Skip to content

[Clang][X86] Add constexpr support for permute4x64_pd and permute4x64_epi64 #170442

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
RKSimon merged 11 commits into from
Dec 3, 2025
Merged

[Clang][X86] Add constexpr support for permute4x64_pd and permute4x64_epi64 #170442

Show file tree
Hide file tree
Changes from 8 commits
Commits
Show all changes
11 commits
Select commit Hold shift + click to select a range
89e3954
feat: Add constexpr support for permdi256 and permdf256
ahmednoursphinx Dec 3, 2025
b949e99
feat: add tests
ahmednoursphinx Dec 3, 2025
962ed40
chore: format files
ahmednoursphinx Dec 3, 2025
4ec125a
refactor: Fix tests and revert changes in intrinsics header file
ahmednoursphinx Dec 3, 2025
10510e2
chore: Fix formatting
ahmednoursphinx Dec 3, 2025
86b263b
chore: revert formatting changes
ahmednoursphinx Dec 3, 2025
358fb5f
refactor: move to avx2 existing container
ahmednoursphinx Dec 3, 2025
7d93e7f
chore: Update formatiing
ahmednoursphinx Dec 3, 2025
b14e20f
refactor: use brace initialization
ahmednoursphinx Dec 3, 2025
4e53bd0
Merge branch 'main' into issue_169304
ahmednoursphinx Dec 3, 2025
cd963f1
Merge branch 'main' into issue_169304
RKSimon Dec 3, 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
7 changes: 4 additions & 3 deletions clang/include/clang/Basic/BuiltinsX86.td
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -577,13 +577,14 @@ let Features = "avx2", Attributes = [NoThrow, Const, RequiredVectorWidth<256>] i
def psadbw256
: X86Builtin<
"_Vector<4, long long int>(_Vector<32, char>, _Vector<32, char>)">;
def permdf256 : X86Builtin<"_Vector<4, double>(_Vector<4, double>, _Constant int)">;
def permti256 : X86Builtin<"_Vector<4, long long int>(_Vector<4, long long int>, _Vector<4, long long int>, _Constant int)">;
def permdi256 : X86Builtin<"_Vector<4, long long int>(_Vector<4, long long int>, _Constant int)">;
}


let Features = "avx2", Attributes = [NoThrow, Const, Constexpr, RequiredVectorWidth<256>] in {
def permdf256
: X86Builtin<"_Vector<4, double>(_Vector<4, double>, _Constant int)">;
def permdi256 : X86Builtin<"_Vector<4, long long int>(_Vector<4, long long "
"int>, _Constant int)">;
def pmovmskb256 : X86Builtin<"int(_Vector<32, char>)">;
def pavgb256 : X86Builtin<"_Vector<32, unsigned char>(_Vector<32, unsigned char>, _Vector<32, unsigned char>)">;
def pavgw256 : X86Builtin<"_Vector<16, unsigned short>(_Vector<16, unsigned short>, _Vector<16, unsigned short>)">;
Expand Down
10 changes: 10 additions & 0 deletions clang/lib/AST/ByteCode/InterpBuiltin.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4944,6 +4944,16 @@ bool InterpretBuiltin(InterpState &S, CodePtr OpPC, const CallExpr *Call,
return std::make_pair(0, static_cast<int>(LaneOffset + Index));
});

case X86::BI__builtin_ia32_permdf256:
case X86::BI__builtin_ia32_permdi256:
return interp__builtin_ia32_shuffle_generic(
S, OpPC, Call, [](unsigned DstIdx, unsigned Control) {
// permute4x64 operates on 4 64-bit elements
// For element i (0-3), extract bits [2*i+1:2*i] from Control
unsigned Index = (Control >> (2 * DstIdx)) & 0x3;
return std::make_pair(0, static_cast<int>(Index));
});

case X86::BI__builtin_ia32_vpmultishiftqb128:
case X86::BI__builtin_ia32_vpmultishiftqb256:
case X86::BI__builtin_ia32_vpmultishiftqb512:
Expand Down
13 changes: 13 additions & 0 deletions clang/lib/AST/ExprConstant.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13122,6 +13122,19 @@ bool VectorExprEvaluator::VisitCallExpr(const CallExpr *E) {
return Success(R, E);
}

case X86::BI__builtin_ia32_permdf256:
case X86::BI__builtin_ia32_permdi256: {
APValue R;
if (!evalShuffleGeneric(Info, E, R, [](unsigned DstIdx, unsigned Control) {
// permute4x64 operates on 4 64-bit elements
// For element i (0-3), extract bits [2*i+1:2*i] from Control
unsigned Index = (Control >> (2 * DstIdx)) & 0x3;
return std::make_pair(0, static_cast<int>(Index));
}))
return false;
return Success(R, E);
}

case X86::BI__builtin_ia32_vpermilvarps:
case X86::BI__builtin_ia32_vpermilvarps256:
case X86::BI__builtin_ia32_vpermilvarps512: {
Expand Down
22 changes: 22 additions & 0 deletions clang/test/CodeGen/X86/avx2-builtins.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1111,12 +1111,34 @@ __m256i test_mm256_permute4x64_epi64(__m256i a) {
// CHECK: shufflevector <4 x i64> %{{.*}}, <4 x i64> poison, <4 x i32> <i32 3, i32 0, i32 2, i32 0>
return _mm256_permute4x64_epi64(a, 35);
}
// Control value 0x00: [0,0,0,0] -> broadcast element 0
TEST_CONSTEXPR(match_v4di(_mm256_permute4x64_epi64(_mm256_set_epi64x(40LL, 30LL, 20LL, 10LL), 0x00), 10LL, 10LL, 10LL, 10LL));
// Control value 0x1B: [0,1,2,3] -> reverse order [3,2,1,0] = [D,C,B,A]
TEST_CONSTEXPR(match_v4di(_mm256_permute4x64_epi64(_mm256_set_epi64x(40LL, 30LL, 20LL, 10LL), 0x1B), 40LL, 30LL, 20LL, 10LL));
// Control value 0x39: [1,2,3,0] -> rotate left [B,C,D,A]
TEST_CONSTEXPR(match_v4di(_mm256_permute4x64_epi64(_mm256_set_epi64x(40LL, 30LL, 20LL, 10LL), 0x39), 20LL, 30LL, 40LL, 10LL));
// Control value 0x12: [2,0,1,0] -> [C,A,B,A]
TEST_CONSTEXPR(match_v4di(_mm256_permute4x64_epi64(_mm256_set_epi64x(40LL, 30LL, 20LL, 10LL), 0x12), 30LL, 10LL, 20LL, 10LL));
// Control value 0xE4: [3,2,1,0] -> identity [A,B,C,D]
TEST_CONSTEXPR(match_v4di(_mm256_permute4x64_epi64(_mm256_set_epi64x(40LL, 30LL, 20LL, 10LL), 0xE4), 10LL, 20LL, 30LL, 40LL));
// Test with negative values
TEST_CONSTEXPR(match_v4di(_mm256_permute4x64_epi64(_mm256_set_epi64x(-40LL, -30LL, -20LL, -10LL), 0x1B), -40LL, -30LL, -20LL, -10LL));
Copy link
Collaborator

@RKSimon RKSimon Dec 3, 2025

Choose a reason for hiding this comment

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

Don't use _mm256_set_epi64x - use brace initialisation


__m256d test_mm256_permute4x64_pd(__m256d a) {
// CHECK-LABEL: test_mm256_permute4x64_pd
// CHECK: shufflevector <4 x double> %{{.*}}, <4 x double> poison, <4 x i32> <i32 1, i32 2, i32 1, i32 0>
return _mm256_permute4x64_pd(a, 25);
}
// Control value 0x00: [0,0,0,0] -> broadcast element 0
TEST_CONSTEXPR(match_m256d(_mm256_permute4x64_pd(_mm256_set_pd(4.0, 3.0, 2.0, 1.0), 0x00), 1.0, 1.0, 1.0, 1.0));
// Control value 0x1B: [0,1,2,3] -> reverse order [3,2,1,0] = [D,C,B,A]
TEST_CONSTEXPR(match_m256d(_mm256_permute4x64_pd(_mm256_set_pd(4.0, 3.0, 2.0, 1.0), 0x1B), 4.0, 3.0, 2.0, 1.0));
// Control value 0x39: [1,2,3,0] -> rotate left [B,C,D,A]
TEST_CONSTEXPR(match_m256d(_mm256_permute4x64_pd(_mm256_set_pd(4.0, 3.0, 2.0, 1.0), 0x39), 2.0, 3.0, 4.0, 1.0));
// Control value 0x12: [2,0,1,0] -> [C,A,B,A]
TEST_CONSTEXPR(match_m256d(_mm256_permute4x64_pd(_mm256_set_pd(4.0, 3.0, 2.0, 1.0), 0x12), 3.0, 1.0, 2.0, 1.0));
// Control value 0xE4: [3,2,1,0] -> identity [A,B,C,D]
TEST_CONSTEXPR(match_m256d(_mm256_permute4x64_pd(_mm256_set_pd(4.0, 3.0, 2.0, 1.0), 0xE4), 1.0, 2.0, 3.0, 4.0));
Copy link
Collaborator

@RKSimon RKSimon Dec 3, 2025

Choose a reason for hiding this comment

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

Don't use _mm256_set_pd - use brace initialisation


__m256i test_mm256_permutevar8x32_epi32(__m256i a, __m256i b) {
// CHECK-LABEL: test_mm256_permutevar8x32_epi32
Expand Down