[clang] [Headers][X86] Allow AVX512fp16 initialization intrinsics to be used in constexpr (PR #159929)

[via cfe-commits]

llvmbot wrote:


<!--LLVM PR SUMMARY COMMENT-->

@llvm/pr-subscribers-backend-x86

Author: Shreeyash Pandey (bojle)

<details>
<summary>Changes</summary>

Fixes https://github.com/llvm/llvm-project/issues/156866

---
Full diff: https://github.com/llvm/llvm-project/pull/159929.diff


2 Files Affected:

- (modified) clang/lib/Headers/avx512vlfp16intrin.h (+8-7) 
- (modified) clang/test/CodeGen/X86/avx512vlfp16-builtins.c (+9) 


``````````diff
diff --git a/clang/lib/Headers/avx512vlfp16intrin.h b/clang/lib/Headers/avx512vlfp16intrin.h
index 8eb31eae6173b..1036a87c50109 100644
--- a/clang/lib/Headers/avx512vlfp16intrin.h
+++ b/clang/lib/Headers/avx512vlfp16intrin.h
@@ -42,7 +42,8 @@ static __inline__ _Float16 __DEFAULT_FN_ATTRS256 _mm256_cvtsh_h(__m256h __a) {
   return __a[0];
 }
 
-static __inline__ __m128h __DEFAULT_FN_ATTRS128 _mm_set_sh(_Float16 __h) {
+static __inline__ __m128h __DEFAULT_FN_ATTRS128_CONSTEXPR
+_mm_set_sh(_Float16 __h) {
   return __extension__(__m128h){__h, 0, 0, 0, 0, 0, 0, 0};
 }
 
@@ -57,23 +58,23 @@ _mm256_set1_ph(_Float16 __h) {
                             __h, __h, __h, __h, __h, __h, __h, __h};
 }
 
-static __inline __m128h __DEFAULT_FN_ATTRS128
+static __inline __m128h __DEFAULT_FN_ATTRS128_CONSTEXPR
 _mm_set_ph(_Float16 __h1, _Float16 __h2, _Float16 __h3, _Float16 __h4,
            _Float16 __h5, _Float16 __h6, _Float16 __h7, _Float16 __h8) {
   return (__m128h)(__v8hf){__h8, __h7, __h6, __h5, __h4, __h3, __h2, __h1};
 }
 
-static __inline __m256h __DEFAULT_FN_ATTRS256
+static __inline __m256h __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_set1_pch(_Float16 _Complex h) {
   return (__m256h)_mm256_set1_ps(__builtin_bit_cast(float, h));
 }
 
-static __inline __m128h __DEFAULT_FN_ATTRS128
+static __inline __m128h __DEFAULT_FN_ATTRS128_CONSTEXPR
 _mm_set1_pch(_Float16 _Complex h) {
   return (__m128h)_mm_set1_ps(__builtin_bit_cast(float, h));
 }
 
-static __inline __m256h __DEFAULT_FN_ATTRS256
+static __inline __m256h __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_set_ph(_Float16 __h1, _Float16 __h2, _Float16 __h3, _Float16 __h4,
               _Float16 __h5, _Float16 __h6, _Float16 __h7, _Float16 __h8,
               _Float16 __h9, _Float16 __h10, _Float16 __h11, _Float16 __h12,
@@ -83,13 +84,13 @@ _mm256_set_ph(_Float16 __h1, _Float16 __h2, _Float16 __h3, _Float16 __h4,
                             __h4,  __h3,  __h2,  __h1};
 }
 
-static __inline__ __m128h __DEFAULT_FN_ATTRS128
+static __inline__ __m128h __DEFAULT_FN_ATTRS128_CONSTEXPR
 _mm_setr_ph(_Float16 e0, _Float16 e1, _Float16 e2, _Float16 e3, _Float16 e4,
             _Float16 e5, _Float16 e6, _Float16 e7) {
   return _mm_set_ph(e7, e6, e5, e4, e3, e2, e1, e0);
 }
 
-static __inline__ __m256h __DEFAULT_FN_ATTRS256
+static __inline__ __m256h __DEFAULT_FN_ATTRS256_CONSTEXPR
 _mm256_setr_ph(_Float16 e0, _Float16 e1, _Float16 e2, _Float16 e3, _Float16 e4,
                _Float16 e5, _Float16 e6, _Float16 e7, _Float16 e8, _Float16 e9,
                _Float16 e10, _Float16 e11, _Float16 e12, _Float16 e13,
diff --git a/clang/test/CodeGen/X86/avx512vlfp16-builtins.c b/clang/test/CodeGen/X86/avx512vlfp16-builtins.c
index fd6ea8fe6056d..fe3f425bbf89c 100644
--- a/clang/test/CodeGen/X86/avx512vlfp16-builtins.c
+++ b/clang/test/CodeGen/X86/avx512vlfp16-builtins.c
@@ -37,6 +37,8 @@ __m128h test_mm_set_sh(_Float16 __h) {
   return _mm_set_sh(__h);
 }
 
+TEST_CONSTEXPR(match_m128h(_mm_set_sh(2.0), 2.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0));
+
 __m128h test_mm_set1_ph(_Float16 h) {
   // CHECK-LABEL: test_mm_set1_ph
   // CHECK: insertelement <8 x half> {{.*}}, i32 0
@@ -110,6 +112,7 @@ __m128h test_mm_set_ph(_Float16 __h1, _Float16 __h2, _Float16 __h3, _Float16 __h
   // CHECK: insertelement <8 x half> {{.*}}, i32 7
   return _mm_set_ph(__h1, __h2, __h3, __h4, __h5, __h6, __h7, __h8);
 }
+TEST_CONSTEXPR(match_m128h(_mm_set_ph(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0), 8.0, 7.0, 6.0, 5.0, 4.0, 3.0, 2.0, 1.0));
 
 __m256h test_mm256_set_ph(_Float16 __h1, _Float16 __h2, _Float16 __h3, _Float16 __h4,
                           _Float16 __h5, _Float16 __h6, _Float16 __h7, _Float16 __h8,
@@ -136,6 +139,8 @@ __m256h test_mm256_set_ph(_Float16 __h1, _Float16 __h2, _Float16 __h3, _Float16
                        __h9, __h10, __h11, __h12, __h13, __h14, __h15, __h16);
 }
 
+TEST_CONSTEXPR(match_m256h(_mm256_set_ph(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0), 8.0, 7.0, 6.0, 5.0, 4.0, 3.0, 2.0, 1.0, 8.0, 7.0, 6.0, 5.0, 4.0, 3.0, 2.0, 1.0));
+
 __m128h test_mm_setr_ph(_Float16 __h1, _Float16 __h2, _Float16 __h3, _Float16 __h4,
                         _Float16 __h5, _Float16 __h6, _Float16 __h7, _Float16 __h8) {
   // CHECK-LABEL: test_mm_setr_ph
@@ -150,6 +155,8 @@ __m128h test_mm_setr_ph(_Float16 __h1, _Float16 __h2, _Float16 __h3, _Float16 __
   return _mm_setr_ph(__h1, __h2, __h3, __h4, __h5, __h6, __h7, __h8);
 }
 
+TEST_CONSTEXPR(match_m128h(_mm_setr_ph(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0), 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0));
+
 __m256h test_mm256_setr_ph(_Float16 __h1, _Float16 __h2, _Float16 __h3, _Float16 __h4,
                            _Float16 __h5, _Float16 __h6, _Float16 __h7, _Float16 __h8,
                            _Float16 __h9, _Float16 __h10, _Float16 __h11, _Float16 __h12,
@@ -175,6 +182,8 @@ __m256h test_mm256_setr_ph(_Float16 __h1, _Float16 __h2, _Float16 __h3, _Float16
                         __h9, __h10, __h11, __h12, __h13, __h14, __h15, __h16);
 }
 
+TEST_CONSTEXPR(match_m256h(_mm256_setr_ph(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0), 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0));
+
 __m256h test_mm256_add_ph(__m256h __A, __m256h __B) {
   // CHECK-LABEL: test_mm256_add_ph
   // CHECK: %{{.*}} = fadd <16 x half> %{{.*}}, %{{.*}}

``````````

</details>


https://github.com/llvm/llvm-project/pull/159929