-- v3: d3d11/tests: Teste UpdateSubresource() for NV12 textures. d3d11/tests: Test without render target. tests/d3d11: Do not check pitches. d3d11/tests: Check for NV12 texture support before testing them. d3d11/tests: Check the result of compiling HLSL shaders.
From: Giovanni Mascellani gmascellani@codeweavers.com
--- dlls/d3d11/tests/d3d11.c | 3 +++ 1 file changed, 3 insertions(+)
diff --git a/dlls/d3d11/tests/d3d11.c b/dlls/d3d11/tests/d3d11.c index c8112cb8442..7053d32f0db 100644 --- a/dlls/d3d11/tests/d3d11.c +++ b/dlls/d3d11/tests/d3d11.c @@ -36290,16 +36290,19 @@ static void test_nv12(void) bytecode = compile_shader(cs_code, sizeof(cs_code) - 1, "cs_5_0"); hr = ID3D11Device_CreateComputeShader(device, ID3D10Blob_GetBufferPointer(bytecode), ID3D10Blob_GetBufferSize(bytecode), NULL, &cs); + ok(hr == S_OK, "Got hr %#lx.\n", hr); ID3D10Blob_Release(bytecode);
bytecode = compile_shader(luma_ps_code, sizeof(luma_ps_code) - 1, "ps_4_0"); hr = ID3D11Device_CreatePixelShader(device, ID3D10Blob_GetBufferPointer(bytecode), ID3D10Blob_GetBufferSize(bytecode), NULL, &luma_ps); + ok(hr == S_OK, "Got hr %#lx.\n", hr); ID3D10Blob_Release(bytecode);
bytecode = compile_shader(chroma_ps_code, sizeof(chroma_ps_code) - 1, "ps_4_0"); hr = ID3D11Device_CreatePixelShader(device, ID3D10Blob_GetBufferPointer(bytecode), ID3D10Blob_GetBufferSize(bytecode), NULL, &chroma_ps); + ok(hr == S_OK, "Got hr %#lx.\n", hr); ID3D10Blob_Release(bytecode);
for (test_idx = 0; test_idx < ARRAY_SIZE(tests); ++test_idx)
From: Giovanni Mascellani gmascellani@codeweavers.com
--- dlls/d3d11/tests/d3d11.c | 11 +++++++++++ 1 file changed, 11 insertions(+)
diff --git a/dlls/d3d11/tests/d3d11.c b/dlls/d3d11/tests/d3d11.c index 7053d32f0db..43b8f1494e8 100644 --- a/dlls/d3d11/tests/d3d11.c +++ b/dlls/d3d11/tests/d3d11.c @@ -36225,6 +36225,7 @@ static void test_nv12(void) unsigned int test_idx; ID3D10Blob *bytecode; ID3D11Device *device; + UINT support; HRESULT hr;
static const uint32_t clear_values[4] = {0xabcdef00, 0xabcdef00, 0xabcdef00, 0xabcdef00}; @@ -36287,6 +36288,16 @@ static void test_nv12(void) device = test_context.device; device_context = test_context.immediate_context;
+ hr = ID3D11Device_CheckFormatSupport(device, DXGI_FORMAT_NV12, &support); + ok(hr == S_OK, "Got hr %#lx.\n", hr); + + if (!(support & D3D11_FORMAT_SUPPORT_TEXTURE2D)) + { + skip("NV12 textures are not supported.\n"); + release_test_context(&test_context); + return; + } + bytecode = compile_shader(cs_code, sizeof(cs_code) - 1, "cs_5_0"); hr = ID3D11Device_CreateComputeShader(device, ID3D10Blob_GetBufferPointer(bytecode), ID3D10Blob_GetBufferSize(bytecode), NULL, &cs);
From: Giovanni Mascellani gmascellani@codeweavers.com
Implementations might use different values. For example, AMD GPUs seem to use 768 here. --- dlls/d3d11/tests/d3d11.c | 24 ++++++++---------------- 1 file changed, 8 insertions(+), 16 deletions(-)
diff --git a/dlls/d3d11/tests/d3d11.c b/dlls/d3d11/tests/d3d11.c index 43b8f1494e8..46bdff09bca 100644 --- a/dlls/d3d11/tests/d3d11.c +++ b/dlls/d3d11/tests/d3d11.c @@ -1169,11 +1169,6 @@ static void check_readback_data_u8_with_buffer_(unsigned int line, struct resour unsigned int x = 0, y = 0, z = 0; BOOL all_match = FALSE;
- ok_(__FILE__, line)(rb->map_desc.RowPitch == depth_pitch, "Got row pitch %u instead of %u.\n", - rb->map_desc.RowPitch, depth_pitch); - ok_(__FILE__, line)(rb->map_desc.DepthPitch == slice_pitch, "Got depth pitch %u instead of %u.\n", - rb->map_desc.DepthPitch, depth_pitch); - for (z = 0; z < rb->depth; ++z) { for (y = 0; y < rb->height; ++y) @@ -36319,7 +36314,6 @@ static void test_nv12(void) for (test_idx = 0; test_idx < ARRAY_SIZE(tests); ++test_idx) { /* I need only two uints in the cbuffer, but the size must be a multiple of 16. */ - uint32_t cbuffer_data[4], expected_row_pitch, expected_depth_pitch; D3D11_UNORDERED_ACCESS_VIEW_DESC uav_desc = {0}; D3D11_SHADER_RESOURCE_VIEW_DESC srv_desc = {0}; D3D11_SUBRESOURCE_DATA subresource_data = {0}; @@ -36330,6 +36324,7 @@ static void test_nv12(void) ID3D11ShaderResourceView *srvs[2]; D3D11_TEXTURE2D_DESC desc = {0}; struct resource_readback rb; + uint32_t cbuffer_data[4]; ID3D11Buffer *cbuffer; HRESULT expected_hr; unsigned int i, j; @@ -36340,9 +36335,6 @@ static void test_nv12(void)
winetest_push_context("test %u (%ux%u)", test_idx, width, height);
- expected_row_pitch = (width + 3) & ~3; - expected_depth_pitch = expected_row_pitch * height * 3 / 2; - desc.Width = width; desc.Height = height; desc.MipLevels = 1; @@ -36352,14 +36344,14 @@ static void test_nv12(void) desc.Usage = D3D11_USAGE_DEFAULT; desc.BindFlags = D3D11_BIND_SHADER_RESOURCE | D3D11_BIND_RENDER_TARGET;
- content = calloc(expected_depth_pitch, 1); + content = calloc(width * height * 3 / 2, 1); ok(!!content, "Failed to allocate memory.\n");
for (i = 0; i < height; ++i) { for (j = 0; j < width; ++j) { - unsigned int idx = i * expected_row_pitch + j; + unsigned int idx = i * width + j;
content[idx] = (i & 7) << 3 | (j & 7); } @@ -36369,7 +36361,7 @@ static void test_nv12(void) { for (j = 0; j < width / 2; ++j) { - unsigned int idx = expected_row_pitch * (height + i) + j * 2; + unsigned int idx = width * (height + i) + j * 2;
content[idx] = 1 << 6 | (i & 7) << 3 | (j & 7); content[idx + 1] = 1 << 7 | (i & 7) << 3 | (j & 7); @@ -36377,8 +36369,8 @@ static void test_nv12(void) }
subresource_data.pSysMem = content; - subresource_data.SysMemPitch = expected_row_pitch; - subresource_data.SysMemSlicePitch = expected_depth_pitch; + subresource_data.SysMemPitch = width; + subresource_data.SysMemSlicePitch = 0;
expected_hr = (width & 1 || height & 1) ? E_INVALIDARG : S_OK; hr = ID3D11Device_CreateTexture2D(device, &desc, &subresource_data, &texture); @@ -36432,7 +36424,7 @@ static void test_nv12(void) ID3D11DeviceContext_Dispatch(device_context, width, height, 1);
get_texture_readback(check_texture, 0, &rb); - check_readback_data_u8_with_buffer(&rb, content, expected_row_pitch, expected_depth_pitch); + check_readback_data_u8_with_buffer(&rb, content, width, 0); release_resource_readback(&rb);
rtv_desc.Format = DXGI_FORMAT_R8_UINT; @@ -36469,7 +36461,7 @@ static void test_nv12(void) ID3D11DeviceContext_Dispatch(device_context, width, height, 1);
get_texture_readback(check_texture, 0, &rb); - check_readback_data_u8_with_buffer(&rb, content, expected_row_pitch, expected_depth_pitch); + check_readback_data_u8_with_buffer(&rb, content, width, 0); release_resource_readback(&rb);
ID3D11RenderTargetView_Release(rtv2);
From: Giovanni Mascellani gmascellani@codeweavers.com
--- dlls/d3d11/tests/d3d11.c | 31 ++++++++++++++++++++++++++++++- 1 file changed, 30 insertions(+), 1 deletion(-)
diff --git a/dlls/d3d11/tests/d3d11.c b/dlls/d3d11/tests/d3d11.c index 46bdff09bca..8e46d97d439 100644 --- a/dlls/d3d11/tests/d3d11.c +++ b/dlls/d3d11/tests/d3d11.c @@ -36335,6 +36335,8 @@ static void test_nv12(void)
winetest_push_context("test %u (%ux%u)", test_idx, width, height);
+ /* Apparently no Vulkan implementation supports rendering to a NV12 texture, so here we do + * not request D3D11_BIND_RENDER_TARGET. We will recreate it later for render target usage. */ desc.Width = width; desc.Height = height; desc.MipLevels = 1; @@ -36342,7 +36344,7 @@ static void test_nv12(void) desc.Format = DXGI_FORMAT_NV12; desc.SampleDesc.Count = 1; desc.Usage = D3D11_USAGE_DEFAULT; - desc.BindFlags = D3D11_BIND_SHADER_RESOURCE | D3D11_BIND_RENDER_TARGET; + desc.BindFlags = D3D11_BIND_SHADER_RESOURCE;
content = calloc(width * height * 3 / 2, 1); ok(!!content, "Failed to allocate memory.\n"); @@ -36427,6 +36429,31 @@ static void test_nv12(void) check_readback_data_u8_with_buffer(&rb, content, width, 0); release_resource_readback(&rb);
+ ID3D11ShaderResourceView_Release(srvs[0]); + ID3D11ShaderResourceView_Release(srvs[1]); + ID3D11Texture2D_Release(texture); + + desc.Height = height; + desc.Format = DXGI_FORMAT_NV12; + desc.BindFlags = D3D11_BIND_SHADER_RESOURCE | D3D11_BIND_RENDER_TARGET; + + hr = ID3D11Device_CreateTexture2D(device, &desc, NULL, &texture); + todo_wine + ok(hr == S_OK, "Got hr %#lx.\n", hr); + + if (FAILED(hr)) + goto no_render_target; + + srv_desc.Format = DXGI_FORMAT_R8_UINT; + + hr = ID3D11Device_CreateShaderResourceView(device, (ID3D11Resource *)texture, &srv_desc, &srvs[0]); + ok(hr == S_OK, "Got hr %#lx.\n", hr); + + srv_desc.Format = DXGI_FORMAT_R8G8_UINT; + + hr = ID3D11Device_CreateShaderResourceView(device, (ID3D11Resource *)texture, &srv_desc, &srvs[1]); + ok(hr == S_OK, "Got hr %#lx.\n", hr); + rtv_desc.Format = DXGI_FORMAT_R8_UINT; rtv_desc.ViewDimension = D3D11_RTV_DIMENSION_TEXTURE2D; rtv_desc.Texture2D.MipSlice = 0; @@ -36466,6 +36493,8 @@ static void test_nv12(void)
ID3D11RenderTargetView_Release(rtv2); ID3D11RenderTargetView_Release(rtv1); + + no_render_target: ID3D11Buffer_Release(cbuffer); ID3D11UnorderedAccessView_Release(check_uav); ID3D11ShaderResourceView_Release(srvs[1]);
From: Giovanni Mascellani gmascellani@codeweavers.com
--- dlls/d3d11/tests/d3d11.c | 94 +++++++++++++++++++++++++++++++++++----- 1 file changed, 82 insertions(+), 12 deletions(-)
diff --git a/dlls/d3d11/tests/d3d11.c b/dlls/d3d11/tests/d3d11.c index 8e46d97d439..4b6c54fc71e 100644 --- a/dlls/d3d11/tests/d3d11.c +++ b/dlls/d3d11/tests/d3d11.c @@ -36264,18 +36264,30 @@ static void test_nv12(void) { uint32_t width; uint32_t height; + uint32_t copy_x; + uint32_t copy_y; + uint32_t copy_width; + uint32_t copy_height; } tests[] = { - {640, 480}, - {640, 481}, - {641, 480}, - {641, 481}, - {642, 480}, - {642, 481}, - {642, 482}, - {644, 482}, - {644, 484}, + {640, 480, 10, 20, 4, 6}, + {640, 480, 10, 20, 4, 7}, + {640, 480, 10, 20, 5, 6}, + {640, 480, 10, 20, 5, 7}, + + {640, 480, 10, 21, 4, 6}, + {640, 480, 11, 20, 4, 6}, + {640, 480, 11, 21, 4, 6}, + + {640, 481, 10, 20, 4, 6}, + {641, 480, 10, 20, 4, 6}, + {641, 481, 10, 20, 4, 6}, + {642, 480, 10, 20, 4, 6}, + {642, 481, 10, 20, 4, 6}, + {642, 482, 10, 20, 4, 6}, + {644, 482, 10, 20, 4, 6}, + {644, 484, 10, 20, 4, 6}, };
if (!init_test_context(&test_context, NULL)) @@ -36319,6 +36331,7 @@ static void test_nv12(void) D3D11_SUBRESOURCE_DATA subresource_data = {0}; D3D11_RENDER_TARGET_VIEW_DESC rtv_desc = {0}; ID3D11Texture2D *texture, *check_texture; + char *content, *content2, *copy_source; ID3D11UnorderedAccessView *check_uav; ID3D11RenderTargetView *rtv1, *rtv2; ID3D11ShaderResourceView *srvs[2]; @@ -36328,12 +36341,17 @@ static void test_nv12(void) ID3D11Buffer *cbuffer; HRESULT expected_hr; unsigned int i, j; - char *content; + D3D11_BOX box;
const uint32_t width = tests[test_idx].width; const uint32_t height = tests[test_idx].height; + const uint32_t copy_x = tests[test_idx].copy_x; + const uint32_t copy_y = tests[test_idx].copy_y; + const uint32_t copy_width = tests[test_idx].copy_width; + const uint32_t copy_height = tests[test_idx].copy_height;
- winetest_push_context("test %u (%ux%u)", test_idx, width, height); + winetest_push_context("test %u (%ux%u, %u,%u,%ux%u)", test_idx, width, height, + copy_x, copy_y, copy_width, copy_height);
/* Apparently no Vulkan implementation supports rendering to a NV12 texture, so here we do * not request D3D11_BIND_RENDER_TARGET. We will recreate it later for render target usage. */ @@ -36347,7 +36365,9 @@ static void test_nv12(void) desc.BindFlags = D3D11_BIND_SHADER_RESOURCE;
content = calloc(width * height * 3 / 2, 1); - ok(!!content, "Failed to allocate memory.\n"); + content2 = calloc(width * height * 3 / 2, 1); + copy_source = calloc(copy_width * copy_height * 3 / 2, 1); + ok(content && content2 && copy_source, "Failed to allocate memory.\n");
for (i = 0; i < height; ++i) { @@ -36429,6 +36449,56 @@ static void test_nv12(void) check_readback_data_u8_with_buffer(&rb, content, width, 0); release_resource_readback(&rb);
+ memcpy(content2, content, width * height * 3 / 2); + if (copy_x % 2 == 0 && copy_y % 2 == 0 && copy_width % 2 == 0 && copy_height % 2 == 0) + { + for (i = copy_y; i < copy_y + copy_height; ++i) + { + for (j = copy_x; j < copy_x + copy_width; ++j) + { + content2[i * width + j] = 0xab; + if (i % 2 == 0 && j % 2 == 0) + { + content2[width * (height + i / 2) + j] = 0xcd; + content2[width * (height + i / 2) + j + 1] = 0xef; + } + } + } + for (i = 0; i < copy_height; ++i) + { + for (j = 0; j < copy_width; ++j) + { + copy_source[i * copy_width + j] = 0xab; + if (i % 2 == 0 && j % 2 == 0) + { + copy_source[copy_width * (copy_height + i / 2) + j] = 0xcd; + copy_source[copy_width * (copy_height + i / 2) + j + 1] = 0xef; + } + } + } + } + + /* UpdateSubresource() copies the specified box on the luma plane and also the corresponding + * box on the chroma plane. It does nothing as soon as any coordinate of the box is not a + * multiple of 2. AMD seems to have a bug and copies data with the wrong pitch. */ + if (!is_amd_device(device)) + { + set_box(&box, copy_x, copy_y, 0, copy_x + copy_width, copy_y + copy_height, 1); + ID3D11DeviceContext_UpdateSubresource(device_context, + (ID3D11Resource *)texture, 0, &box, copy_source, copy_width, 0); + + ID3D11DeviceContext_ClearUnorderedAccessViewUint(device_context, check_uav, clear_values); + ID3D11DeviceContext_CSSetShader(device_context, cs, NULL, 0); + ID3D11DeviceContext_CSSetShaderResources(device_context, 0, ARRAY_SIZE(srvs), srvs); + ID3D11DeviceContext_CSSetUnorderedAccessViews(device_context, 1, 1, &check_uav, NULL); + ID3D11DeviceContext_CSSetConstantBuffers(device_context, 0, 1, &cbuffer); + ID3D11DeviceContext_Dispatch(device_context, width, height, 1); + + get_texture_readback(check_texture, 0, &rb); + check_readback_data_u8_with_buffer(&rb, content2, width, 0); + release_resource_readback(&rb); + } + ID3D11ShaderResourceView_Release(srvs[0]); ID3D11ShaderResourceView_Release(srvs[1]); ID3D11Texture2D_Release(texture);
wrt 1/4, is it actually broken? Do the relevant machines report support for nv12 via CheckFormatSupport()?
Right, I completely forgot to check that. It should now do that, let's see whether it works.
Also, I added a number of small improvements.
This merge request was approved by Elizabeth Figura.
-
Elizabeth Figura (＠zfigura) -
Giovanni Mascellani
-
Giovanni Mascellani (＠giomasce)