Update SSE/AVX integer comparison intrinsics to be used in constexpr #155656

smoke-y · 2025-08-27T17:11:42Z

github-actions · 2025-08-27T17:12:03Z

Thank you for submitting a Pull Request (PR) to the LLVM Project!

This PR will be automatically labeled and the relevant teams will be notified.

If you wish to, you can add reviewers by using the "Reviewers" section on this page.

If this is not working for you, it is probably because you do not have write permissions for the repository. In which case you can instead tag reviewers by name in a comment by using @ followed by their GitHub username.

If you have received no comments on your PR for a week, you can request a review by "ping"ing the PR by adding a comment “Ping”. The common courtesy "ping" rate is once a week. Please remember that you are asking for valuable time from other developers.

If you have further questions, they may be answered by the LLVM GitHub User Guide.

You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums.

llvmbot · 2025-08-27T17:12:34Z

@llvm/pr-subscribers-backend-x86

Author: Smokey (smoke-y)

Changes

Closes #155413

Full diff: https://github.com/llvm/llvm-project/pull/155656.diff

7 Files Affected:

(modified) clang/lib/Headers/avx2intrin.h (+8-8)
(modified) clang/lib/Headers/emmintrin.h (+9-9)
(modified) clang/lib/Headers/smmintrin.h (+2-2)
(modified) clang/test/CodeGen/X86/avx2-builtins.c (+8)
(modified) clang/test/CodeGen/X86/sse2-builtins.c (+9)
(modified) clang/test/CodeGen/X86/sse41-builtins.c (+1)
(modified) clang/test/CodeGen/X86/sse42-builtins.c (+2)

diff --git a/clang/lib/Headers/avx2intrin.h b/clang/lib/Headers/avx2intrin.h
index ce5b2b7544d8c..05bd15385d149 100644
--- a/clang/lib/Headers/avx2intrin.h
+++ b/clang/lib/Headers/avx2intrin.h
@@ -637,7 +637,7 @@ _mm256_blendv_epi8(__m256i __V1, __m256i __V2, __m256i __M)
 /// \param __b
 ///    A 256-bit integer vector containing one of the inputs.
 /// \returns A 256-bit integer vector containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_cmpeq_epi8(__m256i __a, __m256i __b)
 {
   return (__m256i)((__v32qi)__a == (__v32qi)__b);
@@ -663,7 +663,7 @@ _mm256_cmpeq_epi8(__m256i __a, __m256i __b)
 /// \param __b
 ///    A 256-bit vector of [16 x i16] containing one of the inputs.
 /// \returns A 256-bit vector of [16 x i16] containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_cmpeq_epi16(__m256i __a, __m256i __b)
 {
   return (__m256i)((__v16hi)__a == (__v16hi)__b);
@@ -689,7 +689,7 @@ _mm256_cmpeq_epi16(__m256i __a, __m256i __b)
 /// \param __b
 ///    A 256-bit vector of [8 x i32] containing one of the inputs.
 /// \returns A 256-bit vector of [8 x i32] containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_cmpeq_epi32(__m256i __a, __m256i __b)
 {
   return (__m256i)((__v8si)__a == (__v8si)__b);
@@ -715,7 +715,7 @@ _mm256_cmpeq_epi32(__m256i __a, __m256i __b)
 /// \param __b
 ///    A 256-bit vector of [4 x i64] containing one of the inputs.
 /// \returns A 256-bit vector of [4 x i64] containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_cmpeq_epi64(__m256i __a, __m256i __b)
 {
   return (__m256i)((__v4di)__a == (__v4di)__b);
@@ -741,7 +741,7 @@ _mm256_cmpeq_epi64(__m256i __a, __m256i __b)
 /// \param __b
 ///    A 256-bit integer vector containing one of the inputs.
 /// \returns A 256-bit integer vector containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_cmpgt_epi8(__m256i __a, __m256i __b)
 {
   /* This function always performs a signed comparison, but __v32qi is a char
@@ -769,7 +769,7 @@ _mm256_cmpgt_epi8(__m256i __a, __m256i __b)
 /// \param __b
 ///    A 256-bit vector of [16 x i16] containing one of the inputs.
 /// \returns A 256-bit vector of [16 x i16] containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_cmpgt_epi16(__m256i __a, __m256i __b)
 {
   return (__m256i)((__v16hi)__a > (__v16hi)__b);
@@ -795,7 +795,7 @@ _mm256_cmpgt_epi16(__m256i __a, __m256i __b)
 /// \param __b
 ///    A 256-bit vector of [8 x i32] containing one of the inputs.
 /// \returns A 256-bit vector of [8 x i32] containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_cmpgt_epi32(__m256i __a, __m256i __b)
 {
   return (__m256i)((__v8si)__a > (__v8si)__b);
@@ -821,7 +821,7 @@ _mm256_cmpgt_epi32(__m256i __a, __m256i __b)
 /// \param __b
 ///    A 256-bit vector of [4 x i64] containing one of the inputs.
 /// \returns A 256-bit vector of [4 x i64] containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_cmpgt_epi64(__m256i __a, __m256i __b)
 {
   return (__m256i)((__v4di)__a > (__v4di)__b);
diff --git a/clang/lib/Headers/emmintrin.h b/clang/lib/Headers/emmintrin.h
index 8b6b62458dac1..7f69019e01b06 100644
--- a/clang/lib/Headers/emmintrin.h
+++ b/clang/lib/Headers/emmintrin.h
@@ -3090,7 +3090,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_srl_epi64(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi8(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpeq_epi8(__m128i __a,
                                                             __m128i __b) {
   return (__m128i)((__v16qi)__a == (__v16qi)__b);
 }
@@ -3109,7 +3109,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi8(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi16(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpeq_epi16(__m128i __a,
                                                              __m128i __b) {
   return (__m128i)((__v8hi)__a == (__v8hi)__b);
 }
@@ -3128,7 +3128,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi16(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi32(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpeq_epi32(__m128i __a,
                                                              __m128i __b) {
   return (__m128i)((__v4si)__a == (__v4si)__b);
 }
@@ -3148,7 +3148,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi32(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi8(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpgt_epi8(__m128i __a,
                                                             __m128i __b) {
   /* This function always performs a signed comparison, but __v16qi is a char
      which may be signed or unsigned, so use __v16qs. */
@@ -3170,7 +3170,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi8(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi16(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpgt_epi16(__m128i __a,
                                                              __m128i __b) {
   return (__m128i)((__v8hi)__a > (__v8hi)__b);
 }
@@ -3190,7 +3190,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi16(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi32(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpgt_epi32(__m128i __a,
                                                              __m128i __b) {
   return (__m128i)((__v4si)__a > (__v4si)__b);
 }
@@ -3210,7 +3210,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi32(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmplt_epi8(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmplt_epi8(__m128i __a,
                                                             __m128i __b) {
   return _mm_cmpgt_epi8(__b, __a);
 }
@@ -3230,7 +3230,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmplt_epi8(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmplt_epi16(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmplt_epi16(__m128i __a,
                                                              __m128i __b) {
   return _mm_cmpgt_epi16(__b, __a);
 }
@@ -3250,7 +3250,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmplt_epi16(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmplt_epi32(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmplt_epi32(__m128i __a,
                                                              __m128i __b) {
   return _mm_cmpgt_epi32(__b, __a);
 }
diff --git a/clang/lib/Headers/smmintrin.h b/clang/lib/Headers/smmintrin.h
index 57d0d329312af..3f44c786fb75f 100644
--- a/clang/lib/Headers/smmintrin.h
+++ b/clang/lib/Headers/smmintrin.h
@@ -1211,7 +1211,7 @@ static __inline__ int __DEFAULT_FN_ATTRS _mm_testnzc_si128(__m128i __M,
 /// \param __V2
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi64(__m128i __V1,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpeq_epi64(__m128i __V1,
                                                              __m128i __V2) {
   return (__m128i)((__v2di)__V1 == (__v2di)__V2);
 }
@@ -2338,7 +2338,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_minpos_epu16(__m128i __V) {
 /// \param __V2
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi64(__m128i __V1,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpgt_epi64(__m128i __V1,
                                                              __m128i __V2) {
   return (__m128i)((__v2di)__V1 > (__v2di)__V2);
 }
diff --git a/clang/test/CodeGen/X86/avx2-builtins.c b/clang/test/CodeGen/X86/avx2-builtins.c
index 49e35230ba225..84a4db9695b88 100644
--- a/clang/test/CodeGen/X86/avx2-builtins.c
+++ b/clang/test/CodeGen/X86/avx2-builtins.c
@@ -287,48 +287,56 @@ __m256i test_mm256_cmpeq_epi8(__m256i a, __m256i b) {
   // CHECK: icmp eq <32 x i8>
   return _mm256_cmpeq_epi8(a, b);
 }
+TEST_CONSTEXPR(match_v32qi(_mm256_cmpeq_epi8((__m256i)(__v32qi){1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32}, (__m256i)(__v32qi){10,2,6,4,5,12,14,8,9,20,22,12,26,14,30,16,17,36,38,20,42,22,46,24,50,26,54,28,58,30,62,32}), 0,-1,0,-1,-1,0,0,-1,-1,0,0,-1,0,-1,0,-1,-1,0,0,-1,0,-1,0,-1,0,-1,0,-1,0,-1,0,-1));
 
 __m256i test_mm256_cmpeq_epi16(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_cmpeq_epi16
   // CHECK: icmp eq <16 x i16>
   return _mm256_cmpeq_epi16(a, b);
 }
+TEST_CONSTEXPR(match_v16hi(_mm256_cmpeq_epi16((__m256i)(__v16hi){+1, -2, +3, -4, +5, -6, +7, -8, +9, -10, +11, -12, +13, -14, +15, -16}, (__m256i)(__v16hi){-10, -2, +6, -4, +5, -12, +14, -8, +9, -20, +22, -12, +26, -14, +30, -16}), 0, -1, 0, -1, -1, 0, 0, -1, -1, 0, 0, -1, 0, -1, 0, -1));
 
 __m256i test_mm256_cmpeq_epi32(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_cmpeq_epi32
   // CHECK: icmp eq <8 x i32>
   return _mm256_cmpeq_epi32(a, b);
 }
+TEST_CONSTEXPR(match_v8si(_mm256_cmpeq_epi32((__m256i)(__v8si){+1, -2, +3, -4, +5, -6, +7, -8}, (__m256i)(__v8si){-10, -2, +6, -4, +5, -12, +14, -8}), 0, -1, 0, -1, -1, 0, 0, -1));
 
 __m256i test_mm256_cmpeq_epi64(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_cmpeq_epi64
   // CHECK: icmp eq <4 x i64>
   return _mm256_cmpeq_epi64(a, b);
 }
+TEST_CONSTEXPR(match_v4di(_mm256_cmpeq_epi64((__m256i)(__v4di){+1, -2, +3, -4}, (__m256i)(__v4di){-10, -2, +6, -4}), 0, -1, 0, -1));
 
 __m256i test_mm256_cmpgt_epi8(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_cmpgt_epi8
   // CHECK: icmp sgt <32 x i8>
   return _mm256_cmpgt_epi8(a, b);
 }
+TEST_CONSTEXPR(match_v32qi(_mm256_cmpgt_epi8((__m256i)(__v32qi){1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16}, (__m256i)(__v32qi){10, 2, 6, 5, 30, 7, 8, 1, 20, 3, 12, 8, 25, 10, 9, 2, 10, 2, 6, 5, 30, 7, 8, 1, 20, 3, 12, 8, 25, 10, 9, 2}), 0, 0, 0, 0, 0, 0, 0, -1, 0, -1, 0, -1, 0, -1, -1, -1, 0, 0, 0, 0, 0, 0, 0, -1, 0, -1, 0, -1, 0, -1, -1, -1));
 
 __m256i test_mm256_cmpgt_epi16(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_cmpgt_epi16
   // CHECK: icmp sgt <16 x i16>
   return _mm256_cmpgt_epi16(a, b);
 }
+TEST_CONSTEXPR(match_v16hi(_mm256_cmpgt_epi16((__m256i)(__v16hi){+1, -2, +3, -4, +5, -6, +7, -8, +1, -2, +3, -4, +5, -6, +7, -8}, (__m256i)(__v16hi){-10, -2, +6, -5, +30, -7, +8, -1, -10, -2, +6, -5, +30, -7, +8, -1}), -1, 0, 0, -1, 0, -1, 0, 0, -1, 0, 0, -1, 0, -1, 0, 0));
 
 __m256i test_mm256_cmpgt_epi32(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_cmpgt_epi32
   // CHECK: icmp sgt <8 x i32>
   return _mm256_cmpgt_epi32(a, b);
 }
+TEST_CONSTEXPR(match_v8si(_mm256_cmpgt_epi32((__m256i)(__v8si){+1, -2, +3, -4, +5, -6, +7, -8}, (__m256i)(__v8si){-10, -2, +6, -5, +30, -7, +8, -1}), -1, 0, 0, -1, 0, -1, 0, 0));
 
 __m256i test_mm256_cmpgt_epi64(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_cmpgt_epi64
   // CHECK: icmp sgt <4 x i64>
   return _mm256_cmpgt_epi64(a, b);
 }
+TEST_CONSTEXPR(match_v4di(_mm256_cmpgt_epi64((__m256i)(__v4di){+1, -2, +3, -4}, (__m256i)(__v4di){-10, -2, +6, -5}), -1, 0, 0, -1));
 
 __m256i test_mm256_cvtepi8_epi16(__m128i a) {
   // CHECK-LABEL: test_mm256_cvtepi8_epi16
diff --git a/clang/test/CodeGen/X86/sse2-builtins.c b/clang/test/CodeGen/X86/sse2-builtins.c
index 38d5e877a5036..a578f04c97c71 100644
--- a/clang/test/CodeGen/X86/sse2-builtins.c
+++ b/clang/test/CodeGen/X86/sse2-builtins.c
@@ -243,18 +243,21 @@ __m128i test_mm_cmpeq_epi8(__m128i A, __m128i B) {
   // CHECK: icmp eq <16 x i8>
   return _mm_cmpeq_epi8(A, B);
 }
+TEST_CONSTEXPR(match_v16qi(_mm_cmpeq_epi8((__m128i)(__v16qi){1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16}, (__m128i)(__v16qi){10,2,6,4,5,12,14,8,9,20,22,12,26,14,30,16}), 0,-1,0,-1,-1,0,0,-1,-1,0,0,-1,0,-1,0,-1));
 
 __m128i test_mm_cmpeq_epi16(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_cmpeq_epi16
   // CHECK: icmp eq <8 x i16>
   return _mm_cmpeq_epi16(A, B);
 }
+TEST_CONSTEXPR(match_v8hi(_mm_cmpeq_epi16((__m128i)(__v8hi){+1, -2, +3, -4, +5, -6, +7, -8}, (__m128i)(__v8hi){-10, -2, +6, -4, +5, -12, +14, -8}), 0, -1, 0, -1, -1, 0, 0, -1));
 
 __m128i test_mm_cmpeq_epi32(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_cmpeq_epi32
   // CHECK: icmp eq <4 x i32>
   return _mm_cmpeq_epi32(A, B);
 }
+TEST_CONSTEXPR(match_v4si(_mm_cmpeq_epi32((__m128i)(__v4si){+1, -2, +3, -4}, (__m128i)(__v4si){-10, -2, +6, -4}), 0, -1, 0, -1));
 
 __m128d test_mm_cmpeq_pd(__m128d A, __m128d B) {
   // CHECK-LABEL: test_mm_cmpeq_pd
@@ -293,18 +296,21 @@ __m128i test_mm_cmpgt_epi8(__m128i A, __m128i B) {
   // CHECK: icmp sgt <16 x i8>
   return _mm_cmpgt_epi8(A, B);
 }
+TEST_CONSTEXPR(match_v16qi(_mm_cmpgt_epi8((__m128i)(__v16qi){15,2,8,4,12,6,20,8,25,10,30,12,35,14,40,16}, (__m128i)(__v16qi){10,2,6,4,5,12,14,8,9,20,22,12,26,14,30,16}), -1,0,-1,0,-1,0,-1,0,-1,0,-1,0,-1,0,-1,0));
 
 __m128i test_mm_cmpgt_epi16(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_cmpgt_epi16
   // CHECK: icmp sgt <8 x i16>
   return _mm_cmpgt_epi16(A, B);
 }
+TEST_CONSTEXPR(match_v8hi(_mm_cmpgt_epi16((__m128i)(__v8hi){15,2,8,4,12,6,20,8}, (__m128i)(__v8hi){10,2,6,4,5,12,14,8}), -1,0,-1,0,-1,0,-1,0));
 
 __m128i test_mm_cmpgt_epi32(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_cmpgt_epi32
   // CHECK: icmp sgt <4 x i32>
   return _mm_cmpgt_epi32(A, B);
 }
+TEST_CONSTEXPR(match_v4si(_mm_cmpgt_epi32((__m128i)(__v4si){15,2,8,4}, (__m128i)(__v4si){10,2,6,4}), -1,0,-1,0));
 
 __m128d test_mm_cmpgt_pd(__m128d A, __m128d B) {
   // CHECK-LABEL: test_mm_cmpgt_pd
@@ -343,18 +349,21 @@ __m128i test_mm_cmplt_epi8(__m128i A, __m128i B) {
   // CHECK: icmp sgt <16 x i8>
   return _mm_cmplt_epi8(A, B);
 }
+TEST_CONSTEXPR(match_v16qi(_mm_cmplt_epi8((__m128i)(__v16qi){5,2,3,4,1,6,7,8,9,5,11,12,13,10,15,8}, (__m128i)(__v16qi){10,2,6,4,5,12,14,8,9,20,22,12,26,14,30,16}), -1, 0, -1, 0, -1, -1, -1, 0, 0, -1, -1, 0, -1, -1, -1, -1));
 
 __m128i test_mm_cmplt_epi16(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_cmplt_epi16
   // CHECK: icmp sgt <8 x i16>
   return _mm_cmplt_epi16(A, B);
 }
+TEST_CONSTEXPR(match_v8hi(_mm_cmplt_epi16((__m128i)(__v8hi){5,2,3,4,1,6,7,8}, (__m128i)(__v8hi){10,2,6,4,5,12,14,8}), -1, 0, -1, 0, -1, -1, -1, 0));
 
 __m128i test_mm_cmplt_epi32(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_cmplt_epi32
   // CHECK: icmp sgt <4 x i32>
   return _mm_cmplt_epi32(A, B);
 }
+TEST_CONSTEXPR(match_v4si(_mm_cmplt_epi32((__m128i)(__v4si){5,2,3,4}, (__m128i)(__v4si){10,2,6,4}), -1,0,-1,0));
 
 __m128d test_mm_cmplt_pd(__m128d A, __m128d B) {
   // CHECK-LABEL: test_mm_cmplt_pd
diff --git a/clang/test/CodeGen/X86/sse41-builtins.c b/clang/test/CodeGen/X86/sse41-builtins.c
index 500b780d49057..4f9f1530ce427 100644
--- a/clang/test/CodeGen/X86/sse41-builtins.c
+++ b/clang/test/CodeGen/X86/sse41-builtins.c
@@ -79,6 +79,7 @@ __m128i test_mm_cmpeq_epi64(__m128i A, __m128i B) {
   // CHECK: sext <2 x i1> %{{.*}} to <2 x i64>
   return _mm_cmpeq_epi64(A, B);
 }
+TEST_CONSTEXPR(match_v2di(_mm_cmpeq_epi64((__m128i)(__v2di){+1, -8}, (__m128i)(__v2di){-10, -8}), 0, -1));
 
 __m128i test_mm_cvtepi8_epi16(__m128i a) {
   // CHECK-LABEL: test_mm_cvtepi8_epi16
diff --git a/clang/test/CodeGen/X86/sse42-builtins.c b/clang/test/CodeGen/X86/sse42-builtins.c
index d0c0cce33e1d0..aa598b8f78069 100644
--- a/clang/test/CodeGen/X86/sse42-builtins.c
+++ b/clang/test/CodeGen/X86/sse42-builtins.c
@@ -9,6 +9,7 @@
 
 
 #include <immintrin.h>
+#include "builtin_test_helpers.h"
 
 // NOTE: This should match the tests in llvm/test/CodeGen/X86/sse42-intrinsics-fast-isel.ll
 
@@ -59,6 +60,7 @@ __m128i test_mm_cmpgt_epi64(__m128i A, __m128i B) {
   // CHECK: icmp sgt <2 x i64>
   return _mm_cmpgt_epi64(A, B);
 }
+TEST_CONSTEXPR(match_v2di(_mm_cmpgt_epi64((__m128i)(__v2di){+1, -8}, (__m128i)(__v2di){-10, -8}), -1, 0));
 
 int test_mm_cmpistra(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_cmpistra

llvmbot · 2025-08-27T17:12:35Z

@llvm/pr-subscribers-clang

Author: Smokey (smoke-y)

Changes

Closes #155413

Full diff: https://github.com/llvm/llvm-project/pull/155656.diff

7 Files Affected:

(modified) clang/lib/Headers/avx2intrin.h (+8-8)
(modified) clang/lib/Headers/emmintrin.h (+9-9)
(modified) clang/lib/Headers/smmintrin.h (+2-2)
(modified) clang/test/CodeGen/X86/avx2-builtins.c (+8)
(modified) clang/test/CodeGen/X86/sse2-builtins.c (+9)
(modified) clang/test/CodeGen/X86/sse41-builtins.c (+1)
(modified) clang/test/CodeGen/X86/sse42-builtins.c (+2)

diff --git a/clang/lib/Headers/avx2intrin.h b/clang/lib/Headers/avx2intrin.h
index ce5b2b7544d8c..05bd15385d149 100644
--- a/clang/lib/Headers/avx2intrin.h
+++ b/clang/lib/Headers/avx2intrin.h
@@ -637,7 +637,7 @@ _mm256_blendv_epi8(__m256i __V1, __m256i __V2, __m256i __M)
 /// \param __b
 ///    A 256-bit integer vector containing one of the inputs.
 /// \returns A 256-bit integer vector containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_cmpeq_epi8(__m256i __a, __m256i __b)
 {
   return (__m256i)((__v32qi)__a == (__v32qi)__b);
@@ -663,7 +663,7 @@ _mm256_cmpeq_epi8(__m256i __a, __m256i __b)
 /// \param __b
 ///    A 256-bit vector of [16 x i16] containing one of the inputs.
 /// \returns A 256-bit vector of [16 x i16] containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_cmpeq_epi16(__m256i __a, __m256i __b)
 {
   return (__m256i)((__v16hi)__a == (__v16hi)__b);
@@ -689,7 +689,7 @@ _mm256_cmpeq_epi16(__m256i __a, __m256i __b)
 /// \param __b
 ///    A 256-bit vector of [8 x i32] containing one of the inputs.
 /// \returns A 256-bit vector of [8 x i32] containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_cmpeq_epi32(__m256i __a, __m256i __b)
 {
   return (__m256i)((__v8si)__a == (__v8si)__b);
@@ -715,7 +715,7 @@ _mm256_cmpeq_epi32(__m256i __a, __m256i __b)
 /// \param __b
 ///    A 256-bit vector of [4 x i64] containing one of the inputs.
 /// \returns A 256-bit vector of [4 x i64] containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_cmpeq_epi64(__m256i __a, __m256i __b)
 {
   return (__m256i)((__v4di)__a == (__v4di)__b);
@@ -741,7 +741,7 @@ _mm256_cmpeq_epi64(__m256i __a, __m256i __b)
 /// \param __b
 ///    A 256-bit integer vector containing one of the inputs.
 /// \returns A 256-bit integer vector containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_cmpgt_epi8(__m256i __a, __m256i __b)
 {
   /* This function always performs a signed comparison, but __v32qi is a char
@@ -769,7 +769,7 @@ _mm256_cmpgt_epi8(__m256i __a, __m256i __b)
 /// \param __b
 ///    A 256-bit vector of [16 x i16] containing one of the inputs.
 /// \returns A 256-bit vector of [16 x i16] containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_cmpgt_epi16(__m256i __a, __m256i __b)
 {
   return (__m256i)((__v16hi)__a > (__v16hi)__b);
@@ -795,7 +795,7 @@ _mm256_cmpgt_epi16(__m256i __a, __m256i __b)
 /// \param __b
 ///    A 256-bit vector of [8 x i32] containing one of the inputs.
 /// \returns A 256-bit vector of [8 x i32] containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_cmpgt_epi32(__m256i __a, __m256i __b)
 {
   return (__m256i)((__v8si)__a > (__v8si)__b);
@@ -821,7 +821,7 @@ _mm256_cmpgt_epi32(__m256i __a, __m256i __b)
 /// \param __b
 ///    A 256-bit vector of [4 x i64] containing one of the inputs.
 /// \returns A 256-bit vector of [4 x i64] containing the result.
-static __inline__ __m256i __DEFAULT_FN_ATTRS256
+static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_cmpgt_epi64(__m256i __a, __m256i __b)
 {
   return (__m256i)((__v4di)__a > (__v4di)__b);
diff --git a/clang/lib/Headers/emmintrin.h b/clang/lib/Headers/emmintrin.h
index 8b6b62458dac1..7f69019e01b06 100644
--- a/clang/lib/Headers/emmintrin.h
+++ b/clang/lib/Headers/emmintrin.h
@@ -3090,7 +3090,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_srl_epi64(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi8(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpeq_epi8(__m128i __a,
                                                             __m128i __b) {
   return (__m128i)((__v16qi)__a == (__v16qi)__b);
 }
@@ -3109,7 +3109,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi8(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi16(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpeq_epi16(__m128i __a,
                                                              __m128i __b) {
   return (__m128i)((__v8hi)__a == (__v8hi)__b);
 }
@@ -3128,7 +3128,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi16(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi32(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpeq_epi32(__m128i __a,
                                                              __m128i __b) {
   return (__m128i)((__v4si)__a == (__v4si)__b);
 }
@@ -3148,7 +3148,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi32(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi8(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpgt_epi8(__m128i __a,
                                                             __m128i __b) {
   /* This function always performs a signed comparison, but __v16qi is a char
      which may be signed or unsigned, so use __v16qs. */
@@ -3170,7 +3170,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi8(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi16(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpgt_epi16(__m128i __a,
                                                              __m128i __b) {
   return (__m128i)((__v8hi)__a > (__v8hi)__b);
 }
@@ -3190,7 +3190,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi16(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi32(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpgt_epi32(__m128i __a,
                                                              __m128i __b) {
   return (__m128i)((__v4si)__a > (__v4si)__b);
 }
@@ -3210,7 +3210,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi32(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmplt_epi8(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmplt_epi8(__m128i __a,
                                                             __m128i __b) {
   return _mm_cmpgt_epi8(__b, __a);
 }
@@ -3230,7 +3230,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmplt_epi8(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmplt_epi16(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmplt_epi16(__m128i __a,
                                                              __m128i __b) {
   return _mm_cmpgt_epi16(__b, __a);
 }
@@ -3250,7 +3250,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmplt_epi16(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmplt_epi32(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmplt_epi32(__m128i __a,
                                                              __m128i __b) {
   return _mm_cmpgt_epi32(__b, __a);
 }
diff --git a/clang/lib/Headers/smmintrin.h b/clang/lib/Headers/smmintrin.h
index 57d0d329312af..3f44c786fb75f 100644
--- a/clang/lib/Headers/smmintrin.h
+++ b/clang/lib/Headers/smmintrin.h
@@ -1211,7 +1211,7 @@ static __inline__ int __DEFAULT_FN_ATTRS _mm_testnzc_si128(__m128i __M,
 /// \param __V2
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi64(__m128i __V1,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpeq_epi64(__m128i __V1,
                                                              __m128i __V2) {
   return (__m128i)((__v2di)__V1 == (__v2di)__V2);
 }
@@ -2338,7 +2338,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_minpos_epu16(__m128i __V) {
 /// \param __V2
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpgt_epi64(__m128i __V1,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpgt_epi64(__m128i __V1,
                                                              __m128i __V2) {
   return (__m128i)((__v2di)__V1 > (__v2di)__V2);
 }
diff --git a/clang/test/CodeGen/X86/avx2-builtins.c b/clang/test/CodeGen/X86/avx2-builtins.c
index 49e35230ba225..84a4db9695b88 100644
--- a/clang/test/CodeGen/X86/avx2-builtins.c
+++ b/clang/test/CodeGen/X86/avx2-builtins.c
@@ -287,48 +287,56 @@ __m256i test_mm256_cmpeq_epi8(__m256i a, __m256i b) {
   // CHECK: icmp eq <32 x i8>
   return _mm256_cmpeq_epi8(a, b);
 }
+TEST_CONSTEXPR(match_v32qi(_mm256_cmpeq_epi8((__m256i)(__v32qi){1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32}, (__m256i)(__v32qi){10,2,6,4,5,12,14,8,9,20,22,12,26,14,30,16,17,36,38,20,42,22,46,24,50,26,54,28,58,30,62,32}), 0,-1,0,-1,-1,0,0,-1,-1,0,0,-1,0,-1,0,-1,-1,0,0,-1,0,-1,0,-1,0,-1,0,-1,0,-1,0,-1));
 
 __m256i test_mm256_cmpeq_epi16(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_cmpeq_epi16
   // CHECK: icmp eq <16 x i16>
   return _mm256_cmpeq_epi16(a, b);
 }
+TEST_CONSTEXPR(match_v16hi(_mm256_cmpeq_epi16((__m256i)(__v16hi){+1, -2, +3, -4, +5, -6, +7, -8, +9, -10, +11, -12, +13, -14, +15, -16}, (__m256i)(__v16hi){-10, -2, +6, -4, +5, -12, +14, -8, +9, -20, +22, -12, +26, -14, +30, -16}), 0, -1, 0, -1, -1, 0, 0, -1, -1, 0, 0, -1, 0, -1, 0, -1));
 
 __m256i test_mm256_cmpeq_epi32(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_cmpeq_epi32
   // CHECK: icmp eq <8 x i32>
   return _mm256_cmpeq_epi32(a, b);
 }
+TEST_CONSTEXPR(match_v8si(_mm256_cmpeq_epi32((__m256i)(__v8si){+1, -2, +3, -4, +5, -6, +7, -8}, (__m256i)(__v8si){-10, -2, +6, -4, +5, -12, +14, -8}), 0, -1, 0, -1, -1, 0, 0, -1));
 
 __m256i test_mm256_cmpeq_epi64(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_cmpeq_epi64
   // CHECK: icmp eq <4 x i64>
   return _mm256_cmpeq_epi64(a, b);
 }
+TEST_CONSTEXPR(match_v4di(_mm256_cmpeq_epi64((__m256i)(__v4di){+1, -2, +3, -4}, (__m256i)(__v4di){-10, -2, +6, -4}), 0, -1, 0, -1));
 
 __m256i test_mm256_cmpgt_epi8(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_cmpgt_epi8
   // CHECK: icmp sgt <32 x i8>
   return _mm256_cmpgt_epi8(a, b);
 }
+TEST_CONSTEXPR(match_v32qi(_mm256_cmpgt_epi8((__m256i)(__v32qi){1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16}, (__m256i)(__v32qi){10, 2, 6, 5, 30, 7, 8, 1, 20, 3, 12, 8, 25, 10, 9, 2, 10, 2, 6, 5, 30, 7, 8, 1, 20, 3, 12, 8, 25, 10, 9, 2}), 0, 0, 0, 0, 0, 0, 0, -1, 0, -1, 0, -1, 0, -1, -1, -1, 0, 0, 0, 0, 0, 0, 0, -1, 0, -1, 0, -1, 0, -1, -1, -1));
 
 __m256i test_mm256_cmpgt_epi16(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_cmpgt_epi16
   // CHECK: icmp sgt <16 x i16>
   return _mm256_cmpgt_epi16(a, b);
 }
+TEST_CONSTEXPR(match_v16hi(_mm256_cmpgt_epi16((__m256i)(__v16hi){+1, -2, +3, -4, +5, -6, +7, -8, +1, -2, +3, -4, +5, -6, +7, -8}, (__m256i)(__v16hi){-10, -2, +6, -5, +30, -7, +8, -1, -10, -2, +6, -5, +30, -7, +8, -1}), -1, 0, 0, -1, 0, -1, 0, 0, -1, 0, 0, -1, 0, -1, 0, 0));
 
 __m256i test_mm256_cmpgt_epi32(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_cmpgt_epi32
   // CHECK: icmp sgt <8 x i32>
   return _mm256_cmpgt_epi32(a, b);
 }
+TEST_CONSTEXPR(match_v8si(_mm256_cmpgt_epi32((__m256i)(__v8si){+1, -2, +3, -4, +5, -6, +7, -8}, (__m256i)(__v8si){-10, -2, +6, -5, +30, -7, +8, -1}), -1, 0, 0, -1, 0, -1, 0, 0));
 
 __m256i test_mm256_cmpgt_epi64(__m256i a, __m256i b) {
   // CHECK-LABEL: test_mm256_cmpgt_epi64
   // CHECK: icmp sgt <4 x i64>
   return _mm256_cmpgt_epi64(a, b);
 }
+TEST_CONSTEXPR(match_v4di(_mm256_cmpgt_epi64((__m256i)(__v4di){+1, -2, +3, -4}, (__m256i)(__v4di){-10, -2, +6, -5}), -1, 0, 0, -1));
 
 __m256i test_mm256_cvtepi8_epi16(__m128i a) {
   // CHECK-LABEL: test_mm256_cvtepi8_epi16
diff --git a/clang/test/CodeGen/X86/sse2-builtins.c b/clang/test/CodeGen/X86/sse2-builtins.c
index 38d5e877a5036..a578f04c97c71 100644
--- a/clang/test/CodeGen/X86/sse2-builtins.c
+++ b/clang/test/CodeGen/X86/sse2-builtins.c
@@ -243,18 +243,21 @@ __m128i test_mm_cmpeq_epi8(__m128i A, __m128i B) {
   // CHECK: icmp eq <16 x i8>
   return _mm_cmpeq_epi8(A, B);
 }
+TEST_CONSTEXPR(match_v16qi(_mm_cmpeq_epi8((__m128i)(__v16qi){1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16}, (__m128i)(__v16qi){10,2,6,4,5,12,14,8,9,20,22,12,26,14,30,16}), 0,-1,0,-1,-1,0,0,-1,-1,0,0,-1,0,-1,0,-1));
 
 __m128i test_mm_cmpeq_epi16(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_cmpeq_epi16
   // CHECK: icmp eq <8 x i16>
   return _mm_cmpeq_epi16(A, B);
 }
+TEST_CONSTEXPR(match_v8hi(_mm_cmpeq_epi16((__m128i)(__v8hi){+1, -2, +3, -4, +5, -6, +7, -8}, (__m128i)(__v8hi){-10, -2, +6, -4, +5, -12, +14, -8}), 0, -1, 0, -1, -1, 0, 0, -1));
 
 __m128i test_mm_cmpeq_epi32(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_cmpeq_epi32
   // CHECK: icmp eq <4 x i32>
   return _mm_cmpeq_epi32(A, B);
 }
+TEST_CONSTEXPR(match_v4si(_mm_cmpeq_epi32((__m128i)(__v4si){+1, -2, +3, -4}, (__m128i)(__v4si){-10, -2, +6, -4}), 0, -1, 0, -1));
 
 __m128d test_mm_cmpeq_pd(__m128d A, __m128d B) {
   // CHECK-LABEL: test_mm_cmpeq_pd
@@ -293,18 +296,21 @@ __m128i test_mm_cmpgt_epi8(__m128i A, __m128i B) {
   // CHECK: icmp sgt <16 x i8>
   return _mm_cmpgt_epi8(A, B);
 }
+TEST_CONSTEXPR(match_v16qi(_mm_cmpgt_epi8((__m128i)(__v16qi){15,2,8,4,12,6,20,8,25,10,30,12,35,14,40,16}, (__m128i)(__v16qi){10,2,6,4,5,12,14,8,9,20,22,12,26,14,30,16}), -1,0,-1,0,-1,0,-1,0,-1,0,-1,0,-1,0,-1,0));
 
 __m128i test_mm_cmpgt_epi16(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_cmpgt_epi16
   // CHECK: icmp sgt <8 x i16>
   return _mm_cmpgt_epi16(A, B);
 }
+TEST_CONSTEXPR(match_v8hi(_mm_cmpgt_epi16((__m128i)(__v8hi){15,2,8,4,12,6,20,8}, (__m128i)(__v8hi){10,2,6,4,5,12,14,8}), -1,0,-1,0,-1,0,-1,0));
 
 __m128i test_mm_cmpgt_epi32(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_cmpgt_epi32
   // CHECK: icmp sgt <4 x i32>
   return _mm_cmpgt_epi32(A, B);
 }
+TEST_CONSTEXPR(match_v4si(_mm_cmpgt_epi32((__m128i)(__v4si){15,2,8,4}, (__m128i)(__v4si){10,2,6,4}), -1,0,-1,0));
 
 __m128d test_mm_cmpgt_pd(__m128d A, __m128d B) {
   // CHECK-LABEL: test_mm_cmpgt_pd
@@ -343,18 +349,21 @@ __m128i test_mm_cmplt_epi8(__m128i A, __m128i B) {
   // CHECK: icmp sgt <16 x i8>
   return _mm_cmplt_epi8(A, B);
 }
+TEST_CONSTEXPR(match_v16qi(_mm_cmplt_epi8((__m128i)(__v16qi){5,2,3,4,1,6,7,8,9,5,11,12,13,10,15,8}, (__m128i)(__v16qi){10,2,6,4,5,12,14,8,9,20,22,12,26,14,30,16}), -1, 0, -1, 0, -1, -1, -1, 0, 0, -1, -1, 0, -1, -1, -1, -1));
 
 __m128i test_mm_cmplt_epi16(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_cmplt_epi16
   // CHECK: icmp sgt <8 x i16>
   return _mm_cmplt_epi16(A, B);
 }
+TEST_CONSTEXPR(match_v8hi(_mm_cmplt_epi16((__m128i)(__v8hi){5,2,3,4,1,6,7,8}, (__m128i)(__v8hi){10,2,6,4,5,12,14,8}), -1, 0, -1, 0, -1, -1, -1, 0));
 
 __m128i test_mm_cmplt_epi32(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_cmplt_epi32
   // CHECK: icmp sgt <4 x i32>
   return _mm_cmplt_epi32(A, B);
 }
+TEST_CONSTEXPR(match_v4si(_mm_cmplt_epi32((__m128i)(__v4si){5,2,3,4}, (__m128i)(__v4si){10,2,6,4}), -1,0,-1,0));
 
 __m128d test_mm_cmplt_pd(__m128d A, __m128d B) {
   // CHECK-LABEL: test_mm_cmplt_pd
diff --git a/clang/test/CodeGen/X86/sse41-builtins.c b/clang/test/CodeGen/X86/sse41-builtins.c
index 500b780d49057..4f9f1530ce427 100644
--- a/clang/test/CodeGen/X86/sse41-builtins.c
+++ b/clang/test/CodeGen/X86/sse41-builtins.c
@@ -79,6 +79,7 @@ __m128i test_mm_cmpeq_epi64(__m128i A, __m128i B) {
   // CHECK: sext <2 x i1> %{{.*}} to <2 x i64>
   return _mm_cmpeq_epi64(A, B);
 }
+TEST_CONSTEXPR(match_v2di(_mm_cmpeq_epi64((__m128i)(__v2di){+1, -8}, (__m128i)(__v2di){-10, -8}), 0, -1));
 
 __m128i test_mm_cvtepi8_epi16(__m128i a) {
   // CHECK-LABEL: test_mm_cvtepi8_epi16
diff --git a/clang/test/CodeGen/X86/sse42-builtins.c b/clang/test/CodeGen/X86/sse42-builtins.c
index d0c0cce33e1d0..aa598b8f78069 100644
--- a/clang/test/CodeGen/X86/sse42-builtins.c
+++ b/clang/test/CodeGen/X86/sse42-builtins.c
@@ -9,6 +9,7 @@
 
 
 #include <immintrin.h>
+#include "builtin_test_helpers.h"
 
 // NOTE: This should match the tests in llvm/test/CodeGen/X86/sse42-intrinsics-fast-isel.ll
 
@@ -59,6 +60,7 @@ __m128i test_mm_cmpgt_epi64(__m128i A, __m128i B) {
   // CHECK: icmp sgt <2 x i64>
   return _mm_cmpgt_epi64(A, B);
 }
+TEST_CONSTEXPR(match_v2di(_mm_cmpgt_epi64((__m128i)(__v2di){+1, -8}, (__m128i)(__v2di){-10, -8}), -1, 0));
 
 int test_mm_cmpistra(__m128i A, __m128i B) {
   // CHECK-LABEL: test_mm_cmpistra

github-actions · 2025-08-27T17:31:41Z

⚠️ C/C++ code formatter, clang-format found issues in your code. ⚠️

You can test this locally with the following command:

git-clang-format --diff origin/main HEAD --extensions c,h -- clang/lib/Headers/avx2intrin.h clang/lib/Headers/emmintrin.h clang/lib/Headers/smmintrin.h clang/test/CodeGen/X86/avx2-builtins.c clang/test/CodeGen/X86/sse2-builtins.c clang/test/CodeGen/X86/sse41-builtins.c clang/test/CodeGen/X86/sse42-builtins.c

⚠️
The reproduction instructions above might return results for more than one PR
in a stack if you are using a stacked PR workflow. You can limit the results by
changing origin/main to the base branch/commit you want to compare against.
⚠️

View the diff from clang-format here.

diff --git a/clang/lib/Headers/avx2intrin.h b/clang/lib/Headers/avx2intrin.h
index bcf6278fa..85a096bd1 100644
--- a/clang/lib/Headers/avx2intrin.h
+++ b/clang/lib/Headers/avx2intrin.h
@@ -635,8 +635,7 @@ _mm256_blendv_epi8(__m256i __V1, __m256i __V2, __m256i __M)
 ///    A 256-bit integer vector containing one of the inputs.
 /// \returns A 256-bit integer vector containing the result.
 static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
-_mm256_cmpeq_epi8(__m256i __a, __m256i __b)
-{
+_mm256_cmpeq_epi8(__m256i __a, __m256i __b) {
   return (__m256i)((__v32qi)__a == (__v32qi)__b);
 }
 
@@ -661,8 +660,7 @@ _mm256_cmpeq_epi8(__m256i __a, __m256i __b)
 ///    A 256-bit vector of [16 x i16] containing one of the inputs.
 /// \returns A 256-bit vector of [16 x i16] containing the result.
 static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
-_mm256_cmpeq_epi16(__m256i __a, __m256i __b)
-{
+_mm256_cmpeq_epi16(__m256i __a, __m256i __b) {
   return (__m256i)((__v16hi)__a == (__v16hi)__b);
 }
 
@@ -687,8 +685,7 @@ _mm256_cmpeq_epi16(__m256i __a, __m256i __b)
 ///    A 256-bit vector of [8 x i32] containing one of the inputs.
 /// \returns A 256-bit vector of [8 x i32] containing the result.
 static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
-_mm256_cmpeq_epi32(__m256i __a, __m256i __b)
-{
+_mm256_cmpeq_epi32(__m256i __a, __m256i __b) {
   return (__m256i)((__v8si)__a == (__v8si)__b);
 }
 
@@ -713,8 +710,7 @@ _mm256_cmpeq_epi32(__m256i __a, __m256i __b)
 ///    A 256-bit vector of [4 x i64] containing one of the inputs.
 /// \returns A 256-bit vector of [4 x i64] containing the result.
 static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
-_mm256_cmpeq_epi64(__m256i __a, __m256i __b)
-{
+_mm256_cmpeq_epi64(__m256i __a, __m256i __b) {
   return (__m256i)((__v4di)__a == (__v4di)__b);
 }
 
@@ -739,8 +735,7 @@ _mm256_cmpeq_epi64(__m256i __a, __m256i __b)
 ///    A 256-bit integer vector containing one of the inputs.
 /// \returns A 256-bit integer vector containing the result.
 static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
-_mm256_cmpgt_epi8(__m256i __a, __m256i __b)
-{
+_mm256_cmpgt_epi8(__m256i __a, __m256i __b) {
   /* This function always performs a signed comparison, but __v32qi is a char
      which may be signed or unsigned, so use __v32qs. */
   return (__m256i)((__v32qs)__a > (__v32qs)__b);
@@ -767,8 +762,7 @@ _mm256_cmpgt_epi8(__m256i __a, __m256i __b)
 ///    A 256-bit vector of [16 x i16] containing one of the inputs.
 /// \returns A 256-bit vector of [16 x i16] containing the result.
 static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
-_mm256_cmpgt_epi16(__m256i __a, __m256i __b)
-{
+_mm256_cmpgt_epi16(__m256i __a, __m256i __b) {
   return (__m256i)((__v16hi)__a > (__v16hi)__b);
 }
 
@@ -793,8 +787,7 @@ _mm256_cmpgt_epi16(__m256i __a, __m256i __b)
 ///    A 256-bit vector of [8 x i32] containing one of the inputs.
 /// \returns A 256-bit vector of [8 x i32] containing the result.
 static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
-_mm256_cmpgt_epi32(__m256i __a, __m256i __b)
-{
+_mm256_cmpgt_epi32(__m256i __a, __m256i __b) {
   return (__m256i)((__v8si)__a > (__v8si)__b);
 }
 
@@ -819,8 +812,7 @@ _mm256_cmpgt_epi32(__m256i __a, __m256i __b)
 ///    A 256-bit vector of [4 x i64] containing one of the inputs.
 /// \returns A 256-bit vector of [4 x i64] containing the result.
 static __inline__ __m256i __DEFAULT_FN_ATTRS256_CONSTEXPR
-_mm256_cmpgt_epi64(__m256i __a, __m256i __b)
-{
+_mm256_cmpgt_epi64(__m256i __a, __m256i __b) {
   return (__m256i)((__v4di)__a > (__v4di)__b);
 }

RKSimon · 2025-08-28T10:37:13Z

clang/lib/Headers/emmintrin.h

@@ -3090,7 +3090,7 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_srl_epi64(__m128i __a,
 /// \param __b
 ///    A 128-bit integer vector.
 /// \returns A 128-bit integer vector containing the comparison results.
-static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_cmpeq_epi8(__m128i __a,
+static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpeq_epi8(__m128i __a,
                                                            __m128i __b) {


these lines needs clang-formating:

static __inline__ __m128i __DEFAULT_FN_ATTRS_CONSTEXPR _mm_cmpeq_epi8(__m128i __a, __m128i __b) {

RKSimon · 2025-08-28T10:38:19Z

clang/test/CodeGen/X86/avx2-builtins.c

@@ -290,48 +290,56 @@ __m256i test_mm256_cmpeq_epi8(__m256i a, __m256i b) {
  // CHECK: icmp eq <32 x i8>
  return _mm256_cmpeq_epi8(a, b);
 }
+TEST_CONSTEXPR(match_v32qi(_mm256_cmpeq_epi8((__m256i)(__v32qi){1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32}, (__m256i)(__v32qi){10,2,6,4,5,12,14,8,9,20,22,12,26,14,30,16,17,36,38,20,42,22,46,24,50,26,54,28,58,30,62,32}), 0,-1,0,-1,-1,0,0,-1,-1,0,0,-1,0,-1,0,-1,-1,0,0,-1,0,-1,0,-1,0,-1,0,-1,0,-1,0,-1));


use __v32qs instead of __v32qi to allow you to test -ve values

RKSimon · 2025-08-28T10:38:32Z

clang/test/CodeGen/X86/sse2-builtins.c

@@ -243,18 +243,21 @@ __m128i test_mm_cmpeq_epi8(__m128i A, __m128i B) {
  // CHECK: icmp eq <16 x i8>
  return _mm_cmpeq_epi8(A, B);
 }
+TEST_CONSTEXPR(match_v16qi(_mm_cmpeq_epi8((__m128i)(__v16qi){1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16}, (__m128i)(__v16qi){10,2,6,4,5,12,14,8,9,20,22,12,26,14,30,16}), 0,-1,0,-1,-1,0,0,-1,-1,0,0,-1,0,-1,0,-1));


__v16qi -> __v16qs

RKSimon

Still a few signed char tests need converting

RKSimon · 2025-08-28T11:15:23Z

clang/test/CodeGen/X86/sse2-builtins.c

@@ -293,18 +296,21 @@ __m128i test_mm_cmpgt_epi8(__m128i A, __m128i B) {
  // CHECK: icmp sgt <16 x i8>
  return _mm_cmpgt_epi8(A, B);
 }
+TEST_CONSTEXPR(match_v16qi(_mm_cmpgt_epi8((__m128i)(__v16qi){15,2,8,4,12,6,20,8,25,10,30,12,35,14,40,16}, (__m128i)(__v16qi){10,2,6,4,5,12,14,8,9,20,22,12,26,14,30,16}), -1,0,-1,0,-1,0,-1,0,-1,0,-1,0,-1,0,-1,0));


RKSimon · 2025-08-28T11:15:31Z

clang/test/CodeGen/X86/sse2-builtins.c

@@ -343,18 +349,21 @@ __m128i test_mm_cmplt_epi8(__m128i A, __m128i B) {
  // CHECK: icmp sgt <16 x i8>
  return _mm_cmplt_epi8(A, B);
 }
+TEST_CONSTEXPR(match_v16qi(_mm_cmplt_epi8((__m128i)(__v16qi){5,2,3,4,1,6,7,8,9,5,11,12,13,10,15,8}, (__m128i)(__v16qi){10,2,6,4,5,12,14,8,9,20,22,12,26,14,30,16}), -1, 0, -1, 0, -1, -1, -1, 0, 0, -1, -1, 0, -1, -1, -1, -1));


RKSimon · 2025-08-28T11:15:49Z

clang/test/CodeGen/X86/avx2-builtins.c


 __m256i test_mm256_cmpgt_epi8(__m256i a, __m256i b) {
  // CHECK-LABEL: test_mm256_cmpgt_epi8
  // CHECK: icmp sgt <32 x i8>
  return _mm256_cmpgt_epi8(a, b);
 }
+TEST_CONSTEXPR(match_v32qi(_mm256_cmpgt_epi8((__m256i)(__v32qi){1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16}, (__m256i)(__v32qi){10, 2, 6, 5, 30, 7, 8, 1, 20, 3, 12, 8, 25, 10, 9, 2, 10, 2, 6, 5, 30, 7, 8, 1, 20, 3, 12, 8, 25, 10, 9, 2}), 0, 0, 0, 0, 0, 0, 0, -1, 0, -1, 0, -1, 0, -1, -1, -1, 0, 0, 0, 0, 0, 0, 0, -1, 0, -1, 0, -1, 0, -1, -1, -1));


smoke-y · 2025-08-28T11:33:16Z

This PR doesn't require so many commits.
Im so sorry. This is my first PR :(

RKSimon

almost there! Just don't clang-format lines of a file unless you're editing them - it makes the patch too noisy

RKSimon · 2025-08-28T11:35:00Z

clang/lib/Headers/avx2intrin.h

-#define _mm256_mpsadbw_epu8(X, Y, M) \
-  ((__m256i)__builtin_ia32_mpsadbw256((__v32qi)(__m256i)(X), \
+#define _mm256_mpsadbw_epu8(X, Y, M)                                           \
+  ((__m256i)__builtin_ia32_mpsadbw256((__v32qi)(__m256i)(X),                   \


(style) don't clang-format lines you aren't changing

RKSimon

avx2intrin.h is very broken - you would probably be better off undoing all changes in that file and redoing it

smoke-y · 2025-08-28T14:36:36Z

Hey I fixed the file(I screwed up some git command which caused it).
Modified qi -> qs.
I don't see any visible errors but if you do, please let me know. Thank you for your patience :)

RKSimon

LGTM - cheers

github-actions · 2025-08-28T15:10:46Z

@smoke-y Congratulations on having your first Pull Request (PR) merged into the LLVM Project!

Your changes will be combined with recent changes from other authors, then tested by our build bots. If there is a problem with a build, you may receive a report in an email or a comment on this PR.

Please check whether problems have been caused by your change specifically, as the builds can include changes from many authors. It is not uncommon for your change to be included in a build that fails due to someone else's changes, or infrastructure issues.

How to do this, and the rest of the post-merge process, is covered in detail here.

If your change does cause a problem, it may be reverted, or you can revert it yourself. This is a normal part of LLVM development. You can fix your changes and open a new PR to merge them again.

If you don't get any reports, no action is required from you. Your changes are working as expected, well done!

Update MMX integer comparison intrinsics to be used in constexpr

032c631

llvmbot added clang Clang issues not falling into any other category backend:X86 clang:headers Headers provided by Clang, e.g. for intrinsics labels Aug 27, 2025

RKSimon self-requested a review August 27, 2025 17:29

RKSimon changed the title ~~Update MMX integer comparison intrinsics to be used in constexpr~~ Update SSE/AVX integer comparison intrinsics to be used in constexpr Aug 27, 2025

RKSimon reviewed Aug 28, 2025

View reviewed changes

qi -> qs

75fc7f0

smoke-y force-pushed the mmxCmp branch from a41c06c to 75fc7f0 Compare August 28, 2025 11:12

RKSimon reviewed Aug 28, 2025

View reviewed changes

smoke-y added 4 commits August 28, 2025 16:53

clang-format on avx2, emmintrin, smmintrin

5b8a7f1

resolved conflict: added CONSTEXPR support back to abs

d6c72ca

resolved conflict: added CONSTEXPR support back to abs

21fee1b

formated resolved conflicts

399b182

RKSimon requested changes Aug 28, 2025

View reviewed changes

formated and changed i->s

65004c2

smoke-y force-pushed the mmxCmp branch from 8bb058d to 65004c2 Compare August 28, 2025 12:05

RKSimon requested changes Aug 28, 2025

View reviewed changes

smoke-y force-pushed the mmxCmp branch from a535dcf to 65004c2 Compare August 28, 2025 14:13

smoke-y and others added 2 commits August 28, 2025 19:56

cleaned avx2intrin.h

4226edf

Merge branch 'main' into mmxCmp

1416a5e

RKSimon approved these changes Aug 28, 2025

View reviewed changes

RKSimon enabled auto-merge (squash) August 28, 2025 15:05

RKSimon merged commit f5694f0 into llvm:main Aug 28, 2025
8 of 9 checks passed

Update SSE/AVX integer comparison intrinsics to be used in constexpr #155656

Update SSE/AVX integer comparison intrinsics to be used in constexpr #155656

Uh oh!

Conversation

smoke-y commented Aug 27, 2025

Uh oh!

github-actions bot commented Aug 27, 2025

Uh oh!

llvmbot commented Aug 27, 2025

Uh oh!

llvmbot commented Aug 27, 2025

Uh oh!

github-actions bot commented Aug 27, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

RKSimon Aug 28, 2025

Choose a reason for hiding this comment

Uh oh!

RKSimon Aug 28, 2025

Choose a reason for hiding this comment

Uh oh!

RKSimon Aug 28, 2025

Choose a reason for hiding this comment

Uh oh!

RKSimon left a comment

Choose a reason for hiding this comment

Uh oh!

RKSimon Aug 28, 2025

Choose a reason for hiding this comment

Uh oh!

RKSimon Aug 28, 2025

Choose a reason for hiding this comment

Uh oh!

RKSimon Aug 28, 2025

Choose a reason for hiding this comment

Uh oh!

smoke-y commented Aug 28, 2025

Uh oh!

RKSimon left a comment

Choose a reason for hiding this comment

Uh oh!

RKSimon Aug 28, 2025

Choose a reason for hiding this comment

Uh oh!

RKSimon left a comment

Choose a reason for hiding this comment

Uh oh!

smoke-y commented Aug 28, 2025

Uh oh!

RKSimon left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

github-actions bot commented Aug 28, 2025

Uh oh!

Uh oh!

github-actions bot commented Aug 27, 2025 •

edited

Loading