diff --git a/dlls/d3d11/tests/d3d11.c b/dlls/d3d11/tests/d3d11.c index 4ac69f95fce..04853c848c8 100644 --- a/dlls/d3d11/tests/d3d11.c +++ b/dlls/d3d11/tests/d3d11.c @@ -12883,18 +12883,19 @@ static void test_ps_cs_uav_binding(void) release_test_context(&test_context); } -static void test_ps_uav_atomic_instructions(void) +static void test_atomic_instructions(void) { + ID3D11UnorderedAccessView *in_uav, *out_uav; + ID3D11Buffer *cb, *in_buffer, *out_buffer; D3D11_UNORDERED_ACCESS_VIEW_DESC uav_desc; struct d3d11_test_context test_context; + struct resource_readback rb, out_rb; D3D11_TEXTURE2D_DESC texture_desc; - ID3D11UnorderedAccessView *uav; D3D11_BUFFER_DESC buffer_desc; ID3D11DeviceContext *context; ID3D11RenderTargetView *rtv; - struct resource_readback rb; - ID3D11Buffer *cb, *raw_uav; ID3D11Texture2D *texture; + ID3D11ComputeShader *cs; ID3D11PixelShader *ps; ID3D11Device *device; D3D11_VIEWPORT vp; @@ -12902,6 +12903,7 @@ static void test_ps_uav_atomic_instructions(void) HRESULT hr; static const D3D_FEATURE_LEVEL feature_level = D3D_FEATURE_LEVEL_11_0; + static const unsigned int zero[4] = {0, 0, 0, 0}; static const DWORD ps_atomics_code[] = { #if 0 @@ -12938,12 +12940,71 @@ static void test_ps_uav_atomic_instructions(void) 0x00000000, 0x00000000, 0x080000ab, 0x0011e000, 0x00000000, 0x00004001, 0x00000020, 0x0020800a, 0x00000000, 0x00000000, 0x0100003e, }; + static const DWORD cs_atomics_code[] = + { +#if 0 + RWByteAddressBuffer u; + RWByteAddressBuffer u2; + + uint4 v; + int4 i; + + [numthreads(1, 1, 1)] + void main() + { + uint r; + u.InterlockedAnd(0 * 4, v.x, r); + u2.Store(0 * 4, r); + u.InterlockedCompareExchange(1 * 4, v.y, v.x, r); + u2.Store(1 * 4, r); + u.InterlockedAdd(2 * 4, v.x, r); + u2.Store(2 * 4, r); + u.InterlockedOr(3 * 4, v.x, r); + u2.Store(3 * 4, r); + u.InterlockedMax(4 * 4, i.x, r); + u2.Store(4 * 4, r); + u.InterlockedMin(5 * 4, i.x, r); + u2.Store(5 * 4, r); + u.InterlockedMax(6 * 4, v.x, r); + u2.Store(6 * 4, r); + u.InterlockedMin(7 * 4, v.x, r); + u2.Store(7 * 4, r); + u.InterlockedXor(8 * 4, v.x, r); + u2.Store(8 * 4, r); + } +#endif + 0x43425844, 0x859a96e3, 0x1a35e463, 0x1e89ce58, 0x5cfe430a, 0x00000001, 0x0000026c, 0x00000003, + 0x0000002c, 0x0000003c, 0x0000004c, 0x4e475349, 0x00000008, 0x00000000, 0x00000008, 0x4e47534f, + 0x00000008, 0x00000000, 0x00000008, 0x58454853, 0x00000218, 0x00050050, 0x00000086, 0x0100086a, + 0x04000059, 0x00208e46, 0x00000000, 0x00000002, 0x0300009d, 0x0011e000, 0x00000000, 0x0300009d, + 0x0011e000, 0x00000001, 0x02000068, 0x00000001, 0x0400009b, 0x00000001, 0x00000001, 0x00000001, + 0x0a0000b5, 0x00100012, 0x00000000, 0x0011e000, 0x00000000, 0x00004001, 0x00000000, 0x0020800a, + 0x00000000, 0x00000000, 0x0d0000b9, 0x00100022, 0x00000000, 0x0011e000, 0x00000000, 0x00004001, + 0x00000004, 0x0020801a, 0x00000000, 0x00000000, 0x0020800a, 0x00000000, 0x00000000, 0x0a0000b4, + 0x00100042, 0x00000000, 0x0011e000, 0x00000000, 0x00004001, 0x00000008, 0x0020800a, 0x00000000, + 0x00000000, 0x0a0000b6, 0x00100082, 0x00000000, 0x0011e000, 0x00000000, 0x00004001, 0x0000000c, + 0x0020800a, 0x00000000, 0x00000000, 0x070000a6, 0x0011e0f2, 0x00000001, 0x00004001, 0x00000000, + 0x00100e46, 0x00000000, 0x0a0000ba, 0x00100012, 0x00000000, 0x0011e000, 0x00000000, 0x00004001, + 0x00000010, 0x0020800a, 0x00000000, 0x00000001, 0x0a0000bb, 0x00100022, 0x00000000, 0x0011e000, + 0x00000000, 0x00004001, 0x00000014, 0x0020800a, 0x00000000, 0x00000001, 0x0a0000bc, 0x00100042, + 0x00000000, 0x0011e000, 0x00000000, 0x00004001, 0x00000018, 0x0020800a, 0x00000000, 0x00000000, + 0x0a0000bd, 0x00100082, 0x00000000, 0x0011e000, 0x00000000, 0x00004001, 0x0000001c, 0x0020800a, + 0x00000000, 0x00000000, 0x070000a6, 0x0011e0f2, 0x00000001, 0x00004001, 0x00000010, 0x00100e46, + 0x00000000, 0x0a0000b7, 0x00100012, 0x00000000, 0x0011e000, 0x00000000, 0x00004001, 0x00000020, + 0x0020800a, 0x00000000, 0x00000000, 0x070000a6, 0x0011e012, 0x00000001, 0x00004001, 0x00000020, + 0x0010000a, 0x00000000, 0x0100003e, + }; static const char * const instructions[] = { "atomic_and", "atomic_cmp_store", "atomic_iadd", "atomic_or", "atomic_imax", "atomic_imin", "atomic_umax", "atomic_umin", "atomic_xor", }; + static const char * const imm_instructions[] = + { + "imm_atomic_and", "imm_atomic_cmp_exch", "imm_atomic_iadd", "imm_atomic_or", + "imm_atomic_imax", "imm_atomic_imin", "imm_atomic_umax", "imm_atomic_umin", "imm_atomic_xor", + }; static const struct test { struct uvec4 v; @@ -12981,13 +13042,16 @@ static void test_ps_uav_atomic_instructions(void) cb = create_buffer(device, D3D11_BIND_CONSTANT_BUFFER, 2 * sizeof(struct uvec4), NULL); ID3D11DeviceContext_PSSetConstantBuffers(context, 0, 1, &cb); + ID3D11DeviceContext_CSSetConstantBuffers(context, 0, 1, &cb); buffer_desc.ByteWidth = sizeof(tests->input); buffer_desc.Usage = D3D11_USAGE_DEFAULT; buffer_desc.BindFlags = D3D11_BIND_UNORDERED_ACCESS; buffer_desc.CPUAccessFlags = 0; buffer_desc.MiscFlags = D3D11_RESOURCE_MISC_BUFFER_ALLOW_RAW_VIEWS; - hr = ID3D11Device_CreateBuffer(device, &buffer_desc, NULL, &raw_uav); + hr = ID3D11Device_CreateBuffer(device, &buffer_desc, NULL, &in_buffer); + ok(SUCCEEDED(hr), "Failed to create buffer, hr %#x.\n", hr); + hr = ID3D11Device_CreateBuffer(device, &buffer_desc, NULL, &out_buffer); ok(SUCCEEDED(hr), "Failed to create buffer, hr %#x.\n", hr); uav_desc.Format = DXGI_FORMAT_R32_TYPELESS; @@ -12995,12 +13059,10 @@ static void test_ps_uav_atomic_instructions(void) U(uav_desc).Buffer.FirstElement = 0; U(uav_desc).Buffer.NumElements = buffer_desc.ByteWidth / sizeof(*tests->input); U(uav_desc).Buffer.Flags = D3D11_BUFFER_UAV_FLAG_RAW; - hr = ID3D11Device_CreateUnorderedAccessView(device, (ID3D11Resource *)raw_uav, &uav_desc, &uav); + hr = ID3D11Device_CreateUnorderedAccessView(device, (ID3D11Resource *)in_buffer, &uav_desc, &in_uav); + ok(SUCCEEDED(hr), "Failed to create unordered access view, hr %#x.\n", hr); + hr = ID3D11Device_CreateUnorderedAccessView(device, (ID3D11Resource *)out_buffer, &uav_desc, &out_uav); ok(SUCCEEDED(hr), "Failed to create unordered access view, hr %#x.\n", hr); - - /* FIXME: Set the render targets to NULL when no attachment draw calls are supported in wined3d. */ - ID3D11DeviceContext_OMSetRenderTargetsAndUnorderedAccessViews(context, 1, &rtv, NULL, - 0, 1, &uav, NULL); vp.TopLeftX = 0.0f; vp.TopLeftY = 0.0f; @@ -13014,6 +13076,10 @@ static void test_ps_uav_atomic_instructions(void) ok(SUCCEEDED(hr), "Failed to create pixel shader, hr %#x.\n", hr); ID3D11DeviceContext_PSSetShader(context, ps, NULL, 0); + hr = ID3D11Device_CreateComputeShader(device, cs_atomics_code, sizeof(cs_atomics_code), NULL, &cs); + ok(SUCCEEDED(hr), "Failed to create compute shader, hr %#x.\n", hr); + ID3D11DeviceContext_CSSetShader(context, cs, NULL, 0); + for (i = 0; i < sizeof(tests) / sizeof(*tests); ++i) { const struct test *test = &tests[i]; @@ -13021,12 +13087,15 @@ static void test_ps_uav_atomic_instructions(void) ID3D11DeviceContext_UpdateSubresource(context, (ID3D11Resource *)cb, 0, NULL, &test->v, 0, 0); - ID3D11DeviceContext_UpdateSubresource(context, (ID3D11Resource *)raw_uav, 0, + ID3D11DeviceContext_UpdateSubresource(context, (ID3D11Resource *)in_buffer, 0, NULL, test->input, 0, 0); - draw_quad(&test_context); + /* FIXME: Set the render targets to NULL when no attachment draw calls are supported in wined3d. */ + ID3D11DeviceContext_OMSetRenderTargetsAndUnorderedAccessViews(context, 1, &rtv, NULL, + 0, 1, &in_uav, NULL); - get_buffer_readback(raw_uav, &rb); + draw_quad(&test_context); + get_buffer_readback(in_buffer, &rb); for (j = 0; j < sizeof(instructions) / sizeof(*instructions); ++j) { unsigned int value = get_readback_color(&rb, j, 0); @@ -13043,14 +13112,50 @@ static void test_ps_uav_atomic_instructions(void) test->v.x, test->v.y, test->i.x, test->input[j], test->input[j]); } release_resource_readback(&rb); + + ID3D11DeviceContext_UpdateSubresource(context, (ID3D11Resource *)in_buffer, 0, + NULL, test->input, 0, 0); + ID3D11DeviceContext_ClearUnorderedAccessViewUint(context, out_uav, zero); + + ID3D11DeviceContext_CSSetUnorderedAccessViews(context, 0, 1, &in_uav, NULL); + ID3D11DeviceContext_CSSetUnorderedAccessViews(context, 1, 1, &out_uav, NULL); + + ID3D11DeviceContext_Dispatch(context, 1, 1, 1); + get_buffer_readback(in_buffer, &rb); + get_buffer_readback(out_buffer, &out_rb); + for (j = 0; j < sizeof(instructions) / sizeof(*instructions); ++j) + { + BOOL todo_instruction = !strcmp(imm_instructions[j], "imm_atomic_imax") + || !strcmp(imm_instructions[j], "imm_atomic_imin") + || !strcmp(imm_instructions[j], "imm_atomic_umax") + || !strcmp(imm_instructions[j], "imm_atomic_umin"); + unsigned int out_value = get_readback_color(&out_rb, j, 0); + unsigned int value = get_readback_color(&rb, j, 0); + unsigned int expected = test->expected_result[j]; + + todo_wine_if(expected != test->input[j] && todo_instruction) + ok(value == expected, "Test %u: Got %#x (%d), expected %#x (%d) for '%s' " + "with inputs (%u, %u), (%d), %#x (%d).\n", + i, value, value, expected, expected, imm_instructions[j], + test->v.x, test->v.y, test->i.x, test->input[j], test->input[j]); + + todo_wine_if(todo_instruction && out_value != test->input[j]) + ok(out_value == test->input[j], "Got original value %u, expected %u for '%s'.\n", + out_value, test->input[j], imm_instructions[j]); + } + release_resource_readback(&out_rb); + release_resource_readback(&rb); } ID3D11Buffer_Release(cb); - ID3D11Buffer_Release(raw_uav); + ID3D11Buffer_Release(in_buffer); + ID3D11Buffer_Release(out_buffer); + ID3D11ComputeShader_Release(cs); ID3D11PixelShader_Release(ps); ID3D11RenderTargetView_Release(rtv); ID3D11Texture2D_Release(texture); - ID3D11UnorderedAccessView_Release(uav); + ID3D11UnorderedAccessView_Release(in_uav); + ID3D11UnorderedAccessView_Release(out_uav); release_test_context(&test_context); } @@ -14287,7 +14392,7 @@ START_TEST(d3d11) test_uav_load(); test_cs_uav_store(); test_ps_cs_uav_binding(); - test_ps_uav_atomic_instructions(); + test_atomic_instructions(); test_sm4_ret_instruction(); test_primitive_restart(); test_resinfo_instruction();