[clang] [llvm] [msan] Implement support for Arm NEON vst{2,3,4} instructions (PR #99360)
Thurston Dang via cfe-commits
cfe-commits at lists.llvm.org
Wed Jul 17 10:55:09 PDT 2024
================
@@ -0,0 +1,2825 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; Test memory sanitizer instrumentation for Arm NEON VST instructions.
+;
+; RUN: opt < %s -passes=msan -msan-track-origins=2 -S | FileCheck %s
+;
+; Forked from llvm/test/CodeGen/AArch64/arm64-st1.ll
+
+target datalayout = "e-m:e-i8:8:32-i16:16:32-i64:64-i128:128-n32:64-S128"
+target triple = "aarch64--linux-android9001"
+
+; -----------------------------------------------------------------------------------------------------------------------------------------------
+
+define void @st2_8b(<8 x i8> %A, <8 x i8> %B, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st2_8b(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], ptr [[P:%.*]]) #[[ATTR0:[0-9]+]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> [[TMP3]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP8:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP9:%.*]] = xor i64 [[TMP8]], 193514046488576
+; CHECK-NEXT: [[TMP10:%.*]] = inttoptr i64 [[TMP9]] to ptr
+; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[TMP9]], 35184372088832
+; CHECK-NEXT: [[TMP12:%.*]] = and i64 [[TMP11]], -4
+; CHECK-NEXT: [[TMP13:%.*]] = inttoptr i64 [[TMP12]] to ptr
+; CHECK-NEXT: store <16 x i8> [[TMP7]], ptr [[TMP10]], align 1
+; CHECK-NEXT: [[TMP14:%.*]] = bitcast <8 x i8> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP15:%.*]] = icmp ne i64 [[TMP14]], 0
+; CHECK-NEXT: [[TMP16:%.*]] = select i1 [[TMP15]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP6]], i32 [[TMP16]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP19:%.*]], label [[TMP20:%.*]], !prof [[PROF0:![0-9]+]]
+; CHECK: 19:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4:[0-9]+]]
+; CHECK-NEXT: unreachable
+; CHECK: 20:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> [[A]], <8 x i8> [[B]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> %A, <8 x i8> %B, ptr %P)
+ ret void
+}
+
+define void @st2_8b_undefA(<8 x i8> %A, <8 x i8> %B, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st2_8b_undefA(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP1]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP6:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP7:%.*]] = xor i64 [[TMP6]], 193514046488576
+; CHECK-NEXT: [[TMP8:%.*]] = inttoptr i64 [[TMP7]] to ptr
+; CHECK-NEXT: [[TMP9:%.*]] = add i64 [[TMP7]], 35184372088832
+; CHECK-NEXT: [[TMP10:%.*]] = and i64 [[TMP9]], -4
+; CHECK-NEXT: [[TMP11:%.*]] = inttoptr i64 [[TMP10]] to ptr
+; CHECK-NEXT: store <16 x i8> [[TMP5]], ptr [[TMP8]], align 1
+; CHECK-NEXT: [[TMP12:%.*]] = bitcast <8 x i8> [[TMP1]] to i64
+; CHECK-NEXT: [[TMP13:%.*]] = icmp ne i64 [[TMP12]], 0
+; CHECK-NEXT: [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP15:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[TMP16:%.*]] = select i1 [[TMP15]], i32 [[TMP4]], i32 [[TMP14]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP17:%.*]], label [[TMP18:%.*]], !prof [[PROF0]]
+; CHECK: 17:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP4]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 18:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> undef, <8 x i8> [[B]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> undef, <8 x i8> %B, ptr %P)
+ ret void
+}
+
+define void @st2_8b_undefB(<8 x i8> %A, <8 x i8> %B, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st2_8b_undefB(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP6:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP7:%.*]] = xor i64 [[TMP6]], 193514046488576
+; CHECK-NEXT: [[TMP8:%.*]] = inttoptr i64 [[TMP7]] to ptr
+; CHECK-NEXT: [[TMP9:%.*]] = add i64 [[TMP7]], 35184372088832
+; CHECK-NEXT: [[TMP10:%.*]] = and i64 [[TMP9]], -4
+; CHECK-NEXT: [[TMP11:%.*]] = inttoptr i64 [[TMP10]] to ptr
+; CHECK-NEXT: store <16 x i8> [[TMP5]], ptr [[TMP8]], align 1
+; CHECK-NEXT: [[TMP12:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP14:%.*]], label [[TMP15:%.*]], !prof [[PROF0]]
+; CHECK: 14:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP4]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 15:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> [[A]], <8 x i8> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> %A, <8 x i8> undef, ptr %P)
+ ret void
+}
+
+define void @st2_8b_undefAB(<8 x i8> %A, <8 x i8> %B, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st2_8b_undefAB(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP3:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP4:%.*]] = xor i64 [[TMP3]], 193514046488576
+; CHECK-NEXT: [[TMP5:%.*]] = inttoptr i64 [[TMP4]] to ptr
+; CHECK-NEXT: [[TMP6:%.*]] = add i64 [[TMP4]], 35184372088832
+; CHECK-NEXT: [[TMP7:%.*]] = and i64 [[TMP6]], -4
+; CHECK-NEXT: [[TMP8:%.*]] = inttoptr i64 [[TMP7]] to ptr
+; CHECK-NEXT: store <16 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, ptr [[TMP5]], align 1
+; CHECK-NEXT: [[TMP9:%.*]] = icmp ne i64 [[TMP1]], 0
+; CHECK-NEXT: [[TMP10:%.*]] = select i1 [[TMP9]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP11:%.*]], label [[TMP12:%.*]], !prof [[PROF0]]
+; CHECK: 11:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP2]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 12:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> undef, <8 x i8> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8> undef, <8 x i8> undef, ptr %P)
+ ret void
+}
+
+define void @st3_8b(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_8b(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> [[TMP3]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <8 x i8> [[TMP5]], <8 x i8> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <16 x i8> [[TMP9]], <16 x i8> [[TMP10]], <24 x i32> <i32 0, i32 1, i32 8, i32 2, i32 3, i32 10, i32 4, i32 5, i32 12, i32 6, i32 7, i32 14, i32 8, i32 9, i32 16, i32 10, i32 11, i32 18, i32 12, i32 13, i32 20, i32 14, i32 15, i32 22>
+; CHECK-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP13:%.*]] = xor i64 [[TMP12]], 193514046488576
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP13]], 35184372088832
+; CHECK-NEXT: [[TMP16:%.*]] = and i64 [[TMP15]], -4
+; CHECK-NEXT: [[TMP17:%.*]] = inttoptr i64 [[TMP16]] to ptr
+; CHECK-NEXT: store <24 x i8> [[TMP11]], ptr [[TMP14]], align 1
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <8 x i8> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i64 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP21:%.*]] = bitcast <8 x i8> [[TMP5]] to i64
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i64 [[TMP21]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP20]]
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP8]], i32 [[TMP23]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP26:%.*]], label [[TMP27:%.*]], !prof [[PROF0]]
+; CHECK: 26:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP8]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 27:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> [[A]], <8 x i8> [[B]], <8 x i8> [[C]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P)
+ ret void
+}
+
+define void @st3_8b_undefA(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_8b_undefA(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP1]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <8 x i8> [[TMP3]], <8 x i8> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <16 x i8> [[TMP7]], <16 x i8> [[TMP8]], <24 x i32> <i32 0, i32 1, i32 8, i32 2, i32 3, i32 10, i32 4, i32 5, i32 12, i32 6, i32 7, i32 14, i32 8, i32 9, i32 16, i32 10, i32 11, i32 18, i32 12, i32 13, i32 20, i32 14, i32 15, i32 22>
+; CHECK-NEXT: [[TMP10:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: [[TMP13:%.*]] = add i64 [[TMP11]], 35184372088832
+; CHECK-NEXT: [[TMP14:%.*]] = and i64 [[TMP13]], -4
+; CHECK-NEXT: [[TMP15:%.*]] = inttoptr i64 [[TMP14]] to ptr
+; CHECK-NEXT: store <24 x i8> [[TMP9]], ptr [[TMP12]], align 1
+; CHECK-NEXT: [[TMP16:%.*]] = bitcast <8 x i8> [[TMP1]] to i64
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i64 [[TMP16]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP19:%.*]] = bitcast <8 x i8> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP20:%.*]] = icmp ne i64 [[TMP19]], 0
+; CHECK-NEXT: [[TMP21:%.*]] = select i1 [[TMP20]], i32 [[TMP4]], i32 [[TMP18]]
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP21]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP24:%.*]], label [[TMP25:%.*]], !prof [[PROF0]]
+; CHECK: 24:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 25:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> undef, <8 x i8> [[B]], <8 x i8> [[C]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> undef, <8 x i8> %B, <8 x i8> %C, ptr %P)
+ ret void
+}
+
+define void @st3_8b_undefB(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_8b_undefB(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <8 x i8> [[TMP3]], <8 x i8> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <16 x i8> [[TMP7]], <16 x i8> [[TMP8]], <24 x i32> <i32 0, i32 1, i32 8, i32 2, i32 3, i32 10, i32 4, i32 5, i32 12, i32 6, i32 7, i32 14, i32 8, i32 9, i32 16, i32 10, i32 11, i32 18, i32 12, i32 13, i32 20, i32 14, i32 15, i32 22>
+; CHECK-NEXT: [[TMP10:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: [[TMP13:%.*]] = add i64 [[TMP11]], 35184372088832
+; CHECK-NEXT: [[TMP14:%.*]] = and i64 [[TMP13]], -4
+; CHECK-NEXT: [[TMP15:%.*]] = inttoptr i64 [[TMP14]] to ptr
+; CHECK-NEXT: store <24 x i8> [[TMP9]], ptr [[TMP12]], align 1
+; CHECK-NEXT: [[TMP16:%.*]] = bitcast <8 x i8> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i64 [[TMP16]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP6]], i32 [[TMP18]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP21:%.*]], label [[TMP22:%.*]], !prof [[PROF0]]
+; CHECK: 21:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 22:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> [[A]], <8 x i8> undef, <8 x i8> [[C]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> %A, <8 x i8> undef, <8 x i8> %C, ptr %P)
+ ret void
+}
+
+define void @st3_8b_undefC(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_8b_undefC(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> [[TMP3]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <16 x i8> [[TMP7]], <16 x i8> <i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef>, <24 x i32> <i32 0, i32 1, i32 8, i32 2, i32 3, i32 10, i32 4, i32 5, i32 12, i32 6, i32 7, i32 14, i32 8, i32 9, i32 16, i32 10, i32 11, i32 18, i32 12, i32 13, i32 20, i32 14, i32 15, i32 22>
+; CHECK-NEXT: [[TMP9:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP10:%.*]] = xor i64 [[TMP9]], 193514046488576
+; CHECK-NEXT: [[TMP11:%.*]] = inttoptr i64 [[TMP10]] to ptr
+; CHECK-NEXT: [[TMP12:%.*]] = add i64 [[TMP10]], 35184372088832
+; CHECK-NEXT: [[TMP13:%.*]] = and i64 [[TMP12]], -4
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: store <24 x i8> [[TMP8]], ptr [[TMP11]], align 1
+; CHECK-NEXT: [[TMP15:%.*]] = bitcast <8 x i8> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP16:%.*]] = icmp ne i64 [[TMP15]], 0
+; CHECK-NEXT: [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP18:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP19:%.*]] = select i1 [[TMP18]], i32 [[TMP6]], i32 [[TMP17]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP20:%.*]], label [[TMP21:%.*]], !prof [[PROF0]]
+; CHECK: 20:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 21:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> [[A]], <8 x i8> [[B]], <8 x i8> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> %A, <8 x i8> %B, <8 x i8> undef, ptr %P)
+ ret void
+}
+
+define void @st3_8b_undefAB(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_8b_undefAB(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <16 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <16 x i8> [[TMP5]], <24 x i32> <i32 0, i32 1, i32 8, i32 2, i32 3, i32 10, i32 4, i32 5, i32 12, i32 6, i32 7, i32 14, i32 8, i32 9, i32 16, i32 10, i32 11, i32 18, i32 12, i32 13, i32 20, i32 14, i32 15, i32 22>
+; CHECK-NEXT: [[TMP7:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP8:%.*]] = xor i64 [[TMP7]], 193514046488576
+; CHECK-NEXT: [[TMP9:%.*]] = inttoptr i64 [[TMP8]] to ptr
+; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[TMP8]], 35184372088832
+; CHECK-NEXT: [[TMP11:%.*]] = and i64 [[TMP10]], -4
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: store <24 x i8> [[TMP6]], ptr [[TMP9]], align 1
+; CHECK-NEXT: [[TMP13:%.*]] = bitcast <8 x i8> [[TMP1]] to i64
+; CHECK-NEXT: [[TMP14:%.*]] = icmp ne i64 [[TMP13]], 0
+; CHECK-NEXT: [[TMP15:%.*]] = select i1 [[TMP14]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP16:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP4]], i32 [[TMP15]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP18:%.*]], label [[TMP19:%.*]], !prof [[PROF0]]
+; CHECK: 18:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP4]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 19:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> undef, <8 x i8> undef, <8 x i8> [[C]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> undef, <8 x i8> undef, <8 x i8> %C, ptr %P)
+ ret void
+}
+
+define void @st3_8b_undefAC(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_8b_undefAC(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP1]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <16 x i8> [[TMP5]], <16 x i8> <i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef>, <24 x i32> <i32 0, i32 1, i32 8, i32 2, i32 3, i32 10, i32 4, i32 5, i32 12, i32 6, i32 7, i32 14, i32 8, i32 9, i32 16, i32 10, i32 11, i32 18, i32 12, i32 13, i32 20, i32 14, i32 15, i32 22>
+; CHECK-NEXT: [[TMP7:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP8:%.*]] = xor i64 [[TMP7]], 193514046488576
+; CHECK-NEXT: [[TMP9:%.*]] = inttoptr i64 [[TMP8]] to ptr
+; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[TMP8]], 35184372088832
+; CHECK-NEXT: [[TMP11:%.*]] = and i64 [[TMP10]], -4
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: store <24 x i8> [[TMP6]], ptr [[TMP9]], align 1
+; CHECK-NEXT: [[TMP13:%.*]] = bitcast <8 x i8> [[TMP1]] to i64
+; CHECK-NEXT: [[TMP14:%.*]] = icmp ne i64 [[TMP13]], 0
+; CHECK-NEXT: [[TMP15:%.*]] = select i1 [[TMP14]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP16:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP4]], i32 [[TMP15]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP18:%.*]], label [[TMP19:%.*]], !prof [[PROF0]]
+; CHECK: 18:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP4]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 19:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> undef, <8 x i8> [[B]], <8 x i8> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> undef, <8 x i8> %B, <8 x i8> undef, ptr %P)
+ ret void
+}
+
+define void @st3_8b_undefBC(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_8b_undefBC(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <16 x i8> [[TMP5]], <16 x i8> <i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef, i8 -1, i8 undef>, <24 x i32> <i32 0, i32 1, i32 8, i32 2, i32 3, i32 10, i32 4, i32 5, i32 12, i32 6, i32 7, i32 14, i32 8, i32 9, i32 16, i32 10, i32 11, i32 18, i32 12, i32 13, i32 20, i32 14, i32 15, i32 22>
+; CHECK-NEXT: [[TMP7:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP8:%.*]] = xor i64 [[TMP7]], 193514046488576
+; CHECK-NEXT: [[TMP9:%.*]] = inttoptr i64 [[TMP8]] to ptr
+; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[TMP8]], 35184372088832
+; CHECK-NEXT: [[TMP11:%.*]] = and i64 [[TMP10]], -4
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: store <24 x i8> [[TMP6]], ptr [[TMP9]], align 1
+; CHECK-NEXT: [[TMP13:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP15:%.*]], label [[TMP16:%.*]], !prof [[PROF0]]
+; CHECK: 15:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP4]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 16:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> [[A]], <8 x i8> undef, <8 x i8> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> %A, <8 x i8> undef, <8 x i8> undef, ptr %P)
+ ret void
+}
+
+define void @st3_8b_undefABC(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_8b_undefABC(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP3:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP4:%.*]] = xor i64 [[TMP3]], 193514046488576
+; CHECK-NEXT: [[TMP5:%.*]] = inttoptr i64 [[TMP4]] to ptr
+; CHECK-NEXT: [[TMP6:%.*]] = add i64 [[TMP4]], 35184372088832
+; CHECK-NEXT: [[TMP7:%.*]] = and i64 [[TMP6]], -4
+; CHECK-NEXT: [[TMP8:%.*]] = inttoptr i64 [[TMP7]] to ptr
+; CHECK-NEXT: store <24 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, ptr [[TMP5]], align 1
+; CHECK-NEXT: [[TMP9:%.*]] = icmp ne i64 [[TMP1]], 0
+; CHECK-NEXT: [[TMP10:%.*]] = select i1 [[TMP9]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP11:%.*]], label [[TMP12:%.*]], !prof [[PROF0]]
+; CHECK: 11:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP2]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 12:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> undef, <8 x i8> undef, <8 x i8> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8> undef, <8 x i8> undef, <8 x i8> undef, ptr %P)
+ ret void
+}
+
+define void @st4_8b(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_8b(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], <8 x i8> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: [[TMP9:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> [[TMP3]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <8 x i8> [[TMP5]], <8 x i8> [[TMP7]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP13:%.*]] = shufflevector <16 x i8> [[TMP11]], <16 x i8> [[TMP12]], <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP14:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP15:%.*]] = xor i64 [[TMP14]], 193514046488576
+; CHECK-NEXT: [[TMP16:%.*]] = inttoptr i64 [[TMP15]] to ptr
+; CHECK-NEXT: [[TMP17:%.*]] = add i64 [[TMP15]], 35184372088832
+; CHECK-NEXT: [[TMP18:%.*]] = and i64 [[TMP17]], -4
+; CHECK-NEXT: [[TMP19:%.*]] = inttoptr i64 [[TMP18]] to ptr
+; CHECK-NEXT: store <32 x i8> [[TMP13]], ptr [[TMP16]], align 1
+; CHECK-NEXT: [[TMP20:%.*]] = bitcast <8 x i8> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP21:%.*]] = icmp ne i64 [[TMP20]], 0
+; CHECK-NEXT: [[TMP22:%.*]] = select i1 [[TMP21]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP23:%.*]] = bitcast <8 x i8> [[TMP5]] to i64
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP23]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP6]], i32 [[TMP22]]
+; CHECK-NEXT: [[TMP26:%.*]] = bitcast <8 x i8> [[TMP7]] to i64
+; CHECK-NEXT: [[TMP27:%.*]] = icmp ne i64 [[TMP26]], 0
+; CHECK-NEXT: [[TMP28:%.*]] = select i1 [[TMP27]], i32 [[TMP8]], i32 [[TMP25]]
+; CHECK-NEXT: [[TMP29:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: [[TMP30:%.*]] = select i1 [[TMP29]], i32 [[TMP10]], i32 [[TMP28]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP31:%.*]], label [[TMP32:%.*]], !prof [[PROF0]]
+; CHECK: 31:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP10]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 32:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> [[A]], <8 x i8> [[B]], <8 x i8> [[C]], <8 x i8> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P)
+ ret void
+}
+
+define void @st4_8b_undefA(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_8b_undefA(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], <8 x i8> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP1]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <8 x i8> [[TMP3]], <8 x i8> [[TMP5]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <16 x i8> [[TMP9]], <16 x i8> [[TMP10]], <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP13:%.*]] = xor i64 [[TMP12]], 193514046488576
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP13]], 35184372088832
+; CHECK-NEXT: [[TMP16:%.*]] = and i64 [[TMP15]], -4
+; CHECK-NEXT: [[TMP17:%.*]] = inttoptr i64 [[TMP16]] to ptr
+; CHECK-NEXT: store <32 x i8> [[TMP11]], ptr [[TMP14]], align 1
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <8 x i8> [[TMP1]] to i64
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i64 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP21:%.*]] = bitcast <8 x i8> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i64 [[TMP21]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP4]], i32 [[TMP20]]
+; CHECK-NEXT: [[TMP24:%.*]] = bitcast <8 x i8> [[TMP5]] to i64
+; CHECK-NEXT: [[TMP25:%.*]] = icmp ne i64 [[TMP24]], 0
+; CHECK-NEXT: [[TMP26:%.*]] = select i1 [[TMP25]], i32 [[TMP6]], i32 [[TMP23]]
+; CHECK-NEXT: [[TMP27:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[TMP28:%.*]] = select i1 [[TMP27]], i32 [[TMP8]], i32 [[TMP26]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP29:%.*]], label [[TMP30:%.*]], !prof [[PROF0]]
+; CHECK: 29:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP8]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 30:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> undef, <8 x i8> [[B]], <8 x i8> [[C]], <8 x i8> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> undef, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P)
+ ret void
+}
+
+define void @st4_8b_undefB(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_8b_undefB(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], <8 x i8> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <8 x i8> [[TMP3]], <8 x i8> [[TMP5]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <16 x i8> [[TMP9]], <16 x i8> [[TMP10]], <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP13:%.*]] = xor i64 [[TMP12]], 193514046488576
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP13]], 35184372088832
+; CHECK-NEXT: [[TMP16:%.*]] = and i64 [[TMP15]], -4
+; CHECK-NEXT: [[TMP17:%.*]] = inttoptr i64 [[TMP16]] to ptr
+; CHECK-NEXT: store <32 x i8> [[TMP11]], ptr [[TMP14]], align 1
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <8 x i8> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i64 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP21:%.*]] = bitcast <8 x i8> [[TMP5]] to i64
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i64 [[TMP21]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP20]]
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP8]], i32 [[TMP23]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP26:%.*]], label [[TMP27:%.*]], !prof [[PROF0]]
+; CHECK: 26:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP8]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 27:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> [[A]], <8 x i8> undef, <8 x i8> [[C]], <8 x i8> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> %A, <8 x i8> undef, <8 x i8> %C, <8 x i8> %D, ptr %P)
+ ret void
+}
+
+define void @st4_8b_undefC(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_8b_undefC(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], <8 x i8> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> [[TMP3]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP5]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <16 x i8> [[TMP9]], <16 x i8> [[TMP10]], <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP13:%.*]] = xor i64 [[TMP12]], 193514046488576
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP13]], 35184372088832
+; CHECK-NEXT: [[TMP16:%.*]] = and i64 [[TMP15]], -4
+; CHECK-NEXT: [[TMP17:%.*]] = inttoptr i64 [[TMP16]] to ptr
+; CHECK-NEXT: store <32 x i8> [[TMP11]], ptr [[TMP14]], align 1
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <8 x i8> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i64 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP21:%.*]] = bitcast <8 x i8> [[TMP5]] to i64
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i64 [[TMP21]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP20]]
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP8]], i32 [[TMP23]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP26:%.*]], label [[TMP27:%.*]], !prof [[PROF0]]
+; CHECK: 26:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP8]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 27:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> [[A]], <8 x i8> [[B]], <8 x i8> undef, <8 x i8> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> %A, <8 x i8> %B, <8 x i8> undef, <8 x i8> %D, ptr %P)
+ ret void
+}
+
+define void @st4_8b_undefD(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_8b_undefD(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], <8 x i8> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> [[TMP3]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <8 x i8> [[TMP5]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <16 x i8> [[TMP9]], <16 x i8> [[TMP10]], <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP13:%.*]] = xor i64 [[TMP12]], 193514046488576
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP13]], 35184372088832
+; CHECK-NEXT: [[TMP16:%.*]] = and i64 [[TMP15]], -4
+; CHECK-NEXT: [[TMP17:%.*]] = inttoptr i64 [[TMP16]] to ptr
+; CHECK-NEXT: store <32 x i8> [[TMP11]], ptr [[TMP14]], align 1
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <8 x i8> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i64 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP21:%.*]] = bitcast <8 x i8> [[TMP5]] to i64
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i64 [[TMP21]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP20]]
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP8]], i32 [[TMP23]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP26:%.*]], label [[TMP27:%.*]], !prof [[PROF0]]
+; CHECK: 26:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP8]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 27:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> [[A]], <8 x i8> [[B]], <8 x i8> [[C]], <8 x i8> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> undef, ptr %P)
+ ret void
+}
+
+define void @st4_8b_undefAB(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_8b_undefAB(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], <8 x i8> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> [[TMP3]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <16 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <16 x i8> [[TMP7]], <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP9:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP10:%.*]] = xor i64 [[TMP9]], 193514046488576
+; CHECK-NEXT: [[TMP11:%.*]] = inttoptr i64 [[TMP10]] to ptr
+; CHECK-NEXT: [[TMP12:%.*]] = add i64 [[TMP10]], 35184372088832
+; CHECK-NEXT: [[TMP13:%.*]] = and i64 [[TMP12]], -4
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: store <32 x i8> [[TMP8]], ptr [[TMP11]], align 1
+; CHECK-NEXT: [[TMP15:%.*]] = bitcast <8 x i8> [[TMP1]] to i64
+; CHECK-NEXT: [[TMP16:%.*]] = icmp ne i64 [[TMP15]], 0
+; CHECK-NEXT: [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <8 x i8> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i64 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP4]], i32 [[TMP17]]
+; CHECK-NEXT: [[TMP21:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP22:%.*]] = select i1 [[TMP21]], i32 [[TMP6]], i32 [[TMP20]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP23:%.*]], label [[TMP24:%.*]], !prof [[PROF0]]
+; CHECK: 23:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 24:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> undef, <8 x i8> undef, <8 x i8> [[C]], <8 x i8> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> undef, <8 x i8> undef, <8 x i8> %C, <8 x i8> %D, ptr %P)
+ ret void
+}
+
+define void @st4_8b_undefAC(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_8b_undefAC(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], <8 x i8> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP1]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP3]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <16 x i8> [[TMP7]], <16 x i8> [[TMP8]], <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP10:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: [[TMP13:%.*]] = add i64 [[TMP11]], 35184372088832
+; CHECK-NEXT: [[TMP14:%.*]] = and i64 [[TMP13]], -4
+; CHECK-NEXT: [[TMP15:%.*]] = inttoptr i64 [[TMP14]] to ptr
+; CHECK-NEXT: store <32 x i8> [[TMP9]], ptr [[TMP12]], align 1
+; CHECK-NEXT: [[TMP16:%.*]] = bitcast <8 x i8> [[TMP1]] to i64
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i64 [[TMP16]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP19:%.*]] = bitcast <8 x i8> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP20:%.*]] = icmp ne i64 [[TMP19]], 0
+; CHECK-NEXT: [[TMP21:%.*]] = select i1 [[TMP20]], i32 [[TMP4]], i32 [[TMP18]]
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP21]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP24:%.*]], label [[TMP25:%.*]], !prof [[PROF0]]
+; CHECK: 24:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 25:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> undef, <8 x i8> [[B]], <8 x i8> undef, <8 x i8> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> undef, <8 x i8> %B, <8 x i8> undef, <8 x i8> %D, ptr %P)
+ ret void
+}
+
+define void @st4_8b_undefBC(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_8b_undefBC(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], <8 x i8> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP3]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <16 x i8> [[TMP7]], <16 x i8> [[TMP8]], <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP10:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: [[TMP13:%.*]] = add i64 [[TMP11]], 35184372088832
+; CHECK-NEXT: [[TMP14:%.*]] = and i64 [[TMP13]], -4
+; CHECK-NEXT: [[TMP15:%.*]] = inttoptr i64 [[TMP14]] to ptr
+; CHECK-NEXT: store <32 x i8> [[TMP9]], ptr [[TMP12]], align 1
+; CHECK-NEXT: [[TMP16:%.*]] = bitcast <8 x i8> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i64 [[TMP16]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP6]], i32 [[TMP18]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP21:%.*]], label [[TMP22:%.*]], !prof [[PROF0]]
+; CHECK: 21:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 22:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> [[A]], <8 x i8> undef, <8 x i8> undef, <8 x i8> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> %A, <8 x i8> undef, <8 x i8> undef, <8 x i8> %D, ptr %P)
+ ret void
+}
+
+define void @st4_8b_undefBD(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_8b_undefBD(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], <8 x i8> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <8 x i8> [[TMP3]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <16 x i8> [[TMP7]], <16 x i8> [[TMP8]], <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP10:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: [[TMP13:%.*]] = add i64 [[TMP11]], 35184372088832
+; CHECK-NEXT: [[TMP14:%.*]] = and i64 [[TMP13]], -4
+; CHECK-NEXT: [[TMP15:%.*]] = inttoptr i64 [[TMP14]] to ptr
+; CHECK-NEXT: store <32 x i8> [[TMP9]], ptr [[TMP12]], align 1
+; CHECK-NEXT: [[TMP16:%.*]] = bitcast <8 x i8> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i64 [[TMP16]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP6]], i32 [[TMP18]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP21:%.*]], label [[TMP22:%.*]], !prof [[PROF0]]
+; CHECK: 21:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 22:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> [[A]], <8 x i8> undef, <8 x i8> [[C]], <8 x i8> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> %A, <8 x i8> undef, <8 x i8> %C, <8 x i8> undef, ptr %P)
+ ret void
+}
+
+define void @st4_8b_undefABC(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_8b_undefABC(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], <8 x i8> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP1]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <16 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <16 x i8> [[TMP5]], <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP7:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP8:%.*]] = xor i64 [[TMP7]], 193514046488576
+; CHECK-NEXT: [[TMP9:%.*]] = inttoptr i64 [[TMP8]] to ptr
+; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[TMP8]], 35184372088832
+; CHECK-NEXT: [[TMP11:%.*]] = and i64 [[TMP10]], -4
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: store <32 x i8> [[TMP6]], ptr [[TMP9]], align 1
+; CHECK-NEXT: [[TMP13:%.*]] = bitcast <8 x i8> [[TMP1]] to i64
+; CHECK-NEXT: [[TMP14:%.*]] = icmp ne i64 [[TMP13]], 0
+; CHECK-NEXT: [[TMP15:%.*]] = select i1 [[TMP14]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP16:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP4]], i32 [[TMP15]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP18:%.*]], label [[TMP19:%.*]], !prof [[PROF0]]
+; CHECK: 18:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP4]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 19:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> undef, <8 x i8> undef, <8 x i8> undef, <8 x i8> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> undef, <8 x i8> undef, <8 x i8> undef, <8 x i8> %D, ptr %P)
+ ret void
+}
+
+define void @st4_8b_undefABD(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_8b_undefABD(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], <8 x i8> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <16 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <16 x i8> [[TMP5]], <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP7:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP8:%.*]] = xor i64 [[TMP7]], 193514046488576
+; CHECK-NEXT: [[TMP9:%.*]] = inttoptr i64 [[TMP8]] to ptr
+; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[TMP8]], 35184372088832
+; CHECK-NEXT: [[TMP11:%.*]] = and i64 [[TMP10]], -4
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: store <32 x i8> [[TMP6]], ptr [[TMP9]], align 1
+; CHECK-NEXT: [[TMP13:%.*]] = bitcast <8 x i8> [[TMP1]] to i64
+; CHECK-NEXT: [[TMP14:%.*]] = icmp ne i64 [[TMP13]], 0
+; CHECK-NEXT: [[TMP15:%.*]] = select i1 [[TMP14]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP16:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP4]], i32 [[TMP15]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP18:%.*]], label [[TMP19:%.*]], !prof [[PROF0]]
+; CHECK: 18:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP4]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 19:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> undef, <8 x i8> undef, <8 x i8> [[C]], <8 x i8> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> undef, <8 x i8> undef, <8 x i8> %C, <8 x i8> undef, ptr %P)
+ ret void
+}
+
+define void @st4_8b_undefACD(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_8b_undefACD(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], <8 x i8> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <8 x i8> [[TMP1]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <16 x i8> [[TMP5]], <16 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP7:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP8:%.*]] = xor i64 [[TMP7]], 193514046488576
+; CHECK-NEXT: [[TMP9:%.*]] = inttoptr i64 [[TMP8]] to ptr
+; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[TMP8]], 35184372088832
+; CHECK-NEXT: [[TMP11:%.*]] = and i64 [[TMP10]], -4
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: store <32 x i8> [[TMP6]], ptr [[TMP9]], align 1
+; CHECK-NEXT: [[TMP13:%.*]] = bitcast <8 x i8> [[TMP1]] to i64
+; CHECK-NEXT: [[TMP14:%.*]] = icmp ne i64 [[TMP13]], 0
+; CHECK-NEXT: [[TMP15:%.*]] = select i1 [[TMP14]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP16:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP4]], i32 [[TMP15]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP18:%.*]], label [[TMP19:%.*]], !prof [[PROF0]]
+; CHECK: 18:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP4]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 19:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> undef, <8 x i8> [[B]], <8 x i8> undef, <8 x i8> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> undef, <8 x i8> %B, <8 x i8> undef, <8 x i8> undef, ptr %P)
+ ret void
+}
+
+define void @st4_8b_undefBCD(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_8b_undefBCD(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], <8 x i8> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i8>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <16 x i8> [[TMP5]], <16 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP7:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP8:%.*]] = xor i64 [[TMP7]], 193514046488576
+; CHECK-NEXT: [[TMP9:%.*]] = inttoptr i64 [[TMP8]] to ptr
+; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[TMP8]], 35184372088832
+; CHECK-NEXT: [[TMP11:%.*]] = and i64 [[TMP10]], -4
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: store <32 x i8> [[TMP6]], ptr [[TMP9]], align 1
+; CHECK-NEXT: [[TMP13:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP15:%.*]], label [[TMP16:%.*]], !prof [[PROF0]]
+; CHECK: 15:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP4]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 16:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> [[A]], <8 x i8> undef, <8 x i8> undef, <8 x i8> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> %A, <8 x i8> undef, <8 x i8> undef, <8 x i8> undef, ptr %P)
+ ret void
+}
+
+define void @st4_8b_undefABCD(<8 x i8> %A, <8 x i8> %B, <8 x i8> %C, <8 x i8> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_8b_undefABCD(
+; CHECK-SAME: <8 x i8> [[A:%.*]], <8 x i8> [[B:%.*]], <8 x i8> [[C:%.*]], <8 x i8> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP3:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP4:%.*]] = xor i64 [[TMP3]], 193514046488576
+; CHECK-NEXT: [[TMP5:%.*]] = inttoptr i64 [[TMP4]] to ptr
+; CHECK-NEXT: [[TMP6:%.*]] = add i64 [[TMP4]], 35184372088832
+; CHECK-NEXT: [[TMP7:%.*]] = and i64 [[TMP6]], -4
+; CHECK-NEXT: [[TMP8:%.*]] = inttoptr i64 [[TMP7]] to ptr
+; CHECK-NEXT: store <32 x i8> <i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1, i8 -1>, ptr [[TMP5]], align 1
+; CHECK-NEXT: [[TMP9:%.*]] = icmp ne i64 [[TMP1]], 0
+; CHECK-NEXT: [[TMP10:%.*]] = select i1 [[TMP9]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP11:%.*]], label [[TMP12:%.*]], !prof [[PROF0]]
+; CHECK: 11:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP2]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 12:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> undef, <8 x i8> undef, <8 x i8> undef, <8 x i8> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8> undef, <8 x i8> undef, <8 x i8> undef, <8 x i8> undef, ptr %P)
+ ret void
+}
+
+; -----------------------------------------------------------------------------------------------------------------------------------------------
+
+declare void @llvm.aarch64.neon.st2.v8i8.p0(<8 x i8>, <8 x i8>, ptr) nounwind sanitize_memory readonly
+declare void @llvm.aarch64.neon.st3.v8i8.p0(<8 x i8>, <8 x i8>, <8 x i8>, ptr) nounwind sanitize_memory readonly
+declare void @llvm.aarch64.neon.st4.v8i8.p0(<8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, ptr) nounwind sanitize_memory readonly
+
+define void @st2_16b(<16 x i8> %A, <16 x i8> %B, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st2_16b(
+; CHECK-SAME: <16 x i8> [[A:%.*]], <16 x i8> [[B:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <16 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <16 x i8> [[TMP1]], <16 x i8> [[TMP3]], <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP8:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP9:%.*]] = xor i64 [[TMP8]], 193514046488576
+; CHECK-NEXT: [[TMP10:%.*]] = inttoptr i64 [[TMP9]] to ptr
+; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[TMP9]], 35184372088832
+; CHECK-NEXT: [[TMP12:%.*]] = and i64 [[TMP11]], -4
+; CHECK-NEXT: [[TMP13:%.*]] = inttoptr i64 [[TMP12]] to ptr
+; CHECK-NEXT: store <32 x i8> [[TMP7]], ptr [[TMP10]], align 1
+; CHECK-NEXT: [[TMP14:%.*]] = bitcast <16 x i8> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP15:%.*]] = icmp ne i128 [[TMP14]], 0
+; CHECK-NEXT: [[TMP16:%.*]] = select i1 [[TMP15]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP6]], i32 [[TMP16]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP19:%.*]], label [[TMP20:%.*]], !prof [[PROF0]]
+; CHECK: 19:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 20:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v16i8.p0(<16 x i8> [[A]], <16 x i8> [[B]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st2.v16i8.p0(<16 x i8> %A, <16 x i8> %B, ptr %P)
+ ret void
+}
+
+define void @st3_16b(<16 x i8> %A, <16 x i8> %B, <16 x i8> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_16b(
+; CHECK-SAME: <16 x i8> [[A:%.*]], <16 x i8> [[B:%.*]], <16 x i8> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <16 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <16 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <16 x i8> [[TMP1]], <16 x i8> [[TMP3]], <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <16 x i8> [[TMP5]], <16 x i8> undef, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <32 x i8> [[TMP9]], <32 x i8> [[TMP10]], <48 x i32> <i32 0, i32 1, i32 16, i32 2, i32 3, i32 18, i32 4, i32 5, i32 20, i32 6, i32 7, i32 22, i32 8, i32 9, i32 24, i32 10, i32 11, i32 26, i32 12, i32 13, i32 28, i32 14, i32 15, i32 30, i32 16, i32 17, i32 32, i32 18, i32 19, i32 34, i32 20, i32 21, i32 36, i32 22, i32 23, i32 38, i32 24, i32 25, i32 40, i32 26, i32 27, i32 42, i32 28, i32 29, i32 44, i32 30, i32 31, i32 46>
+; CHECK-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP13:%.*]] = xor i64 [[TMP12]], 193514046488576
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP13]], 35184372088832
+; CHECK-NEXT: [[TMP16:%.*]] = and i64 [[TMP15]], -4
+; CHECK-NEXT: [[TMP17:%.*]] = inttoptr i64 [[TMP16]] to ptr
+; CHECK-NEXT: store <48 x i8> [[TMP11]], ptr [[TMP14]], align 1
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <16 x i8> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i128 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP21:%.*]] = bitcast <16 x i8> [[TMP5]] to i128
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i128 [[TMP21]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP20]]
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP8]], i32 [[TMP23]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP26:%.*]], label [[TMP27:%.*]], !prof [[PROF0]]
+; CHECK: 26:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP8]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 27:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v16i8.p0(<16 x i8> [[A]], <16 x i8> [[B]], <16 x i8> [[C]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v16i8.p0(<16 x i8> %A, <16 x i8> %B, <16 x i8> %C, ptr %P)
+ ret void
+}
+
+define void @st4_16b(<16 x i8> %A, <16 x i8> %B, <16 x i8> %C, <16 x i8> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_16b(
+; CHECK-SAME: <16 x i8> [[A:%.*]], <16 x i8> [[B:%.*]], <16 x i8> [[C:%.*]], <16 x i8> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <16 x i8>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <16 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <16 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load <16 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: [[TMP9:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 64) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <16 x i8> [[TMP1]], <16 x i8> [[TMP3]], <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <16 x i8> [[TMP5]], <16 x i8> [[TMP7]], <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP13:%.*]] = shufflevector <32 x i8> [[TMP11]], <32 x i8> [[TMP12]], <64 x i32> <i32 0, i32 32, i32 1, i32 33, i32 2, i32 34, i32 3, i32 35, i32 4, i32 36, i32 5, i32 37, i32 6, i32 38, i32 7, i32 39, i32 8, i32 40, i32 9, i32 41, i32 10, i32 42, i32 11, i32 43, i32 12, i32 44, i32 13, i32 45, i32 14, i32 46, i32 15, i32 47, i32 16, i32 48, i32 17, i32 49, i32 18, i32 50, i32 19, i32 51, i32 20, i32 52, i32 21, i32 53, i32 22, i32 54, i32 23, i32 55, i32 24, i32 56, i32 25, i32 57, i32 26, i32 58, i32 27, i32 59, i32 28, i32 60, i32 29, i32 61, i32 30, i32 62, i32 31, i32 63>
+; CHECK-NEXT: [[TMP14:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP15:%.*]] = xor i64 [[TMP14]], 193514046488576
+; CHECK-NEXT: [[TMP16:%.*]] = inttoptr i64 [[TMP15]] to ptr
+; CHECK-NEXT: [[TMP17:%.*]] = add i64 [[TMP15]], 35184372088832
+; CHECK-NEXT: [[TMP18:%.*]] = and i64 [[TMP17]], -4
+; CHECK-NEXT: [[TMP19:%.*]] = inttoptr i64 [[TMP18]] to ptr
+; CHECK-NEXT: store <64 x i8> [[TMP13]], ptr [[TMP16]], align 1
+; CHECK-NEXT: [[TMP20:%.*]] = bitcast <16 x i8> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP21:%.*]] = icmp ne i128 [[TMP20]], 0
+; CHECK-NEXT: [[TMP22:%.*]] = select i1 [[TMP21]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP23:%.*]] = bitcast <16 x i8> [[TMP5]] to i128
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i128 [[TMP23]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP6]], i32 [[TMP22]]
+; CHECK-NEXT: [[TMP26:%.*]] = bitcast <16 x i8> [[TMP7]] to i128
+; CHECK-NEXT: [[TMP27:%.*]] = icmp ne i128 [[TMP26]], 0
+; CHECK-NEXT: [[TMP28:%.*]] = select i1 [[TMP27]], i32 [[TMP8]], i32 [[TMP25]]
+; CHECK-NEXT: [[TMP29:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: [[TMP30:%.*]] = select i1 [[TMP29]], i32 [[TMP10]], i32 [[TMP28]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP31:%.*]], label [[TMP32:%.*]], !prof [[PROF0]]
+; CHECK: 31:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP10]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 32:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v16i8.p0(<16 x i8> [[A]], <16 x i8> [[B]], <16 x i8> [[C]], <16 x i8> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v16i8.p0(<16 x i8> %A, <16 x i8> %B, <16 x i8> %C, <16 x i8> %D, ptr %P)
+ ret void
+}
+
+; -----------------------------------------------------------------------------------------------------------------------------------------------
+
+declare void @llvm.aarch64.neon.st2.v16i8.p0(<16 x i8>, <16 x i8>, ptr) nounwind sanitize_memory readonly
+declare void @llvm.aarch64.neon.st3.v16i8.p0(<16 x i8>, <16 x i8>, <16 x i8>, ptr) nounwind sanitize_memory readonly
+declare void @llvm.aarch64.neon.st4.v16i8.p0(<16 x i8>, <16 x i8>, <16 x i8>, <16 x i8>, ptr) nounwind sanitize_memory readonly
+
+define void @st2_4h(<4 x i16> %A, <4 x i16> %B, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st2_4h(
+; CHECK-SAME: <4 x i16> [[A:%.*]], <4 x i16> [[B:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i16>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP3]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP8:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP9:%.*]] = xor i64 [[TMP8]], 193514046488576
+; CHECK-NEXT: [[TMP10:%.*]] = inttoptr i64 [[TMP9]] to ptr
+; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[TMP9]], 35184372088832
+; CHECK-NEXT: [[TMP12:%.*]] = and i64 [[TMP11]], -4
+; CHECK-NEXT: [[TMP13:%.*]] = inttoptr i64 [[TMP12]] to ptr
+; CHECK-NEXT: store <8 x i16> [[TMP7]], ptr [[TMP10]], align 1
+; CHECK-NEXT: [[TMP14:%.*]] = bitcast <4 x i16> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP15:%.*]] = icmp ne i64 [[TMP14]], 0
+; CHECK-NEXT: [[TMP16:%.*]] = select i1 [[TMP15]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP6]], i32 [[TMP16]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP19:%.*]], label [[TMP20:%.*]], !prof [[PROF0]]
+; CHECK: 19:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 20:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v4i16.p0(<4 x i16> [[A]], <4 x i16> [[B]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st2.v4i16.p0(<4 x i16> %A, <4 x i16> %B, ptr %P)
+ ret void
+}
+
+define void @st3_4h(<4 x i16> %A, <4 x i16> %B, <4 x i16> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_4h(
+; CHECK-SAME: <4 x i16> [[A:%.*]], <4 x i16> [[B:%.*]], <4 x i16> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i16>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <4 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP3]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <4 x i16> [[TMP5]], <4 x i16> undef, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <8 x i16> [[TMP9]], <8 x i16> [[TMP10]], <12 x i32> <i32 0, i32 1, i32 4, i32 2, i32 3, i32 6, i32 4, i32 5, i32 8, i32 6, i32 7, i32 10>
+; CHECK-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP13:%.*]] = xor i64 [[TMP12]], 193514046488576
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP13]], 35184372088832
+; CHECK-NEXT: [[TMP16:%.*]] = and i64 [[TMP15]], -4
+; CHECK-NEXT: [[TMP17:%.*]] = inttoptr i64 [[TMP16]] to ptr
+; CHECK-NEXT: store <12 x i16> [[TMP11]], ptr [[TMP14]], align 1
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <4 x i16> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i64 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP21:%.*]] = bitcast <4 x i16> [[TMP5]] to i64
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i64 [[TMP21]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP20]]
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP8]], i32 [[TMP23]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP26:%.*]], label [[TMP27:%.*]], !prof [[PROF0]]
+; CHECK: 26:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP8]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 27:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v4i16.p0(<4 x i16> [[A]], <4 x i16> [[B]], <4 x i16> [[C]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v4i16.p0(<4 x i16> %A, <4 x i16> %B, <4 x i16> %C, ptr %P)
+ ret void
+}
+
+define void @st4_4h(<4 x i16> %A, <4 x i16> %B, <4 x i16> %C, <4 x i16> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_4h(
+; CHECK-SAME: <4 x i16> [[A:%.*]], <4 x i16> [[B:%.*]], <4 x i16> [[C:%.*]], <4 x i16> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i16>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <4 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load <4 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: [[TMP9:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <4 x i16> [[TMP1]], <4 x i16> [[TMP3]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <4 x i16> [[TMP5]], <4 x i16> [[TMP7]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP13:%.*]] = shufflevector <8 x i16> [[TMP11]], <8 x i16> [[TMP12]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP14:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP15:%.*]] = xor i64 [[TMP14]], 193514046488576
+; CHECK-NEXT: [[TMP16:%.*]] = inttoptr i64 [[TMP15]] to ptr
+; CHECK-NEXT: [[TMP17:%.*]] = add i64 [[TMP15]], 35184372088832
+; CHECK-NEXT: [[TMP18:%.*]] = and i64 [[TMP17]], -4
+; CHECK-NEXT: [[TMP19:%.*]] = inttoptr i64 [[TMP18]] to ptr
+; CHECK-NEXT: store <16 x i16> [[TMP13]], ptr [[TMP16]], align 1
+; CHECK-NEXT: [[TMP20:%.*]] = bitcast <4 x i16> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP21:%.*]] = icmp ne i64 [[TMP20]], 0
+; CHECK-NEXT: [[TMP22:%.*]] = select i1 [[TMP21]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP23:%.*]] = bitcast <4 x i16> [[TMP5]] to i64
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP23]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP6]], i32 [[TMP22]]
+; CHECK-NEXT: [[TMP26:%.*]] = bitcast <4 x i16> [[TMP7]] to i64
+; CHECK-NEXT: [[TMP27:%.*]] = icmp ne i64 [[TMP26]], 0
+; CHECK-NEXT: [[TMP28:%.*]] = select i1 [[TMP27]], i32 [[TMP8]], i32 [[TMP25]]
+; CHECK-NEXT: [[TMP29:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: [[TMP30:%.*]] = select i1 [[TMP29]], i32 [[TMP10]], i32 [[TMP28]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP31:%.*]], label [[TMP32:%.*]], !prof [[PROF0]]
+; CHECK: 31:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP10]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 32:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v4i16.p0(<4 x i16> [[A]], <4 x i16> [[B]], <4 x i16> [[C]], <4 x i16> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v4i16.p0(<4 x i16> %A, <4 x i16> %B, <4 x i16> %C, <4 x i16> %D, ptr %P)
+ ret void
+}
+
+; -----------------------------------------------------------------------------------------------------------------------------------------------
+
+declare void @llvm.aarch64.neon.st2.v4i16.p0(<4 x i16>, <4 x i16>, ptr) nounwind sanitize_memory readonly
+declare void @llvm.aarch64.neon.st3.v4i16.p0(<4 x i16>, <4 x i16>, <4 x i16>, ptr) nounwind sanitize_memory readonly
+declare void @llvm.aarch64.neon.st4.v4i16.p0(<4 x i16>, <4 x i16>, <4 x i16>, <4 x i16>, ptr) nounwind sanitize_memory readonly
+
+define void @st2_8h(<8 x i16> %A, <8 x i16> %B, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st2_8h(
+; CHECK-SAME: <8 x i16> [[A:%.*]], <8 x i16> [[B:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i16>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP3]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP8:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP9:%.*]] = xor i64 [[TMP8]], 193514046488576
+; CHECK-NEXT: [[TMP10:%.*]] = inttoptr i64 [[TMP9]] to ptr
+; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[TMP9]], 35184372088832
+; CHECK-NEXT: [[TMP12:%.*]] = and i64 [[TMP11]], -4
+; CHECK-NEXT: [[TMP13:%.*]] = inttoptr i64 [[TMP12]] to ptr
+; CHECK-NEXT: store <16 x i16> [[TMP7]], ptr [[TMP10]], align 1
+; CHECK-NEXT: [[TMP14:%.*]] = bitcast <8 x i16> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP15:%.*]] = icmp ne i128 [[TMP14]], 0
+; CHECK-NEXT: [[TMP16:%.*]] = select i1 [[TMP15]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP6]], i32 [[TMP16]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP19:%.*]], label [[TMP20:%.*]], !prof [[PROF0]]
+; CHECK: 19:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 20:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v8i16.p0(<8 x i16> [[A]], <8 x i16> [[B]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st2.v8i16.p0(<8 x i16> %A, <8 x i16> %B, ptr %P)
+ ret void
+}
+
+define void @st3_8h(<8 x i16> %A, <8 x i16> %B, <8 x i16> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_8h(
+; CHECK-SAME: <8 x i16> [[A:%.*]], <8 x i16> [[B:%.*]], <8 x i16> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i16>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <8 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP3]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <8 x i16> [[TMP5]], <8 x i16> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <16 x i16> [[TMP9]], <16 x i16> [[TMP10]], <24 x i32> <i32 0, i32 1, i32 8, i32 2, i32 3, i32 10, i32 4, i32 5, i32 12, i32 6, i32 7, i32 14, i32 8, i32 9, i32 16, i32 10, i32 11, i32 18, i32 12, i32 13, i32 20, i32 14, i32 15, i32 22>
+; CHECK-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP13:%.*]] = xor i64 [[TMP12]], 193514046488576
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP13]], 35184372088832
+; CHECK-NEXT: [[TMP16:%.*]] = and i64 [[TMP15]], -4
+; CHECK-NEXT: [[TMP17:%.*]] = inttoptr i64 [[TMP16]] to ptr
+; CHECK-NEXT: store <24 x i16> [[TMP11]], ptr [[TMP14]], align 1
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <8 x i16> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i128 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP21:%.*]] = bitcast <8 x i16> [[TMP5]] to i128
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i128 [[TMP21]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP20]]
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP8]], i32 [[TMP23]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP26:%.*]], label [[TMP27:%.*]], !prof [[PROF0]]
+; CHECK: 26:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP8]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 27:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v8i16.p0(<8 x i16> [[A]], <8 x i16> [[B]], <8 x i16> [[C]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v8i16.p0(<8 x i16> %A, <8 x i16> %B, <8 x i16> %C, ptr %P)
+ ret void
+}
+
+define void @st4_8h(<8 x i16> %A, <8 x i16> %B, <8 x i16> %C, <8 x i16> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_8h(
+; CHECK-SAME: <8 x i16> [[A:%.*]], <8 x i16> [[B:%.*]], <8 x i16> [[C:%.*]], <8 x i16> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <8 x i16>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <8 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load <8 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: [[TMP9:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 64) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> [[TMP3]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <8 x i16> [[TMP5]], <8 x i16> [[TMP7]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP13:%.*]] = shufflevector <16 x i16> [[TMP11]], <16 x i16> [[TMP12]], <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT: [[TMP14:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP15:%.*]] = xor i64 [[TMP14]], 193514046488576
+; CHECK-NEXT: [[TMP16:%.*]] = inttoptr i64 [[TMP15]] to ptr
+; CHECK-NEXT: [[TMP17:%.*]] = add i64 [[TMP15]], 35184372088832
+; CHECK-NEXT: [[TMP18:%.*]] = and i64 [[TMP17]], -4
+; CHECK-NEXT: [[TMP19:%.*]] = inttoptr i64 [[TMP18]] to ptr
+; CHECK-NEXT: store <32 x i16> [[TMP13]], ptr [[TMP16]], align 1
+; CHECK-NEXT: [[TMP20:%.*]] = bitcast <8 x i16> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP21:%.*]] = icmp ne i128 [[TMP20]], 0
+; CHECK-NEXT: [[TMP22:%.*]] = select i1 [[TMP21]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP23:%.*]] = bitcast <8 x i16> [[TMP5]] to i128
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i128 [[TMP23]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP6]], i32 [[TMP22]]
+; CHECK-NEXT: [[TMP26:%.*]] = bitcast <8 x i16> [[TMP7]] to i128
+; CHECK-NEXT: [[TMP27:%.*]] = icmp ne i128 [[TMP26]], 0
+; CHECK-NEXT: [[TMP28:%.*]] = select i1 [[TMP27]], i32 [[TMP8]], i32 [[TMP25]]
+; CHECK-NEXT: [[TMP29:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: [[TMP30:%.*]] = select i1 [[TMP29]], i32 [[TMP10]], i32 [[TMP28]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP31:%.*]], label [[TMP32:%.*]], !prof [[PROF0]]
+; CHECK: 31:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP10]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 32:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v8i16.p0(<8 x i16> [[A]], <8 x i16> [[B]], <8 x i16> [[C]], <8 x i16> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v8i16.p0(<8 x i16> %A, <8 x i16> %B, <8 x i16> %C, <8 x i16> %D, ptr %P)
+ ret void
+}
+
+; -----------------------------------------------------------------------------------------------------------------------------------------------
+
+declare void @llvm.aarch64.neon.st2.v8i16.p0(<8 x i16>, <8 x i16>, ptr) nounwind sanitize_memory readonly
+declare void @llvm.aarch64.neon.st3.v8i16.p0(<8 x i16>, <8 x i16>, <8 x i16>, ptr) nounwind sanitize_memory readonly
+declare void @llvm.aarch64.neon.st4.v8i16.p0(<8 x i16>, <8 x i16>, <8 x i16>, <8 x i16>, ptr) nounwind sanitize_memory readonly
+
+define void @st2_2s(<2 x i32> %A, <2 x i32> %B, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st2_2s(
+; CHECK-SAME: <2 x i32> [[A:%.*]], <2 x i32> [[B:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i32>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP3]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP8:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP9:%.*]] = xor i64 [[TMP8]], 193514046488576
+; CHECK-NEXT: [[TMP10:%.*]] = inttoptr i64 [[TMP9]] to ptr
+; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[TMP9]], 35184372088832
+; CHECK-NEXT: [[TMP12:%.*]] = and i64 [[TMP11]], -4
+; CHECK-NEXT: [[TMP13:%.*]] = inttoptr i64 [[TMP12]] to ptr
+; CHECK-NEXT: store <4 x i32> [[TMP7]], ptr [[TMP10]], align 1
+; CHECK-NEXT: [[TMP14:%.*]] = bitcast <2 x i32> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP15:%.*]] = icmp ne i64 [[TMP14]], 0
+; CHECK-NEXT: [[TMP16:%.*]] = select i1 [[TMP15]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP6]], i32 [[TMP16]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP19:%.*]], label [[TMP20:%.*]], !prof [[PROF0]]
+; CHECK: 19:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 20:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v2i32.p0(<2 x i32> [[A]], <2 x i32> [[B]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st2.v2i32.p0(<2 x i32> %A, <2 x i32> %B, ptr %P)
+ ret void
+}
+
+define void @st3_2s(<2 x i32> %A, <2 x i32> %B, <2 x i32> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_2s(
+; CHECK-SAME: <2 x i32> [[A:%.*]], <2 x i32> [[B:%.*]], <2 x i32> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i32>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <2 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP3]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <2 x i32> [[TMP5]], <2 x i32> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <4 x i32> [[TMP9]], <4 x i32> [[TMP10]], <6 x i32> <i32 0, i32 1, i32 2, i32 2, i32 3, i32 4>
+; CHECK-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP13:%.*]] = xor i64 [[TMP12]], 193514046488576
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP13]], 35184372088832
+; CHECK-NEXT: [[TMP16:%.*]] = and i64 [[TMP15]], -4
+; CHECK-NEXT: [[TMP17:%.*]] = inttoptr i64 [[TMP16]] to ptr
+; CHECK-NEXT: store <6 x i32> [[TMP11]], ptr [[TMP14]], align 1
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <2 x i32> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i64 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP21:%.*]] = bitcast <2 x i32> [[TMP5]] to i64
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i64 [[TMP21]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP20]]
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP8]], i32 [[TMP23]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP26:%.*]], label [[TMP27:%.*]], !prof [[PROF0]]
+; CHECK: 26:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP8]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 27:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v2i32.p0(<2 x i32> [[A]], <2 x i32> [[B]], <2 x i32> [[C]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v2i32.p0(<2 x i32> %A, <2 x i32> %B, <2 x i32> %C, ptr %P)
+ ret void
+}
+
+define void @st4_2s(<2 x i32> %A, <2 x i32> %B, <2 x i32> %C, <2 x i32> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_2s(
+; CHECK-SAME: <2 x i32> [[A:%.*]], <2 x i32> [[B:%.*]], <2 x i32> [[C:%.*]], <2 x i32> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i32>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <2 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load <2 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: [[TMP9:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP3]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <2 x i32> [[TMP5]], <2 x i32> [[TMP7]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP13:%.*]] = shufflevector <4 x i32> [[TMP11]], <4 x i32> [[TMP12]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP14:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP15:%.*]] = xor i64 [[TMP14]], 193514046488576
+; CHECK-NEXT: [[TMP16:%.*]] = inttoptr i64 [[TMP15]] to ptr
+; CHECK-NEXT: [[TMP17:%.*]] = add i64 [[TMP15]], 35184372088832
+; CHECK-NEXT: [[TMP18:%.*]] = and i64 [[TMP17]], -4
+; CHECK-NEXT: [[TMP19:%.*]] = inttoptr i64 [[TMP18]] to ptr
+; CHECK-NEXT: store <8 x i32> [[TMP13]], ptr [[TMP16]], align 1
+; CHECK-NEXT: [[TMP20:%.*]] = bitcast <2 x i32> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP21:%.*]] = icmp ne i64 [[TMP20]], 0
+; CHECK-NEXT: [[TMP22:%.*]] = select i1 [[TMP21]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP23:%.*]] = bitcast <2 x i32> [[TMP5]] to i64
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP23]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP6]], i32 [[TMP22]]
+; CHECK-NEXT: [[TMP26:%.*]] = bitcast <2 x i32> [[TMP7]] to i64
+; CHECK-NEXT: [[TMP27:%.*]] = icmp ne i64 [[TMP26]], 0
+; CHECK-NEXT: [[TMP28:%.*]] = select i1 [[TMP27]], i32 [[TMP8]], i32 [[TMP25]]
+; CHECK-NEXT: [[TMP29:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: [[TMP30:%.*]] = select i1 [[TMP29]], i32 [[TMP10]], i32 [[TMP28]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP31:%.*]], label [[TMP32:%.*]], !prof [[PROF0]]
+; CHECK: 31:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP10]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 32:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v2i32.p0(<2 x i32> [[A]], <2 x i32> [[B]], <2 x i32> [[C]], <2 x i32> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v2i32.p0(<2 x i32> %A, <2 x i32> %B, <2 x i32> %C, <2 x i32> %D, ptr %P)
+ ret void
+}
+
+declare void @llvm.aarch64.neon.st2.v2i32.p0(<2 x i32>, <2 x i32>, ptr) nounwind sanitize_memory readonly
+declare void @llvm.aarch64.neon.st3.v2i32.p0(<2 x i32>, <2 x i32>, <2 x i32>, ptr) nounwind sanitize_memory readonly
+declare void @llvm.aarch64.neon.st4.v2i32.p0(<2 x i32>, <2 x i32>, <2 x i32>, <2 x i32>, ptr) nounwind sanitize_memory readonly
+
+define void @st2_4s(<4 x i32> %A, <4 x i32> %B, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st2_4s(
+; CHECK-SAME: <4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP3]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP8:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP9:%.*]] = xor i64 [[TMP8]], 193514046488576
+; CHECK-NEXT: [[TMP10:%.*]] = inttoptr i64 [[TMP9]] to ptr
+; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[TMP9]], 35184372088832
+; CHECK-NEXT: [[TMP12:%.*]] = and i64 [[TMP11]], -4
+; CHECK-NEXT: [[TMP13:%.*]] = inttoptr i64 [[TMP12]] to ptr
+; CHECK-NEXT: store <8 x i32> [[TMP7]], ptr [[TMP10]], align 1
+; CHECK-NEXT: [[TMP14:%.*]] = bitcast <4 x i32> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP15:%.*]] = icmp ne i128 [[TMP14]], 0
+; CHECK-NEXT: [[TMP16:%.*]] = select i1 [[TMP15]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP6]], i32 [[TMP16]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP19:%.*]], label [[TMP20:%.*]], !prof [[PROF0]]
+; CHECK: 19:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 20:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v4i32.p0(<4 x i32> [[A]], <4 x i32> [[B]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st2.v4i32.p0(<4 x i32> %A, <4 x i32> %B, ptr %P)
+ ret void
+}
+
+define void @st3_4s(<4 x i32> %A, <4 x i32> %B, <4 x i32> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_4s(
+; CHECK-SAME: <4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i32> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <4 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP3]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <4 x i32> [[TMP5]], <4 x i32> undef, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <8 x i32> [[TMP9]], <8 x i32> [[TMP10]], <12 x i32> <i32 0, i32 1, i32 4, i32 2, i32 3, i32 6, i32 4, i32 5, i32 8, i32 6, i32 7, i32 10>
+; CHECK-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP13:%.*]] = xor i64 [[TMP12]], 193514046488576
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP13]], 35184372088832
+; CHECK-NEXT: [[TMP16:%.*]] = and i64 [[TMP15]], -4
+; CHECK-NEXT: [[TMP17:%.*]] = inttoptr i64 [[TMP16]] to ptr
+; CHECK-NEXT: store <12 x i32> [[TMP11]], ptr [[TMP14]], align 1
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <4 x i32> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i128 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP21:%.*]] = bitcast <4 x i32> [[TMP5]] to i128
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i128 [[TMP21]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP20]]
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP8]], i32 [[TMP23]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP26:%.*]], label [[TMP27:%.*]], !prof [[PROF0]]
+; CHECK: 26:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP8]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 27:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v4i32.p0(<4 x i32> [[A]], <4 x i32> [[B]], <4 x i32> [[C]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v4i32.p0(<4 x i32> %A, <4 x i32> %B, <4 x i32> %C, ptr %P)
+ ret void
+}
+
+define void @st4_4s(<4 x i32> %A, <4 x i32> %B, <4 x i32> %C, <4 x i32> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_4s(
+; CHECK-SAME: <4 x i32> [[A:%.*]], <4 x i32> [[B:%.*]], <4 x i32> [[C:%.*]], <4 x i32> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <4 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load <4 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: [[TMP9:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 64) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> [[TMP3]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <4 x i32> [[TMP5]], <4 x i32> [[TMP7]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP13:%.*]] = shufflevector <8 x i32> [[TMP11]], <8 x i32> [[TMP12]], <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT: [[TMP14:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP15:%.*]] = xor i64 [[TMP14]], 193514046488576
+; CHECK-NEXT: [[TMP16:%.*]] = inttoptr i64 [[TMP15]] to ptr
+; CHECK-NEXT: [[TMP17:%.*]] = add i64 [[TMP15]], 35184372088832
+; CHECK-NEXT: [[TMP18:%.*]] = and i64 [[TMP17]], -4
+; CHECK-NEXT: [[TMP19:%.*]] = inttoptr i64 [[TMP18]] to ptr
+; CHECK-NEXT: store <16 x i32> [[TMP13]], ptr [[TMP16]], align 1
+; CHECK-NEXT: [[TMP20:%.*]] = bitcast <4 x i32> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP21:%.*]] = icmp ne i128 [[TMP20]], 0
+; CHECK-NEXT: [[TMP22:%.*]] = select i1 [[TMP21]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP23:%.*]] = bitcast <4 x i32> [[TMP5]] to i128
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i128 [[TMP23]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP6]], i32 [[TMP22]]
+; CHECK-NEXT: [[TMP26:%.*]] = bitcast <4 x i32> [[TMP7]] to i128
+; CHECK-NEXT: [[TMP27:%.*]] = icmp ne i128 [[TMP26]], 0
+; CHECK-NEXT: [[TMP28:%.*]] = select i1 [[TMP27]], i32 [[TMP8]], i32 [[TMP25]]
+; CHECK-NEXT: [[TMP29:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: [[TMP30:%.*]] = select i1 [[TMP29]], i32 [[TMP10]], i32 [[TMP28]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP31:%.*]], label [[TMP32:%.*]], !prof [[PROF0]]
+; CHECK: 31:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP10]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 32:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v4i32.p0(<4 x i32> [[A]], <4 x i32> [[B]], <4 x i32> [[C]], <4 x i32> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v4i32.p0(<4 x i32> %A, <4 x i32> %B, <4 x i32> %C, <4 x i32> %D, ptr %P)
+ ret void
+}
+
+; -----------------------------------------------------------------------------------------------------------------------------------------------
+
+declare void @llvm.aarch64.neon.st2.v4i32.p0(<4 x i32>, <4 x i32>, ptr) nounwind sanitize_memory readonly
+declare void @llvm.aarch64.neon.st3.v4i32.p0(<4 x i32>, <4 x i32>, <4 x i32>, ptr) nounwind sanitize_memory readonly
+declare void @llvm.aarch64.neon.st4.v4i32.p0(<4 x i32>, <4 x i32>, <4 x i32>, <4 x i32>, ptr) nounwind sanitize_memory readonly
+
+; If there's only one element, st2/3/4 don't make much sense, stick to st1.
+define void @st2_1d(<1 x i64> %A, <1 x i64> %B, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st2_1d(
+; CHECK-SAME: <1 x i64> [[A:%.*]], <1 x i64> [[B:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <1 x i64>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <1 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <1 x i64> [[TMP1]], <1 x i64> [[TMP3]], <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT: [[TMP8:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP9:%.*]] = xor i64 [[TMP8]], 193514046488576
+; CHECK-NEXT: [[TMP10:%.*]] = inttoptr i64 [[TMP9]] to ptr
+; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[TMP9]], 35184372088832
+; CHECK-NEXT: [[TMP12:%.*]] = and i64 [[TMP11]], -4
+; CHECK-NEXT: [[TMP13:%.*]] = inttoptr i64 [[TMP12]] to ptr
+; CHECK-NEXT: store <2 x i64> [[TMP7]], ptr [[TMP10]], align 1
+; CHECK-NEXT: [[TMP14:%.*]] = bitcast <1 x i64> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP15:%.*]] = icmp ne i64 [[TMP14]], 0
+; CHECK-NEXT: [[TMP16:%.*]] = select i1 [[TMP15]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP6]], i32 [[TMP16]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP19:%.*]], label [[TMP20:%.*]], !prof [[PROF0]]
+; CHECK: 19:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 20:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v1i64.p0(<1 x i64> [[A]], <1 x i64> [[B]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st2.v1i64.p0(<1 x i64> %A, <1 x i64> %B, ptr %P)
+ ret void
+}
+
+define void @st3_1d(<1 x i64> %A, <1 x i64> %B, <1 x i64> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_1d(
+; CHECK-SAME: <1 x i64> [[A:%.*]], <1 x i64> [[B:%.*]], <1 x i64> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <1 x i64>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <1 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <1 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <1 x i64> [[TMP1]], <1 x i64> [[TMP3]], <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <1 x i64> [[TMP5]], <1 x i64> undef, <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <2 x i64> [[TMP9]], <2 x i64> [[TMP10]], <3 x i32> <i32 0, i32 1, i32 1>
+; CHECK-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP13:%.*]] = xor i64 [[TMP12]], 193514046488576
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP13]], 35184372088832
+; CHECK-NEXT: [[TMP16:%.*]] = and i64 [[TMP15]], -4
+; CHECK-NEXT: [[TMP17:%.*]] = inttoptr i64 [[TMP16]] to ptr
+; CHECK-NEXT: store <3 x i64> [[TMP11]], ptr [[TMP14]], align 1
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <1 x i64> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i64 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP21:%.*]] = bitcast <1 x i64> [[TMP5]] to i64
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i64 [[TMP21]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP20]]
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP8]], i32 [[TMP23]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP26:%.*]], label [[TMP27:%.*]], !prof [[PROF0]]
+; CHECK: 26:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP8]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 27:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v1i64.p0(<1 x i64> [[A]], <1 x i64> [[B]], <1 x i64> [[C]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v1i64.p0(<1 x i64> %A, <1 x i64> %B, <1 x i64> %C, ptr %P)
+ ret void
+}
+
+define void @st4_1d(<1 x i64> %A, <1 x i64> %B, <1 x i64> %C, <1 x i64> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_1d(
+; CHECK-SAME: <1 x i64> [[A:%.*]], <1 x i64> [[B:%.*]], <1 x i64> [[C:%.*]], <1 x i64> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <1 x i64>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <1 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 8) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <1 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load <1 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 24) to ptr), align 4
+; CHECK-NEXT: [[TMP9:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <1 x i64> [[TMP1]], <1 x i64> [[TMP3]], <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <1 x i64> [[TMP5]], <1 x i64> [[TMP7]], <2 x i32> <i32 0, i32 1>
+; CHECK-NEXT: [[TMP13:%.*]] = shufflevector <2 x i64> [[TMP11]], <2 x i64> [[TMP12]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP14:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP15:%.*]] = xor i64 [[TMP14]], 193514046488576
+; CHECK-NEXT: [[TMP16:%.*]] = inttoptr i64 [[TMP15]] to ptr
+; CHECK-NEXT: [[TMP17:%.*]] = add i64 [[TMP15]], 35184372088832
+; CHECK-NEXT: [[TMP18:%.*]] = and i64 [[TMP17]], -4
+; CHECK-NEXT: [[TMP19:%.*]] = inttoptr i64 [[TMP18]] to ptr
+; CHECK-NEXT: store <4 x i64> [[TMP13]], ptr [[TMP16]], align 1
+; CHECK-NEXT: [[TMP20:%.*]] = bitcast <1 x i64> [[TMP3]] to i64
+; CHECK-NEXT: [[TMP21:%.*]] = icmp ne i64 [[TMP20]], 0
+; CHECK-NEXT: [[TMP22:%.*]] = select i1 [[TMP21]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP23:%.*]] = bitcast <1 x i64> [[TMP5]] to i64
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP23]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP6]], i32 [[TMP22]]
+; CHECK-NEXT: [[TMP26:%.*]] = bitcast <1 x i64> [[TMP7]] to i64
+; CHECK-NEXT: [[TMP27:%.*]] = icmp ne i64 [[TMP26]], 0
+; CHECK-NEXT: [[TMP28:%.*]] = select i1 [[TMP27]], i32 [[TMP8]], i32 [[TMP25]]
+; CHECK-NEXT: [[TMP29:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: [[TMP30:%.*]] = select i1 [[TMP29]], i32 [[TMP10]], i32 [[TMP28]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP31:%.*]], label [[TMP32:%.*]], !prof [[PROF0]]
+; CHECK: 31:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP10]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 32:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v1i64.p0(<1 x i64> [[A]], <1 x i64> [[B]], <1 x i64> [[C]], <1 x i64> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v1i64.p0(<1 x i64> %A, <1 x i64> %B, <1 x i64> %C, <1 x i64> %D, ptr %P)
+ ret void
+}
+
+; -----------------------------------------------------------------------------------------------------------------------------------------------
+
+declare void @llvm.aarch64.neon.st2.v1i64.p0(<1 x i64>, <1 x i64>, ptr) nounwind sanitize_memory readonly
+declare void @llvm.aarch64.neon.st3.v1i64.p0(<1 x i64>, <1 x i64>, <1 x i64>, ptr) nounwind sanitize_memory readonly
+declare void @llvm.aarch64.neon.st4.v1i64.p0(<1 x i64>, <1 x i64>, <1 x i64>, <1 x i64>, ptr) nounwind sanitize_memory readonly
+
+define void @st2_2d(<2 x i64> %A, <2 x i64> %B, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st2_2d(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> [[TMP3]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP8:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP9:%.*]] = xor i64 [[TMP8]], 193514046488576
+; CHECK-NEXT: [[TMP10:%.*]] = inttoptr i64 [[TMP9]] to ptr
+; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[TMP9]], 35184372088832
+; CHECK-NEXT: [[TMP12:%.*]] = and i64 [[TMP11]], -4
+; CHECK-NEXT: [[TMP13:%.*]] = inttoptr i64 [[TMP12]] to ptr
+; CHECK-NEXT: store <4 x i64> [[TMP7]], ptr [[TMP10]], align 1
+; CHECK-NEXT: [[TMP14:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP15:%.*]] = icmp ne i128 [[TMP14]], 0
+; CHECK-NEXT: [[TMP16:%.*]] = select i1 [[TMP15]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP6]], i32 [[TMP16]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP19:%.*]], label [[TMP20:%.*]], !prof [[PROF0]]
+; CHECK: 19:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 20:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v2i64.p0(<2 x i64> [[A]], <2 x i64> [[B]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st2.v2i64.p0(<2 x i64> %A, <2 x i64> %B, ptr %P)
+ ret void
+}
+
+define void @st2_2d_undefA(<2 x i64> %A, <2 x i64> %B, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st2_2d_undefA(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <2 x i64> <i64 -1, i64 -1>, <2 x i64> [[TMP1]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP6:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP7:%.*]] = xor i64 [[TMP6]], 193514046488576
+; CHECK-NEXT: [[TMP8:%.*]] = inttoptr i64 [[TMP7]] to ptr
+; CHECK-NEXT: [[TMP9:%.*]] = add i64 [[TMP7]], 35184372088832
+; CHECK-NEXT: [[TMP10:%.*]] = and i64 [[TMP9]], -4
+; CHECK-NEXT: [[TMP11:%.*]] = inttoptr i64 [[TMP10]] to ptr
+; CHECK-NEXT: store <4 x i64> [[TMP5]], ptr [[TMP8]], align 1
+; CHECK-NEXT: [[TMP12:%.*]] = bitcast <2 x i64> [[TMP1]] to i128
+; CHECK-NEXT: [[TMP13:%.*]] = icmp ne i128 [[TMP12]], 0
+; CHECK-NEXT: [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP15:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[TMP16:%.*]] = select i1 [[TMP15]], i32 [[TMP4]], i32 [[TMP14]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP17:%.*]], label [[TMP18:%.*]], !prof [[PROF0]]
+; CHECK: 17:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP4]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 18:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v2i64.p0(<2 x i64> undef, <2 x i64> [[B]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st2.v2i64.p0(<2 x i64> undef, <2 x i64> %B, ptr %P)
+ ret void
+}
+
+define void @st2_2d_undefB(<2 x i64> %A, <2 x i64> %B, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st2_2d_undefB(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> <i64 -1, i64 -1>, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP6:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP7:%.*]] = xor i64 [[TMP6]], 193514046488576
+; CHECK-NEXT: [[TMP8:%.*]] = inttoptr i64 [[TMP7]] to ptr
+; CHECK-NEXT: [[TMP9:%.*]] = add i64 [[TMP7]], 35184372088832
+; CHECK-NEXT: [[TMP10:%.*]] = and i64 [[TMP9]], -4
+; CHECK-NEXT: [[TMP11:%.*]] = inttoptr i64 [[TMP10]] to ptr
+; CHECK-NEXT: store <4 x i64> [[TMP5]], ptr [[TMP8]], align 1
+; CHECK-NEXT: [[TMP12:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[TMP13:%.*]] = select i1 [[TMP12]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP14:%.*]], label [[TMP15:%.*]], !prof [[PROF0]]
+; CHECK: 14:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP4]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 15:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v2i64.p0(<2 x i64> [[A]], <2 x i64> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st2.v2i64.p0(<2 x i64> %A, <2 x i64> undef, ptr %P)
+ ret void
+}
+
+define void @st2_2d_undefAB(<2 x i64> %A, <2 x i64> %B, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st2_2d_undefAB(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP3:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP4:%.*]] = xor i64 [[TMP3]], 193514046488576
+; CHECK-NEXT: [[TMP5:%.*]] = inttoptr i64 [[TMP4]] to ptr
+; CHECK-NEXT: [[TMP6:%.*]] = add i64 [[TMP4]], 35184372088832
+; CHECK-NEXT: [[TMP7:%.*]] = and i64 [[TMP6]], -4
+; CHECK-NEXT: [[TMP8:%.*]] = inttoptr i64 [[TMP7]] to ptr
+; CHECK-NEXT: store <4 x i64> <i64 -1, i64 -1, i64 -1, i64 -1>, ptr [[TMP5]], align 1
+; CHECK-NEXT: [[TMP9:%.*]] = icmp ne i64 [[TMP1]], 0
+; CHECK-NEXT: [[TMP10:%.*]] = select i1 [[TMP9]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP11:%.*]], label [[TMP12:%.*]], !prof [[PROF0]]
+; CHECK: 11:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP2]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 12:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st2.v2i64.p0(<2 x i64> undef, <2 x i64> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st2.v2i64.p0(<2 x i64> undef, <2 x i64> undef, ptr %P)
+ ret void
+}
+
+define void @st3_2d(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_2d(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> [[TMP3]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <2 x i64> [[TMP5]], <2 x i64> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <4 x i64> [[TMP9]], <4 x i64> [[TMP10]], <6 x i32> <i32 0, i32 1, i32 2, i32 2, i32 3, i32 4>
+; CHECK-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP13:%.*]] = xor i64 [[TMP12]], 193514046488576
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP13]], 35184372088832
+; CHECK-NEXT: [[TMP16:%.*]] = and i64 [[TMP15]], -4
+; CHECK-NEXT: [[TMP17:%.*]] = inttoptr i64 [[TMP16]] to ptr
+; CHECK-NEXT: store <6 x i64> [[TMP11]], ptr [[TMP14]], align 1
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i128 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP21:%.*]] = bitcast <2 x i64> [[TMP5]] to i128
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i128 [[TMP21]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP20]]
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP8]], i32 [[TMP23]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP26:%.*]], label [[TMP27:%.*]], !prof [[PROF0]]
+; CHECK: 26:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP8]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 27:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v2i64.p0(<2 x i64> [[A]], <2 x i64> [[B]], <2 x i64> [[C]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v2i64.p0(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, ptr %P)
+ ret void
+}
+
+define void @st3_2d_undefA(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_2d_undefA(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <2 x i64> <i64 -1, i64 -1>, <2 x i64> [[TMP1]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <2 x i64> [[TMP3]], <2 x i64> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <4 x i64> [[TMP7]], <4 x i64> [[TMP8]], <6 x i32> <i32 0, i32 1, i32 2, i32 2, i32 3, i32 4>
+; CHECK-NEXT: [[TMP10:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: [[TMP13:%.*]] = add i64 [[TMP11]], 35184372088832
+; CHECK-NEXT: [[TMP14:%.*]] = and i64 [[TMP13]], -4
+; CHECK-NEXT: [[TMP15:%.*]] = inttoptr i64 [[TMP14]] to ptr
+; CHECK-NEXT: store <6 x i64> [[TMP9]], ptr [[TMP12]], align 1
+; CHECK-NEXT: [[TMP16:%.*]] = bitcast <2 x i64> [[TMP1]] to i128
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i128 [[TMP16]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP19:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP20:%.*]] = icmp ne i128 [[TMP19]], 0
+; CHECK-NEXT: [[TMP21:%.*]] = select i1 [[TMP20]], i32 [[TMP4]], i32 [[TMP18]]
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP21]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP24:%.*]], label [[TMP25:%.*]], !prof [[PROF0]]
+; CHECK: 24:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 25:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v2i64.p0(<2 x i64> undef, <2 x i64> [[B]], <2 x i64> [[C]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v2i64.p0(<2 x i64> undef, <2 x i64> %B, <2 x i64> %C, ptr %P)
+ ret void
+}
+
+define void @st3_2d_undefB(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_2d_undefB(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> <i64 -1, i64 -1>, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <2 x i64> [[TMP3]], <2 x i64> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <4 x i64> [[TMP7]], <4 x i64> [[TMP8]], <6 x i32> <i32 0, i32 1, i32 2, i32 2, i32 3, i32 4>
+; CHECK-NEXT: [[TMP10:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: [[TMP13:%.*]] = add i64 [[TMP11]], 35184372088832
+; CHECK-NEXT: [[TMP14:%.*]] = and i64 [[TMP13]], -4
+; CHECK-NEXT: [[TMP15:%.*]] = inttoptr i64 [[TMP14]] to ptr
+; CHECK-NEXT: store <6 x i64> [[TMP9]], ptr [[TMP12]], align 1
+; CHECK-NEXT: [[TMP16:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i128 [[TMP16]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP6]], i32 [[TMP18]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP21:%.*]], label [[TMP22:%.*]], !prof [[PROF0]]
+; CHECK: 21:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 22:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v2i64.p0(<2 x i64> [[A]], <2 x i64> undef, <2 x i64> [[C]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v2i64.p0(<2 x i64> %A, <2 x i64> undef, <2 x i64> %C, ptr %P)
+ ret void
+}
+
+define void @st3_2d_undefC(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_2d_undefC(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> [[TMP3]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <4 x i64> [[TMP7]], <4 x i64> <i64 -1, i64 undef, i64 -1, i64 undef>, <6 x i32> <i32 0, i32 1, i32 2, i32 2, i32 3, i32 4>
+; CHECK-NEXT: [[TMP9:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP10:%.*]] = xor i64 [[TMP9]], 193514046488576
+; CHECK-NEXT: [[TMP11:%.*]] = inttoptr i64 [[TMP10]] to ptr
+; CHECK-NEXT: [[TMP12:%.*]] = add i64 [[TMP10]], 35184372088832
+; CHECK-NEXT: [[TMP13:%.*]] = and i64 [[TMP12]], -4
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: store <6 x i64> [[TMP8]], ptr [[TMP11]], align 1
+; CHECK-NEXT: [[TMP15:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP16:%.*]] = icmp ne i128 [[TMP15]], 0
+; CHECK-NEXT: [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP18:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP19:%.*]] = select i1 [[TMP18]], i32 [[TMP6]], i32 [[TMP17]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP20:%.*]], label [[TMP21:%.*]], !prof [[PROF0]]
+; CHECK: 20:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 21:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v2i64.p0(<2 x i64> [[A]], <2 x i64> [[B]], <2 x i64> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v2i64.p0(<2 x i64> %A, <2 x i64> %B, <2 x i64> undef, ptr %P)
+ ret void
+}
+
+define void @st3_2d_undefAB(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_2d_undefAB(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <4 x i64> <i64 -1, i64 -1, i64 -1, i64 -1>, <4 x i64> [[TMP5]], <6 x i32> <i32 0, i32 1, i32 2, i32 2, i32 3, i32 4>
+; CHECK-NEXT: [[TMP7:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP8:%.*]] = xor i64 [[TMP7]], 193514046488576
+; CHECK-NEXT: [[TMP9:%.*]] = inttoptr i64 [[TMP8]] to ptr
+; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[TMP8]], 35184372088832
+; CHECK-NEXT: [[TMP11:%.*]] = and i64 [[TMP10]], -4
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: store <6 x i64> [[TMP6]], ptr [[TMP9]], align 1
+; CHECK-NEXT: [[TMP13:%.*]] = bitcast <2 x i64> [[TMP1]] to i128
+; CHECK-NEXT: [[TMP14:%.*]] = icmp ne i128 [[TMP13]], 0
+; CHECK-NEXT: [[TMP15:%.*]] = select i1 [[TMP14]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP16:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP4]], i32 [[TMP15]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP18:%.*]], label [[TMP19:%.*]], !prof [[PROF0]]
+; CHECK: 18:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP4]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 19:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v2i64.p0(<2 x i64> undef, <2 x i64> undef, <2 x i64> [[C]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v2i64.p0(<2 x i64> undef, <2 x i64> undef, <2 x i64> %C, ptr %P)
+ ret void
+}
+
+define void @st3_2d_undefAC(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_2d_undefAC(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <2 x i64> <i64 -1, i64 -1>, <2 x i64> [[TMP1]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <4 x i64> [[TMP5]], <4 x i64> <i64 -1, i64 undef, i64 -1, i64 undef>, <6 x i32> <i32 0, i32 1, i32 2, i32 2, i32 3, i32 4>
+; CHECK-NEXT: [[TMP7:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP8:%.*]] = xor i64 [[TMP7]], 193514046488576
+; CHECK-NEXT: [[TMP9:%.*]] = inttoptr i64 [[TMP8]] to ptr
+; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[TMP8]], 35184372088832
+; CHECK-NEXT: [[TMP11:%.*]] = and i64 [[TMP10]], -4
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: store <6 x i64> [[TMP6]], ptr [[TMP9]], align 1
+; CHECK-NEXT: [[TMP13:%.*]] = bitcast <2 x i64> [[TMP1]] to i128
+; CHECK-NEXT: [[TMP14:%.*]] = icmp ne i128 [[TMP13]], 0
+; CHECK-NEXT: [[TMP15:%.*]] = select i1 [[TMP14]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP16:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP4]], i32 [[TMP15]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP18:%.*]], label [[TMP19:%.*]], !prof [[PROF0]]
+; CHECK: 18:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP4]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 19:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v2i64.p0(<2 x i64> undef, <2 x i64> [[B]], <2 x i64> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v2i64.p0(<2 x i64> undef, <2 x i64> %B, <2 x i64> undef, ptr %P)
+ ret void
+}
+
+define void @st3_2d_undefBC(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_2d_undefBC(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> <i64 -1, i64 -1>, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <4 x i64> [[TMP5]], <4 x i64> <i64 -1, i64 undef, i64 -1, i64 undef>, <6 x i32> <i32 0, i32 1, i32 2, i32 2, i32 3, i32 4>
+; CHECK-NEXT: [[TMP7:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP8:%.*]] = xor i64 [[TMP7]], 193514046488576
+; CHECK-NEXT: [[TMP9:%.*]] = inttoptr i64 [[TMP8]] to ptr
+; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[TMP8]], 35184372088832
+; CHECK-NEXT: [[TMP11:%.*]] = and i64 [[TMP10]], -4
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: store <6 x i64> [[TMP6]], ptr [[TMP9]], align 1
+; CHECK-NEXT: [[TMP13:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[TMP14:%.*]] = select i1 [[TMP13]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP15:%.*]], label [[TMP16:%.*]], !prof [[PROF0]]
+; CHECK: 15:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP4]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 16:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v2i64.p0(<2 x i64> [[A]], <2 x i64> undef, <2 x i64> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v2i64.p0(<2 x i64> %A, <2 x i64> undef, <2 x i64> undef, ptr %P)
+ ret void
+}
+
+define void @st3_2d_undefABC(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st3_2d_undefABC(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP3:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP4:%.*]] = xor i64 [[TMP3]], 193514046488576
+; CHECK-NEXT: [[TMP5:%.*]] = inttoptr i64 [[TMP4]] to ptr
+; CHECK-NEXT: [[TMP6:%.*]] = add i64 [[TMP4]], 35184372088832
+; CHECK-NEXT: [[TMP7:%.*]] = and i64 [[TMP6]], -4
+; CHECK-NEXT: [[TMP8:%.*]] = inttoptr i64 [[TMP7]] to ptr
+; CHECK-NEXT: store <6 x i64> <i64 -1, i64 -1, i64 -1, i64 -1, i64 -1, i64 -1>, ptr [[TMP5]], align 1
+; CHECK-NEXT: [[TMP9:%.*]] = icmp ne i64 [[TMP1]], 0
+; CHECK-NEXT: [[TMP10:%.*]] = select i1 [[TMP9]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP1]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP11:%.*]], label [[TMP12:%.*]], !prof [[PROF0]]
+; CHECK: 11:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP2]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 12:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st3.v2i64.p0(<2 x i64> undef, <2 x i64> undef, <2 x i64> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st3.v2i64.p0(<2 x i64> undef, <2 x i64> undef, <2 x i64> undef, ptr %P)
+ ret void
+}
+
+define void @st4_2d(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, <2 x i64> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_2d(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], <2 x i64> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: [[TMP9:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: [[TMP10:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 64) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> [[TMP3]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP12:%.*]] = shufflevector <2 x i64> [[TMP5]], <2 x i64> [[TMP7]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP13:%.*]] = shufflevector <4 x i64> [[TMP11]], <4 x i64> [[TMP12]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP14:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP15:%.*]] = xor i64 [[TMP14]], 193514046488576
+; CHECK-NEXT: [[TMP16:%.*]] = inttoptr i64 [[TMP15]] to ptr
+; CHECK-NEXT: [[TMP17:%.*]] = add i64 [[TMP15]], 35184372088832
+; CHECK-NEXT: [[TMP18:%.*]] = and i64 [[TMP17]], -4
+; CHECK-NEXT: [[TMP19:%.*]] = inttoptr i64 [[TMP18]] to ptr
+; CHECK-NEXT: store <8 x i64> [[TMP13]], ptr [[TMP16]], align 1
+; CHECK-NEXT: [[TMP20:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP21:%.*]] = icmp ne i128 [[TMP20]], 0
+; CHECK-NEXT: [[TMP22:%.*]] = select i1 [[TMP21]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP23:%.*]] = bitcast <2 x i64> [[TMP5]] to i128
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i128 [[TMP23]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP6]], i32 [[TMP22]]
+; CHECK-NEXT: [[TMP26:%.*]] = bitcast <2 x i64> [[TMP7]] to i128
+; CHECK-NEXT: [[TMP27:%.*]] = icmp ne i128 [[TMP26]], 0
+; CHECK-NEXT: [[TMP28:%.*]] = select i1 [[TMP27]], i32 [[TMP8]], i32 [[TMP25]]
+; CHECK-NEXT: [[TMP29:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: [[TMP30:%.*]] = select i1 [[TMP29]], i32 [[TMP10]], i32 [[TMP28]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP31:%.*]], label [[TMP32:%.*]], !prof [[PROF0]]
+; CHECK: 31:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP10]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 32:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> [[A]], <2 x i64> [[B]], <2 x i64> [[C]], <2 x i64> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, <2 x i64> %D, ptr %P)
+ ret void
+}
+
+declare void @llvm.aarch64.neon.st2.v2i64.p0(<2 x i64>, <2 x i64>, ptr) nounwind sanitize_memory readonly
+declare void @llvm.aarch64.neon.st3.v2i64.p0(<2 x i64>, <2 x i64>, <2 x i64>, ptr) nounwind sanitize_memory readonly
+declare void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64>, <2 x i64>, <2 x i64>, <2 x i64>, ptr) nounwind sanitize_memory readonly
+
+define void @st4_2d_undefA(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, <2 x i64> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_2d_undefA(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], <2 x i64> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 64) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <2 x i64> <i64 -1, i64 -1>, <2 x i64> [[TMP1]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <2 x i64> [[TMP3]], <2 x i64> [[TMP5]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <4 x i64> [[TMP9]], <4 x i64> [[TMP10]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP13:%.*]] = xor i64 [[TMP12]], 193514046488576
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP13]], 35184372088832
+; CHECK-NEXT: [[TMP16:%.*]] = and i64 [[TMP15]], -4
+; CHECK-NEXT: [[TMP17:%.*]] = inttoptr i64 [[TMP16]] to ptr
+; CHECK-NEXT: store <8 x i64> [[TMP11]], ptr [[TMP14]], align 1
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <2 x i64> [[TMP1]] to i128
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i128 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP21:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i128 [[TMP21]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP4]], i32 [[TMP20]]
+; CHECK-NEXT: [[TMP24:%.*]] = bitcast <2 x i64> [[TMP5]] to i128
+; CHECK-NEXT: [[TMP25:%.*]] = icmp ne i128 [[TMP24]], 0
+; CHECK-NEXT: [[TMP26:%.*]] = select i1 [[TMP25]], i32 [[TMP6]], i32 [[TMP23]]
+; CHECK-NEXT: [[TMP27:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[TMP28:%.*]] = select i1 [[TMP27]], i32 [[TMP8]], i32 [[TMP26]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP29:%.*]], label [[TMP30:%.*]], !prof [[PROF0]]
+; CHECK: 29:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP8]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 30:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> undef, <2 x i64> [[B]], <2 x i64> [[C]], <2 x i64> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> undef, <2 x i64> %B, <2 x i64> %C, <2 x i64> %D, ptr %P)
+ ret void
+}
+
+define void @st4_2d_undefB(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, <2 x i64> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_2d_undefB(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], <2 x i64> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 64) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> <i64 -1, i64 -1>, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <2 x i64> [[TMP3]], <2 x i64> [[TMP5]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <4 x i64> [[TMP9]], <4 x i64> [[TMP10]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP13:%.*]] = xor i64 [[TMP12]], 193514046488576
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP13]], 35184372088832
+; CHECK-NEXT: [[TMP16:%.*]] = and i64 [[TMP15]], -4
+; CHECK-NEXT: [[TMP17:%.*]] = inttoptr i64 [[TMP16]] to ptr
+; CHECK-NEXT: store <8 x i64> [[TMP11]], ptr [[TMP14]], align 1
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i128 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP21:%.*]] = bitcast <2 x i64> [[TMP5]] to i128
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i128 [[TMP21]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP20]]
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP8]], i32 [[TMP23]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP26:%.*]], label [[TMP27:%.*]], !prof [[PROF0]]
+; CHECK: 26:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP8]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 27:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> [[A]], <2 x i64> undef, <2 x i64> [[C]], <2 x i64> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> %A, <2 x i64> undef, <2 x i64> %C, <2 x i64> %D, ptr %P)
+ ret void
+}
+
+define void @st4_2d_undefC(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, <2 x i64> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_2d_undefC(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], <2 x i64> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 64) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> [[TMP3]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <2 x i64> <i64 -1, i64 -1>, <2 x i64> [[TMP5]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <4 x i64> [[TMP9]], <4 x i64> [[TMP10]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP13:%.*]] = xor i64 [[TMP12]], 193514046488576
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP13]], 35184372088832
+; CHECK-NEXT: [[TMP16:%.*]] = and i64 [[TMP15]], -4
+; CHECK-NEXT: [[TMP17:%.*]] = inttoptr i64 [[TMP16]] to ptr
+; CHECK-NEXT: store <8 x i64> [[TMP11]], ptr [[TMP14]], align 1
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i128 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP21:%.*]] = bitcast <2 x i64> [[TMP5]] to i128
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i128 [[TMP21]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP20]]
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP8]], i32 [[TMP23]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP26:%.*]], label [[TMP27:%.*]], !prof [[PROF0]]
+; CHECK: 26:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP8]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 27:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> [[A]], <2 x i64> [[B]], <2 x i64> undef, <2 x i64> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> %A, <2 x i64> %B, <2 x i64> undef, <2 x i64> %D, ptr %P)
+ ret void
+}
+
+define void @st4_2d_undefD(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, <2 x i64> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_2d_undefD(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], <2 x i64> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP7:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: [[TMP8:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 64) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> [[TMP3]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <2 x i64> [[TMP5]], <2 x i64> <i64 -1, i64 -1>, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP11:%.*]] = shufflevector <4 x i64> [[TMP9]], <4 x i64> [[TMP10]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP12:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP13:%.*]] = xor i64 [[TMP12]], 193514046488576
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP13]], 35184372088832
+; CHECK-NEXT: [[TMP16:%.*]] = and i64 [[TMP15]], -4
+; CHECK-NEXT: [[TMP17:%.*]] = inttoptr i64 [[TMP16]] to ptr
+; CHECK-NEXT: store <8 x i64> [[TMP11]], ptr [[TMP14]], align 1
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i128 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP21:%.*]] = bitcast <2 x i64> [[TMP5]] to i128
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i128 [[TMP21]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP20]]
+; CHECK-NEXT: [[TMP24:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[TMP25:%.*]] = select i1 [[TMP24]], i32 [[TMP8]], i32 [[TMP23]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP26:%.*]], label [[TMP27:%.*]], !prof [[PROF0]]
+; CHECK: 26:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP8]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 27:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> [[A]], <2 x i64> [[B]], <2 x i64> [[C]], <2 x i64> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, <2 x i64> undef, ptr %P)
+ ret void
+}
+
+define void @st4_2d_undefAB(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, <2 x i64> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_2d_undefAB(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], <2 x i64> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 64) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> [[TMP3]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <4 x i64> <i64 -1, i64 -1, i64 -1, i64 -1>, <4 x i64> [[TMP7]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP9:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP10:%.*]] = xor i64 [[TMP9]], 193514046488576
+; CHECK-NEXT: [[TMP11:%.*]] = inttoptr i64 [[TMP10]] to ptr
+; CHECK-NEXT: [[TMP12:%.*]] = add i64 [[TMP10]], 35184372088832
+; CHECK-NEXT: [[TMP13:%.*]] = and i64 [[TMP12]], -4
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: store <8 x i64> [[TMP8]], ptr [[TMP11]], align 1
+; CHECK-NEXT: [[TMP15:%.*]] = bitcast <2 x i64> [[TMP1]] to i128
+; CHECK-NEXT: [[TMP16:%.*]] = icmp ne i128 [[TMP15]], 0
+; CHECK-NEXT: [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP18:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i128 [[TMP18]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP4]], i32 [[TMP17]]
+; CHECK-NEXT: [[TMP21:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP22:%.*]] = select i1 [[TMP21]], i32 [[TMP6]], i32 [[TMP20]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP23:%.*]], label [[TMP24:%.*]], !prof [[PROF0]]
+; CHECK: 23:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 24:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> undef, <2 x i64> undef, <2 x i64> [[C]], <2 x i64> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> undef, <2 x i64> undef, <2 x i64> %C, <2 x i64> %D, ptr %P)
+ ret void
+}
+
+define void @st4_2d_undefAC(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, <2 x i64> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_2d_undefAC(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], <2 x i64> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 64) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <2 x i64> <i64 -1, i64 -1>, <2 x i64> [[TMP1]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <2 x i64> <i64 -1, i64 -1>, <2 x i64> [[TMP3]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <4 x i64> [[TMP7]], <4 x i64> [[TMP8]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP10:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: [[TMP13:%.*]] = add i64 [[TMP11]], 35184372088832
+; CHECK-NEXT: [[TMP14:%.*]] = and i64 [[TMP13]], -4
+; CHECK-NEXT: [[TMP15:%.*]] = inttoptr i64 [[TMP14]] to ptr
+; CHECK-NEXT: store <8 x i64> [[TMP9]], ptr [[TMP12]], align 1
+; CHECK-NEXT: [[TMP16:%.*]] = bitcast <2 x i64> [[TMP1]] to i128
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i128 [[TMP16]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP19:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP20:%.*]] = icmp ne i128 [[TMP19]], 0
+; CHECK-NEXT: [[TMP21:%.*]] = select i1 [[TMP20]], i32 [[TMP4]], i32 [[TMP18]]
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP21]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP24:%.*]], label [[TMP25:%.*]], !prof [[PROF0]]
+; CHECK: 24:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 25:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> undef, <2 x i64> [[B]], <2 x i64> undef, <2 x i64> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> undef, <2 x i64> %B, <2 x i64> undef, <2 x i64> %D, ptr %P)
+ ret void
+}
+
+define void @st4_2d_undefAD(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, <2 x i64> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_2d_undefAD(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], <2 x i64> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 64) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <2 x i64> <i64 -1, i64 -1>, <2 x i64> [[TMP1]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <2 x i64> [[TMP3]], <2 x i64> <i64 -1, i64 -1>, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <4 x i64> [[TMP7]], <4 x i64> [[TMP8]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP10:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: [[TMP13:%.*]] = add i64 [[TMP11]], 35184372088832
+; CHECK-NEXT: [[TMP14:%.*]] = and i64 [[TMP13]], -4
+; CHECK-NEXT: [[TMP15:%.*]] = inttoptr i64 [[TMP14]] to ptr
+; CHECK-NEXT: store <8 x i64> [[TMP9]], ptr [[TMP12]], align 1
+; CHECK-NEXT: [[TMP16:%.*]] = bitcast <2 x i64> [[TMP1]] to i128
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i128 [[TMP16]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP19:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP20:%.*]] = icmp ne i128 [[TMP19]], 0
+; CHECK-NEXT: [[TMP21:%.*]] = select i1 [[TMP20]], i32 [[TMP4]], i32 [[TMP18]]
+; CHECK-NEXT: [[TMP22:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP23:%.*]] = select i1 [[TMP22]], i32 [[TMP6]], i32 [[TMP21]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP24:%.*]], label [[TMP25:%.*]], !prof [[PROF0]]
+; CHECK: 24:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 25:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> undef, <2 x i64> [[B]], <2 x i64> [[C]], <2 x i64> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> undef, <2 x i64> %B, <2 x i64> %C, <2 x i64> undef, ptr %P)
+ ret void
+}
+
+define void @st4_2d_undefBC(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, <2 x i64> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_2d_undefBC(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], <2 x i64> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 64) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> <i64 -1, i64 -1>, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <2 x i64> <i64 -1, i64 -1>, <2 x i64> [[TMP3]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <4 x i64> [[TMP7]], <4 x i64> [[TMP8]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP10:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: [[TMP13:%.*]] = add i64 [[TMP11]], 35184372088832
+; CHECK-NEXT: [[TMP14:%.*]] = and i64 [[TMP13]], -4
+; CHECK-NEXT: [[TMP15:%.*]] = inttoptr i64 [[TMP14]] to ptr
+; CHECK-NEXT: store <8 x i64> [[TMP9]], ptr [[TMP12]], align 1
+; CHECK-NEXT: [[TMP16:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i128 [[TMP16]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP6]], i32 [[TMP18]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP21:%.*]], label [[TMP22:%.*]], !prof [[PROF0]]
+; CHECK: 21:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 22:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> [[A]], <2 x i64> undef, <2 x i64> undef, <2 x i64> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> %A, <2 x i64> undef, <2 x i64> undef, <2 x i64> %D, ptr %P)
+ ret void
+}
+
+define void @st4_2d_undefBD(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, <2 x i64> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_2d_undefBD(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], <2 x i64> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 64) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> <i64 -1, i64 -1>, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <2 x i64> [[TMP3]], <2 x i64> <i64 -1, i64 -1>, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP9:%.*]] = shufflevector <4 x i64> [[TMP7]], <4 x i64> [[TMP8]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP10:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP11:%.*]] = xor i64 [[TMP10]], 193514046488576
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: [[TMP13:%.*]] = add i64 [[TMP11]], 35184372088832
+; CHECK-NEXT: [[TMP14:%.*]] = and i64 [[TMP13]], -4
+; CHECK-NEXT: [[TMP15:%.*]] = inttoptr i64 [[TMP14]] to ptr
+; CHECK-NEXT: store <8 x i64> [[TMP9]], ptr [[TMP12]], align 1
+; CHECK-NEXT: [[TMP16:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP17:%.*]] = icmp ne i128 [[TMP16]], 0
+; CHECK-NEXT: [[TMP18:%.*]] = select i1 [[TMP17]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP19:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP20:%.*]] = select i1 [[TMP19]], i32 [[TMP6]], i32 [[TMP18]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP21:%.*]], label [[TMP22:%.*]], !prof [[PROF0]]
+; CHECK: 21:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 22:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> [[A]], <2 x i64> undef, <2 x i64> [[C]], <2 x i64> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> %A, <2 x i64> undef, <2 x i64> %C, <2 x i64> undef, ptr %P)
+ ret void
+}
+
+define void @st4_2d_undefCD(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, <2 x i64> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_2d_undefCD(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], <2 x i64> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr @__msan_param_tls, align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr @__msan_param_origin_tls, align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 16) to ptr), align 4
+; CHECK-NEXT: [[TMP5:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: [[TMP6:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 64) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> [[TMP3]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <4 x i64> [[TMP7]], <4 x i64> <i64 -1, i64 -1, i64 -1, i64 -1>, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP9:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP10:%.*]] = xor i64 [[TMP9]], 193514046488576
+; CHECK-NEXT: [[TMP11:%.*]] = inttoptr i64 [[TMP10]] to ptr
+; CHECK-NEXT: [[TMP12:%.*]] = add i64 [[TMP10]], 35184372088832
+; CHECK-NEXT: [[TMP13:%.*]] = and i64 [[TMP12]], -4
+; CHECK-NEXT: [[TMP14:%.*]] = inttoptr i64 [[TMP13]] to ptr
+; CHECK-NEXT: store <8 x i64> [[TMP8]], ptr [[TMP11]], align 1
+; CHECK-NEXT: [[TMP15:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[TMP16:%.*]] = icmp ne i128 [[TMP15]], 0
+; CHECK-NEXT: [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP4]], i32 [[TMP2]]
+; CHECK-NEXT: [[TMP18:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP19:%.*]] = select i1 [[TMP18]], i32 [[TMP6]], i32 [[TMP17]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP20:%.*]], label [[TMP21:%.*]], !prof [[PROF0]]
+; CHECK: 20:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP6]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 21:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> [[A]], <2 x i64> [[B]], <2 x i64> undef, <2 x i64> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> %A, <2 x i64> %B, <2 x i64> undef, <2 x i64> undef, ptr %P)
+ ret void
+}
+
+define void @st4_2d_undefABC(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, <2 x i64> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_2d_undefABC(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], <2 x i64> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 48) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 64) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <2 x i64> <i64 -1, i64 -1>, <2 x i64> [[TMP1]], <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <4 x i64> <i64 -1, i64 -1, i64 -1, i64 -1>, <4 x i64> [[TMP5]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP7:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP8:%.*]] = xor i64 [[TMP7]], 193514046488576
+; CHECK-NEXT: [[TMP9:%.*]] = inttoptr i64 [[TMP8]] to ptr
+; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[TMP8]], 35184372088832
+; CHECK-NEXT: [[TMP11:%.*]] = and i64 [[TMP10]], -4
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: store <8 x i64> [[TMP6]], ptr [[TMP9]], align 1
+; CHECK-NEXT: [[TMP13:%.*]] = bitcast <2 x i64> [[TMP1]] to i128
+; CHECK-NEXT: [[TMP14:%.*]] = icmp ne i128 [[TMP13]], 0
+; CHECK-NEXT: [[TMP15:%.*]] = select i1 [[TMP14]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP16:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP4]], i32 [[TMP15]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP18:%.*]], label [[TMP19:%.*]], !prof [[PROF0]]
+; CHECK: 18:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP4]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 19:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> undef, <2 x i64> undef, <2 x i64> undef, <2 x i64> [[D]], ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> undef, <2 x i64> undef, <2 x i64> undef, <2 x i64> %D, ptr %P)
+ ret void
+}
+
+define void @st4_2d_undefABD(<2 x i64> %A, <2 x i64> %B, <2 x i64> %C, <2 x i64> %D, ptr %P) nounwind sanitize_memory {
+; CHECK-LABEL: define void @st4_2d_undefABD(
+; CHECK-SAME: <2 x i64> [[A:%.*]], <2 x i64> [[B:%.*]], <2 x i64> [[C:%.*]], <2 x i64> [[D:%.*]], ptr [[P:%.*]]) #[[ATTR0]] {
+; CHECK-NEXT: [[TMP1:%.*]] = load <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP2:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 32) to ptr), align 4
+; CHECK-NEXT: [[TMP3:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i32, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_origin_tls to i64), i64 64) to ptr), align 4
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = shufflevector <2 x i64> [[TMP1]], <2 x i64> <i64 -1, i64 -1>, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <4 x i64> <i64 -1, i64 -1, i64 -1, i64 -1>, <4 x i64> [[TMP5]], <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT: [[TMP7:%.*]] = ptrtoint ptr [[P]] to i64
+; CHECK-NEXT: [[TMP8:%.*]] = xor i64 [[TMP7]], 193514046488576
+; CHECK-NEXT: [[TMP9:%.*]] = inttoptr i64 [[TMP8]] to ptr
+; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[TMP8]], 35184372088832
+; CHECK-NEXT: [[TMP11:%.*]] = and i64 [[TMP10]], -4
+; CHECK-NEXT: [[TMP12:%.*]] = inttoptr i64 [[TMP11]] to ptr
+; CHECK-NEXT: store <8 x i64> [[TMP6]], ptr [[TMP9]], align 1
+; CHECK-NEXT: [[TMP13:%.*]] = bitcast <2 x i64> [[TMP1]] to i128
+; CHECK-NEXT: [[TMP14:%.*]] = icmp ne i128 [[TMP13]], 0
+; CHECK-NEXT: [[TMP15:%.*]] = select i1 [[TMP14]], i32 [[TMP2]], i32 0
+; CHECK-NEXT: [[TMP16:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[TMP17:%.*]] = select i1 [[TMP16]], i32 [[TMP4]], i32 [[TMP15]]
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: br i1 [[_MSCMP]], label [[TMP18:%.*]], label [[TMP19:%.*]], !prof [[PROF0]]
+; CHECK: 18:
+; CHECK-NEXT: call void @__msan_warning_with_origin_noreturn(i32 [[TMP4]]) #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 19:
+; CHECK-NEXT: call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> undef, <2 x i64> undef, <2 x i64> [[C]], <2 x i64> undef, ptr [[P]])
+; CHECK-NEXT: ret void
+;
+ call void @llvm.aarch64.neon.st4.v2i64.p0(<2 x i64> undef, <2 x i64> undef, <2 x i64> %C, <2 x i64> undef, ptr %P)
----------------
thurstond wrote:
There are multiple versions of the test, some with undefs, and some without
e.g.,
void @st4_2d: fully defined
void @st4_2d_undefA: A is undefined
void @st4_2d_undefABD: A, B, D are undefined
...
https://github.com/llvm/llvm-project/pull/99360
More information about the cfe-commits
mailing list