[clang] [Clang] VectorExprEvaluator::VisitCallExpr / InterpretBuiltin - Allow AVX512 conflict intrinsics to be used in constexpr (PR #163293)

[Simon Pilgrim via cfe-commits]

================
@@ -14,37 +14,53 @@
 __m512i test_mm512_conflict_epi64(__m512i __A) {
   // CHECK-LABEL: test_mm512_conflict_epi64
   // CHECK: call {{.*}}<8 x i64> @llvm.x86.avx512.conflict.q.512(<8 x i64> %{{.*}})
-  return _mm512_conflict_epi64(__A); 
+  return _mm512_conflict_epi64(__A);
 }
+
+TEST_CONSTEXPR(match_v8di(_mm512_conflict_epi64((__m512i)(__v8di){1, 2, 1, 3, 2, 4, 1, 5}), 0, 0, 1, 0, 2, 0, 5, 0));
+TEST_CONSTEXPR(match_v8di(_mm512_conflict_epi64((__m512i)(__v8di){5, 5, 5, 5, 5, 5, 5, 5}), 0, 1, 3, 7, 15, 31, 63, 127));
+TEST_CONSTEXPR(match_v8di(_mm512_conflict_epi64((__m512i)(__v8di){1, 2, 3, 4, 5, 6, 7, 8}), 0, 0, 0, 0, 0, 0, 0, 0));
 __m512i test_mm512_mask_conflict_epi64(__m512i __W, __mmask8 __U, __m512i __A) {
   // CHECK-LABEL: test_mm512_mask_conflict_epi64
   // CHECK: call {{.*}}<8 x i64> @llvm.x86.avx512.conflict.q.512(<8 x i64> %{{.*}})
   // CHECK: select <8 x i1> %{{.*}}, <8 x i64> %{{.*}}, <8 x i64> %{{.*}}
-  return _mm512_mask_conflict_epi64(__W,__U,__A); 
+  return _mm512_mask_conflict_epi64(__W,__U,__A);
 }
+
+TEST_CONSTEXPR(match_v8di(_mm512_mask_conflict_epi64(_mm512_set1_epi64((long long)0xFF), 0x55, (__m512i)(__v8di){1, 2, 1, 3, 2, 4, 1, 5}), 0, 0xFF, 1, 0xFF, 2, 0xFF, 5, 0xFF));
----------------
RKSimon wrote:

(style) don't use _mm512_set1_epi64 - use raw vector initialization (same for other cases below) 

https://github.com/llvm/llvm-project/pull/163293