September 2018 - Wine-Devel

[PATCH vkd3d 8/8] vkd3d: Fix potential memory leak in vkd3d_create_vk_device().
by Józef Kucia Sept. 28, 2018

Sept. 28, 2018

From: Józef Kucia <jkucia(a)codeweavers.com> Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- libs/vkd3d/device.c | 19 ++++++++++--------- 1 file changed, 10 insertions(+), 9 deletions(-) diff --git a/libs/vkd3d/device.c b/libs/vkd3d/device.c index 69b6a30fe217..673c54a889cf 100644 --- a/libs/vkd3d/device.c +++ b/libs/vkd3d/device.c @@ -1025,7 +1025,7 @@ static HRESULT vkd3d_create_vk_device(struct d3d12_device *device, if (!(queue_info = vkd3d_calloc(queue_family_count, sizeof(*queue_info)))) { hr = E_OUTOFMEMORY; - goto fail; + goto done; } direct_queue_family_index = ~0u; @@ -1062,13 +1062,13 @@ static HRESULT vkd3d_create_vk_device(struct d3d12_device *device, { FIXME("Could not find a suitable queue family for a direct command queue.\n"); hr = E_FAIL; - goto fail; + goto done; } if (copy_queue_family_index == ~0u) { FIXME("Could not find a suitable queue family for a copy command queue.\n"); hr = E_FAIL; - goto fail; + goto done; } if (compute_queue_family_index == ~0u) { @@ -1085,12 +1085,12 @@ static HRESULT vkd3d_create_vk_device(struct d3d12_device *device, VK_CALL(vkGetPhysicalDeviceFeatures(physical_device, &device_features)); device->vk_physical_device = physical_device; if (FAILED(hr = vkd3d_init_device_caps(device, create_info, &device_features, &extension_count))) - goto fail; + goto done; if (!(extensions = vkd3d_calloc(extension_count, sizeof(*extensions)))) { hr = E_OUTOFMEMORY; - goto fail; + goto done; } /* Create device */ @@ -1114,7 +1114,8 @@ static HRESULT vkd3d_create_vk_device(struct d3d12_device *device, if (vr < 0) { ERR("Failed to create Vulkan device, vr %d.\n", vr); - return hresult_from_vk_result(vr); + hr = hresult_from_vk_result(vr); + goto done; } if (FAILED(hr = vkd3d_load_vk_device_procs(&device->vk_procs, vk_procs, vk_device))) @@ -1122,7 +1123,7 @@ static HRESULT vkd3d_create_vk_device(struct d3d12_device *device, ERR("Failed to load device procs, hr %#x.\n", hr); if (device->vk_procs.vkDestroyDevice) device->vk_procs.vkDestroyDevice(vk_device, NULL); - goto fail; + goto done; } device->vk_device = vk_device; @@ -1132,12 +1133,12 @@ static HRESULT vkd3d_create_vk_device(struct d3d12_device *device, { ERR("Failed to create queues, hr %#x.\n", hr); device->vk_procs.vkDestroyDevice(vk_device, NULL); - goto fail; + goto done; } TRACE("Created Vulkan device %p.\n", vk_device); -fail: +done: if (queue_properties) vkd3d_free(queue_properties); if (queue_info) -- 2.16.4

2 1

[PATCH vkd3d 7/8] vkd3d: Add requirements for maxPerStageDescriptorUniformBuffers.
by Józef Kucia Sept. 28, 2018

Sept. 28, 2018

From: Józef Kucia <jkucia(a)codeweavers.com> Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- include/vkd3d_d3d12.idl | 2 ++ libs/vkd3d/device.c | 4 ++++ 2 files changed, 6 insertions(+) diff --git a/include/vkd3d_d3d12.idl b/include/vkd3d_d3d12.idl index d647b48cbbe4..29f4ab3cbf47 100644 --- a/include/vkd3d_d3d12.idl +++ b/include/vkd3d_d3d12.idl @@ -30,6 +30,8 @@ const UINT D3D12_MAX_ROOT_COST = 64; const UINT D3D12_VIEWPORT_BOUNDS_MAX = 32767; const UINT D3D12_VIEWPORT_BOUNDS_MIN = -32768; +const UINT D3D12_COMMONSHADER_CONSTANT_BUFFER_REGISTER_COUNT = 15; + const UINT D3D12_APPEND_ALIGNED_ELEMENT = 0xffffffff; cpp_quote("#define D3D12_DEFAULT_BLEND_FACTOR_ALPHA (1.0f)") cpp_quote("#define D3D12_DEFAULT_BLEND_FACTOR_BLUE (1.0f)") diff --git a/libs/vkd3d/device.c b/libs/vkd3d/device.c index 3a9dad03ef91..69b6a30fe217 100644 --- a/libs/vkd3d/device.c +++ b/libs/vkd3d/device.c @@ -777,10 +777,14 @@ static void vkd3d_check_feature_level_11_requirements(const VkPhysicalDeviceLimi CHECK_MIN_REQUIREMENT(maxPushConstantsSize, D3D12_MAX_ROOT_COST * sizeof(uint32_t)); CHECK_MIN_REQUIREMENT(maxComputeSharedMemorySize, D3D12_CS_TGSM_REGISTER_COUNT * sizeof(uint32_t)); + CHECK_MAX_REQUIREMENT(viewportBoundsRange[0], D3D12_VIEWPORT_BOUNDS_MIN); CHECK_MIN_REQUIREMENT(viewportBoundsRange[1], D3D12_VIEWPORT_BOUNDS_MAX); CHECK_MIN_REQUIREMENT(viewportSubPixelBits, 8); + CHECK_MIN_REQUIREMENT(maxPerStageDescriptorUniformBuffers, + D3D12_COMMONSHADER_CONSTANT_BUFFER_REGISTER_COUNT); + CHECK_FEATURE(imageCubeArray); CHECK_FEATURE(independentBlend); CHECK_FEATURE(geometryShader); -- 2.16.4

2 1

[PATCH vkd3d 6/8] include: Synchronize d3d12.idl with Wine.
by Józef Kucia Sept. 28, 2018

Sept. 28, 2018

From: Józef Kucia <jkucia(a)codeweavers.com> Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- include/vkd3d_d3d12.idl | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/include/vkd3d_d3d12.idl b/include/vkd3d_d3d12.idl index 4b967d489b77..d647b48cbbe4 100644 --- a/include/vkd3d_d3d12.idl +++ b/include/vkd3d_d3d12.idl @@ -2213,7 +2213,12 @@ interface ID3D12RootSignatureDeserializer : IUnknown const D3D12_VERSIONED_ROOT_SIGNATURE_DESC *root_signature_desc, ID3DBlob **blob, ID3DBlob **error_blob); +typedef HRESULT (__stdcall *PFN_D3D12_CREATE_DEVICE)(IUnknown *adapter, + D3D_FEATURE_LEVEL minimum_feature_level, REFIID iid, void **device); + [local] HRESULT __stdcall D3D12CreateDevice(IUnknown *adapter, D3D_FEATURE_LEVEL minimum_feature_level, REFIID iid, void **device); +typedef HRESULT (__stdcall *PFN_D3D12_GET_DEBUG_INTERFACE)(REFIID iid, void **debug); + [local] HRESULT __stdcall D3D12GetDebugInterface(REFIID iid, void **debug); -- 2.16.4

2 1

[PATCH vkd3d 5/8] tests: Add test for blend factor.
by Józef Kucia Sept. 28, 2018

Sept. 28, 2018

From: Józef Kucia <jkucia(a)codeweavers.com> Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- tests/d3d12.c | 69 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 68 insertions(+), 1 deletion(-) diff --git a/tests/d3d12.c b/tests/d3d12.c index 0528ceb4ba6a..c40ce12719b6 100644 --- a/tests/d3d12.c +++ b/tests/d3d12.c @@ -19970,7 +19970,6 @@ static void test_resource_allocation_info(void) } } - refcount = ID3D12Device_Release(device); ok(!refcount, "ID3D12Device has %u references left.\n", (unsigned int)refcount); } @@ -20134,6 +20133,73 @@ static void test_command_list_initial_pipeline_state(void) destroy_test_context(&context); } +static void test_blend_factor(void) +{ + static const float white[] = {1.0f, 1.0f, 1.0f, 1.0f}; + D3D12_GRAPHICS_PIPELINE_STATE_DESC pso_desc; + ID3D12GraphicsCommandList *command_list; + struct test_context_desc desc; + struct test_context context; + ID3D12CommandQueue *queue; + unsigned int i; + HRESULT hr; + + static const struct + { + float blend_factor[4]; + unsigned int expected_color; + } + tests[] = + { + {{0.0f, 0.0f, 0.0f, 0.0f}, 0xffffffff}, + {{0.0f, 1.0f, 0.0f, 1.0f}, 0xffffffff}, + {{0.5f, 0.5f, 0.5f, 0.5f}, 0xff80ff80}, + {{1.0f, 1.0f, 1.0f, 1.0f}, 0xff00ff00}, + }; + + memset(&desc, 0, sizeof(desc)); + desc.no_pipeline = true; + if (!init_test_context(&context, &desc)) + return; + command_list = context.list; + queue = context.queue; + + init_pipeline_state_desc(&pso_desc, context.root_signature, + context.render_target_desc.Format, NULL, NULL, NULL); + pso_desc.BlendState.RenderTarget[0].BlendEnable = TRUE; + pso_desc.BlendState.RenderTarget[0].SrcBlend = D3D12_BLEND_BLEND_FACTOR; + pso_desc.BlendState.RenderTarget[0].DestBlend = D3D12_BLEND_INV_BLEND_FACTOR; + pso_desc.BlendState.RenderTarget[0].BlendOp = D3D12_BLEND_OP_ADD; + pso_desc.BlendState.RenderTarget[0].SrcBlendAlpha = D3D12_BLEND_BLEND_FACTOR; + pso_desc.BlendState.RenderTarget[0].DestBlendAlpha = D3D12_BLEND_INV_BLEND_FACTOR; + pso_desc.BlendState.RenderTarget[0].BlendOpAlpha = D3D12_BLEND_OP_ADD; + hr = ID3D12Device_CreateGraphicsPipelineState(context.device, &pso_desc, + &IID_ID3D12PipelineState, (void **)&context.pipeline_state); + ok(hr == S_OK, "Failed to create pipeline, hr %#x.\n", hr); + + for (i = 0; i < ARRAY_SIZE(tests); ++i) + { + 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_OMSetBlendFactor(command_list, tests[i].blend_factor); + 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); + check_sub_resource_uint(context.render_target, 0, queue, command_list, tests[i].expected_color, 1); + + reset_command_list(command_list, context.allocator); + transition_resource_state(command_list, context.render_target, + D3D12_RESOURCE_STATE_COPY_SOURCE, D3D12_RESOURCE_STATE_RENDER_TARGET); + } + + destroy_test_context(&context); +} + START_TEST(d3d12) { bool enable_debug_layer = false; @@ -20255,4 +20321,5 @@ START_TEST(d3d12) run_test(test_resource_allocation_info); run_test(test_suballocate_small_textures); run_test(test_command_list_initial_pipeline_state); + run_test(test_blend_factor); } -- 2.16.4

2 1

[PATCH vkd3d 4/8] vkd3d: Implement d3d12_command_list_OMSetBlendFactor().
by Józef Kucia Sept. 28, 2018

Sept. 28, 2018

From: Józef Kucia <jkucia(a)codeweavers.com> Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- libs/vkd3d/command.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/libs/vkd3d/command.c b/libs/vkd3d/command.c index b4d8ecd75179..5a2baef8a7da 100644 --- a/libs/vkd3d/command.c +++ b/libs/vkd3d/command.c @@ -1849,6 +1849,7 @@ static VkPipeline d3d12_command_list_get_or_create_pipeline(struct d3d12_command { VK_DYNAMIC_STATE_VIEWPORT, VK_DYNAMIC_STATE_SCISSOR, + VK_DYNAMIC_STATE_BLEND_CONSTANTS, VK_DYNAMIC_STATE_STENCIL_REFERENCE, }; static const struct VkPipelineDynamicStateCreateInfo dynamic_desc = @@ -2972,7 +2973,13 @@ static void STDMETHODCALLTYPE d3d12_command_list_RSSetScissorRects(ID3D12Graphic static void STDMETHODCALLTYPE d3d12_command_list_OMSetBlendFactor(ID3D12GraphicsCommandList *iface, const FLOAT blend_factor[4]) { - FIXME("iface %p, blend_factor %p stub!\n", iface, blend_factor); + struct d3d12_command_list *list = impl_from_ID3D12GraphicsCommandList(iface); + const struct vkd3d_vk_device_procs *vk_procs; + + TRACE("iface %p, blend_factor %p.\n", iface, blend_factor); + + vk_procs = &list->device->vk_procs; + VK_CALL(vkCmdSetBlendConstants(list->vk_command_buffer, blend_factor)); } static void STDMETHODCALLTYPE d3d12_command_list_OMSetStencilRef(ID3D12GraphicsCommandList *iface, -- 2.16.4

2 1

[PATCH vkd3d 1/8] vkd3d-shader: Use TRACE() for skipped DXBC chunks.
by Józef Kucia Sept. 28, 2018

Sept. 28, 2018

From: Józef Kucia <jkucia(a)codeweavers.com> Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- libs/vkd3d-shader/dxbc.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/libs/vkd3d-shader/dxbc.c b/libs/vkd3d-shader/dxbc.c index 3be3ca917ac3..5daf07a65f00 100644 --- a/libs/vkd3d-shader/dxbc.c +++ b/libs/vkd3d-shader/dxbc.c @@ -2040,7 +2040,7 @@ static int shdr_handler(const char *data, DWORD data_size, DWORD tag, void *cont break; default: - FIXME("Unhandled chunk %#x.\n", tag); + TRACE("Skipping chunk %#x.\n", tag); break; } -- 2.16.4

2 1

[PATCH vkd3d 3/8] tests: Add test for clearing 2D array DSVs.
by Józef Kucia Sept. 28, 2018

Sept. 28, 2018

From: Józef Kucia <jkucia(a)codeweavers.com> Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- tests/d3d12.c | 59 +++++++++++++++++++++++++++++++++++++++++++++++------------ 1 file changed, 47 insertions(+), 12 deletions(-) diff --git a/tests/d3d12.c b/tests/d3d12.c index 83010674d5cf..0528ceb4ba6a 100644 --- a/tests/d3d12.c +++ b/tests/d3d12.c @@ -1058,9 +1058,9 @@ struct depth_stencil_resource D3D12_CPU_DESCRIPTOR_HANDLE dsv_handle; }; -#define init_depth_stencil(a, b, c, d, e, f, g) init_depth_stencil_(__LINE__, a, b, c, d, e, f, g) +#define init_depth_stencil(a, b, c, d, e, f, g, h) init_depth_stencil_(__LINE__, a, b, c, d, e, f, g, h) static void init_depth_stencil_(unsigned int line, struct depth_stencil_resource *ds, - ID3D12Device *device, unsigned int width, unsigned int height, + ID3D12Device *device, unsigned int width, unsigned int height, unsigned int array_size, DXGI_FORMAT format, DXGI_FORMAT view_format, const D3D12_CLEAR_VALUE *clear_value) { D3D12_DEPTH_STENCIL_VIEW_DESC dsv_desc, *view_desc; @@ -1078,7 +1078,7 @@ static void init_depth_stencil_(unsigned int line, struct depth_stencil_resource resource_desc.Alignment = 0; resource_desc.Width = width; resource_desc.Height = height; - resource_desc.DepthOrArraySize = 1; + resource_desc.DepthOrArraySize = array_size; resource_desc.MipLevels = 1; resource_desc.Format = format; resource_desc.SampleDesc.Count = 1; @@ -3289,7 +3289,9 @@ static void test_multithread_fence_wait(void) static void test_clear_depth_stencil_view(void) { + static const float expected_values[] = {0.5f, 0.1f, 0.1f, 0.6, 1.0f, 0.5f}; ID3D12GraphicsCommandList *command_list; + D3D12_DEPTH_STENCIL_VIEW_DESC dsv_desc; struct depth_stencil_resource ds; unsigned int dsv_increment_size; D3D12_CLEAR_VALUE clear_value; @@ -3297,6 +3299,7 @@ static void test_clear_depth_stencil_view(void) struct test_context context; ID3D12CommandQueue *queue; ID3D12Device *device; + unsigned int i; memset(&desc, 0, sizeof(desc)); desc.no_render_target = true; @@ -3314,13 +3317,45 @@ static void test_clear_depth_stencil_view(void) clear_value.Format = DXGI_FORMAT_D32_FLOAT; clear_value.DepthStencil.Depth = 0.5f; clear_value.DepthStencil.Stencil = 0x3; - init_depth_stencil(&ds, device, 32, 32, DXGI_FORMAT_D32_FLOAT, 0, &clear_value); + init_depth_stencil(&ds, device, 32, 32, 1, DXGI_FORMAT_D32_FLOAT, 0, &clear_value); ID3D12GraphicsCommandList_ClearDepthStencilView(command_list, ds.dsv_handle, D3D12_CLEAR_FLAG_DEPTH | D3D12_CLEAR_FLAG_STENCIL, 0.75f, 0x7, 0, NULL); transition_resource_state(command_list, ds.texture, D3D12_RESOURCE_STATE_DEPTH_WRITE, D3D12_RESOURCE_STATE_COPY_SOURCE); - check_sub_resource_uint(ds.texture, 0, queue, command_list, 0x3f400000, 0); + check_sub_resource_float(ds.texture, 0, queue, command_list, 0.75f, 1); + + destroy_depth_stencil(&ds); + reset_command_list(command_list, context.allocator); + init_depth_stencil(&ds, device, 32, 32, 6, DXGI_FORMAT_D32_FLOAT, 0, &clear_value); + + ID3D12GraphicsCommandList_ClearDepthStencilView(command_list, ds.dsv_handle, + D3D12_CLEAR_FLAG_DEPTH, expected_values[0], 0, 0, NULL); + memset(&dsv_desc, 0, sizeof(dsv_desc)); + dsv_desc.Format = DXGI_FORMAT_D32_FLOAT; + dsv_desc.ViewDimension = D3D12_DSV_DIMENSION_TEXTURE2DARRAY; + dsv_desc.Texture2DArray.FirstArraySlice = 1; + dsv_desc.Texture2DArray.ArraySize = 2; + ID3D12Device_CreateDepthStencilView(device, ds.texture, &dsv_desc, ds.dsv_handle); + ID3D12GraphicsCommandList_ClearDepthStencilView(command_list, ds.dsv_handle, + D3D12_CLEAR_FLAG_DEPTH, expected_values[1], 0, 0, NULL); + dsv_desc.Texture2DArray.FirstArraySlice = 3; + dsv_desc.Texture2DArray.ArraySize = 1; + ID3D12Device_CreateDepthStencilView(device, ds.texture, &dsv_desc, ds.dsv_handle); + ID3D12GraphicsCommandList_ClearDepthStencilView(command_list, ds.dsv_handle, + D3D12_CLEAR_FLAG_DEPTH, expected_values[3], 0, 0, NULL); + dsv_desc.Texture2DArray.FirstArraySlice = 4; + ID3D12Device_CreateDepthStencilView(device, ds.texture, &dsv_desc, ds.dsv_handle); + ID3D12GraphicsCommandList_ClearDepthStencilView(command_list, ds.dsv_handle, + D3D12_CLEAR_FLAG_DEPTH, expected_values[4], 0, 0, NULL); + + transition_resource_state(command_list, ds.texture, + D3D12_RESOURCE_STATE_DEPTH_WRITE, D3D12_RESOURCE_STATE_COPY_SOURCE); + for (i = 0; i < ARRAY_SIZE(expected_values); ++i) + { + check_sub_resource_float(ds.texture, i, queue, command_list, expected_values[i], 1); + reset_command_list(command_list, context.allocator); + } destroy_depth_stencil(&ds); destroy_test_context(&context); @@ -4546,7 +4581,7 @@ static void test_draw_depth_only(void) command_list = context.list; queue = context.queue; - init_depth_stencil(&ds, context.device, 640, 480, DXGI_FORMAT_D32_FLOAT, 0, NULL); + init_depth_stencil(&ds, context.device, 640, 480, 1, DXGI_FORMAT_D32_FLOAT, 0, NULL); set_viewport(&context.viewport, 0.0f, 0.0f, 640.0f, 480.0f, 0.0f, 1.0f); set_rect(&context.scissor_rect, 0, 0, 640, 480); @@ -12156,7 +12191,7 @@ static void test_copy_descriptors(void) transition_resource_state(command_list, t[5], D3D12_RESOURCE_STATE_COPY_DEST, D3D12_RESOURCE_STATE_NON_PIXEL_SHADER_RESOURCE); - init_depth_stencil(&ds, device, 32, 32, DXGI_FORMAT_R32_TYPELESS, DXGI_FORMAT_D32_FLOAT, NULL); + init_depth_stencil(&ds, device, 32, 32, 1, DXGI_FORMAT_R32_TYPELESS, DXGI_FORMAT_D32_FLOAT, NULL); t[6] = ds.texture; ID3D12Resource_AddRef(t[6]); ID3D12GraphicsCommandList_ClearDepthStencilView(command_list, ds.dsv_handle, @@ -13509,7 +13544,7 @@ static void test_depth_stencil_sampling(void) reset_command_list(command_list, context.allocator); init_depth_stencil(&ds, device, context.render_target_desc.Width, - context.render_target_desc.Height, tests[i].typeless_format, + context.render_target_desc.Height, 1, tests[i].typeless_format, tests[i].dsv_format, NULL); texture = ds.texture; dsv_handle = ds.dsv_handle; @@ -13722,7 +13757,7 @@ static void test_depth_load(void) heap = create_gpu_descriptor_heap(device, D3D12_DESCRIPTOR_HEAP_TYPE_CBV_SRV_UAV, 2); init_depth_stencil(&ds, device, context.render_target_desc.Width, - context.render_target_desc.Height, DXGI_FORMAT_R32_TYPELESS, + context.render_target_desc.Height, 1, DXGI_FORMAT_R32_TYPELESS, DXGI_FORMAT_D32_FLOAT, NULL); memset(&srv_desc, 0, sizeof(srv_desc)); srv_desc.Format = DXGI_FORMAT_R32_FLOAT; @@ -13854,7 +13889,7 @@ static void test_depth_read_only_view(void) clear_value.DepthStencil.Depth = 0.5f; clear_value.DepthStencil.Stencil = 0; init_depth_stencil(&ds, device, context.render_target_desc.Width, - context.render_target_desc.Height, DXGI_FORMAT_R32_TYPELESS, + context.render_target_desc.Height, 1, DXGI_FORMAT_R32_TYPELESS, DXGI_FORMAT_D32_FLOAT, &clear_value); memset(&dsv_desc, 0, sizeof(dsv_desc)); dsv_desc.Format = DXGI_FORMAT_D32_FLOAT; @@ -16705,7 +16740,7 @@ static void test_query_occlusion(void) command_list = context.list; queue = context.queue; - init_depth_stencil(&ds, context.device, 640, 480, DXGI_FORMAT_D32_FLOAT, 0, NULL); + init_depth_stencil(&ds, context.device, 640, 480, 1, DXGI_FORMAT_D32_FLOAT, 0, NULL); set_viewport(&context.viewport, 0.0f, 0.0f, 640.0f, 480.0f, 0.0f, 1.0f); set_rect(&context.scissor_rect, 0, 0, 640, 480); @@ -17632,7 +17667,7 @@ static void test_copy_texture_region(void) for (i = 0; i < ARRAY_SIZE(depth_values); ++i) { init_depth_stencil(&ds, device, context.render_target_desc.Width, - context.render_target_desc.Height, DXGI_FORMAT_D32_FLOAT, + context.render_target_desc.Height, 1, DXGI_FORMAT_D32_FLOAT, DXGI_FORMAT_D32_FLOAT, NULL); ID3D12GraphicsCommandList_ClearDepthStencilView(command_list, ds.dsv_handle, D3D12_CLEAR_FLAG_DEPTH, depth_values[i], 0, 0, NULL); -- 2.16.4

2 1

[PATCH vkd3d 2/8] vkd3d: Fix 2D array depth stencil views.
by Józef Kucia Sept. 28, 2018

Sept. 28, 2018

From: Józef Kucia <jkucia(a)codeweavers.com> Signed-off-by: Józef Kucia <jkucia(a)codeweavers.com> --- libs/vkd3d/command.c | 5 +++-- libs/vkd3d/resource.c | 1 + libs/vkd3d/vkd3d_private.h | 1 + 3 files changed, 5 insertions(+), 2 deletions(-) diff --git a/libs/vkd3d/command.c b/libs/vkd3d/command.c index 5fd093617e3b..b4d8ecd75179 100644 --- a/libs/vkd3d/command.c +++ b/libs/vkd3d/command.c @@ -3681,7 +3681,7 @@ static void STDMETHODCALLTYPE d3d12_command_list_OMSetRenderTargets(ID3D12Graphi list->views[0] = view->u.vk_image_view; list->fb_width = max(list->fb_width, dsv_desc->width); list->fb_height = max(list->fb_height, dsv_desc->height); - list->fb_layer_count = max(list->fb_layer_count, 1); + list->fb_layer_count = max(list->fb_layer_count, dsv_desc->layer_count); } d3d12_command_list_invalidate_current_framebuffer(list); @@ -3841,7 +3841,8 @@ static void STDMETHODCALLTYPE d3d12_command_list_ClearDepthStencilView(ID3D12Gra ds_reference.layout = VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL; d3d12_command_list_clear(list, &attachment_desc, NULL, &ds_reference, - dsv_desc->view, dsv_desc->width, dsv_desc->height, 1, &clear_value, rect_count, rects); + dsv_desc->view, dsv_desc->width, dsv_desc->height, dsv_desc->layer_count, + &clear_value, rect_count, rects); } static void STDMETHODCALLTYPE d3d12_command_list_ClearRenderTargetView(ID3D12GraphicsCommandList *iface, diff --git a/libs/vkd3d/resource.c b/libs/vkd3d/resource.c index 98c703a9313d..d6330941127a 100644 --- a/libs/vkd3d/resource.c +++ b/libs/vkd3d/resource.c @@ -2131,6 +2131,7 @@ void d3d12_dsv_desc_create_dsv(struct d3d12_dsv_desc *dsv_desc, struct d3d12_dev dsv_desc->format = vkd3d_desc.format->vk_format; dsv_desc->width = d3d12_resource_desc_get_width(&resource->desc, vkd3d_desc.miplevel_idx); dsv_desc->height = d3d12_resource_desc_get_height(&resource->desc, vkd3d_desc.miplevel_idx); + dsv_desc->layer_count = vkd3d_desc.layer_count; dsv_desc->view = view; dsv_desc->resource = resource; } diff --git a/libs/vkd3d/vkd3d_private.h b/libs/vkd3d/vkd3d_private.h index bffae35fb830..d744af5a5bac 100644 --- a/libs/vkd3d/vkd3d_private.h +++ b/libs/vkd3d/vkd3d_private.h @@ -358,6 +358,7 @@ struct d3d12_dsv_desc VkFormat format; uint64_t width; unsigned int height; + unsigned int layer_count; struct vkd3d_view *view; struct d3d12_resource *resource; }; -- 2.16.4

2 1

[PATCH] wined3d: Use query buffer objects for occlusion queries.
by Henri Verbeet Sept. 28, 2018

Sept. 28, 2018

From: Andrew Wesie <awesie(a)gmail.com> Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=45723 Signed-off-by: Andrew Wesie <awesie(a)gmail.com> Signed-off-by: Henri Verbeet <hverbeet(a)codeweavers.com> --- dlls/wined3d/adapter_gl.c | 6 +++ dlls/wined3d/cs.c | 5 +- dlls/wined3d/query.c | 106 +++++++++++++++++++++++++++++++++++++++-- dlls/wined3d/wined3d_gl.h | 2 + dlls/wined3d/wined3d_private.h | 3 ++ 5 files changed, 116 insertions(+), 6 deletions(-) diff --git a/dlls/wined3d/adapter_gl.c b/dlls/wined3d/adapter_gl.c index 87418e1e44e..81c84244764 100644 --- a/dlls/wined3d/adapter_gl.c +++ b/dlls/wined3d/adapter_gl.c @@ -58,6 +58,7 @@ static const struct wined3d_extension_map gl_extension_map[] = /* ARB */ {"GL_ARB_base_instance", ARB_BASE_INSTANCE }, {"GL_ARB_blend_func_extended", ARB_BLEND_FUNC_EXTENDED }, + {"GL_ARB_buffer_storage", ARB_BUFFER_STORAGE }, {"GL_ARB_clear_buffer_object", ARB_CLEAR_BUFFER_OBJECT }, {"GL_ARB_clear_texture", ARB_CLEAR_TEXTURE }, {"GL_ARB_clip_control", ARB_CLIP_CONTROL }, @@ -103,6 +104,7 @@ static const struct wined3d_extension_map gl_extension_map[] = {"GL_ARB_point_parameters", ARB_POINT_PARAMETERS }, {"GL_ARB_point_sprite", ARB_POINT_SPRITE }, {"GL_ARB_provoking_vertex", ARB_PROVOKING_VERTEX }, + {"GL_ARB_query_buffer_object", ARB_QUERY_BUFFER_OBJECT }, {"GL_ARB_sample_shading", ARB_SAMPLE_SHADING }, {"GL_ARB_sampler_objects", ARB_SAMPLER_OBJECTS }, {"GL_ARB_seamless_cube_map", ARB_SEAMLESS_CUBE_MAP }, @@ -2061,6 +2063,8 @@ static void load_gl_funcs(struct wined3d_gl_info *gl_info) /* GL_ARB_blend_func_extended */ USE_GL_FUNC(glBindFragDataLocationIndexed) USE_GL_FUNC(glGetFragDataIndex) + /* GL_ARB_buffer_storage */ + USE_GL_FUNC(glBufferStorage) /* GL_ARB_clear_buffer_object */ USE_GL_FUNC(glClearBufferData) USE_GL_FUNC(glClearBufferSubData) @@ -3301,7 +3305,9 @@ static BOOL wined3d_adapter_init_gl_caps(struct wined3d_adapter *adapter, {ARB_TEXTURE_STORAGE_MULTISAMPLE, MAKEDWORD_VERSION(4, 2)}, {ARB_TEXTURE_VIEW, MAKEDWORD_VERSION(4, 3)}, + {ARB_BUFFER_STORAGE, MAKEDWORD_VERSION(4, 4)}, {ARB_CLEAR_TEXTURE, MAKEDWORD_VERSION(4, 4)}, + {ARB_QUERY_BUFFER_OBJECT, MAKEDWORD_VERSION(4, 4)}, {ARB_CLIP_CONTROL, MAKEDWORD_VERSION(4, 5)}, {ARB_CULL_DISTANCE, MAKEDWORD_VERSION(4, 5)}, diff --git a/dlls/wined3d/cs.c b/dlls/wined3d/cs.c index ff4d562f106..2aae559a6cb 100644 --- a/dlls/wined3d/cs.c +++ b/dlls/wined3d/cs.c @@ -1986,7 +1986,10 @@ static void wined3d_cs_exec_query_issue(struct wined3d_cs *cs, const void *data) if (poll && list_empty(&query->poll_list_entry)) { - list_add_tail(&cs->query_poll_list, &query->poll_list_entry); + if (query->buffer_object) + InterlockedIncrement(&query->counter_retrieved); + else + list_add_tail(&cs->query_poll_list, &query->poll_list_entry); return; } diff --git a/dlls/wined3d/query.c b/dlls/wined3d/query.c index 01e6bcb9905..26c32b1f13b 100644 --- a/dlls/wined3d/query.c +++ b/dlls/wined3d/query.c @@ -25,6 +25,88 @@ WINE_DEFAULT_DEBUG_CHANNEL(d3d); +static void wined3d_query_buffer_invalidate(struct wined3d_query *query) +{ + /* map[0] != map[1]: exact values do not have any significance. */ + query->map_ptr[0] = 0; + query->map_ptr[1] = ~(UINT64)0; +} + +static BOOL wined3d_query_buffer_is_valid(struct wined3d_query *query) +{ + return query->map_ptr[0] == query->map_ptr[1]; +} + +static void wined3d_query_create_buffer_object(struct wined3d_context *context, struct wined3d_query *query) +{ + const struct wined3d_gl_info *gl_info = context->gl_info; + const GLuint map_flags = GL_MAP_READ_BIT | GL_MAP_WRITE_BIT | GL_MAP_PERSISTENT_BIT | GL_MAP_COHERENT_BIT; + GLuint buffer_object; + + GL_EXTCALL(glGenBuffers(1, &buffer_object)); + GL_EXTCALL(glBindBuffer(GL_QUERY_BUFFER, buffer_object)); + GL_EXTCALL(glBufferStorage(GL_QUERY_BUFFER, sizeof(query->map_ptr[0]) * 2, NULL, map_flags)); + query->map_ptr = GL_EXTCALL(glMapBufferRange(GL_QUERY_BUFFER, 0, sizeof(query->map_ptr[0]) * 2, map_flags)); + GL_EXTCALL(glBindBuffer(GL_QUERY_BUFFER, 0)); + checkGLcall("query buffer object creation"); + + wined3d_query_buffer_invalidate(query); + query->buffer_object = buffer_object; +} + +static void wined3d_query_destroy_buffer_object(struct wined3d_context *context, struct wined3d_query *query) +{ + const struct wined3d_gl_info *gl_info = context->gl_info; + + GL_EXTCALL(glDeleteBuffers(1, &query->buffer_object)); + checkGLcall("query buffer object destruction"); + + query->buffer_object = 0; + query->map_ptr = NULL; +} + +/* From ARB_occlusion_query: "Querying the state for a given occlusion query + * forces that occlusion query to complete within a finite amount of time." + * In practice, that means drivers flush when retrieving + * GL_QUERY_RESULT_AVAILABLE, which can be undesirable when applications use a + * significant number of queries. Using a persistently mapped query buffer + * object allows us to avoid these implicit flushes. An additional benefit is + * that it allows us to poll the query status from the application-thread + * instead of from the csmt-thread. */ +static BOOL wined3d_query_buffer_queue_result(struct wined3d_context *context, struct wined3d_query *query, GLuint id) +{ + const struct wined3d_gl_info *gl_info = context->gl_info; + + if (!gl_info->supported[ARB_QUERY_BUFFER_OBJECT] || !gl_info->supported[ARB_BUFFER_STORAGE]) + return FALSE; + /* Don't use query buffers without CSMT, mainly for simplicity. */ + if (!context->device->cs->thread) + return FALSE; + + if (query->buffer_object) + { + /* If there's still a query result in-flight for the existing buffer + * object (i.e., the query was restarted before we received its + * result), we can't reuse the existing buffer object. */ + if (wined3d_query_buffer_is_valid(query)) + wined3d_query_buffer_invalidate(query); + else + wined3d_query_destroy_buffer_object(context, query); + } + + if (!query->buffer_object) + wined3d_query_create_buffer_object(context, query); + + GL_EXTCALL(glBindBuffer(GL_QUERY_BUFFER, query->buffer_object)); + /* Read the same value twice. We know we have the result if map_ptr[0] == map_ptr[1]. */ + GL_EXTCALL(glGetQueryObjectui64v(id, GL_QUERY_RESULT, (void *)0)); + GL_EXTCALL(glGetQueryObjectui64v(id, GL_QUERY_RESULT, (void *)sizeof(query->map_ptr[0]))); + GL_EXTCALL(glBindBuffer(GL_QUERY_BUFFER, 0)); + checkGLcall("queue query result"); + + return TRUE; +} + static UINT64 get_query_result64(GLuint id, const struct wined3d_gl_info *gl_info) { if (gl_info->supported[ARB_TIMER_QUERY]) @@ -339,6 +421,14 @@ static void wined3d_query_destroy_object(void *object) if (!list_empty(&query->poll_list_entry)) list_remove(&query->poll_list_entry); + if (query->buffer_object) + { + struct wined3d_context *context; + context = context_acquire(query->device, NULL, 0); + wined3d_query_destroy_buffer_object(context, query); + context_release(context); + } + /* Queries are specific to the GL context that created them. Not * deleting the query will obviously leak it, but that's still better * than potentially deleting a different query with the same id in this @@ -382,15 +472,20 @@ HRESULT CDECL wined3d_query_get_data(struct wined3d_query *query, return WINED3DERR_INVALIDCALL; } - if (!query->device->cs->thread) + if (query->device->cs->thread) { - if (!query->query_ops->query_poll(query, flags)) + if (query->counter_main != query->counter_retrieved + || (query->buffer_object && !wined3d_query_buffer_is_valid(query))) + { + if (flags & WINED3DGETDATA_FLUSH && !query->device->cs->queries_flushed) + wined3d_cs_emit_flush(query->device->cs); return S_FALSE; + } + if (query->buffer_object) + query->data = query->map_ptr; } - else if (query->counter_main != query->counter_retrieved) + else if (!query->query_ops->query_poll(query, flags)) { - if (flags & WINED3DGETDATA_FLUSH && !query->device->cs->queries_flushed) - wined3d_cs_emit_flush(query->device->cs); return S_FALSE; } @@ -579,6 +674,7 @@ static BOOL wined3d_occlusion_query_ops_issue(struct wined3d_query *query, DWORD gl_info = context->gl_info; GL_EXTCALL(glEndQuery(GL_SAMPLES_PASSED)); checkGLcall("glEndQuery()"); + wined3d_query_buffer_queue_result(context, query, oq->id); context_release(context); poll = TRUE; diff --git a/dlls/wined3d/wined3d_gl.h b/dlls/wined3d/wined3d_gl.h index 06aa21c33fa..2a985da9faf 100644 --- a/dlls/wined3d/wined3d_gl.h +++ b/dlls/wined3d/wined3d_gl.h @@ -44,6 +44,7 @@ enum wined3d_gl_extension /* ARB */ ARB_BASE_INSTANCE, ARB_BLEND_FUNC_EXTENDED, + ARB_BUFFER_STORAGE, ARB_CLEAR_BUFFER_OBJECT, ARB_CLEAR_TEXTURE, ARB_CLIP_CONTROL, @@ -89,6 +90,7 @@ enum wined3d_gl_extension ARB_POINT_PARAMETERS, ARB_POINT_SPRITE, ARB_PROVOKING_VERTEX, + ARB_QUERY_BUFFER_OBJECT, ARB_SAMPLE_SHADING, ARB_SAMPLER_OBJECTS, ARB_SEAMLESS_CUBE_MAP, diff --git a/dlls/wined3d/wined3d_private.h b/dlls/wined3d/wined3d_private.h index 941bfa68fad..8cfa4c37ef8 100644 --- a/dlls/wined3d/wined3d_private.h +++ b/dlls/wined3d/wined3d_private.h @@ -1730,6 +1730,9 @@ struct wined3d_query LONG counter_main, counter_retrieved; struct list poll_list_entry; + + GLuint buffer_object; + UINT64 *map_ptr; }; struct wined3d_event_query -- 2.11.0

1 0

[PATCH 1/2] msvcrt: Use isnan instead of isnanf
by Alex Henrie Sept. 28, 2018

Sept. 28, 2018

This resolves several warnings when compiling with MinGW because isnanf is not in MinGW's math.h (it's not a standard C function). On the other hand, isnan is a widely available C99 macro designed for both floats and doubles. We also have an isnan implementation in libs/port/isnan.c for pre-C99 compilers. Signed-off-by: Alex Henrie <alexhenrie24(a)gmail.com> --- configure.ac | 1 - dlls/msvcrt/math.c | 30 +++++++++++------------------- 2 files changed, 11 insertions(+), 20 deletions(-) diff --git a/configure.ac b/configure.ac index 6936a71b6f..2cc481630c 100644 --- a/configure.ac +++ b/configure.ac @@ -2163,7 +2163,6 @@ AC_CHECK_FUNCS(\ getpwuid \ gettimeofday \ getuid \ - isnanf \ kqueue \ lstat \ memmove \ diff --git a/dlls/msvcrt/math.c b/dlls/msvcrt/math.c index a1ba552632..dbfaa24756 100644 --- a/dlls/msvcrt/math.c +++ b/dlls/msvcrt/math.c @@ -38,14 +38,6 @@ WINE_DEFAULT_DEBUG_CHANNEL(msvcrt); #define finitef(x) isfinite(x) #endif -#ifndef HAVE_ISNANF -#ifdef HAVE_ISNAN -#define isnanf(x) isnan(x) -#else -#define isnanf(x) 0 -#endif -#endif - /* FIXME: Does not work with -NAN and -0. */ #ifndef signbit #define signbit(x) ((x) < 0) @@ -190,7 +182,7 @@ INT CDECL MSVCRT__isnanf( float num ) /* Some implementations return -1 for true(glibc), msvcrt/crtdll return 1. * Do the same, as the result may be used in calculations */ - return isnanf(num) != 0; + return isnan(num) != 0; } /********************************************************************* @@ -199,7 +191,7 @@ INT CDECL MSVCRT__isnanf( float num ) float CDECL MSVCRT__logbf( float num ) { float ret = logbf(num); - if (isnanf(num)) math_error(_DOMAIN, "_logbf", num, 0, ret); + if (isnan(num)) math_error(_DOMAIN, "_logbf", num, 0, ret); else if (!num) math_error(_SING, "_logbf", num, 0, ret); return ret; } @@ -245,7 +237,7 @@ float CDECL MSVCRT_atanf( float x ) float CDECL MSVCRT_atan2f( float x, float y ) { float ret = atan2f(x, y); - if (isnanf(x)) math_error(_DOMAIN, "atan2f", x, y, ret); + if (isnan(x)) math_error(_DOMAIN, "atan2f", x, y, ret); return ret; } @@ -265,7 +257,7 @@ float CDECL MSVCRT_cosf( float x ) float CDECL MSVCRT_coshf( float x ) { float ret = coshf(x); - if (isnanf(x)) math_error(_DOMAIN, "coshf", x, 0, ret); + if (isnan(x)) math_error(_DOMAIN, "coshf", x, 0, ret); return ret; } @@ -275,7 +267,7 @@ float CDECL MSVCRT_coshf( float x ) float CDECL MSVCRT_expf( float x ) { float ret = expf(x); - if (isnanf(x)) math_error(_DOMAIN, "expf", x, 0, ret); + if (isnan(x)) math_error(_DOMAIN, "expf", x, 0, ret); else if (finitef(x) && !ret) math_error(_UNDERFLOW, "expf", x, 0, ret); else if (finitef(x) && !finitef(ret)) math_error(_OVERFLOW, "expf", x, 0, ret); return ret; @@ -342,7 +334,7 @@ float CDECL MSVCRT_sinf( float x ) float CDECL MSVCRT_sinhf( float x ) { float ret = sinhf(x); - if (isnanf(x)) math_error(_DOMAIN, "sinhf", x, 0, ret); + if (isnan(x)) math_error(_DOMAIN, "sinhf", x, 0, ret); return ret; } @@ -2943,9 +2935,9 @@ LDOUBLE CDECL MSVCR120_erfcl(LDOUBLE x) */ float CDECL MSVCR120_fmaxf(float x, float y) { - if(isnanf(x)) + if(isnan(x)) return y; - if(isnanf(y)) + if(isnan(y)) return x; if(x==0 && y==0) return signbit(x) ? y : x; @@ -3008,9 +3000,9 @@ int CDECL MSVCR120__fdpcomp(float x, float y) */ float CDECL MSVCR120_fminf(float x, float y) { - if(isnanf(x)) + if(isnan(x)) return y; - if(isnanf(y)) + if(isnan(y)) return x; if(x==0 && y==0) return signbit(x) ? x : y; @@ -3232,7 +3224,7 @@ float CDECL MSVCR120_remainderf(float x, float y) #ifdef HAVE_REMAINDERF /* this matches 64-bit Windows. 32-bit Windows is slightly different */ if(!finitef(x)) *MSVCRT__errno() = MSVCRT_EDOM; - if(isnanf(y) || y==0.0f) *MSVCRT__errno() = MSVCRT_EDOM; + if(isnan(y) || y==0.0f) *MSVCRT__errno() = MSVCRT_EDOM; return remainderf(x, y); #else FIXME( "not implemented\n" ); -- 2.19.0

3 3