[llvm] ff08215 - [msan] Precommit MSan Arm NEON vst tests (#98247)
via llvm-commits
llvm-commits at lists.llvm.org
Wed Jul 17 12:14:39 PDT 2024
Author: Thurston Dang
Date: 2024-07-17T12:14:36-07:00
New Revision: ff0821583eab1651ff126bbf4f881e6163b67435
URL: https://github.com/llvm/llvm-project/commit/ff0821583eab1651ff126bbf4f881e6163b67435
DIFF: https://github.com/llvm/llvm-project/commit/ff0821583eab1651ff126bbf4f881e6163b67435.diff
LOG: [msan] Precommit MSan Arm NEON vst tests (#98247)
These tests show that MSan currently does not handle vst (or vld)
correctly.
Added:
llvm/test/Instrumentation/MemorySanitizer/AArch64/neon_vst.ll
Modified:
Removed:
################################################################################
diff --git a/llvm/test/Instrumentation/MemorySanitizer/AArch64/neon_vst.ll b/llvm/test/Instrumentation/MemorySanitizer/AArch64/neon_vst.ll
new file mode 100644
index 0000000000000..1c4ca47b60c13
--- /dev/null
+++ b/llvm/test/Instrumentation/MemorySanitizer/AArch64/neon_vst.ll
@@ -0,0 +1,1515 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 2
+; Test memory sanitizer instrumentation for Arm NEON VST instructions.
+;
+; RUN: opt < %s -passes=msan -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 <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast <8 x i8> [[TMP1]] to i64
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP4]], 0
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <8 x i8> [[TMP2]] to i64
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: br i1 [[_MSOR3]], label [[TMP6:%.*]], label [[TMP7:%.*]], !prof [[PROF0:![0-9]+]]
+; CHECK: 6:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4:[0-9]+]]
+; CHECK-NEXT: unreachable
+; CHECK: 7:
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <8 x i8> [[TMP1]] to i64
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <8 x i8> [[TMP2]] to i64
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i64 [[TMP6]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <8 x i8> [[TMP3]] to i64
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: [[_MSCMP4:%.*]] = icmp ne i64 [[TMP4]], 0
+; CHECK-NEXT: [[_MSOR5:%.*]] = or i1 [[_MSOR3]], [[_MSCMP4]]
+; CHECK-NEXT: br i1 [[_MSOR5]], label [[TMP8:%.*]], label [[TMP9:%.*]], !prof [[PROF0]]
+; CHECK: 8:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 9:
+; 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 <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast <8 x i8> [[TMP1]] to i64
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP4]], 0
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <8 x i8> [[TMP2]] to i64
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; 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 <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <8 x i8> [[TMP1]] to i64
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP6]], 0
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <8 x i8> [[TMP2]] to i64
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP8:%.*]] = bitcast <8 x i8> [[TMP3]] to i64
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i64 [[TMP8]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: [[TMP9:%.*]] = bitcast <8 x i8> [[TMP4]] to i64
+; CHECK-NEXT: [[_MSCMP4:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: [[_MSOR5:%.*]] = or i1 [[_MSOR3]], [[_MSCMP4]]
+; CHECK-NEXT: [[_MSCMP6:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR7:%.*]] = or i1 [[_MSOR5]], [[_MSCMP6]]
+; CHECK-NEXT: br i1 [[_MSOR7]], label [[TMP10:%.*]], label [[TMP11:%.*]], !prof [[PROF0]]
+; CHECK: 10:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 11:
+; 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 <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <8 x i8> [[TMP1]] to i64
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <8 x i8> [[TMP2]] to i64
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i64 [[TMP6]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <8 x i8> [[TMP1]] to i64
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <8 x i8> [[TMP2]] to i64
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i64 [[TMP6]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <8 x i8> [[TMP3]] to i64
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <8 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <16 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast <16 x i8> [[TMP1]] to i128
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i128 [[TMP4]], 0
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <16 x i8> [[TMP2]] to i128
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i128 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: br i1 [[_MSOR3]], label [[TMP6:%.*]], label [[TMP7:%.*]], !prof [[PROF0]]
+; CHECK: 6:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 7:
+; 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 <16 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP3:%.*]] = load <16 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <16 x i8> [[TMP1]] to i128
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i128 [[TMP5]], 0
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <16 x i8> [[TMP2]] to i128
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i128 [[TMP6]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <16 x i8> [[TMP3]] to i128
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i128 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: [[_MSCMP4:%.*]] = icmp ne i64 [[TMP4]], 0
+; CHECK-NEXT: [[_MSOR5:%.*]] = or i1 [[_MSOR3]], [[_MSCMP4]]
+; CHECK-NEXT: br i1 [[_MSOR5]], label [[TMP8:%.*]], label [[TMP9:%.*]], !prof [[PROF0]]
+; CHECK: 8:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 9:
+; 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 <16 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP3:%.*]] = load <16 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load <16 x i8>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <16 x i8> [[TMP1]] to i128
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i128 [[TMP6]], 0
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <16 x i8> [[TMP2]] to i128
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i128 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP8:%.*]] = bitcast <16 x i8> [[TMP3]] to i128
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i128 [[TMP8]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: [[TMP9:%.*]] = bitcast <16 x i8> [[TMP4]] to i128
+; CHECK-NEXT: [[_MSCMP4:%.*]] = icmp ne i128 [[TMP9]], 0
+; CHECK-NEXT: [[_MSOR5:%.*]] = or i1 [[_MSOR3]], [[_MSCMP4]]
+; CHECK-NEXT: [[_MSCMP6:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR7:%.*]] = or i1 [[_MSOR5]], [[_MSCMP6]]
+; CHECK-NEXT: br i1 [[_MSOR7]], label [[TMP10:%.*]], label [[TMP11:%.*]], !prof [[PROF0]]
+; CHECK: 10:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 11:
+; 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 <4 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast <4 x i16> [[TMP1]] to i64
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP4]], 0
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <4 x i16> [[TMP2]] to i64
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: br i1 [[_MSOR3]], label [[TMP6:%.*]], label [[TMP7:%.*]], !prof [[PROF0]]
+; CHECK: 6:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 7:
+; 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 <4 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <4 x i16> [[TMP1]] to i64
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <4 x i16> [[TMP2]] to i64
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i64 [[TMP6]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <4 x i16> [[TMP3]] to i64
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: [[_MSCMP4:%.*]] = icmp ne i64 [[TMP4]], 0
+; CHECK-NEXT: [[_MSOR5:%.*]] = or i1 [[_MSOR3]], [[_MSCMP4]]
+; CHECK-NEXT: br i1 [[_MSOR5]], label [[TMP8:%.*]], label [[TMP9:%.*]], !prof [[PROF0]]
+; CHECK: 8:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 9:
+; 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 <4 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load <4 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <4 x i16> [[TMP1]] to i64
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP6]], 0
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <4 x i16> [[TMP2]] to i64
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP8:%.*]] = bitcast <4 x i16> [[TMP3]] to i64
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i64 [[TMP8]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: [[TMP9:%.*]] = bitcast <4 x i16> [[TMP4]] to i64
+; CHECK-NEXT: [[_MSCMP4:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: [[_MSOR5:%.*]] = or i1 [[_MSOR3]], [[_MSCMP4]]
+; CHECK-NEXT: [[_MSCMP6:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR7:%.*]] = or i1 [[_MSOR5]], [[_MSCMP6]]
+; CHECK-NEXT: br i1 [[_MSOR7]], label [[TMP10:%.*]], label [[TMP11:%.*]], !prof [[PROF0]]
+; CHECK: 10:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 11:
+; 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 <8 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast <8 x i16> [[TMP1]] to i128
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i128 [[TMP4]], 0
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <8 x i16> [[TMP2]] to i128
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i128 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: br i1 [[_MSOR3]], label [[TMP6:%.*]], label [[TMP7:%.*]], !prof [[PROF0]]
+; CHECK: 6:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 7:
+; 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 <8 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <8 x i16> [[TMP1]] to i128
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i128 [[TMP5]], 0
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <8 x i16> [[TMP2]] to i128
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i128 [[TMP6]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <8 x i16> [[TMP3]] to i128
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i128 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: [[_MSCMP4:%.*]] = icmp ne i64 [[TMP4]], 0
+; CHECK-NEXT: [[_MSOR5:%.*]] = or i1 [[_MSOR3]], [[_MSCMP4]]
+; CHECK-NEXT: br i1 [[_MSOR5]], label [[TMP8:%.*]], label [[TMP9:%.*]], !prof [[PROF0]]
+; CHECK: 8:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 9:
+; 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 <8 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP3:%.*]] = load <8 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load <8 x i16>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <8 x i16> [[TMP1]] to i128
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i128 [[TMP6]], 0
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <8 x i16> [[TMP2]] to i128
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i128 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP8:%.*]] = bitcast <8 x i16> [[TMP3]] to i128
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i128 [[TMP8]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: [[TMP9:%.*]] = bitcast <8 x i16> [[TMP4]] to i128
+; CHECK-NEXT: [[_MSCMP4:%.*]] = icmp ne i128 [[TMP9]], 0
+; CHECK-NEXT: [[_MSOR5:%.*]] = or i1 [[_MSOR3]], [[_MSCMP4]]
+; CHECK-NEXT: [[_MSCMP6:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR7:%.*]] = or i1 [[_MSOR5]], [[_MSCMP6]]
+; CHECK-NEXT: br i1 [[_MSOR7]], label [[TMP10:%.*]], label [[TMP11:%.*]], !prof [[PROF0]]
+; CHECK: 10:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 11:
+; 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 <2 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast <2 x i32> [[TMP1]] to i64
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP4]], 0
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <2 x i32> [[TMP2]] to i64
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: br i1 [[_MSOR3]], label [[TMP6:%.*]], label [[TMP7:%.*]], !prof [[PROF0]]
+; CHECK: 6:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 7:
+; 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 <2 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <2 x i32> [[TMP1]] to i64
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <2 x i32> [[TMP2]] to i64
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i64 [[TMP6]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <2 x i32> [[TMP3]] to i64
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: [[_MSCMP4:%.*]] = icmp ne i64 [[TMP4]], 0
+; CHECK-NEXT: [[_MSOR5:%.*]] = or i1 [[_MSOR3]], [[_MSCMP4]]
+; CHECK-NEXT: br i1 [[_MSOR5]], label [[TMP8:%.*]], label [[TMP9:%.*]], !prof [[PROF0]]
+; CHECK: 8:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 9:
+; 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 <2 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load <2 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <2 x i32> [[TMP1]] to i64
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP6]], 0
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <2 x i32> [[TMP2]] to i64
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP8:%.*]] = bitcast <2 x i32> [[TMP3]] to i64
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i64 [[TMP8]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: [[TMP9:%.*]] = bitcast <2 x i32> [[TMP4]] to i64
+; CHECK-NEXT: [[_MSCMP4:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: [[_MSOR5:%.*]] = or i1 [[_MSOR3]], [[_MSCMP4]]
+; CHECK-NEXT: [[_MSCMP6:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR7:%.*]] = or i1 [[_MSOR5]], [[_MSCMP6]]
+; CHECK-NEXT: br i1 [[_MSOR7]], label [[TMP10:%.*]], label [[TMP11:%.*]], !prof [[PROF0]]
+; CHECK: 10:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 11:
+; 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 <4 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast <4 x i32> [[TMP1]] to i128
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i128 [[TMP4]], 0
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <4 x i32> [[TMP2]] to i128
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i128 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: br i1 [[_MSOR3]], label [[TMP6:%.*]], label [[TMP7:%.*]], !prof [[PROF0]]
+; CHECK: 6:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 7:
+; 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 <4 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <4 x i32> [[TMP1]] to i128
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i128 [[TMP5]], 0
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <4 x i32> [[TMP2]] to i128
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i128 [[TMP6]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <4 x i32> [[TMP3]] to i128
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i128 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: [[_MSCMP4:%.*]] = icmp ne i64 [[TMP4]], 0
+; CHECK-NEXT: [[_MSOR5:%.*]] = or i1 [[_MSOR3]], [[_MSCMP4]]
+; CHECK-NEXT: br i1 [[_MSOR5]], label [[TMP8:%.*]], label [[TMP9:%.*]], !prof [[PROF0]]
+; CHECK: 8:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 9:
+; 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 <4 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <4 x i32> [[TMP1]] to i128
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i128 [[TMP6]], 0
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <4 x i32> [[TMP2]] to i128
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i128 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP8:%.*]] = bitcast <4 x i32> [[TMP3]] to i128
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i128 [[TMP8]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: [[TMP9:%.*]] = bitcast <4 x i32> [[TMP4]] to i128
+; CHECK-NEXT: [[_MSCMP4:%.*]] = icmp ne i128 [[TMP9]], 0
+; CHECK-NEXT: [[_MSOR5:%.*]] = or i1 [[_MSOR3]], [[_MSCMP4]]
+; CHECK-NEXT: [[_MSCMP6:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR7:%.*]] = or i1 [[_MSOR5]], [[_MSCMP6]]
+; CHECK-NEXT: br i1 [[_MSOR7]], label [[TMP10:%.*]], label [[TMP11:%.*]], !prof [[PROF0]]
+; CHECK: 10:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 11:
+; 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 <1 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast <1 x i64> [[TMP1]] to i64
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP4]], 0
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <1 x i64> [[TMP2]] to i64
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: br i1 [[_MSOR3]], label [[TMP6:%.*]], label [[TMP7:%.*]], !prof [[PROF0]]
+; CHECK: 6:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 7:
+; 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 <1 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP3:%.*]] = load <1 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <1 x i64> [[TMP1]] to i64
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <1 x i64> [[TMP2]] to i64
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i64 [[TMP6]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <1 x i64> [[TMP3]] to i64
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: [[_MSCMP4:%.*]] = icmp ne i64 [[TMP4]], 0
+; CHECK-NEXT: [[_MSOR5:%.*]] = or i1 [[_MSOR3]], [[_MSCMP4]]
+; CHECK-NEXT: br i1 [[_MSOR5]], label [[TMP8:%.*]], label [[TMP9:%.*]], !prof [[PROF0]]
+; CHECK: 8:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 9:
+; 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 <1 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 8) to ptr), align 8
+; CHECK-NEXT: [[TMP3:%.*]] = load <1 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; CHECK-NEXT: [[TMP4:%.*]] = load <1 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 24) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <1 x i64> [[TMP1]] to i64
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i64 [[TMP6]], 0
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <1 x i64> [[TMP2]] to i64
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i64 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP8:%.*]] = bitcast <1 x i64> [[TMP3]] to i64
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i64 [[TMP8]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: [[TMP9:%.*]] = bitcast <1 x i64> [[TMP4]] to i64
+; CHECK-NEXT: [[_MSCMP4:%.*]] = icmp ne i64 [[TMP9]], 0
+; CHECK-NEXT: [[_MSOR5:%.*]] = or i1 [[_MSOR3]], [[_MSCMP4]]
+; CHECK-NEXT: [[_MSCMP6:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR7:%.*]] = or i1 [[_MSOR5]], [[_MSCMP6]]
+; CHECK-NEXT: br i1 [[_MSOR7]], label [[TMP10:%.*]], label [[TMP11:%.*]], !prof [[PROF0]]
+; CHECK: 10:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 11:
+; 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 <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast <2 x i64> [[TMP1]] to i128
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i128 [[TMP4]], 0
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <2 x i64> [[TMP2]] to i128
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i128 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i64 [[TMP3]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: br i1 [[_MSOR3]], label [[TMP6:%.*]], label [[TMP7:%.*]], !prof [[PROF0]]
+; CHECK: 6:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 7:
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <2 x i64> [[TMP1]] to i128
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i128 [[TMP5]], 0
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <2 x i64> [[TMP2]] to i128
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i128 [[TMP6]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i128 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: [[_MSCMP4:%.*]] = icmp ne i64 [[TMP4]], 0
+; CHECK-NEXT: [[_MSOR5:%.*]] = or i1 [[_MSOR3]], [[_MSCMP4]]
+; CHECK-NEXT: br i1 [[_MSOR5]], label [[TMP8:%.*]], label [[TMP9:%.*]], !prof [[PROF0]]
+; CHECK: 8:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 9:
+; 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 <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast <2 x i64> [[TMP1]] to i128
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i128 [[TMP4]], 0
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <2 x i64> [[TMP2]] to i128
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i128 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; 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 <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <2 x i64> [[TMP1]] to i128
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i128 [[TMP6]], 0
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <2 x i64> [[TMP2]] to i128
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i128 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP8:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i128 [[TMP8]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: [[TMP9:%.*]] = bitcast <2 x i64> [[TMP4]] to i128
+; CHECK-NEXT: [[_MSCMP4:%.*]] = icmp ne i128 [[TMP9]], 0
+; CHECK-NEXT: [[_MSOR5:%.*]] = or i1 [[_MSOR3]], [[_MSCMP4]]
+; CHECK-NEXT: [[_MSCMP6:%.*]] = icmp ne i64 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR7:%.*]] = or i1 [[_MSOR5]], [[_MSCMP6]]
+; CHECK-NEXT: br i1 [[_MSOR7]], label [[TMP10:%.*]], label [[TMP11:%.*]], !prof [[PROF0]]
+; CHECK: 10:
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; CHECK-NEXT: unreachable
+; CHECK: 11:
+; 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 <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <2 x i64> [[TMP1]] to i128
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i128 [[TMP5]], 0
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <2 x i64> [[TMP2]] to i128
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i128 [[TMP6]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <2 x i64> [[TMP1]] to i128
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i128 [[TMP5]], 0
+; CHECK-NEXT: [[TMP6:%.*]] = bitcast <2 x i64> [[TMP2]] to i128
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i128 [[TMP6]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: [[TMP7:%.*]] = bitcast <2 x i64> [[TMP3]] to i128
+; CHECK-NEXT: [[_MSCMP2:%.*]] = icmp ne i128 [[TMP7]], 0
+; CHECK-NEXT: [[_MSOR3:%.*]] = or i1 [[_MSOR]], [[_MSCMP2]]
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 48) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 32) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 <2 x i64>, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 16) to ptr), align 8
+; 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: call void @llvm.donothing()
+; CHECK-NEXT: [[TMP4:%.*]] = bitcast <2 x i64> [[TMP1]] to i128
+; CHECK-NEXT: [[_MSCMP:%.*]] = icmp ne i128 [[TMP4]], 0
+; CHECK-NEXT: [[TMP5:%.*]] = bitcast <2 x i64> [[TMP2]] to i128
+; CHECK-NEXT: [[_MSCMP1:%.*]] = icmp ne i128 [[TMP5]], 0
+; CHECK-NEXT: [[_MSOR:%.*]] = or i1 [[_MSCMP]], [[_MSCMP1]]
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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)
+ ret void
+}
+
+define void @st4_2d_undefACD(<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_undefACD
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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> 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> undef, <2 x i64> undef, ptr %P)
+ ret void
+}
+
+define void @st4_2d_undefBCD(<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_undefBCD
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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> 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> undef, <2 x i64> undef, ptr %P)
+ ret void
+}
+
+define void @st4_2d_undefABCD(<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_undefABCD
+; 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 i64, ptr inttoptr (i64 add (i64 ptrtoint (ptr @__msan_param_tls to i64), i64 64) to ptr), align 8
+; CHECK-NEXT: call void @llvm.donothing()
+; CHECK-NEXT: call void @__msan_warning_noreturn() #[[ATTR4]]
+; 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> 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> undef, <2 x i64> undef, ptr %P)
+ ret void
+}
More information about the llvm-commits
mailing list