[PATCH 6/6] d3d11/tests: Add test for thread group shared memory.
Józef Kucia
jkucia at codeweavers.com
Wed Mar 1 17:55:31 CST 2017
Signed-off-by: Józef Kucia <jkucia at codeweavers.com>
---
dlls/d3d11/tests/d3d11.c | 284 +++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 284 insertions(+)
diff --git a/dlls/d3d11/tests/d3d11.c b/dlls/d3d11/tests/d3d11.c
index e7ac5a0..af51575 100644
--- a/dlls/d3d11/tests/d3d11.c
+++ b/dlls/d3d11/tests/d3d11.c
@@ -14854,6 +14854,289 @@ static void test_compute_shader_registers(void)
release_test_context(&test_context);
}
+static void test_tgsm(void)
+{
+ D3D11_UNORDERED_ACCESS_VIEW_DESC uav_desc;
+ struct d3d11_test_context test_context;
+ ID3D11UnorderedAccessView *uav, *uav2;
+ struct resource_readback rb, rb2;
+ unsigned int i, data, expected;
+ ID3D11Buffer *buffer, *buffer2;
+ D3D11_BUFFER_DESC buffer_desc;
+ ID3D11DeviceContext *context;
+ ID3D11ComputeShader *cs;
+ ID3D11Device *device;
+ float float_data;
+ HRESULT hr;
+
+ static const DWORD raw_tgsm_code[] =
+ {
+#if 0
+ RWByteAddressBuffer u;
+ groupshared uint m;
+
+ [numthreads(32, 1, 1)]
+ void main(uint local_idx : SV_GroupIndex, uint group_id : SV_GroupID)
+ {
+ if (!local_idx)
+ m = group_id.x;
+ GroupMemoryBarrierWithGroupSync();
+ InterlockedAdd(m, group_id.x);
+ GroupMemoryBarrierWithGroupSync();
+ if (!local_idx)
+ u.Store(4 * group_id.x, m);
+ }
+#endif
+ 0x43425844, 0x467df6d9, 0x5f56edda, 0x5c96b787, 0x60c91fb8, 0x00000001, 0x00000148, 0x00000003,
+ 0x0000002c, 0x0000003c, 0x0000004c, 0x4e475349, 0x00000008, 0x00000000, 0x00000008, 0x4e47534f,
+ 0x00000008, 0x00000000, 0x00000008, 0x58454853, 0x000000f4, 0x00050050, 0x0000003d, 0x0100086a,
+ 0x0300009d, 0x0011e000, 0x00000000, 0x0200005f, 0x00024000, 0x0200005f, 0x00021012, 0x02000068,
+ 0x00000001, 0x0400009f, 0x0011f000, 0x00000000, 0x00000004, 0x0400009b, 0x00000020, 0x00000001,
+ 0x00000001, 0x0200001f, 0x0002400a, 0x060000a6, 0x0011f012, 0x00000000, 0x00004001, 0x00000000,
+ 0x0002100a, 0x01000015, 0x010018be, 0x060000ad, 0x0011f000, 0x00000000, 0x00004001, 0x00000000,
+ 0x0002100a, 0x010018be, 0x0200001f, 0x0002400a, 0x06000029, 0x00100012, 0x00000000, 0x0002100a,
+ 0x00004001, 0x00000002, 0x070000a5, 0x00100022, 0x00000000, 0x00004001, 0x00000000, 0x0011f006,
+ 0x00000000, 0x070000a6, 0x0011e012, 0x00000000, 0x0010000a, 0x00000000, 0x0010001a, 0x00000000,
+ 0x01000015, 0x0100003e,
+ };
+ static const DWORD structured_tgsm_code[] =
+ {
+#if 0
+ #define GROUP_SIZE 32
+
+ RWByteAddressBuffer u;
+ RWByteAddressBuffer u2;
+ groupshared uint m[GROUP_SIZE];
+
+ [numthreads(GROUP_SIZE, 1, 1)]
+ void main(uint local_idx : SV_GroupIndex, uint group_id : SV_GroupID)
+ {
+ uint sum, original, i;
+
+ if (!local_idx)
+ {
+ for (i = 0; i < GROUP_SIZE; ++i)
+ m[i] = 2 * group_id.x;
+ }
+ GroupMemoryBarrierWithGroupSync();
+ InterlockedAdd(m[local_idx], 1);
+ GroupMemoryBarrierWithGroupSync();
+ for (i = 0, sum = 0; i < GROUP_SIZE; sum += m[i++]);
+ u.InterlockedExchange(4 * group_id.x, sum, original);
+ u2.Store(4 * group_id.x, original);
+ }
+#endif
+ 0x43425844, 0x9d906c94, 0x81f5ad92, 0x11e860b2, 0x3623c824, 0x00000001, 0x000002c0, 0x00000003,
+ 0x0000002c, 0x0000003c, 0x0000004c, 0x4e475349, 0x00000008, 0x00000000, 0x00000008, 0x4e47534f,
+ 0x00000008, 0x00000000, 0x00000008, 0x58454853, 0x0000026c, 0x00050050, 0x0000009b, 0x0100086a,
+ 0x0300009d, 0x0011e000, 0x00000000, 0x0300009d, 0x0011e000, 0x00000001, 0x0200005f, 0x00024000,
+ 0x0200005f, 0x00021012, 0x02000068, 0x00000002, 0x050000a0, 0x0011f000, 0x00000000, 0x00000004,
+ 0x00000020, 0x0400009b, 0x00000020, 0x00000001, 0x00000001, 0x0200001f, 0x0002400a, 0x06000029,
+ 0x00100012, 0x00000000, 0x0002100a, 0x00004001, 0x00000001, 0x05000036, 0x00100022, 0x00000000,
+ 0x00004001, 0x00000000, 0x01000030, 0x07000050, 0x00100042, 0x00000000, 0x0010001a, 0x00000000,
+ 0x00004001, 0x00000020, 0x03040003, 0x0010002a, 0x00000000, 0x090000a8, 0x0011f012, 0x00000000,
+ 0x0010001a, 0x00000000, 0x00004001, 0x00000000, 0x0010000a, 0x00000000, 0x0700001e, 0x00100022,
+ 0x00000000, 0x0010001a, 0x00000000, 0x00004001, 0x00000001, 0x01000016, 0x01000015, 0x010018be,
+ 0x04000036, 0x00100012, 0x00000000, 0x0002400a, 0x05000036, 0x00100022, 0x00000000, 0x00004001,
+ 0x00000000, 0x070000ad, 0x0011f000, 0x00000000, 0x00100046, 0x00000000, 0x00004001, 0x00000001,
+ 0x010018be, 0x08000036, 0x00100032, 0x00000000, 0x00004002, 0x00000000, 0x00000000, 0x00000000,
+ 0x00000000, 0x01000030, 0x07000050, 0x00100042, 0x00000000, 0x0010001a, 0x00000000, 0x00004001,
+ 0x00000020, 0x03040003, 0x0010002a, 0x00000000, 0x0700001e, 0x00100022, 0x00000001, 0x0010001a,
+ 0x00000000, 0x00004001, 0x00000001, 0x090000a7, 0x00100042, 0x00000000, 0x0010001a, 0x00000000,
+ 0x00004001, 0x00000000, 0x0011f006, 0x00000000, 0x0700001e, 0x00100012, 0x00000001, 0x0010000a,
+ 0x00000000, 0x0010002a, 0x00000000, 0x05000036, 0x00100032, 0x00000000, 0x00100046, 0x00000001,
+ 0x01000016, 0x06000029, 0x00100022, 0x00000000, 0x0002100a, 0x00004001, 0x00000002, 0x090000b8,
+ 0x00100012, 0x00000001, 0x0011e000, 0x00000000, 0x0010001a, 0x00000000, 0x0010000a, 0x00000000,
+ 0x070000a6, 0x0011e012, 0x00000001, 0x0010001a, 0x00000000, 0x0010000a, 0x00000001, 0x0100003e,
+ };
+ static const DWORD structured_tgsm_float_code[] =
+ {
+#if 0
+ #define GROUP_SIZE 32
+
+ struct data
+ {
+ float f;
+ uint u;
+ };
+
+ RWBuffer<float> u;
+ RWBuffer<uint> u2;
+ groupshared data m[GROUP_SIZE];
+
+ [numthreads(GROUP_SIZE, 1, 1)]
+ void main(uint local_idx : SV_GroupIndex, uint group_id : SV_GroupID,
+ uint thread_id : SV_DispatchThreadID)
+ {
+ uint i;
+ if (!local_idx)
+ {
+ for (i = 0; i < GROUP_SIZE; ++i)
+ {
+ m[i].f = group_id.x;
+ m[i].u = group_id.x;
+ }
+ }
+ GroupMemoryBarrierWithGroupSync();
+ for (i = 0; i < local_idx; ++i)
+ {
+ m[local_idx].f += group_id.x;
+ m[local_idx].u += group_id.x;
+ }
+ u[thread_id.x] = m[local_idx].f;
+ u2[thread_id.x] = m[local_idx].u;
+ }
+#endif
+ 0x43425844, 0xaadf1a71, 0x16f60224, 0x89b6ce76, 0xb66fb96f, 0x00000001, 0x000002ac, 0x00000003,
+ 0x0000002c, 0x0000003c, 0x0000004c, 0x4e475349, 0x00000008, 0x00000000, 0x00000008, 0x4e47534f,
+ 0x00000008, 0x00000000, 0x00000008, 0x58454853, 0x00000258, 0x00050050, 0x00000096, 0x0100086a,
+ 0x0400089c, 0x0011e000, 0x00000000, 0x00005555, 0x0400089c, 0x0011e000, 0x00000001, 0x00004444,
+ 0x0200005f, 0x00024000, 0x0200005f, 0x00021012, 0x0200005f, 0x00020012, 0x02000068, 0x00000002,
+ 0x050000a0, 0x0011f000, 0x00000000, 0x00000008, 0x00000020, 0x0400009b, 0x00000020, 0x00000001,
+ 0x00000001, 0x0200001f, 0x0002400a, 0x04000056, 0x00100012, 0x00000000, 0x0002100a, 0x04000036,
+ 0x00100022, 0x00000000, 0x0002100a, 0x05000036, 0x00100042, 0x00000000, 0x00004001, 0x00000000,
+ 0x01000030, 0x07000050, 0x00100082, 0x00000000, 0x0010002a, 0x00000000, 0x00004001, 0x00000020,
+ 0x03040003, 0x0010003a, 0x00000000, 0x090000a8, 0x0011f032, 0x00000000, 0x0010002a, 0x00000000,
+ 0x00004001, 0x00000000, 0x00100046, 0x00000000, 0x0700001e, 0x00100042, 0x00000000, 0x0010002a,
+ 0x00000000, 0x00004001, 0x00000001, 0x01000016, 0x01000015, 0x010018be, 0x04000056, 0x00100012,
+ 0x00000000, 0x0002100a, 0x05000036, 0x00100022, 0x00000000, 0x00004001, 0x00000000, 0x01000030,
+ 0x06000050, 0x00100042, 0x00000000, 0x0010001a, 0x00000000, 0x0002400a, 0x03040003, 0x0010002a,
+ 0x00000000, 0x080000a7, 0x001000c2, 0x00000000, 0x0002400a, 0x00004001, 0x00000000, 0x0011f406,
+ 0x00000000, 0x07000000, 0x00100012, 0x00000001, 0x0010000a, 0x00000000, 0x0010002a, 0x00000000,
+ 0x0600001e, 0x00100022, 0x00000001, 0x0010003a, 0x00000000, 0x0002100a, 0x080000a8, 0x0011f032,
+ 0x00000000, 0x0002400a, 0x00004001, 0x00000000, 0x00100046, 0x00000001, 0x0700001e, 0x00100022,
+ 0x00000000, 0x0010001a, 0x00000000, 0x00004001, 0x00000001, 0x01000016, 0x080000a7, 0x00100032,
+ 0x00000000, 0x0002400a, 0x00004001, 0x00000000, 0x0011f046, 0x00000000, 0x060000a4, 0x0011e0f2,
+ 0x00000000, 0x00020006, 0x00100006, 0x00000000, 0x060000a4, 0x0011e0f2, 0x00000001, 0x00020006,
+ 0x00100556, 0x00000000, 0x0100003e,
+ };
+ static const D3D_FEATURE_LEVEL feature_level = D3D_FEATURE_LEVEL_11_0;
+ static const unsigned int zero[4] = {0};
+
+ if (!init_test_context(&test_context, &feature_level))
+ return;
+
+ device = test_context.device;
+ context = test_context.immediate_context;
+
+ buffer_desc.ByteWidth = 1024;
+ 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, &buffer);
+ ok(SUCCEEDED(hr), "Failed to create buffer, hr %#x.\n", hr);
+
+ uav_desc.Format = DXGI_FORMAT_R32_TYPELESS;
+ uav_desc.ViewDimension = D3D11_UAV_DIMENSION_BUFFER;
+ U(uav_desc).Buffer.FirstElement = 0;
+ U(uav_desc).Buffer.NumElements = buffer_desc.ByteWidth / sizeof(unsigned int);
+ U(uav_desc).Buffer.Flags = D3D11_BUFFER_UAV_FLAG_RAW;
+ hr = ID3D11Device_CreateUnorderedAccessView(device, (ID3D11Resource *)buffer, &uav_desc, &uav);
+ ok(SUCCEEDED(hr), "Failed to create unordered access view, hr %#x.\n", hr);
+
+ hr = ID3D11Device_CreateComputeShader(device, raw_tgsm_code, sizeof(raw_tgsm_code), NULL, &cs);
+ ok(SUCCEEDED(hr), "Failed to create compute shader, hr %#x.\n", hr);
+
+ ID3D11DeviceContext_CSSetShader(context, cs, NULL, 0);
+ ID3D11DeviceContext_CSSetUnorderedAccessViews(context, 0, 1, &uav, NULL);
+
+ ID3D11DeviceContext_ClearUnorderedAccessViewUint(context, uav, zero);
+ ID3D11DeviceContext_Dispatch(context, 64, 1, 1);
+ get_buffer_readback(buffer, &rb);
+ for (i = 0; i < 64; ++i)
+ {
+ data = get_readback_color(&rb, i, 0);
+ expected = 33 * i;
+ ok(data == expected, "Got %u, expected %u (index %u).\n", data, expected, i);
+ }
+ release_resource_readback(&rb);
+
+ ID3D11Buffer_Release(buffer);
+ ID3D11ComputeShader_Release(cs);
+ ID3D11UnorderedAccessView_Release(uav);
+
+ hr = ID3D11Device_CreateBuffer(device, &buffer_desc, NULL, &buffer);
+ ok(SUCCEEDED(hr), "Failed to create buffer, hr %#x.\n", hr);
+ hr = ID3D11Device_CreateUnorderedAccessView(device, (ID3D11Resource *)buffer, &uav_desc, &uav);
+ ok(SUCCEEDED(hr), "Failed to create unordered access view, hr %#x.\n", hr);
+ hr = ID3D11Device_CreateBuffer(device, &buffer_desc, NULL, &buffer2);
+ ok(SUCCEEDED(hr), "Failed to create buffer, hr %#x.\n", hr);
+ hr = ID3D11Device_CreateUnorderedAccessView(device, (ID3D11Resource *)buffer2, &uav_desc, &uav2);
+ ok(SUCCEEDED(hr), "Failed to create unordered access view, hr %#x.\n", hr);
+ hr = ID3D11Device_CreateComputeShader(device, structured_tgsm_code, sizeof(structured_tgsm_code), NULL, &cs);
+ ok(SUCCEEDED(hr), "Failed to create compute shader, hr %#x.\n", hr);
+
+ ID3D11DeviceContext_CSSetShader(context, cs, NULL, 0);
+ ID3D11DeviceContext_CSSetUnorderedAccessViews(context, 0, 1, &uav, NULL);
+ ID3D11DeviceContext_CSSetUnorderedAccessViews(context, 1, 1, &uav2, NULL);
+
+ ID3D11DeviceContext_ClearUnorderedAccessViewUint(context, uav, zero);
+ ID3D11DeviceContext_ClearUnorderedAccessViewUint(context, uav2, zero);
+ ID3D11DeviceContext_Dispatch(context, 32, 1, 1);
+ get_buffer_readback(buffer, &rb);
+ get_buffer_readback(buffer2, &rb2);
+ for (i = 0; i < 32; ++i)
+ {
+ expected = 64 * i + 32;
+ data = get_readback_color(&rb, i, 0);
+ ok(data == expected, "Got %u, expected %u (index %u).\n", data, expected, i);
+ data = get_readback_color(&rb2, i, 0);
+ ok(data == expected, "Got %u, expected %u (index %u).\n", data, expected, i);
+ }
+ release_resource_readback(&rb);
+ release_resource_readback(&rb2);
+
+ ID3D11Buffer_Release(buffer);
+ ID3D11Buffer_Release(buffer2);
+ ID3D11ComputeShader_Release(cs);
+ ID3D11UnorderedAccessView_Release(uav);
+ ID3D11UnorderedAccessView_Release(uav2);
+
+ buffer_desc.MiscFlags = 0;
+ U(uav_desc).Buffer.Flags = 0;
+ hr = ID3D11Device_CreateBuffer(device, &buffer_desc, NULL, &buffer);
+ ok(SUCCEEDED(hr), "Failed to create buffer, hr %#x.\n", hr);
+ uav_desc.Format = DXGI_FORMAT_R32_FLOAT;
+ hr = ID3D11Device_CreateUnorderedAccessView(device, (ID3D11Resource *)buffer, &uav_desc, &uav);
+ ok(SUCCEEDED(hr), "Failed to create unordered access view, hr %#x.\n", hr);
+ hr = ID3D11Device_CreateBuffer(device, &buffer_desc, NULL, &buffer2);
+ ok(SUCCEEDED(hr), "Failed to create buffer, hr %#x.\n", hr);
+ uav_desc.Format = DXGI_FORMAT_R32_UINT;
+ hr = ID3D11Device_CreateUnorderedAccessView(device, (ID3D11Resource *)buffer2, &uav_desc, &uav2);
+ ok(SUCCEEDED(hr), "Failed to create unordered access view, hr %#x.\n", hr);
+ hr = ID3D11Device_CreateComputeShader(device, structured_tgsm_float_code,
+ sizeof(structured_tgsm_float_code), NULL, &cs);
+ ok(SUCCEEDED(hr), "Failed to create compute shader, hr %#x.\n", hr);
+
+ ID3D11DeviceContext_CSSetShader(context, cs, NULL, 0);
+ ID3D11DeviceContext_CSSetUnorderedAccessViews(context, 0, 1, &uav, NULL);
+ ID3D11DeviceContext_CSSetUnorderedAccessViews(context, 1, 1, &uav2, NULL);
+
+ ID3D11DeviceContext_ClearUnorderedAccessViewUint(context, uav, zero);
+ ID3D11DeviceContext_ClearUnorderedAccessViewUint(context, uav2, zero);
+ ID3D11DeviceContext_Dispatch(context, 3, 1, 1);
+ get_buffer_readback(buffer, &rb);
+ get_buffer_readback(buffer2, &rb2);
+ for (i = 0; i < 96; ++i)
+ {
+ expected = (i % 32 + 1) * (i / 32);
+ float_data = get_readback_float(&rb, i, 0);
+ ok(float_data == expected, "Got %.8e, expected %u (index %u).\n", float_data, expected, i);
+ data = get_readback_color(&rb2, i, 0);
+ ok(data == expected, "Got %u, expected %u (index %u).\n", data, expected, i);
+ }
+ release_resource_readback(&rb);
+ release_resource_readback(&rb2);
+
+ ID3D11Buffer_Release(buffer);
+ ID3D11Buffer_Release(buffer2);
+ ID3D11ComputeShader_Release(cs);
+ ID3D11UnorderedAccessView_Release(uav);
+ ID3D11UnorderedAccessView_Release(uav2);
+ release_test_context(&test_context);
+}
+
START_TEST(d3d11)
{
test_create_device();
@@ -14931,4 +15214,5 @@ START_TEST(d3d11)
run_for_each_feature_level_in_range(D3D_FEATURE_LEVEL_10_0, D3D_FEATURE_LEVEL_11_0,
test_unaligned_raw_buffer_access);
test_compute_shader_registers();
+ test_tgsm();
}
--
2.10.2
More information about the wine-patches
mailing list