From 15e97642cec6b169f580ddb03c27925b40b3fb58 Mon Sep 17 00:00:00 2001
From: Thurston Dang <thurston@google.com>
Date: Wed, 5 Mar 2025 04:45:35 +0000
Subject: [PATCH 1/2] [msan] Handle Arm NEON pairwise min/max instructions

Change the handling of:
- llvm.aarch64.neon.fmaxp
- llvm.aarch64.neon.fminp
- llvm.aarch64.neon.fmaxnmp
- llvm.aarch64.neon.fminnmp
- llvm.aarch64.neon.smaxp
- llvm.aarch64.neon.sminp
- llvm.aarch64.neon.umaxp
- llvm.aarch64.neon.uminp
from the incorrect heuristic handler (maybeHandleSimpleNomemIntrinsic)
to handlePairwiseShadowOrIntrinsic.

Updates the tests from https://github.com/llvm/llvm-project/pull/129760
---
 .../Instrumentation/MemorySanitizer.cpp       |  12 ++
 .../MemorySanitizer/AArch64/arm64-vmax.ll     | 150 +++++++++++++-----
 2 files changed, 122 insertions(+), 40 deletions(-)

diff --git a/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp b/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
index 52e42932fc751..6f1d1a97867a7 100644
--- a/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
+++ b/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
@@ -4871,6 +4871,18 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
       setOrigin(&I, getCleanOrigin());
       break;
 
+    // TODO: handling max/min similarly to AND/OR may be more precise
+    // Floating-Point Maximum/Minimum Pairwise
+    case Intrinsic::aarch64_neon_fmaxp:
+    case Intrinsic::aarch64_neon_fminp:
+    // Floating-Point Maximum/Minimum Number Pairwise
+    case Intrinsic::aarch64_neon_fmaxnmp:
+    case Intrinsic::aarch64_neon_fminnmp:
+    // Signed/Unsigned Maximum/Minimum Pairwise
+    case Intrinsic::aarch64_neon_smaxp:
+    case Intrinsic::aarch64_neon_sminp:
+    case Intrinsic::aarch64_neon_umaxp:
+    case Intrinsic::aarch64_neon_uminp:
     // Add Pairwise
     case Intrinsic::aarch64_neon_addp:
     // Floating-point Add Pairwise
diff --git a/llvm/test/Instrumentation/MemorySanitizer/AArch64/arm64-vmax.ll b/llvm/test/Instrumentation/MemorySanitizer/AArch64/arm64-vmax.ll
index a46a5b1c747db..e2457c0a51d46 100644
--- a/llvm/test/Instrumentation/MemorySanitizer/AArch64/arm64-vmax.ll
+++ b/llvm/test/Instrumentation/MemorySanitizer/AArch64/arm64-vmax.ll
@@ -3,7 +3,8 @@
 ;
 ; Forked from llvm/test/CodeGen/AArch64/arm64-vmax.ll
 ;
-; Pairwise instructions which are handled incorrectly by heuristics:
+; Pairwise instructions which are explicitly handled by
+; handlePairwiseShadowOrIntrinsic:
 ; - llvm.aarch64.neon.fmaxp (floating-point maximum pairwise)
 ; - llvm.aarch64.neon.fminp
 ; - llvm.aarch64.neon.fmaxnmp (floating-point maximum number pairwise)
@@ -1017,7 +1018,9 @@ define <8 x i8> @smaxp_8b(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <8 x i8>, ptr [[TMP12]], align 8
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <8 x i8> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <8 x i8> [[_MSLD]], <8 x i8> [[_MSLD1]], <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <8 x i8> [[_MSLD]], <8 x i8> [[_MSLD1]], <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <8 x i8> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <8 x i8> @llvm.aarch64.neon.smaxp.v8i8(<8 x i8> [[TMPVAR1]], <8 x i8> [[TMPVAR2]])
 ; CHECK-NEXT:    store <8 x i8> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <8 x i8> [[TMPVAR3]]
@@ -1056,7 +1059,9 @@ define <16 x i8> @smaxp_16b(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <16 x i8>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <16 x i8> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <16 x i8> [[_MSLD]], <16 x i8> [[_MSLD1]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <16 x i8> [[_MSLD]], <16 x i8> [[_MSLD1]], <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <16 x i8> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <16 x i8> @llvm.aarch64.neon.smaxp.v16i8(<16 x i8> [[TMPVAR1]], <16 x i8> [[TMPVAR2]])
 ; CHECK-NEXT:    store <16 x i8> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <16 x i8> [[TMPVAR3]]
@@ -1095,7 +1100,9 @@ define <4 x i16> @smaxp_4h(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <4 x i16>, ptr [[TMP12]], align 8
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i16> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i16> [[_MSLD]], <4 x i16> [[_MSLD1]], <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i16> [[_MSLD]], <4 x i16> [[_MSLD1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i16> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <4 x i16> @llvm.aarch64.neon.smaxp.v4i16(<4 x i16> [[TMPVAR1]], <4 x i16> [[TMPVAR2]])
 ; CHECK-NEXT:    store <4 x i16> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <4 x i16> [[TMPVAR3]]
@@ -1134,7 +1141,9 @@ define <8 x i16> @smaxp_8h(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <8 x i16>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <8 x i16> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <8 x i16> [[_MSLD]], <8 x i16> [[_MSLD1]], <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <8 x i16> [[_MSLD]], <8 x i16> [[_MSLD1]], <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <8 x i16> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <8 x i16> @llvm.aarch64.neon.smaxp.v8i16(<8 x i16> [[TMPVAR1]], <8 x i16> [[TMPVAR2]])
 ; CHECK-NEXT:    store <8 x i16> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <8 x i16> [[TMPVAR3]]
@@ -1173,7 +1182,9 @@ define <2 x i32> @smaxp_2s(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <2 x i32>, ptr [[TMP12]], align 8
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i32> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <2 x i32> [[_MSLD]], <2 x i32> [[_MSLD1]], <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <2 x i32> [[_MSLD]], <2 x i32> [[_MSLD1]], <2 x i32> <i32 1, i32 3>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i32> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <2 x i32> @llvm.aarch64.neon.smaxp.v2i32(<2 x i32> [[TMPVAR1]], <2 x i32> [[TMPVAR2]])
 ; CHECK-NEXT:    store <2 x i32> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <2 x i32> [[TMPVAR3]]
@@ -1212,7 +1223,9 @@ define <4 x i32> @smaxp_4s(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <4 x i32>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i32> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i32> [[_MSLD]], <4 x i32> [[_MSLD1]], <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i32> [[_MSLD]], <4 x i32> [[_MSLD1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i32> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <4 x i32> @llvm.aarch64.neon.smaxp.v4i32(<4 x i32> [[TMPVAR1]], <4 x i32> [[TMPVAR2]])
 ; CHECK-NEXT:    store <4 x i32> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <4 x i32> [[TMPVAR3]]
@@ -1258,7 +1271,9 @@ define <8 x i8> @umaxp_8b(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <8 x i8>, ptr [[TMP12]], align 8
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <8 x i8> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <8 x i8> [[_MSLD]], <8 x i8> [[_MSLD1]], <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <8 x i8> [[_MSLD]], <8 x i8> [[_MSLD1]], <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <8 x i8> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <8 x i8> @llvm.aarch64.neon.umaxp.v8i8(<8 x i8> [[TMPVAR1]], <8 x i8> [[TMPVAR2]])
 ; CHECK-NEXT:    store <8 x i8> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <8 x i8> [[TMPVAR3]]
@@ -1297,7 +1312,9 @@ define <16 x i8> @umaxp_16b(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <16 x i8>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <16 x i8> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <16 x i8> [[_MSLD]], <16 x i8> [[_MSLD1]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <16 x i8> [[_MSLD]], <16 x i8> [[_MSLD1]], <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <16 x i8> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <16 x i8> @llvm.aarch64.neon.umaxp.v16i8(<16 x i8> [[TMPVAR1]], <16 x i8> [[TMPVAR2]])
 ; CHECK-NEXT:    store <16 x i8> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <16 x i8> [[TMPVAR3]]
@@ -1336,7 +1353,9 @@ define <4 x i16> @umaxp_4h(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <4 x i16>, ptr [[TMP12]], align 8
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i16> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i16> [[_MSLD]], <4 x i16> [[_MSLD1]], <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i16> [[_MSLD]], <4 x i16> [[_MSLD1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i16> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <4 x i16> @llvm.aarch64.neon.umaxp.v4i16(<4 x i16> [[TMPVAR1]], <4 x i16> [[TMPVAR2]])
 ; CHECK-NEXT:    store <4 x i16> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <4 x i16> [[TMPVAR3]]
@@ -1375,7 +1394,9 @@ define <8 x i16> @umaxp_8h(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <8 x i16>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <8 x i16> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <8 x i16> [[_MSLD]], <8 x i16> [[_MSLD1]], <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <8 x i16> [[_MSLD]], <8 x i16> [[_MSLD1]], <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <8 x i16> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <8 x i16> @llvm.aarch64.neon.umaxp.v8i16(<8 x i16> [[TMPVAR1]], <8 x i16> [[TMPVAR2]])
 ; CHECK-NEXT:    store <8 x i16> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <8 x i16> [[TMPVAR3]]
@@ -1414,7 +1435,9 @@ define <2 x i32> @umaxp_2s(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <2 x i32>, ptr [[TMP12]], align 8
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i32> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <2 x i32> [[_MSLD]], <2 x i32> [[_MSLD1]], <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <2 x i32> [[_MSLD]], <2 x i32> [[_MSLD1]], <2 x i32> <i32 1, i32 3>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i32> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <2 x i32> @llvm.aarch64.neon.umaxp.v2i32(<2 x i32> [[TMPVAR1]], <2 x i32> [[TMPVAR2]])
 ; CHECK-NEXT:    store <2 x i32> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <2 x i32> [[TMPVAR3]]
@@ -1453,7 +1476,9 @@ define <4 x i32> @umaxp_4s(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <4 x i32>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i32> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i32> [[_MSLD]], <4 x i32> [[_MSLD1]], <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i32> [[_MSLD]], <4 x i32> [[_MSLD1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i32> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <4 x i32> @llvm.aarch64.neon.umaxp.v4i32(<4 x i32> [[TMPVAR1]], <4 x i32> [[TMPVAR2]])
 ; CHECK-NEXT:    store <4 x i32> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <4 x i32> [[TMPVAR3]]
@@ -1500,7 +1525,9 @@ define <8 x i8> @sminp_8b(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <8 x i8>, ptr [[TMP12]], align 8
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <8 x i8> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <8 x i8> [[_MSLD]], <8 x i8> [[_MSLD1]], <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <8 x i8> [[_MSLD]], <8 x i8> [[_MSLD1]], <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <8 x i8> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <8 x i8> @llvm.aarch64.neon.sminp.v8i8(<8 x i8> [[TMPVAR1]], <8 x i8> [[TMPVAR2]])
 ; CHECK-NEXT:    store <8 x i8> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <8 x i8> [[TMPVAR3]]
@@ -1539,7 +1566,9 @@ define <16 x i8> @sminp_16b(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <16 x i8>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <16 x i8> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <16 x i8> [[_MSLD]], <16 x i8> [[_MSLD1]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <16 x i8> [[_MSLD]], <16 x i8> [[_MSLD1]], <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <16 x i8> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <16 x i8> @llvm.aarch64.neon.sminp.v16i8(<16 x i8> [[TMPVAR1]], <16 x i8> [[TMPVAR2]])
 ; CHECK-NEXT:    store <16 x i8> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <16 x i8> [[TMPVAR3]]
@@ -1578,7 +1607,9 @@ define <4 x i16> @sminp_4h(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <4 x i16>, ptr [[TMP12]], align 8
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i16> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i16> [[_MSLD]], <4 x i16> [[_MSLD1]], <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i16> [[_MSLD]], <4 x i16> [[_MSLD1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i16> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <4 x i16> @llvm.aarch64.neon.sminp.v4i16(<4 x i16> [[TMPVAR1]], <4 x i16> [[TMPVAR2]])
 ; CHECK-NEXT:    store <4 x i16> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <4 x i16> [[TMPVAR3]]
@@ -1617,7 +1648,9 @@ define <8 x i16> @sminp_8h(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <8 x i16>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <8 x i16> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <8 x i16> [[_MSLD]], <8 x i16> [[_MSLD1]], <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <8 x i16> [[_MSLD]], <8 x i16> [[_MSLD1]], <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <8 x i16> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <8 x i16> @llvm.aarch64.neon.sminp.v8i16(<8 x i16> [[TMPVAR1]], <8 x i16> [[TMPVAR2]])
 ; CHECK-NEXT:    store <8 x i16> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <8 x i16> [[TMPVAR3]]
@@ -1656,7 +1689,9 @@ define <2 x i32> @sminp_2s(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <2 x i32>, ptr [[TMP12]], align 8
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i32> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <2 x i32> [[_MSLD]], <2 x i32> [[_MSLD1]], <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <2 x i32> [[_MSLD]], <2 x i32> [[_MSLD1]], <2 x i32> <i32 1, i32 3>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i32> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <2 x i32> @llvm.aarch64.neon.sminp.v2i32(<2 x i32> [[TMPVAR1]], <2 x i32> [[TMPVAR2]])
 ; CHECK-NEXT:    store <2 x i32> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <2 x i32> [[TMPVAR3]]
@@ -1695,7 +1730,9 @@ define <4 x i32> @sminp_4s(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <4 x i32>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i32> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i32> [[_MSLD]], <4 x i32> [[_MSLD1]], <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i32> [[_MSLD]], <4 x i32> [[_MSLD1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i32> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <4 x i32> @llvm.aarch64.neon.sminp.v4i32(<4 x i32> [[TMPVAR1]], <4 x i32> [[TMPVAR2]])
 ; CHECK-NEXT:    store <4 x i32> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <4 x i32> [[TMPVAR3]]
@@ -1741,7 +1778,9 @@ define <8 x i8> @uminp_8b(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <8 x i8>, ptr [[TMP12]], align 8
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <8 x i8> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <8 x i8> [[_MSLD]], <8 x i8> [[_MSLD1]], <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <8 x i8> [[_MSLD]], <8 x i8> [[_MSLD1]], <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <8 x i8> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <8 x i8> @llvm.aarch64.neon.uminp.v8i8(<8 x i8> [[TMPVAR1]], <8 x i8> [[TMPVAR2]])
 ; CHECK-NEXT:    store <8 x i8> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <8 x i8> [[TMPVAR3]]
@@ -1780,7 +1819,9 @@ define <16 x i8> @uminp_16b(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <16 x i8>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <16 x i8> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <16 x i8> [[_MSLD]], <16 x i8> [[_MSLD1]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <16 x i8> [[_MSLD]], <16 x i8> [[_MSLD1]], <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <16 x i8> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <16 x i8> @llvm.aarch64.neon.uminp.v16i8(<16 x i8> [[TMPVAR1]], <16 x i8> [[TMPVAR2]])
 ; CHECK-NEXT:    store <16 x i8> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <16 x i8> [[TMPVAR3]]
@@ -1819,7 +1860,9 @@ define <4 x i16> @uminp_4h(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <4 x i16>, ptr [[TMP12]], align 8
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i16> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i16> [[_MSLD]], <4 x i16> [[_MSLD1]], <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i16> [[_MSLD]], <4 x i16> [[_MSLD1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i16> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <4 x i16> @llvm.aarch64.neon.uminp.v4i16(<4 x i16> [[TMPVAR1]], <4 x i16> [[TMPVAR2]])
 ; CHECK-NEXT:    store <4 x i16> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <4 x i16> [[TMPVAR3]]
@@ -1858,7 +1901,9 @@ define <8 x i16> @uminp_8h(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <8 x i16>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <8 x i16> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <8 x i16> [[_MSLD]], <8 x i16> [[_MSLD1]], <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <8 x i16> [[_MSLD]], <8 x i16> [[_MSLD1]], <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <8 x i16> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <8 x i16> @llvm.aarch64.neon.uminp.v8i16(<8 x i16> [[TMPVAR1]], <8 x i16> [[TMPVAR2]])
 ; CHECK-NEXT:    store <8 x i16> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <8 x i16> [[TMPVAR3]]
@@ -1897,7 +1942,9 @@ define <2 x i32> @uminp_2s(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <2 x i32>, ptr [[TMP12]], align 8
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i32> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <2 x i32> [[_MSLD]], <2 x i32> [[_MSLD1]], <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <2 x i32> [[_MSLD]], <2 x i32> [[_MSLD1]], <2 x i32> <i32 1, i32 3>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i32> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <2 x i32> @llvm.aarch64.neon.uminp.v2i32(<2 x i32> [[TMPVAR1]], <2 x i32> [[TMPVAR2]])
 ; CHECK-NEXT:    store <2 x i32> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <2 x i32> [[TMPVAR3]]
@@ -1936,7 +1983,9 @@ define <4 x i32> @uminp_4s(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <4 x i32>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i32> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i32> [[_MSLD]], <4 x i32> [[_MSLD1]], <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i32> [[_MSLD]], <4 x i32> [[_MSLD1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i32> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <4 x i32> @llvm.aarch64.neon.uminp.v4i32(<4 x i32> [[TMPVAR1]], <4 x i32> [[TMPVAR2]])
 ; CHECK-NEXT:    store <4 x i32> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <4 x i32> [[TMPVAR3]]
@@ -2103,7 +2152,9 @@ define <2 x float> @fmaxp_2s(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <2 x i32>, ptr [[TMP12]], align 8
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i32> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <2 x i32> [[_MSLD]], <2 x i32> [[_MSLD1]], <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <2 x i32> [[_MSLD]], <2 x i32> [[_MSLD1]], <2 x i32> <i32 1, i32 3>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i32> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <2 x float> @llvm.aarch64.neon.fmaxp.v2f32(<2 x float> [[TMPVAR1]], <2 x float> [[TMPVAR2]])
 ; CHECK-NEXT:    store <2 x i32> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <2 x float> [[TMPVAR3]]
@@ -2142,7 +2193,9 @@ define <4 x float> @fmaxp_4s(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <4 x i32>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i32> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i32> [[_MSLD]], <4 x i32> [[_MSLD1]], <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i32> [[_MSLD]], <4 x i32> [[_MSLD1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i32> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <4 x float> @llvm.aarch64.neon.fmaxp.v4f32(<4 x float> [[TMPVAR1]], <4 x float> [[TMPVAR2]])
 ; CHECK-NEXT:    store <4 x i32> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <4 x float> [[TMPVAR3]]
@@ -2181,7 +2234,9 @@ define <2 x double> @fmaxp_2d(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <2 x i64>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i64> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <2 x i64> [[_MSLD]], <2 x i64> [[_MSLD1]], <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <2 x i64> [[_MSLD]], <2 x i64> [[_MSLD1]], <2 x i32> <i32 1, i32 3>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i64> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <2 x double> @llvm.aarch64.neon.fmaxp.v2f64(<2 x double> [[TMPVAR1]], <2 x double> [[TMPVAR2]])
 ; CHECK-NEXT:    store <2 x i64> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <2 x double> [[TMPVAR3]]
@@ -2345,7 +2400,9 @@ define <2 x float> @fminp_2s(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <2 x i32>, ptr [[TMP12]], align 8
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i32> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <2 x i32> [[_MSLD]], <2 x i32> [[_MSLD1]], <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <2 x i32> [[_MSLD]], <2 x i32> [[_MSLD1]], <2 x i32> <i32 1, i32 3>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i32> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <2 x float> @llvm.aarch64.neon.fminp.v2f32(<2 x float> [[TMPVAR1]], <2 x float> [[TMPVAR2]])
 ; CHECK-NEXT:    store <2 x i32> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <2 x float> [[TMPVAR3]]
@@ -2384,7 +2441,9 @@ define <4 x float> @fminp_4s(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <4 x i32>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i32> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i32> [[_MSLD]], <4 x i32> [[_MSLD1]], <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i32> [[_MSLD]], <4 x i32> [[_MSLD1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i32> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <4 x float> @llvm.aarch64.neon.fminp.v4f32(<4 x float> [[TMPVAR1]], <4 x float> [[TMPVAR2]])
 ; CHECK-NEXT:    store <4 x i32> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <4 x float> [[TMPVAR3]]
@@ -2423,7 +2482,9 @@ define <2 x double> @fminp_2d(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <2 x i64>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i64> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <2 x i64> [[_MSLD]], <2 x i64> [[_MSLD1]], <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <2 x i64> [[_MSLD]], <2 x i64> [[_MSLD1]], <2 x i32> <i32 1, i32 3>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i64> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <2 x double> @llvm.aarch64.neon.fminp.v2f64(<2 x double> [[TMPVAR1]], <2 x double> [[TMPVAR2]])
 ; CHECK-NEXT:    store <2 x i64> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <2 x double> [[TMPVAR3]]
@@ -2466,7 +2527,9 @@ define <2 x float> @fminnmp_2s(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <2 x i32>, ptr [[TMP12]], align 8
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i32> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <2 x i32> [[_MSLD]], <2 x i32> [[_MSLD1]], <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <2 x i32> [[_MSLD]], <2 x i32> [[_MSLD1]], <2 x i32> <i32 1, i32 3>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i32> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <2 x float> @llvm.aarch64.neon.fminnmp.v2f32(<2 x float> [[TMPVAR1]], <2 x float> [[TMPVAR2]])
 ; CHECK-NEXT:    store <2 x i32> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <2 x float> [[TMPVAR3]]
@@ -2505,7 +2568,9 @@ define <4 x float> @fminnmp_4s(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <4 x i32>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i32> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i32> [[_MSLD]], <4 x i32> [[_MSLD1]], <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i32> [[_MSLD]], <4 x i32> [[_MSLD1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i32> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <4 x float> @llvm.aarch64.neon.fminnmp.v4f32(<4 x float> [[TMPVAR1]], <4 x float> [[TMPVAR2]])
 ; CHECK-NEXT:    store <4 x i32> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <4 x float> [[TMPVAR3]]
@@ -2544,7 +2609,9 @@ define <2 x double> @fminnmp_2d(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <2 x i64>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i64> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <2 x i64> [[_MSLD]], <2 x i64> [[_MSLD1]], <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <2 x i64> [[_MSLD]], <2 x i64> [[_MSLD1]], <2 x i32> <i32 1, i32 3>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i64> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <2 x double> @llvm.aarch64.neon.fminnmp.v2f64(<2 x double> [[TMPVAR1]], <2 x double> [[TMPVAR2]])
 ; CHECK-NEXT:    store <2 x i64> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <2 x double> [[TMPVAR3]]
@@ -2587,7 +2654,9 @@ define <2 x float> @fmaxnmp_2s(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <2 x i32>, ptr [[TMP12]], align 8
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i32> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <2 x i32> [[_MSLD]], <2 x i32> [[_MSLD1]], <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <2 x i32> [[_MSLD]], <2 x i32> [[_MSLD1]], <2 x i32> <i32 1, i32 3>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i32> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <2 x float> @llvm.aarch64.neon.fmaxnmp.v2f32(<2 x float> [[TMPVAR1]], <2 x float> [[TMPVAR2]])
 ; CHECK-NEXT:    store <2 x i32> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <2 x float> [[TMPVAR3]]
@@ -2626,7 +2695,9 @@ define <4 x float> @fmaxnmp_4s(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <4 x i32>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i32> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i32> [[_MSLD]], <4 x i32> [[_MSLD1]], <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i32> [[_MSLD]], <4 x i32> [[_MSLD1]], <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <4 x i32> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <4 x float> @llvm.aarch64.neon.fmaxnmp.v4f32(<4 x float> [[TMPVAR1]], <4 x float> [[TMPVAR2]])
 ; CHECK-NEXT:    store <4 x i32> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <4 x float> [[TMPVAR3]]
@@ -2665,7 +2736,9 @@ define <2 x double> @fmaxnmp_2d(ptr %A, ptr %B) nounwind #0 {
 ; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
 ; CHECK-NEXT:    [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
 ; CHECK-NEXT:    [[_MSLD1:%.*]] = load <2 x i64>, ptr [[TMP12]], align 16
-; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i64> [[_MSLD]], [[_MSLD1]]
+; CHECK-NEXT:    [[TMP13:%.*]] = shufflevector <2 x i64> [[_MSLD]], <2 x i64> [[_MSLD1]], <2 x i32> <i32 0, i32 2>
+; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <2 x i64> [[_MSLD]], <2 x i64> [[_MSLD1]], <2 x i32> <i32 1, i32 3>
+; CHECK-NEXT:    [[_MSPROP:%.*]] = or <2 x i64> [[TMP13]], [[TMP14]]
 ; CHECK-NEXT:    [[TMPVAR3:%.*]] = call <2 x double> @llvm.aarch64.neon.fmaxnmp.v2f64(<2 x double> [[TMPVAR1]], <2 x double> [[TMPVAR2]])
 ; CHECK-NEXT:    store <2 x i64> [[_MSPROP]], ptr @__msan_retval_tls, align 8
 ; CHECK-NEXT:    ret <2 x double> [[TMPVAR3]]
@@ -2681,6 +2754,3 @@ declare <4 x float> @llvm.aarch64.neon.fmaxnmp.v4f32(<4 x float>, <4 x float>) n
 declare <2 x double> @llvm.aarch64.neon.fmaxnmp.v2f64(<2 x double>, <2 x double>) nounwind readnone
 
 attributes #0 = { sanitize_memory }
-;.
-; CHECK: [[PROF1]] = !{!"branch_weights", i32 1, i32 1048575}
-;.

From ad0b18674ab364c42d7cc0d2b0e275b2c0ef69e5 Mon Sep 17 00:00:00 2001
From: Thurston Dang <thurston@google.com>
Date: Wed, 5 Mar 2025 18:38:44 +0000
Subject: [PATCH 2/2] Add note that maybeHandleSimpleNomemIntrinsic may
 incorrectly match horizontal/pairwise intrinsics

---
 llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp b/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
index 6f1d1a97867a7..de7f023e2c528 100644
--- a/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
+++ b/llvm/lib/Transforms/Instrumentation/MemorySanitizer.cpp
@@ -3129,6 +3129,9 @@ struct MemorySanitizerVisitor : public InstVisitor<MemorySanitizerVisitor> {
   /// fine).
   ///
   /// Caller guarantees that this intrinsic does not access memory.
+  ///
+  /// TODO: "horizontal"/"pairwise" intrinsics are often incorrectly matched by
+  ///       by this handler.
   [[maybe_unused]] bool
   maybeHandleSimpleNomemIntrinsic(IntrinsicInst &I,
                                   unsigned int trailingFlags) {