hresult_from_vk_result(vr); +} + +HRESULT vkd3d_render_pass_cache_find(struct vkd3d_render_pass_cache *cache, + struct d3d12_device *device, const struct vkd3d_render_pass_key *key, VkRenderPass *vk_render_pass) +{ + bool found = false; + HRESULT hr = S_OK; + unsigned int i; + int rc; + + if ((rc = pthread_mutex_lock(&device->mutex))) + { + ERR("Failed to lock mutex, error %d.\n", rc); + *vk_render_pass = VK_NULL_HANDLE; + return hresult_from_errno(rc); + } + + for (i = 0; i < cache->render_pass_count; ++i) + { + struct vkd3d_render_pass_entry *current = &cache->render_passes[i]; + + if (!memcmp(&current->key, key, sizeof(*key))) + { + *vk_render_pass = current->vk_render_pass; + found = true; + break; + } + } + + if (!found) + hr = vkd3d_render_pass_cache_create_pass_locked(cache, device, key, vk_render_pass); + + pthread_mutex_unlock(&device->mutex); + + return hr; +} + +void vkd3d_render_pass_cache_init(struct vkd3d_render_pass_cache *cache) +{ + cache->render_passes = NULL; + cache->render_pass_count = 0; + cache->render_passes_size = 0; +} + +void vkd3d_render_pass_cache_cleanup(struct vkd3d_render_pass_cache *cache, + struct d3d12_device *device) +{ + const struct vkd3d_vk_device_procs *vk_procs = &device->vk_procs; + unsigned int i; + + for (i = 0; i < cache->render_pass_count; ++i) + { + struct vkd3d_render_pass_entry *current = &cache->render_passes[i]; + VK_CALL(vkDestroyRenderPass(device->vk_device, current->vk_render_pass, NULL)); + } + + vkd3d_free(cache->render_passes); + cache->render_passes = NULL; +} + struct vkd3d_pipeline_key { D3D12_PRIMITIVE_TOPOLOGY topology; @@ -1096,7 +1286,6 @@ static void d3d12_pipeline_state_destroy_graphics(struct d3d12_pipeline_state *s { VK_CALL(vkDestroyShaderModule(device->vk_device, graphics->stages[i].module, NULL)); } - VK_CALL(vkDestroyRenderPass(device->vk_device, graphics->render_pass, NULL)); LIST_FOR_EACH_ENTRY_SAFE(current, e, &graphics->compiled_pipelines, struct vkd3d_compiled_pipeline, entry) { @@ -1883,17 +2072,15 @@ static HRESULT d3d12_pipeline_state_init_graphics(struct d3d12_pipeline_state *s struct vkd3d_shader_interface_info shader_interface; const struct d3d12_root_signature *root_signature; struct vkd3d_shader_signature input_signature; + struct vkd3d_render_pass_key render_pass_key; VkShaderStageFlagBits xfb_stage = 0; VkSampleCountFlagBits sample_count; - VkSubpassDescription sub_pass_desc; const struct vkd3d_format *format; - VkRenderPassCreateInfo pass_desc; unsigned int instance_divisor; VkVertexInputRate input_rate; unsigned int i, j; size_t rt_count; uint32_t mask; - VkResult vr; HRESULT hr; int ret; @@ -1983,13 +2170,20 @@ static HRESULT d3d12_pipeline_state_init_graphics(struct d3d12_pipeline_state *s if ((ds_desc->DepthEnable && ds_desc->DepthWriteMask) || (ds_desc->StencilEnable && ds_desc->StencilWriteMask)) + { depth_layout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL; + render_pass_key.depth_stencil_write = true; + } else + { depth_layout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_READ_ONLY_OPTIMAL; + render_pass_key.depth_stencil_write = false; + } graphics->attachments[0].flags = 0; graphics->attachments[0].format = format->vk_format; graphics->attachments[0].samples = sample_count; + render_pass_key.depth_enable = desc->DepthStencilState.DepthEnable; if (desc->DepthStencilState.DepthEnable) { graphics->attachments[0].loadOp = VK_ATTACHMENT_LOAD_OP_LOAD; @@ -2000,6 +2194,7 @@ static HRESULT d3d12_pipeline_state_init_graphics(struct d3d12_pipeline_state *s graphics->attachments[0].loadOp = VK_ATTACHMENT_LOAD_OP_DONT_CARE; graphics->attachments[0].storeOp = VK_ATTACHMENT_STORE_OP_DONT_CARE; } + render_pass_key.stencil_enable = desc->DepthStencilState.StencilEnable; if (desc->DepthStencilState.StencilEnable) { graphics->attachments[0].stencilLoadOp = VK_ATTACHMENT_LOAD_OP_LOAD; @@ -2017,6 +2212,8 @@ static HRESULT d3d12_pipeline_state_init_graphics(struct d3d12_pipeline_state *s graphics->attachment_references[0].layout = depth_layout; ++graphics->rt_idx; + render_pass_key.vk_formats[0] = format->vk_format; + if (!desc->PS.pShaderBytecode) { if (FAILED(hr = create_shader_stage(device, &graphics->stages[graphics->stage_count], @@ -2026,6 +2223,12 @@ static HRESULT d3d12_pipeline_state_init_graphics(struct d3d12_pipeline_state *s ++graphics->stage_count; } } + else + { + render_pass_key.depth_enable = false; + render_pass_key.stencil_enable = false; + render_pass_key.depth_stencil_write = false; + } for (i = 0; i < rt_count; ++i) { @@ -2068,10 +2271,18 @@ static HRESULT d3d12_pipeline_state_init_graphics(struct d3d12_pipeline_state *s graphics->attachment_references[idx].attachment = idx; graphics->attachment_references[idx].layout = VK_IMAGE_LAYOUT_COLOR_ATTACHMENT_OPTIMAL; + render_pass_key.vk_formats[idx] = format->vk_format; + blend_attachment_from_d3d12(&graphics->blend_attachments[i], rt_desc); } graphics->attachment_count = graphics->rt_idx + rt_count; + render_pass_key.attachment_count = graphics->rt_idx + rt_count; + render_pass_key.padding = 0; + render_pass_key.sample_count = sample_count; + for (i = render_pass_key.attachment_count; i < ARRAY_SIZE(render_pass_key.vk_formats); ++i) + render_pass_key.vk_formats[i] = VK_FORMAT_UNDEFINED; + ps_shader_parameters[0].name = VKD3D_SHADER_PARAMETER_NAME_RASTERIZER_SAMPLE_COUNT; ps_shader_parameters[0].type = VKD3D_SHADER_PARAMETER_TYPE_IMMEDIATE_CONSTANT; ps_shader_parameters[0].data_type = VKD3D_SHADER_PARAMETER_DATA_TYPE_UINT32; @@ -2329,35 +2540,9 @@ static HRESULT d3d12_pipeline_state_init_graphics(struct d3d12_pipeline_state *s goto fail; } - sub_pass_desc.flags = 0; - sub_pass_desc.pipelineBindPoint = VK_PIPELINE_BIND_POINT_GRAPHICS; - sub_pass_desc.inputAttachmentCount = 0; - sub_pass_desc.pInputAttachments = NULL; - sub_pass_desc.colorAttachmentCount = rt_count; - sub_pass_desc.pColorAttachments = &graphics->attachment_references[graphics->rt_idx]; - sub_pass_desc.pResolveAttachments = NULL; - if (graphics->rt_idx) - sub_pass_desc.pDepthStencilAttachment = &graphics->attachment_references[0]; - else - sub_pass_desc.pDepthStencilAttachment = NULL; - sub_pass_desc.preserveAttachmentCount = 0; - sub_pass_desc.pPreserveAttachments = NULL; - - pass_desc.sType = VK_STRUCTURE_TYPE_RENDER_PASS_CREATE_INFO; - pass_desc.pNext = NULL; - pass_desc.flags = 0; - pass_desc.attachmentCount = graphics->attachment_count; - pass_desc.pAttachments = graphics->attachments; - pass_desc.subpassCount = 1; - pass_desc.pSubpasses = &sub_pass_desc; - pass_desc.dependencyCount = 0; - pass_desc.pDependencies = NULL; - if ((vr = VK_CALL(vkCreateRenderPass(device->vk_device, &pass_desc, NULL, &graphics->render_pass))) < 0) - { - WARN("Failed to create Vulkan render pass, vr %d.\n", vr); - hr = hresult_from_vk_result(vr); + if (FAILED(hr = vkd3d_render_pass_cache_find(&device->render_pass_cache, device, + &render_pass_key, &graphics->render_pass))) goto fail; - } rs_desc_from_d3d12(&graphics->rs_desc, &desc->RasterizerState); if ((!graphics->attachment_count && !(desc->PS.pShaderBytecode && desc->PS.BytecodeLength)) @@ -2392,10 +2577,7 @@ static HRESULT d3d12_pipeline_state_init_graphics(struct d3d12_pipeline_state *s list_init(&graphics->compiled_pipelines); if (FAILED(hr = vkd3d_private_store_init(&state->private_store))) - { - VK_CALL(vkDestroyRenderPass(device->vk_device, graphics->render_pass, NULL)); goto fail; - } state->vk_bind_point = VK_PIPELINE_BIND_POINT_GRAPHICS; state->device = device; @@ -2497,7 +2679,7 @@ static VkPipeline d3d12_pipeline_state_find_compiled_pipeline(const struct d3d12 struct vkd3d_compiled_pipeline *current; int rc; - if (!(rc = pthread_mutex_lock(&device->pipeline_cache_mutex))) + if (!(rc = pthread_mutex_lock(&device->mutex))) { LIST_FOR_EACH_ENTRY(current, &graphics->compiled_pipelines, struct vkd3d_compiled_pipeline, entry) { @@ -2507,7 +2689,7 @@ static VkPipeline d3d12_pipeline_state_find_compiled_pipeline(const struct d3d12 break; } } - pthread_mutex_unlock(&device->pipeline_cache_mutex); + pthread_mutex_unlock(&device->mutex); } else { @@ -2531,7 +2713,7 @@ static bool d3d12_pipeline_state_put_pipeline_to_cache(struct d3d12_pipeline_sta compiled_pipeline->key = *key; compiled_pipeline->vk_pipeline = vk_pipeline; - if ((rc = pthread_mutex_lock(&device->pipeline_cache_mutex))) + if ((rc = pthread_mutex_lock(&device->mutex))) { ERR("Failed to lock mutex, error %d.\n", rc); vkd3d_free(compiled_pipeline); @@ -2551,7 +2733,7 @@ static bool d3d12_pipeline_state_put_pipeline_to_cache(struct d3d12_pipeline_sta if (compiled_pipeline) list_add_tail(&graphics->compiled_pipelines, &compiled_pipeline->entry); - pthread_mutex_unlock(&device->pipeline_cache_mutex); + pthread_mutex_unlock(&device->mutex); return compiled_pipeline; } diff --git a/libs/vkd3d/vkd3d_private.h b/libs/vkd3d/vkd3d_private.h index ba586b44aab2..950be6e56981 100644 --- a/libs/vkd3d/vkd3d_private.h +++ b/libs/vkd3d/vkd3d_private.h @@ -184,6 +184,33 @@ void *vkd3d_gpu_va_allocator_dereference(struct vkd3d_gpu_va_allocator *allocato void vkd3d_gpu_va_allocator_free(struct vkd3d_gpu_va_allocator *allocator, D3D12_GPU_VIRTUAL_ADDRESS address) DECLSPEC_HIDDEN; +struct vkd3d_render_pass_key +{ + unsigned int attachment_count; + bool depth_enable; + bool stencil_enable; + bool depth_stencil_write; + bool padding; + unsigned int sample_count; + VkFormat vk_formats[D3D12_SIMULTANEOUS_RENDER_TARGET_COUNT + 1]; +}; + +struct vkd3d_render_pass_entry; + +struct vkd3d_render_pass_cache +{ + struct vkd3d_render_pass_entry *render_passes; + size_t render_pass_count; + size_t render_passes_size; +}; + +void vkd3d_render_pass_cache_cleanup(struct vkd3d_render_pass_cache *cache, + struct d3d12_device *device) DECLSPEC_HIDDEN; +HRESULT vkd3d_render_pass_cache_find(struct vkd3d_render_pass_cache *cache, + struct d3d12_device *device, const struct vkd3d_render_pass_key *key, + VkRenderPass *vk_render_pass) DECLSPEC_HIDDEN; +void vkd3d_render_pass_cache_init(struct vkd3d_render_pass_cache *cache) DECLSPEC_HIDDEN; + struct vkd3d_private_store { pthread_mutex_t mutex; @@ -925,7 +952,8 @@ struct d3d12_device struct vkd3d_gpu_va_allocator gpu_va_allocator; struct vkd3d_fence_worker fence_worker; - pthread_mutex_t pipeline_cache_mutex; + pthread_mutex_t mutex; + struct vkd3d_render_pass_cache render_pass_cache; VkPipelineCache vk_pipeline_cache; VkPhysicalDeviceMemoryProperties memory_properties; -- 2.21.0

2 1

[PATCH vkd3d 3/9] vkd3d-shader: Improve ID assignment for specialization constants.
by Józef Kucia April 29, 2019

April 29, 2019

From: Józef Kucia <jkucia(a)codeweavers.com> Allows using 0 as SpecId. Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- libs/vkd3d-shader/spirv.c | 24 ++++++++++-------------- 1 file changed, 10 insertions(+), 14 deletions(-) diff --git a/libs/vkd3d-shader/spirv.c b/libs/vkd3d-shader/spirv.c index 22e02f9ffa01..86d5be89b039 100644 --- a/libs/vkd3d-shader/spirv.c +++ b/libs/vkd3d-shader/spirv.c @@ -2565,20 +2565,20 @@ static const struct vkd3d_spec_constant_info *get_spec_constant_info(enum vkd3d_ static uint32_t vkd3d_dxbc_compiler_alloc_spec_constant_id(struct vkd3d_dxbc_compiler *compiler) { - if (!compiler->current_spec_constant_id && compiler->compile_args) + if (!compiler->current_spec_constant_id) { const struct vkd3d_shader_compile_arguments *compile_args = compiler->compile_args; - unsigned int i, id; + unsigned int i, id = 0; - for (i = 0, id = 0; i < compile_args->parameter_count; ++i) + for (i = 0; compiler->compile_args && i < compile_args->parameter_count; ++i) { const struct vkd3d_shader_parameter *current = &compile_args->parameters[i]; if (current->type == VKD3D_SHADER_PARAMETER_TYPE_SPECIALIZATION_CONSTANT) - id = max(current->u.specialization_constant.id, id); + id = max(current->u.specialization_constant.id + 1, id); } - compiler->current_spec_constant_id = id + 1; + compiler->current_spec_constant_id = id; } return compiler->current_spec_constant_id++; @@ -2623,9 +2623,6 @@ static uint32_t vkd3d_dxbc_compiler_get_spec_constant(struct vkd3d_dxbc_compiler return compiler->spec_constants[i].id; } - if (!spec_id) - spec_id = vkd3d_dxbc_compiler_alloc_spec_constant_id(compiler); - return vkd3d_dxbc_compiler_emit_spec_constant(compiler, name, spec_id); } @@ -2633,7 +2630,6 @@ static uint32_t vkd3d_dxbc_compiler_emit_uint_shader_parameter(struct vkd3d_dxbc enum vkd3d_shader_parameter_name name) { const struct vkd3d_shader_parameter *parameter; - uint32_t spec_constant_id = 0; if (!(parameter = vkd3d_dxbc_compiler_get_shader_parameter(compiler, name))) { @@ -2643,14 +2639,14 @@ static uint32_t vkd3d_dxbc_compiler_emit_uint_shader_parameter(struct vkd3d_dxbc if (parameter->type == VKD3D_SHADER_PARAMETER_TYPE_IMMEDIATE_CONSTANT) return vkd3d_dxbc_compiler_get_constant_uint(compiler, parameter->u.immediate_constant.u.u32); - if (parameter->type == VKD3D_SHADER_PARAMETER_TYPE_SPECIALIZATION_CONSTANT) - spec_constant_id = parameter->u.specialization_constant.id; - else - FIXME("Unhandled parameter type %#x.\n", parameter->type); + return vkd3d_dxbc_compiler_get_spec_constant(compiler, name, parameter->u.specialization_constant.id); + + FIXME("Unhandled parameter type %#x.\n", parameter->type); default_parameter: - return vkd3d_dxbc_compiler_get_spec_constant(compiler, name, spec_constant_id); + return vkd3d_dxbc_compiler_get_spec_constant(compiler, + name, vkd3d_dxbc_compiler_alloc_spec_constant_id(compiler)); } static uint32_t vkd3d_dxbc_compiler_emit_construct_vector(struct vkd3d_dxbc_compiler *compiler, -- 2.21.0

2 1

[PATCH vkd3d 2/9] tests: Add test for sample positions.
by Józef Kucia April 29, 2019

April 29, 2019

From: Józef Kucia <jkucia(a)codeweavers.com> Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- tests/d3d12.c | 130 ++++++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 130 insertions(+) diff --git a/tests/d3d12.c b/tests/d3d12.c index 5f2e6b1da365..dd13ddf999b9 100644 --- a/tests/d3d12.c +++ b/tests/d3d12.c @@ -27607,6 +27607,135 @@ static void test_shader_get_render_target_sample_count(void) destroy_test_context(&context); } +static void test_shader_sample_position(void) +{ + D3D12_TEXTURE_COPY_LOCATION src_location, dst_location; + ID3D12Resource *texture, *readback_texture; + ID3D12GraphicsCommandList *command_list; + D3D12_HEAP_PROPERTIES heap_properties; + D3D12_RESOURCE_DESC resource_desc; + struct test_context_desc desc; + struct resource_readback rb; + struct test_context context; + ID3D12DescriptorHeap *heap; + ID3D12CommandQueue *queue; + uint32_t sample_index; + unsigned int i; + D3D12_BOX box; + HRESULT hr; + + static const float white[] = {1.0f, 1.0f, 1.0f, 1.0f}; + static const DWORD ps_code[] = + { +#if 0 + uint index; + Texture2DMS<float4> t; + + float4 main() : SV_Target + { + return float4(t.GetSamplePosition(index), 0, 0); + } +#endif + 0x43425844, 0x89611945, 0x2b7e06f0, 0x953a72bb, 0x1590618f, 0x00000001, 0x000000f8, 0x00000003, + 0x0000002c, 0x0000003c, 0x00000070, 0x4e475349, 0x00000008, 0x00000000, 0x00000008, 0x4e47534f, + 0x0000002c, 0x00000001, 0x00000008, 0x00000020, 0x00000000, 0x00000000, 0x00000003, 0x00000000, + 0x0000000f, 0x545f5653, 0x65677261, 0xabab0074, 0x58454853, 0x00000080, 0x00000050, 0x00000020, + 0x0100086a, 0x04000059, 0x00208e46, 0x00000000, 0x00000001, 0x04002058, 0x00107000, 0x00000000, + 0x00005555, 0x03000065, 0x001020f2, 0x00000000, 0x0900006e, 0x00102032, 0x00000000, 0x00107046, + 0x00000000, 0x0020800a, 0x00000000, 0x00000000, 0x00000000, 0x08000036, 0x001020c2, 0x00000000, + 0x00004002, 0x00000000, 0x00000000, 0x00000000, 0x00000000, 0x0100003e, + }; + static const D3D12_SHADER_BYTECODE ps = {ps_code, sizeof(ps_code)}; + + memset(&desc, 0, sizeof(desc)); + desc.rt_format = DXGI_FORMAT_R32G32B32A32_FLOAT; + desc.rt_width = desc.rt_height = 1; + desc.no_root_signature = true; + if (!init_test_context(&context, &desc)) + return; + command_list = context.list; + queue = context.queue; + + context.root_signature = create_texture_root_signature(context.device, + D3D12_SHADER_VISIBILITY_PIXEL, 1, 0); + context.pipeline_state = create_pipeline_state(context.device, + context.root_signature, context.render_target_desc.Format, NULL, &ps, NULL); + + memset(&heap_properties, 0, sizeof(heap_properties)); + heap_properties.Type = D3D12_HEAP_TYPE_DEFAULT; + resource_desc = ID3D12Resource_GetDesc(context.render_target); + resource_desc.SampleDesc.Count = 4; + hr = ID3D12Device_CreateCommittedResource(context.device, + &heap_properties, D3D12_HEAP_FLAG_NONE, &resource_desc, + D3D12_RESOURCE_STATE_COMMON, NULL, &IID_ID3D12Resource, (void **)&texture); + ok(hr == S_OK, "Failed to create texture, hr %#x.\n", hr); + + heap = create_gpu_descriptor_heap(context.device, D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV, 16); + readback_texture = create_default_texture(context.device, 4, 1, DXGI_FORMAT_R32G32B32A32_FLOAT, + D3D12_RESOURCE_FLAG_NONE, D3D12_RESOURCE_STATE_COPY_DEST); + ID3D12Device_CreateShaderResourceView(context.device, texture, NULL, + ID3D12DescriptorHeap_GetCPUDescriptorHandleForHeapStart(heap)); + + transition_resource_state(command_list, + texture, D3D12_RESOURCE_STATE_COMMON, D3D12_RESOURCE_STATE_PIXEL_SHADER_RESOURCE); + + for (sample_index = 0; sample_index < resource_desc.SampleDesc.Count; ++sample_index) + { + ID3D12GraphicsCommandList_ClearRenderTargetView(command_list, context.rtv, white, 0, NULL); + ID3D12GraphicsCommandList_OMSetRenderTargets(command_list, 1, &context.rtv, FALSE, NULL); + ID3D12GraphicsCommandList_SetGraphicsRootSignature(command_list, context.root_signature); + ID3D12GraphicsCommandList_SetPipelineState(command_list, context.pipeline_state); + ID3D12GraphicsCommandList_IASetPrimitiveTopology(command_list, D3D_PRIMITIVE_TOPOLOGY_TRIANGLELIST); + ID3D12GraphicsCommandList_RSSetViewports(command_list, 1, &context.viewport); + ID3D12GraphicsCommandList_RSSetScissorRects(command_list, 1, &context.scissor_rect); + ID3D12GraphicsCommandList_SetDescriptorHeaps(command_list, 1, &heap); + ID3D12GraphicsCommandList_SetGraphicsRootDescriptorTable(command_list, 0, + get_gpu_descriptor_handle(&context, heap, 0)); + ID3D12GraphicsCommandList_SetGraphicsRoot32BitConstants(command_list, 1, 1, &sample_index, 0); + ID3D12GraphicsCommandList_DrawInstanced(command_list, 3, 1, 0, 0); + + transition_resource_state(command_list, context.render_target, + D3D12_RESOURCE_STATE_RENDER_TARGET, D3D12_RESOURCE_STATE_COPY_SOURCE); + src_location.pResource = context.render_target; + src_location.Type = D3D12_TEXTURE_COPY_TYPE_SUBRESOURCE_INDEX; + src_location.SubresourceIndex = 0; + dst_location.pResource = readback_texture; + dst_location.Type = D3D12_TEXTURE_COPY_TYPE_SUBRESOURCE_INDEX; + dst_location.SubresourceIndex = 0; + set_box(&box, 0, 0, 0, 1, 1, 1); + ID3D12GraphicsCommandList_CopyTextureRegion(command_list, &dst_location, sample_index, 0, 0, &src_location, &box); + transition_resource_state(command_list, context.render_target, + D3D12_RESOURCE_STATE_COPY_SOURCE, D3D12_RESOURCE_STATE_RENDER_TARGET); + } + + transition_resource_state(command_list, readback_texture, + D3D12_RESOURCE_STATE_COPY_DEST, D3D12_RESOURCE_STATE_COPY_SOURCE); + + get_texture_readback_with_command_list(readback_texture, 0, &rb, queue, command_list); + for (i = 0; i < resource_desc.SampleDesc.Count; ++i) + { + const struct vec4 *position = get_readback_vec4(&rb, i, 0); + + vkd3d_test_set_context("Sample %u", i); + + ok(-1.0f <= position->x && position->x <= 1.0f, "Unexpected x %.8e.\n", position->x); + ok(-1.0f <= position->y && position->y <= 1.0f, "Unexpected y %.8e.\n", position->y); + ok(!position->z, "Unexpected z %.8e.\n", position->z); + ok(!position->w, "Unexpected w %.8e.\n", position->w); + + if (vkd3d_test_state.debug_level > 0) + trace("Sample %u position {%.8e, %.8e}.\n", i, position->x, position->y); + + vkd3d_test_set_context(NULL); + } + release_resource_readback(&rb); + + ID3D12DescriptorHeap_Release(heap); + ID3D12Resource_Release(texture); + ID3D12Resource_Release(readback_texture); + destroy_test_context(&context); +} + static void test_shader_eval_attribute(void) { D3D12_GRAPHICS_PIPELINE_STATE_DESC pso_desc; @@ -28623,6 +28752,7 @@ START_TEST(d3d12) run_test(test_sample_mask); run_test(test_coverage); run_test(test_shader_get_render_target_sample_count); + run_test(test_shader_sample_position); run_test(test_shader_eval_attribute); run_test(test_primitive_restart); run_test(test_vertex_shader_stream_output); -- 2.21.0

2 1

[PATCH vkd3d 1/9] vkd3d-shader: Translate samplepos instructions.
by Józef Kucia April 29, 2019

April 29, 2019

From: Józef Kucia <jkucia(a)codeweavers.com> This returns standard sample locations. DXC and glslang do the same thing. https://github.com/khronosgroup/glslang/commit/d4d0b29752c173d2aa8d2d105f2f… In the long term, we could also add a shader parameter to pass sample positions to the shader compiler. Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- libs/vkd3d-shader/spirv.c | 171 ++++++++++++++++++++++++++++++++++---- 1 file changed, 154 insertions(+), 17 deletions(-) diff --git a/libs/vkd3d-shader/spirv.c b/libs/vkd3d-shader/spirv.c index 6332ef50394b..22e02f9ffa01 100644 --- a/libs/vkd3d-shader/spirv.c +++ b/libs/vkd3d-shader/spirv.c @@ -1129,6 +1129,12 @@ static uint32_t vkd3d_spirv_build_op_in_bounds_access_chain(struct vkd3d_spirv_b SpvOpInBoundsAccessChain, result_type, base_id, indexes, index_count); } +static uint32_t vkd3d_spirv_build_op_in_bounds_access_chain1(struct vkd3d_spirv_builder *builder, + uint32_t result_type, uint32_t base_id, uint32_t index) +{ + return vkd3d_spirv_build_op_in_bounds_access_chain(builder, result_type, base_id, &index, 1); +} + static uint32_t vkd3d_spirv_build_op_vector_shuffle(struct vkd3d_spirv_builder *builder, uint32_t result_type, uint32_t vector1_id, uint32_t vector2_id, const uint32_t *components, uint32_t component_count) @@ -1340,6 +1346,27 @@ static uint32_t vkd3d_spirv_build_op_shift_right_logical(struct vkd3d_spirv_buil SpvOpShiftRightLogical, result_type, base, shift); } +static uint32_t vkd3d_spirv_build_op_logical_and(struct vkd3d_spirv_builder *builder, + uint32_t result_type, uint32_t operand0, uint32_t operand1) +{ + return vkd3d_spirv_build_op_tr2(builder, &builder->function_stream, + SpvOpLogicalAnd, result_type, operand0, operand1); +} + +static uint32_t vkd3d_spirv_build_op_uless_than(struct vkd3d_spirv_builder *builder, + uint32_t result_type, uint32_t operand0, uint32_t operand1) +{ + return vkd3d_spirv_build_op_tr2(builder, &builder->function_stream, + SpvOpULessThan, result_type, operand0, operand1); +} + +static uint32_t vkd3d_spirv_build_op_uless_than_equal(struct vkd3d_spirv_builder *builder, + uint32_t result_type, uint32_t operand0, uint32_t operand1) +{ + return vkd3d_spirv_build_op_tr2(builder, &builder->function_stream, + SpvOpULessThanEqual, result_type, operand0, operand1); +} + static uint32_t vkd3d_spirv_build_op_convert_utof(struct vkd3d_spirv_builder *builder, uint32_t result_type, uint32_t unsigned_value) { @@ -1989,6 +2016,7 @@ struct vkd3d_dxbc_compiler uint32_t temp_id; unsigned int temp_count; struct vkd3d_hull_shader_variables hs; + uint32_t sample_positions_id; enum vkd3d_shader_type shader_type; @@ -2917,7 +2945,7 @@ static uint32_t vkd3d_dxbc_compiler_emit_load_scalar(struct vkd3d_dxbc_compiler const struct vkd3d_shader_register_info *reg_info) { struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; - uint32_t type_id, ptr_type_id, indexes[1], reg_id, val_id; + uint32_t type_id, ptr_type_id, index, reg_id, val_id; unsigned int component_idx, reg_component_count; enum vkd3d_component_type component_type; unsigned int skipped_component_mask; @@ -2945,9 +2973,8 @@ static uint32_t vkd3d_dxbc_compiler_emit_load_scalar(struct vkd3d_dxbc_compiler if (reg_component_count != 1) { ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, reg_info->storage_class, type_id); - indexes[0] = vkd3d_dxbc_compiler_get_constant_uint(compiler, component_idx); - reg_id = vkd3d_spirv_build_op_in_bounds_access_chain(builder, - ptr_type_id, reg_id, indexes, ARRAY_SIZE(indexes)); + index = vkd3d_dxbc_compiler_get_constant_uint(compiler, component_idx); + reg_id = vkd3d_spirv_build_op_in_bounds_access_chain1(builder, ptr_type_id, reg_id, index); } val_id = vkd3d_spirv_build_op_load(builder, type_id, reg_id, SpvMemoryAccessMaskNone); @@ -3104,7 +3131,7 @@ static void vkd3d_dxbc_compiler_emit_store_scalar(struct vkd3d_dxbc_compiler *co SpvStorageClass storage_class, unsigned int write_mask, uint32_t val_id) { struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; - uint32_t type_id, ptr_type_id, index[1]; + uint32_t type_id, ptr_type_id, index; unsigned int component_idx; if (vkd3d_write_mask_component_count(dst_write_mask) > 1) @@ -3113,8 +3140,8 @@ static void vkd3d_dxbc_compiler_emit_store_scalar(struct vkd3d_dxbc_compiler *co ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, storage_class, type_id); component_idx = vkd3d_write_mask_get_component_idx(write_mask); component_idx -= vkd3d_write_mask_get_component_idx(dst_write_mask); - index[0] = vkd3d_dxbc_compiler_get_constant_uint(compiler, component_idx); - dst_id = vkd3d_spirv_build_op_in_bounds_access_chain(builder, ptr_type_id, dst_id, index, ARRAY_SIZE(index)); + index = vkd3d_dxbc_compiler_get_constant_uint(compiler, component_idx); + dst_id = vkd3d_spirv_build_op_in_bounds_access_chain1(builder, ptr_type_id, dst_id, index); } vkd3d_spirv_build_op_store(builder, dst_id, val_id, SpvMemoryAccessMaskNone); @@ -3870,8 +3897,7 @@ static uint32_t vkd3d_dxbc_compiler_emit_input(struct vkd3d_dxbc_compiler *compi { ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, SpvStorageClassInput, type_id); index = vkd3d_dxbc_compiler_get_constant_uint(compiler, i); - val_id = vkd3d_spirv_build_op_in_bounds_access_chain(builder, - ptr_type_id, input_id, &index, 1); + val_id = vkd3d_spirv_build_op_in_bounds_access_chain1(builder, ptr_type_id, input_id, index); dst_reg.idx[0].offset = i; } val_id = vkd3d_spirv_build_op_load(builder, type_id, val_id, SpvMemoryAccessMaskNone); @@ -7704,15 +7730,12 @@ static void vkd3d_dxbc_compiler_emit_resinfo(struct vkd3d_dxbc_compiler *compile vkd3d_dxbc_compiler_emit_store_dst(compiler, dst, val_id); } -static void vkd3d_dxbc_compiler_emit_sample_info(struct vkd3d_dxbc_compiler *compiler, - const struct vkd3d_shader_instruction *instruction) +static uint32_t vkd3d_dxbc_compiler_emit_query_sample_count(struct vkd3d_dxbc_compiler *compiler, + const struct vkd3d_shader_src_param *src) { struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; - const struct vkd3d_shader_dst_param *dst = instruction->dst; - const struct vkd3d_shader_src_param *src = instruction->src; - uint32_t constituents[VKD3D_VEC4_SIZE]; + struct vkd3d_shader_image image; uint32_t type_id, val_id; - unsigned int i; if (src->reg.type == VKD3DSPR_RASTERIZER) { @@ -7721,8 +7744,6 @@ static void vkd3d_dxbc_compiler_emit_sample_info(struct vkd3d_dxbc_compiler *com } else { - struct vkd3d_shader_image image; - vkd3d_spirv_enable_capability(builder, SpvCapabilityImageQuery); vkd3d_dxbc_compiler_prepare_image(compiler, &image, &src->reg, NULL, VKD3D_IMAGE_FLAG_NONE); @@ -7730,6 +7751,21 @@ static void vkd3d_dxbc_compiler_emit_sample_info(struct vkd3d_dxbc_compiler *com val_id = vkd3d_spirv_build_op_image_query_samples(builder, type_id, image.image_id); } + return val_id; +} + +static void vkd3d_dxbc_compiler_emit_sample_info(struct vkd3d_dxbc_compiler *compiler, + const struct vkd3d_shader_instruction *instruction) +{ + struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; + const struct vkd3d_shader_dst_param *dst = instruction->dst; + const struct vkd3d_shader_src_param *src = instruction->src; + uint32_t constituents[VKD3D_VEC4_SIZE]; + uint32_t type_id, val_id; + unsigned int i; + + val_id = vkd3d_dxbc_compiler_emit_query_sample_count(compiler, src); + constituents[0] = val_id; for (i = 1; i < VKD3D_VEC4_SIZE; ++i) constituents[i] = vkd3d_dxbc_compiler_get_constant_uint(compiler, 0); @@ -7754,6 +7790,104 @@ static void vkd3d_dxbc_compiler_emit_sample_info(struct vkd3d_dxbc_compiler *com vkd3d_dxbc_compiler_emit_store_dst(compiler, dst, val_id); } +/* XXX: This is correct only when standard sample positions are used. */ +static void vkd3d_dxbc_compiler_emit_sample_position(struct vkd3d_dxbc_compiler *compiler, + const struct vkd3d_shader_instruction *instruction) +{ + /* Standard sample locations from the Vulkan spec. */ + static const float standard_sample_positions[][2] = + { + /* 1 sample */ + { 0.0 / 16.0, 0.0 / 16.0}, + /* 2 samples */ + { 4.0 / 16.0, 4.0 / 16.0}, + {-4.0 / 16.0, -4.0 / 16.0}, + /* 4 samples */ + {-2.0 / 16.0, -6.0 / 16.0}, + { 6.0 / 16.0, -2.0 / 16.0}, + {-6.0 / 16.0, 2.0 / 16.0}, + { 2.0 / 16.0, 6.0 / 16.0}, + /* 8 samples */ + { 1.0 / 16.0, -3.0 / 16.0}, + {-1.0 / 16.0, 3.0 / 16.0}, + { 5.0 / 16.0, 1.0 / 16.0}, + {-3.0 / 16.0, -5.0 / 16.0}, + {-5.0 / 16.0, 5.0 / 16.0}, + {-7.0 / 16.0, -1.0 / 16.0}, + { 3.0 / 16.0, 7.0 / 16.0}, + { 7.0 / 16.0, -7.0 / 16.0}, + /* 16 samples */ + { 1.0 / 16.0, 1.0 / 16.0}, + {-1.0 / 16.0, -3.0 / 16.0}, + {-3.0 / 16.0, 2.0 / 16.0}, + { 4.0 / 16.0, -1.0 / 16.0}, + {-5.0 / 16.0, -2.0 / 16.0}, + { 2.0 / 16.0, 5.0 / 16.0}, + { 5.0 / 16.0, 3.0 / 16.0}, + { 3.0 / 16.0, -5.0 / 16.0}, + {-2.0 / 16.0, 6.0 / 16.0}, + { 0.0 / 16.0, -7.0 / 16.0}, + {-4.0 / 16.0, -6.0 / 16.0}, + {-6.0 / 16.0, 4.0 / 16.0}, + {-8.0 / 16.0, 0.0 / 16.0}, + { 7.0 / 16.0, -4.0 / 16.0}, + { 6.0 / 16.0, 7.0 / 16.0}, + {-7.0 / 16.0, -8.0 / 16.0}, + }; + struct vkd3d_spirv_builder *builder = &compiler->spirv_builder; + uint32_t constituents[ARRAY_SIZE(standard_sample_positions)]; + const struct vkd3d_shader_dst_param *dst = instruction->dst; + uint32_t array_type_id, length_id, index_id, id; + uint32_t sample_count_id, sample_index_id; + uint32_t type_id, bool_id, ptr_type_id; + unsigned int i; + + sample_count_id = vkd3d_dxbc_compiler_emit_query_sample_count(compiler, &instruction->src[0]); + sample_index_id = vkd3d_dxbc_compiler_emit_load_src(compiler, &instruction->src[1], VKD3DSP_WRITEMASK_0); + + type_id = vkd3d_spirv_get_type_id(builder, VKD3D_TYPE_UINT, 1); + index_id = vkd3d_spirv_build_op_iadd(builder, type_id, sample_count_id, sample_index_id); + index_id = vkd3d_spirv_build_op_isub(builder, + type_id, index_id, vkd3d_dxbc_compiler_get_constant_uint(compiler, 1)); + + /* Validate sample index. */ + bool_id = vkd3d_spirv_get_type_id(builder, VKD3D_TYPE_BOOL, 1); + id = vkd3d_spirv_build_op_logical_and(builder, bool_id, + vkd3d_spirv_build_op_uless_than(builder, bool_id, sample_index_id, sample_count_id), + vkd3d_spirv_build_op_uless_than_equal(builder, + bool_id, sample_index_id, vkd3d_dxbc_compiler_get_constant_uint(compiler, 16))); + index_id = vkd3d_spirv_build_op_select(builder, type_id, + id, index_id, vkd3d_dxbc_compiler_get_constant_uint(compiler, 0)); + + type_id = vkd3d_spirv_get_type_id(builder, VKD3D_TYPE_FLOAT, 2); + if (!(id = compiler->sample_positions_id)) + { + length_id = vkd3d_dxbc_compiler_get_constant_uint(compiler, ARRAY_SIZE(standard_sample_positions)); + array_type_id = vkd3d_spirv_get_op_type_array(builder, type_id, length_id); + + for (i = 0; i < ARRAY_SIZE(standard_sample_positions); ++ i) + { + constituents[i] = vkd3d_dxbc_compiler_get_constant(compiler, + VKD3D_TYPE_FLOAT, 2, (const uint32_t *)standard_sample_positions[i]); + } + + id = vkd3d_spirv_build_op_constant_composite(builder, array_type_id, constituents, ARRAY_SIZE(constituents)); + ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, SpvStorageClassPrivate, array_type_id); + id = vkd3d_spirv_build_op_variable(builder, &builder->global_stream, ptr_type_id, SpvStorageClassPrivate, id); + vkd3d_spirv_build_op_name(builder, id, "sample_pos"); + compiler->sample_positions_id = id; + } + + ptr_type_id = vkd3d_spirv_get_op_type_pointer(builder, SpvStorageClassPrivate, type_id); + id = vkd3d_spirv_build_op_in_bounds_access_chain1(builder, ptr_type_id, id, index_id); + id = vkd3d_spirv_build_op_load(builder, type_id, id, SpvMemoryAccessMaskNone); + + id = vkd3d_dxbc_compiler_emit_swizzle(compiler, + id, VKD3DSP_WRITEMASK_0 | VKD3DSP_WRITEMASK_1, VKD3D_TYPE_FLOAT, + instruction->src[0].swizzle, dst->write_mask); + vkd3d_dxbc_compiler_emit_store_dst(compiler, dst, id); +} + static void vkd3d_dxbc_compiler_emit_eval_attrib(struct vkd3d_dxbc_compiler *compiler, const struct vkd3d_shader_instruction *instruction) { @@ -8208,6 +8342,9 @@ int vkd3d_dxbc_compiler_handle_instruction(struct vkd3d_dxbc_compiler *compiler, case VKD3DSIH_SAMPLE_INFO: vkd3d_dxbc_compiler_emit_sample_info(compiler, instruction); break; + case VKD3DSIH_SAMPLE_POS: + vkd3d_dxbc_compiler_emit_sample_position(compiler, instruction); + break; case VKD3DSIH_EVAL_CENTROID: case VKD3DSIH_EVAL_SAMPLE_INDEX: vkd3d_dxbc_compiler_emit_eval_attrib(compiler, instruction); -- 2.21.0

2 1

[PATCH v3 1/3] wrc: Support Asturian
by Alex Henrie April 29, 2019

April 29, 2019

From: Enol Puente <enolp(a)softastur.org> Signed-off-by: Enol Puente <enolp(a)softastur.org> Signed-off-by: Alex Henrie <alexhenrie24(a)gmail.com> --- include/winnt.rh | 1 + tools/wmc/po.c | 2 ++ tools/wrc/po.c | 2 ++ tools/wrc/utils.c | 1 + 4 files changed, 6 insertions(+) diff --git a/include/winnt.rh b/include/winnt.rh index 50aafb1d96..59791bc996 100644 --- a/include/winnt.rh +++ b/include/winnt.rh @@ -185,6 +185,7 @@ #define LANG_VENDA 0x33 /* non standard; keep the number high enough (but < 0xff) */ +#define LANG_ASTURIAN 0xa5 #define LANG_ESPERANTO 0x8f #define LANG_WALON 0x90 #define LANG_CORNISH 0x92 diff --git a/tools/wmc/po.c b/tools/wmc/po.c index 8a8064a25f..bf10c50c63 100644 --- a/tools/wmc/po.c +++ b/tools/wmc/po.c @@ -82,6 +82,8 @@ static const struct { LANG_ARMENIAN, SUBLANG_ARMENIAN_ARMENIA, "hy_AM" }, { LANG_ASSAMESE, SUBLANG_NEUTRAL, "as" }, { LANG_ASSAMESE, SUBLANG_ASSAMESE_INDIA, "as_IN" }, + { LANG_ASTURIAN, SUBLANG_NEUTRAL, "ast" }, + { LANG_ASTURIAN, SUBLANG_DEFAULT, "ast_ES" }, { LANG_AZERBAIJANI, SUBLANG_NEUTRAL, "az" }, { LANG_AZERBAIJANI, SUBLANG_AZERBAIJANI_AZERBAIJAN_LATIN,"az_AZ(a)latin" }, { LANG_AZERBAIJANI, SUBLANG_AZERBAIJANI_AZERBAIJAN_CYRILLIC, "az_AZ(a)cyrillic" }, diff --git a/tools/wrc/po.c b/tools/wrc/po.c index 913a918a69..3f5affd49f 100644 --- a/tools/wrc/po.c +++ b/tools/wrc/po.c @@ -239,6 +239,8 @@ static const struct { LANG_ARMENIAN, SUBLANG_ARMENIAN_ARMENIA, "hy_AM" }, { LANG_ASSAMESE, SUBLANG_NEUTRAL, "as" }, { LANG_ASSAMESE, SUBLANG_ASSAMESE_INDIA, "as_IN" }, + { LANG_ASTURIAN, SUBLANG_NEUTRAL, "ast" }, + { LANG_ASTURIAN, SUBLANG_DEFAULT, "ast_ES" }, { LANG_AZERBAIJANI, SUBLANG_NEUTRAL, "az" }, { LANG_AZERBAIJANI, SUBLANG_AZERBAIJANI_AZERBAIJAN_LATIN,"az_AZ(a)latin" }, { LANG_AZERBAIJANI, SUBLANG_AZERBAIJANI_AZERBAIJAN_CYRILLIC, "az_AZ(a)cyrillic" }, diff --git a/tools/wrc/utils.c b/tools/wrc/utils.c index 5533a531f1..a1e07b85b2 100644 --- a/tools/wrc/utils.c +++ b/tools/wrc/utils.c @@ -457,6 +457,7 @@ static const struct lang2cp lang2cps[] = { LANG_ARABIC, SUBLANG_NEUTRAL, 1256 }, { LANG_ARMENIAN, SUBLANG_NEUTRAL, 0 }, { LANG_ASSAMESE, SUBLANG_NEUTRAL, 0 }, + { LANG_ASTURIAN, SUBLANG_NEUTRAL, 1252 }, { LANG_AZERI, SUBLANG_NEUTRAL, 1254 }, { LANG_AZERI, SUBLANG_AZERI_CYRILLIC, 1251 }, { LANG_BASHKIR, SUBLANG_NEUTRAL, 1251 }, -- 2.21.0

3 6

[PATCH] wintrust: Use the ARRAY_SIZE() macro
by Michael Stefaniuc April 29, 2019

April 29, 2019

Signed-off-by: Michael Stefaniuc <mstefani(a)winehq.org> --- dlls/wintrust/tests/softpub.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/dlls/wintrust/tests/softpub.c b/dlls/wintrust/tests/softpub.c index 914f04d7c0..2a4f4f4b25 100644 --- a/dlls/wintrust/tests/softpub.c +++ b/dlls/wintrust/tests/softpub.c @@ -1211,7 +1211,7 @@ static void test_wintrust_digest(void) BOOL ret; int i, j; - for (i = 0; i < sizeof(tests)/sizeof(tests[0]); i++) + for (i = 0; i < ARRAY_SIZE(tests); i++) { file = create_temp_file(pathW); ok(file != INVALID_HANDLE_VALUE, "failed to create temporary file\n"); -- 2.20.1

1 0

[PATCH] mscoree: Remove a redundant NULL check before heap_free()
by Michael Stefaniuc April 29, 2019

April 29, 2019

Signed-off-by: Michael Stefaniuc <mstefani(a)winehq.org> --- dlls/mscoree/corruntimehost.c | 3 +-- 1 file changed, 1 insertion(+), 2 deletions(-) diff --git a/dlls/mscoree/corruntimehost.c b/dlls/mscoree/corruntimehost.c index 95f6a9a892..13619bdc19 100644 --- a/dlls/mscoree/corruntimehost.c +++ b/dlls/mscoree/corruntimehost.c @@ -1707,8 +1707,7 @@ static BOOL try_create_registration_free_com(REFIID clsid, WCHAR *classname, UIN ret = TRUE; end: - if (assembly_info) - heap_free(assembly_info); + heap_free(assembly_info); if (guid_info.hActCtx) ReleaseActCtx(guid_info.hActCtx); -- 2.20.1

1 0

[PATCH] ntdll/tests: Avoid the comma operator
by Michael Stefaniuc April 29, 2019

April 29, 2019

Signed-off-by: Michael Stefaniuc <mstefani(a)winehq.org> --- dlls/ntdll/tests/exception.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/dlls/ntdll/tests/exception.c b/dlls/ntdll/tests/exception.c index 4f75cc2454..ada28c91eb 100644 --- a/dlls/ntdll/tests/exception.c +++ b/dlls/ntdll/tests/exception.c @@ -3274,7 +3274,7 @@ static void test_suspend_process(void) startup.cb = sizeof(startup); sprintf(path_name, "%s exception suspend_process", argv[0]); - ret = CreateProcessA(NULL, path_name, NULL, NULL, FALSE, 0, NULL, NULL, &startup, &info), + ret = CreateProcessA(NULL, path_name, NULL, NULL, FALSE, 0, NULL, NULL, &startup, &info); ok(ret, "Failed to create target process.\n"); /* New process signals this event. */ -- 2.20.1

1 0

[PATCH] d3dcompiler: Implement D3DReadFileToBlob().
by Jactry Zeng April 29, 2019

April 29, 2019

Signed-off-by: Jactry Zeng <jzeng(a)codeweavers.com> --- dlls/d3dcompiler_43/blob.c | 54 +++++++++++++++- dlls/d3dcompiler_43/tests/blob.c | 106 +++++++++++++++++++++++++++++++ 2 files changed, 158 insertions(+), 2 deletions(-) diff --git a/dlls/d3dcompiler_43/blob.c b/dlls/d3dcompiler_43/blob.c index f22dc7183d..6b1b20317f 100644 --- a/dlls/d3dcompiler_43/blob.c +++ b/dlls/d3dcompiler_43/blob.c @@ -466,9 +466,59 @@ HRESULT WINAPI D3DStripShader(const void *data, SIZE_T data_size, UINT flags, ID HRESULT WINAPI D3DReadFileToBlob(const WCHAR *filename, ID3DBlob **contents) { - FIXME("filename %s, contents %p\n", debugstr_w(filename), contents); + struct d3dcompiler_blob *object; + HANDLE file; + SIZE_T data_size; + DWORD read_size; + HRESULT hr; - return E_NOTIMPL; + TRACE("filename %s, contents %p\n", debugstr_w(filename), contents); + + file = CreateFileW(filename, GENERIC_READ, FILE_SHARE_READ, NULL, + OPEN_EXISTING, FILE_ATTRIBUTE_NORMAL, NULL); + if (file == INVALID_HANDLE_VALUE) + return HRESULT_FROM_WIN32(GetLastError()); + + data_size = GetFileSize(file, NULL); + if (data_size == INVALID_FILE_SIZE) + { + CloseHandle(file); + return HRESULT_FROM_WIN32(GetLastError()); + } + + object = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(*object)); + if (!object) + { + CloseHandle(file); + return E_OUTOFMEMORY; + } + + hr = d3dcompiler_blob_init(object, data_size); + if (FAILED(hr)) + { + WARN("Failed to initialize blob, hr %#x.\n", hr); + CloseHandle(file); + HeapFree(GetProcessHeap(), 0, object); + return hr; + } + + if (!ReadFile(file, object->data, data_size, &read_size, NULL) || + (read_size != data_size)) + { + WARN("Failed to read file contents.\n"); + CloseHandle(file); + HeapFree(GetProcessHeap(), 0, object->data); + HeapFree(GetProcessHeap(), 0, object); + return E_FAIL; + } + CloseHandle(file); + object->size = read_size; + + *contents = &object->ID3DBlob_iface; + + TRACE("Returning ID3DBlob %p.\n", *contents); + + return S_OK; } HRESULT WINAPI D3DWriteBlobToFile(ID3DBlob* blob, const WCHAR *filename, BOOL overwrite) diff --git a/dlls/d3dcompiler_43/tests/blob.c b/dlls/d3dcompiler_43/tests/blob.c index 443ed2754c..6a9fa5de8b 100644 --- a/dlls/d3dcompiler_43/tests/blob.c +++ b/dlls/d3dcompiler_43/tests/blob.c @@ -35,6 +35,7 @@ static HRESULT (WINAPI *pD3DCreateBlob)(SIZE_T, ID3DBlob **); static HRESULT (WINAPI *pD3DGetBlobPart)(const void *, SIZE_T, D3D_BLOB_PART, UINT, ID3DBlob **); +static HRESULT (WINAPI *pD3DReadFileToBlob)(const WCHAR *, ID3DBlob **); static HRESULT (WINAPI *pD3DStripShader)(const void *, SIZE_T, UINT, ID3DBlob **); #define MAKE_TAG(ch0, ch1, ch2, ch3) \ @@ -752,6 +753,103 @@ static BOOL load_d3dcompiler(void) return TRUE; } +static BOOL load_d3dcompiler_47(void) +{ + HMODULE module; + + if (!(module = LoadLibraryA("d3dcompiler_47.dll"))) return FALSE; + + pD3DReadFileToBlob = (void*)GetProcAddress(module, "D3DReadFileToBlob"); + return TRUE; +} + +static void create_cso_file(LPCWSTR pathW, void *data, DWORD data_size) +{ + HANDLE file; + DWORD written; + + file = CreateFileW(pathW, GENERIC_READ | GENERIC_WRITE, 0, + NULL, CREATE_ALWAYS, 0, 0); + ok(file != INVALID_HANDLE_VALUE, "File creation failed, at %s, error 0x%08x.\n", + wine_dbgstr_w(pathW), GetLastError()); + + if (data) + { + WriteFile(file, data, data_size, &written, NULL); + ok(written == data_size, "Couldn't write .cso file.\n" ); + } + CloseHandle(file); +} + +/* .cso file compiled by fxc.exe. + HLSL source: + ``` + struct PSInput + { + float4 value : SV_POSITION; + }; + + PSInput main(float4 position : POSITION) + { + PSInput result; + result.value = position; + return result; + } + ``` + */ +static byte test_cso_data[] = +{ + 0x00,0x02,0xfe,0xff,0xfe,0xff,0x14,0x00,0x43,0x54,0x41,0x42,0x1c,0x00,0x00,0x00,0x23,0x00, + 0x00,0x00,0x00,0x02,0xfe,0xff,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x00,0x01,0x00,0x00, + 0x1c,0x00,0x00,0x00,0x76,0x73,0x5f,0x32,0x5f,0x30,0x00,0x4d,0x69,0x63,0x72,0x6f,0x73,0x6f, + 0x66,0x74,0x20,0x28,0x52,0x29,0x20,0x48,0x4c,0x53,0x4c,0x20,0x53,0x68,0x61,0x64,0x65,0x72, + 0x20,0x43,0x6f,0x6d,0x70,0x69,0x6c,0x65,0x72,0x20,0x31,0x30,0x2e,0x31,0x00,0xab,0x1f,0x00, + 0x00,0x02,0x00,0x00,0x00,0x80,0x00,0x00,0x0f,0x90,0x01,0x00,0x00,0x02,0x00,0x00,0x0f,0xc0, + 0x00,0x00,0xe4,0x90,0xff,0xff,0x00,0x00 +}; + +static void test_D3DReadFileToBlob(void) +{ + ID3DBlob *blob = NULL; + HRESULT hr; + static const WCHAR filenameW[] = {'t','e','s','t','.','c','s','o',0}; + byte *data_enter; + SIZE_T data_size; + + hr = pD3DReadFileToBlob(filenameW, NULL); + ok(hr == HRESULT_FROM_WIN32(ERROR_FILE_NOT_FOUND), + "D3DReadFileToBlob returned: 0x%08x.\n", hr); + + hr = pD3DReadFileToBlob(filenameW, &blob); + ok(hr == HRESULT_FROM_WIN32(ERROR_FILE_NOT_FOUND), + "D3DReadFileToBlob returned: 0x%08x.\n", hr); + + /* Crash on Windows + create_cso_file(filenameW, test_cso_data, sizeof(test_cso_data)); + pD3DReadFileToBlob(filenameW, NULL); + DeleteFileW(filenameW); + */ + + create_cso_file(filenameW, NULL, 0); + hr = pD3DReadFileToBlob(filenameW, &blob); + ok(hr == S_OK, "D3DReadFileToBlob failed: 0x%08x.\n", hr); + data_size = ID3D10Blob_GetBufferSize(blob); + ok(data_size == 0, "got wrong data size: %lu, expected %u.\n", data_size, 0); + DeleteFileW(filenameW); + ID3D10Blob_Release(blob); + + create_cso_file(filenameW, test_cso_data, sizeof(test_cso_data)); + hr = pD3DReadFileToBlob(filenameW, &blob); + ok(hr == S_OK, "D3DReadFileToBlob failed: 0x%08x.\n", hr); + data_size = ID3D10Blob_GetBufferSize(blob); + ok(data_size == sizeof(test_cso_data), "got wrong data size: %lu, expected %u.\n", + data_size, sizeof(test_cso_data)); + data_enter = ID3D10Blob_GetBufferPointer(blob); + ok(!memcmp(data_enter, test_cso_data, sizeof(test_cso_data)), "got wrong data.\n"); + DeleteFileW(filenameW); + ID3D10Blob_Release(blob); +} + START_TEST(blob) { if (!load_d3dcompiler()) @@ -763,4 +861,12 @@ START_TEST(blob) test_create_blob(); test_get_blob_part(); test_get_blob_part2(); + + if (!load_d3dcompiler_47()) + { + win_skip("Could not load d3dcompiler_47.dll\n"); + return; + } + + test_D3DReadFileToBlob(); } -- 2.20.1

2 1

[PATCH] user32: Add DECLSPEC_HOTPATCH to UnRegisterDeviceNotification Wine-bug: https://bugs.winehq.org/show_bug.cgi?id=47098
by Louis Lenders April 29, 2019

April 29, 2019

Signed-off-by: Louis Lenders <xerox.xerox2000x(a)gmail.com> --- dlls/user32/misc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/dlls/user32/misc.c b/dlls/user32/misc.c index bd4795a346..6ff0980ecd 100644 --- a/dlls/user32/misc.c +++ b/dlls/user32/misc.c @@ -404,7 +404,7 @@ HDEVNOTIFY WINAPI DECLSPEC_HOTPATCH RegisterDeviceNotificationW(HANDLE hRecipien * UnregisterDeviceNotification (USER32.@) * */ -BOOL WINAPI UnregisterDeviceNotification(HDEVNOTIFY hnd) +BOOL WINAPI DECLSPEC_HOTPATCH UnregisterDeviceNotification(HDEVNOTIFY hnd) { FIXME("(handle=%p), STUB!\n", hnd); return TRUE; -- 2.20.1

2 1