Skip to content

[HLK] Adding WaveMatch Long Vector Test #7937

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 27 commits into from
Nov 27, 2025
Merged

[HLK] Adding WaveMatch Long Vector Test #7937

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
27 commits
Select commit Hold shift + click to select a range
68ae80f
init test
joaosaffran Nov 20, 2025
fec685c
update test
joaosaffran Nov 20, 2025
42a3127
clean up
joaosaffran Nov 20, 2025
997405c
update test
joaosaffran Nov 21, 2025
40c3814
fixing test
joaosaffran Nov 21, 2025
30709ab
clean up
joaosaffran Nov 21, 2025
74b20df
clean up 2
joaosaffran Nov 21, 2025
77bc0fe
clean test
joaosaffran Nov 24, 2025
53aa9a2
improve text
joaosaffran Nov 25, 2025
f1944d3
format
joaosaffran Nov 25, 2025
d0a8eed
addressing comments from Damyan
joaosaffran Nov 25, 2025
59eb475
format
joaosaffran Nov 25, 2025
4b692ce
fix redundant masking
joaosaffran Nov 25, 2025
1224e9a
Merge branch 'main' into test/wave-match
joaosaffran Nov 25, 2025
1fd9a33
remove oopsi
joaosaffran Nov 25, 2025
c74ccd9
remove oopsi
joaosaffran Nov 25, 2025
f976a91
adding flag back
joaosaffran Nov 25, 2025
8d942fb
clean up
joaosaffran Nov 25, 2025
9e3f3f1
format
joaosaffran Nov 25, 2025
9ca3790
remove unused parameter
joaosaffran-zz Nov 25, 2025
f8869cf
renaming vars
joaosaffran-zz Nov 25, 2025
3efdb38
change code
joaosaffran-zz Nov 25, 2025
882e3e1
format
joaosaffran-zz Nov 25, 2025
4cf8b98
improve code
joaosaffran-zz Nov 25, 2025
71dea03
fix damyan comment
joaosaffran-zz Nov 25, 2025
f51add7
addressing comments from damyan
joaosaffran-zz Nov 25, 2025
360f6ff
updating code to follow llvm guide rules
joaosaffran Nov 26, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
1 change: 1 addition & 0 deletions tools/clang/unittests/HLSLExec/LongVectorOps.def
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -219,5 +219,6 @@ OP(Wave, WaveMultiPrefixProduct, 1, "TestWaveMultiPrefixProduct", "", " -DFUNC_W
OP(Wave, WaveMultiPrefixBitAnd, 1, "TestWaveMultiPrefixBitAnd", "", " -DFUNC_WAVE_MULTI_PREFIX_BIT_AND=1 -DIS_WAVE_PREFIX_OP=1", "LongVectorOp", WaveMultiPrefixBitwise, Default2, Default3)
OP(Wave, WaveMultiPrefixBitOr, 1, "TestWaveMultiPrefixBitOr", "", " -DFUNC_WAVE_MULTI_PREFIX_BIT_OR=1 -DIS_WAVE_PREFIX_OP=1", "LongVectorOp", WaveMultiPrefixBitwise, Default2, Default3)
OP(Wave, WaveMultiPrefixBitXor, 1, "TestWaveMultiPrefixBitXor", "", " -DFUNC_WAVE_MULTI_PREFIX_BIT_XOR=1 -DIS_WAVE_PREFIX_OP=1", "LongVectorOp", WaveMultiPrefixBitwise, Default2, Default3)
OP_DEFAULT_DEFINES(Wave, WaveMatch, 1, "TestWaveMatch", "", " -DFUNC_WAVE_MATCH=1 -DIS_WAVE_PREFIX_OP=1")
Copy link
Collaborator Author

@joaosaffran joaosaffran Nov 25, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

-DIS_WAVE_PREFIX_OP=1 is required, the test function returns void, since it handles writing to the out vector inside the test function, instead of delegating to main


#undef OP
53 changes: 53 additions & 0 deletions tools/clang/unittests/HLSLExec/LongVectors.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1542,6 +1542,49 @@ template <typename T> T waveMultiPrefixProduct(T A, UINT) {
return A * A;
}

template <typename T> struct Op<OpType::WaveMatch, T, 1> : StrictValidation {};

template <typename T> struct ExpectedBuilder<OpType::WaveMatch, T> {
static std::vector<UINT> buildExpected(Op<OpType::WaveMatch, T, 1> &,
const InputSets<T> &,
const UINT WaveSize) {
// For this test, the shader arranges it so that lane 0 is different from
// all the other lanes. Besides that all other lines write their result of
// WaveMatch as well.

std::vector<UINT> Expected;
Expected.assign(WaveSize * 4, 0);

const UINT LowWaves = std::min(64U, WaveSize);
const UINT HighWaves = WaveSize - LowWaves;

const uint64_t LowWaveMask =
(LowWaves < 64) ? (1ULL << LowWaves) - 1 : ~0ULL;

const uint64_t HighWaveMask =
(HighWaves < 64) ? (1ULL << HighWaves) - 1 : ~0ULL;

const uint64_t LowExpected = ~1ULL & LowWaveMask;
const uint64_t HighExpected = ~0ULL & HighWaveMask;

Expected[0] = 1;
Expected[1] = 0;
Expected[2] = 0;
Expected[3] = 0;

// all lanes other than the first one have the same result
for (UINT I = 1; I < WaveSize; ++I) {
const UINT Index = I * 4;
Expected[Index] = static_cast<UINT>(LowExpected);
Expected[Index + 1] = static_cast<UINT>(LowExpected >> 32);
Expected[Index + 2] = static_cast<UINT>(HighExpected);
Expected[Index + 3] = static_cast<UINT>(HighExpected >> 32);
}

return Expected;
}
};

#undef WAVE_OP

//
Expand Down Expand Up @@ -2461,6 +2504,7 @@ class DxilConf_SM69_Vectorized {
HLK_WAVEOP_TEST(WaveActiveAllEqual, HLSLBool_t);
HLK_WAVEOP_TEST(WaveReadLaneAt, HLSLBool_t);
HLK_WAVEOP_TEST(WaveReadLaneFirst, HLSLBool_t);
HLK_WAVEOP_TEST(WaveMatch, HLSLBool_t);

HLK_WAVEOP_TEST(WaveActiveSum, int16_t);
HLK_WAVEOP_TEST(WaveActiveMin, int16_t);
Expand All @@ -2476,6 +2520,7 @@ class DxilConf_SM69_Vectorized {
HLK_WAVEOP_TEST(WaveMultiPrefixBitAnd, int16_t);
HLK_WAVEOP_TEST(WaveMultiPrefixBitOr, int16_t);
HLK_WAVEOP_TEST(WaveMultiPrefixBitXor, int16_t);
HLK_WAVEOP_TEST(WaveMatch, int16_t);
HLK_WAVEOP_TEST(WaveActiveSum, int32_t);
HLK_WAVEOP_TEST(WaveActiveMin, int32_t);
HLK_WAVEOP_TEST(WaveActiveMax, int32_t);
Expand All @@ -2490,6 +2535,7 @@ class DxilConf_SM69_Vectorized {
HLK_WAVEOP_TEST(WaveMultiPrefixBitAnd, int32_t);
HLK_WAVEOP_TEST(WaveMultiPrefixBitOr, int32_t);
HLK_WAVEOP_TEST(WaveMultiPrefixBitXor, int32_t);
HLK_WAVEOP_TEST(WaveMatch, int32_t);
HLK_WAVEOP_TEST(WaveActiveSum, int64_t);
HLK_WAVEOP_TEST(WaveActiveMin, int64_t);
HLK_WAVEOP_TEST(WaveActiveMax, int64_t);
Expand All @@ -2504,6 +2550,7 @@ class DxilConf_SM69_Vectorized {
HLK_WAVEOP_TEST(WaveMultiPrefixBitAnd, int64_t);
HLK_WAVEOP_TEST(WaveMultiPrefixBitOr, int64_t);
HLK_WAVEOP_TEST(WaveMultiPrefixBitXor, int64_t);
HLK_WAVEOP_TEST(WaveMatch, int64_t);

// Note: WaveActiveBit* ops don't support uint16_t in HLSL
// But the WaveMultiPrefixBit ops support all int and uint types
Expand All @@ -2521,6 +2568,7 @@ class DxilConf_SM69_Vectorized {
HLK_WAVEOP_TEST(WaveMultiPrefixBitAnd, uint16_t);
HLK_WAVEOP_TEST(WaveMultiPrefixBitOr, uint16_t);
HLK_WAVEOP_TEST(WaveMultiPrefixBitXor, uint16_t);
HLK_WAVEOP_TEST(WaveMatch, uint16_t);
HLK_WAVEOP_TEST(WaveActiveSum, uint32_t);
HLK_WAVEOP_TEST(WaveActiveMin, uint32_t);
HLK_WAVEOP_TEST(WaveActiveMax, uint32_t);
Expand All @@ -2538,6 +2586,7 @@ class DxilConf_SM69_Vectorized {
HLK_WAVEOP_TEST(WaveMultiPrefixBitAnd, uint32_t);
HLK_WAVEOP_TEST(WaveMultiPrefixBitOr, uint32_t);
HLK_WAVEOP_TEST(WaveMultiPrefixBitXor, uint32_t);
HLK_WAVEOP_TEST(WaveMatch, uint32_t);
HLK_WAVEOP_TEST(WaveActiveSum, uint64_t);
HLK_WAVEOP_TEST(WaveActiveMin, uint64_t);
HLK_WAVEOP_TEST(WaveActiveMax, uint64_t);
Expand All @@ -2555,6 +2604,7 @@ class DxilConf_SM69_Vectorized {
HLK_WAVEOP_TEST(WaveMultiPrefixBitAnd, uint64_t);
HLK_WAVEOP_TEST(WaveMultiPrefixBitOr, uint64_t);
HLK_WAVEOP_TEST(WaveMultiPrefixBitXor, uint64_t);
HLK_WAVEOP_TEST(WaveMatch, uint64_t);

HLK_WAVEOP_TEST(WaveActiveSum, HLSLHalf_t);
HLK_WAVEOP_TEST(WaveActiveMin, HLSLHalf_t);
Expand All @@ -2567,6 +2617,7 @@ class DxilConf_SM69_Vectorized {
HLK_WAVEOP_TEST(WavePrefixProduct, HLSLHalf_t);
HLK_WAVEOP_TEST(WaveMultiPrefixSum, HLSLHalf_t);
HLK_WAVEOP_TEST(WaveMultiPrefixProduct, HLSLHalf_t);
HLK_WAVEOP_TEST(WaveMatch, HLSLHalf_t);
HLK_WAVEOP_TEST(WaveActiveSum, float);
HLK_WAVEOP_TEST(WaveActiveMin, float);
HLK_WAVEOP_TEST(WaveActiveMax, float);
Expand All @@ -2578,6 +2629,7 @@ class DxilConf_SM69_Vectorized {
HLK_WAVEOP_TEST(WavePrefixProduct, float);
HLK_WAVEOP_TEST(WaveMultiPrefixSum, float);
HLK_WAVEOP_TEST(WaveMultiPrefixProduct, float);
HLK_WAVEOP_TEST(WaveMatch, float);
HLK_WAVEOP_TEST(WaveActiveSum, double);
HLK_WAVEOP_TEST(WaveActiveMin, double);
HLK_WAVEOP_TEST(WaveActiveMax, double);
Expand All @@ -2589,6 +2641,7 @@ class DxilConf_SM69_Vectorized {
HLK_WAVEOP_TEST(WavePrefixProduct, double);
HLK_WAVEOP_TEST(WaveMultiPrefixSum, double);
HLK_WAVEOP_TEST(WaveMultiPrefixProduct, double);
HLK_WAVEOP_TEST(WaveMatch, double);

private:
bool Initialized = false;
Expand Down
19 changes: 19 additions & 0 deletions tools/clang/unittests/HLSLExec/ShaderOpArith.xml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4405,6 +4405,25 @@ void MSMain(uint GID : SV_GroupIndex,
}
#endif

#ifdef FUNC_WAVE_MATCH
void TestWaveMatch(vector<TYPE, NUM> Vector)
{
if(WaveGetLaneIndex() == 0)
{
if(Vector[0] == (TYPE)0)
Vector[0] = (TYPE) 1;
else if(Vector[0] == (TYPE)1)
Vector[0] = (TYPE) 0;
else
Vector[0] = (TYPE) 1;
}
uint4 result = WaveMatch(Vector);
uint index = WaveGetLaneIndex();

g_OutputVector.Store<uint4>(index * sizeof(uint4), result);
}
#endif

#ifdef FUNC_TEST_SELECT
vector<OUT_TYPE, NUM> TestSelect(vector<TYPE, NUM> Vector1,
vector<TYPE, NUM> Vector2,
Expand Down