diff --git a/llvm/include/llvm/Analysis/StackSafetyAnalysis.h b/llvm/include/llvm/Analysis/StackSafetyAnalysis.h
--- a/llvm/include/llvm/Analysis/StackSafetyAnalysis.h
+++ b/llvm/include/llvm/Analysis/StackSafetyAnalysis.h
@@ -78,8 +78,8 @@
   // Whether we can prove that all accesses to this Alloca are in-range and
   // during its lifetime.
   bool isSafe(const AllocaInst &AI) const;
-  // Whether we can prove that an instruction only accesses a live alloca in
-  // range.
+  // Whether we can prove that all stack accesses in an instruction only access
+  // live alloca in range.
   bool accessIsSafe(const Instruction &I) const;
   void print(raw_ostream &O) const;
   void dump() const;
diff --git a/llvm/lib/Transforms/Instrumentation/HWAddressSanitizer.cpp b/llvm/lib/Transforms/Instrumentation/HWAddressSanitizer.cpp
--- a/llvm/lib/Transforms/Instrumentation/HWAddressSanitizer.cpp
+++ b/llvm/lib/Transforms/Instrumentation/HWAddressSanitizer.cpp
@@ -285,6 +285,7 @@
   void instrumentMemAccessInline(Value *Ptr, bool IsWrite,
                                  unsigned AccessSizeIndex,
                                  Instruction *InsertBefore);
+  bool ignoreMemIntrinsic(MemIntrinsic *MI);
   void instrumentMemIntrinsic(MemIntrinsic *MI);
   bool instrumentMemAccess(InterestingMemoryOperand &O);
   bool ignoreAccess(Instruction *Inst, Value *Ptr);
@@ -785,7 +786,7 @@
   if (!InstrumentStack) {
     if (findAllocaForValue(Ptr))
       return true;
-  } else if (SSI && SSI->accessIsSafe(*Inst)) {
+  } else if (SSI && SSI->accessIsSafe(*Inst) && findAllocaForValue(Ptr)) {
     return true;
   }
   return false;
@@ -970,6 +971,16 @@
     cast<BranchInst>(CheckFailTerm)->setSuccessor(0, CheckTerm->getParent());
 }
 
+bool HWAddressSanitizer::ignoreMemIntrinsic(MemIntrinsic *MI) {
+  if (MemTransferInst *MTI = dyn_cast<MemTransferInst>(MI)) {
+    return (!ClInstrumentWrites || ignoreAccess(MTI, MTI->getDest())) &&
+           (!ClInstrumentReads || ignoreAccess(MTI, MTI->getSource()));
+  }
+  if (isa<MemSetInst>(MI))
+    return !ClInstrumentWrites || ignoreAccess(MI, MI->getDest());
+  return false;
+}
+
 void HWAddressSanitizer::instrumentMemIntrinsic(MemIntrinsic *MI) {
   IRBuilder<> IRB(MI);
   if (isa<MemTransferInst>(MI)) {
@@ -1516,7 +1527,8 @@
       getInterestingMemoryOperands(&Inst, OperandsToInstrument);
 
       if (MemIntrinsic *MI = dyn_cast<MemIntrinsic>(&Inst))
-        IntrinToInstrument.push_back(MI);
+        if (!ignoreMemIntrinsic(MI))
+          IntrinToInstrument.push_back(MI);
     }
   }
 
diff --git a/llvm/test/Instrumentation/HWAddressSanitizer/mem-intrinsics.ll b/llvm/test/Instrumentation/HWAddressSanitizer/mem-intrinsics.ll
--- a/llvm/test/Instrumentation/HWAddressSanitizer/mem-intrinsics.ll
+++ b/llvm/test/Instrumentation/HWAddressSanitizer/mem-intrinsics.ll
@@ -1,4 +1,4 @@
-; RUN: opt -S -passes=hwasan %s | FileCheck %s
+; RUN: opt -S -passes=hwasan -hwasan-use-stack-safety=0 %s | FileCheck %s
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64-unknown-linux-gnu"
diff --git a/llvm/test/Instrumentation/HWAddressSanitizer/stack-safety-analysis.ll b/llvm/test/Instrumentation/HWAddressSanitizer/stack-safety-analysis.ll
--- a/llvm/test/Instrumentation/HWAddressSanitizer/stack-safety-analysis.ll
+++ b/llvm/test/Instrumentation/HWAddressSanitizer/stack-safety-analysis.ll
@@ -78,6 +78,65 @@
   ret i32 0
 }
 
+define i32 @test_in_range3(i32* %a) sanitize_hwaddress {
+entry:
+  ; CHECK-LABEL: @test_in_range3
+  ; NOSAFETY: call {{.*}}__hwasan_generate_tag
+  ; NOSAFETY: call {{.*}}__hwasan_memset
+  ; SAFETY-NOT: call {{.*}}__hwasan_generate_tag
+  ; SAFETY-NOT: call {{.*}}__hwasan_memset
+  ; NOSTACK-NOT: call {{.*}}__hwasan_generate_tag
+  ; NOSTACK-NOT: call {{.*}}__hwasan_memset
+  %buf.sroa.0 = alloca [10 x i8], align 4
+  %ptr = getelementptr [10 x i8], [10 x i8]* %buf.sroa.0, i32 0, i32 9
+  %x = bitcast [10 x i8]* %buf.sroa.0 to i8*
+  call void @llvm.lifetime.start.p0i8(i64 10, i8* nonnull %x)
+  call void @llvm.memset.p0i8.i32(i8* %ptr, i8 0, i32 1, i1 true)
+  call void @llvm.lifetime.end.p0i8(i64 10, i8* nonnull %x)
+  ret i32 0
+}
+
+define i32 @test_in_range4(i32* %a) sanitize_hwaddress {
+entry:
+  ; CHECK-LABEL: @test_in_range4
+  ; NOSAFETY: call {{.*}}__hwasan_generate_tag
+  ; NOSAFETY: call {{.*}}__hwasan_memmove
+  ; SAFETY-NOT: call {{.*}}__hwasan_generate_tag
+  ; SAFETY-NOT: call {{.*}}__hwasan_memmove
+  ; NOSTACK-NOT: call {{.*}}__hwasan_generate_tag
+  ; NOSTACK-NOT: call {{.*}}__hwasan_memmove
+  %buf.sroa.0 = alloca [10 x i8], align 4
+  %ptr = getelementptr [10 x i8], [10 x i8]* %buf.sroa.0, i32 0, i32 9
+  %x = bitcast [10 x i8]* %buf.sroa.0 to i8*
+  call void @llvm.lifetime.start.p0i8(i64 10, i8* nonnull %x)
+  call void @llvm.memmove.p0i8.p0i8.i32(i8* %ptr, i8* %ptr, i32 1, i1 true)
+  call void @llvm.lifetime.end.p0i8(i64 10, i8* nonnull %x)
+  ret i32 0
+}
+
+define i32 @test_in_range5(i32* %a) sanitize_hwaddress {
+entry:
+  ; CHECK-LABEL: @test_in_range5
+  ; NOSAFETY: call {{.*}}__hwasan_generate_tag
+  ; NOSAFETY: call {{.*}}__hwasan_memmove
+  ; SAFETY-NOT: call {{.*}}__hwasan_generate_tag
+  ; SAFETY-NOT: call {{.*}}__hwasan_memmove
+  ; NOSTACK-NOT: call {{.*}}__hwasan_generate_tag
+  ; NOSTACK-NOT: call {{.*}}__hwasan_memmove
+  %buf.sroa.0 = alloca [10 x i8], align 4
+  %ptr = getelementptr [10 x i8], [10 x i8]* %buf.sroa.0, i32 0, i32 9
+  %x = bitcast [10 x i8]* %buf.sroa.0 to i8*
+  %buf.sroa.1 = alloca [10 x i8], align 4
+  %ptr1 = getelementptr [10 x i8], [10 x i8]* %buf.sroa.0, i32 0, i32 9
+  %y = bitcast [10 x i8]* %buf.sroa.1 to i8*
+  call void @llvm.lifetime.start.p0i8(i64 10, i8* nonnull %x)
+  call void @llvm.lifetime.start.p0i8(i64 10, i8* nonnull %y)
+  call void @llvm.memmove.p0i8.p0i8.i32(i8* %ptr, i8* %ptr1, i32 1, i1 true)
+  call void @llvm.lifetime.end.p0i8(i64 10, i8* nonnull %x)
+  call void @llvm.lifetime.end.p0i8(i64 10, i8* nonnull %y)
+  ret i32 0
+}
+
 ; Check an alloca with out of range GEP to ensure it gets a tag and check.
 define i32 @test_out_of_range(i32* %a) sanitize_hwaddress {
 entry:
@@ -97,6 +156,65 @@
   ret i32 0
 }
 
+define i32 @test_out_of_range3(i32* %a) sanitize_hwaddress {
+entry:
+  ; CHECK-LABEL: @test_out_of_range3
+  ; NOSAFETY: call {{.*}}__hwasan_generate_tag
+  ; NOSAFETY: call {{.*}}__hwasan_memset
+  ; SAFETY: call {{.*}}__hwasan_generate_tag
+  ; SAFETY: call {{.*}}__hwasan_memset
+  ; NOSTACK-NOT: call {{.*}}__hwasan_generate_tag
+  ; NOSTACK-NOT: call {{.*}}__hwasan_memset
+  %buf.sroa.0 = alloca [10 x i8], align 4
+  %ptr = getelementptr [10 x i8], [10 x i8]* %buf.sroa.0, i32 0, i32 9
+  %x = bitcast [10 x i8]* %buf.sroa.0 to i8*
+  call void @llvm.lifetime.start.p0i8(i64 10, i8* nonnull %x)
+  call void @llvm.memset.p0i8.i32(i8* %ptr, i8 0, i32 2, i1 true)
+  call void @llvm.lifetime.end.p0i8(i64 10, i8* nonnull %x)
+  ret i32 0
+}
+
+define i32 @test_out_of_range4(i32* %a) sanitize_hwaddress {
+entry:
+  ; CHECK-LABEL: @test_out_of_range4
+  ; NOSAFETY: call {{.*}}__hwasan_generate_tag
+  ; NOSAFETY: call {{.*}}__hwasan_memmove
+  ; SAFETY: call {{.*}}__hwasan_generate_tag
+  ; SAFETY: call {{.*}}__hwasan_memmove
+  ; NOSTACK-NOT: call {{.*}}__hwasan_generate_tag
+  ; NOSTACK-NOT: call {{.*}}__hwasan_memmove
+  %buf.sroa.0 = alloca [10 x i8], align 4
+  %ptr = getelementptr [10 x i8], [10 x i8]* %buf.sroa.0, i32 0, i32 9
+  %x = bitcast [10 x i8]* %buf.sroa.0 to i8*
+  call void @llvm.lifetime.start.p0i8(i64 10, i8* nonnull %x)
+  call void @llvm.memmove.p0i8.p0i8.i32(i8* %ptr, i8* %ptr, i32 2, i1 true)
+  call void @llvm.lifetime.end.p0i8(i64 10, i8* nonnull %x)
+  ret i32 0
+}
+
+define i32 @test_out_of_range5(i32* %a) sanitize_hwaddress {
+entry:
+  ; CHECK-LABEL: @test_out_of_range5
+  ; NOSAFETY: call {{.*}}__hwasan_generate_tag
+  ; NOSAFETY: call {{.*}}__hwasan_memmove
+  ; SAFETY: call {{.*}}__hwasan_generate_tag
+  ; SAFETY: call {{.*}}__hwasan_memmove
+  ; NOSTACK-NOT: call {{.*}}__hwasan_generate_tag
+  ; NOSTACK-NOT: call {{.*}}__hwasan_memmove
+  %buf.sroa.0 = alloca [10 x i8], align 4
+  %ptr = getelementptr [10 x i8], [10 x i8]* %buf.sroa.0, i32 0, i32 9
+  %x = bitcast [10 x i8]* %buf.sroa.0 to i8*
+  %buf.sroa.1 = alloca [10 x i8], align 4
+  %ptr1 = getelementptr [10 x i8], [10 x i8]* %buf.sroa.0, i32 0, i32 9
+  %y = bitcast [10 x i8]* %buf.sroa.1 to i8*
+  call void @llvm.lifetime.start.p0i8(i64 10, i8* nonnull %x)
+  call void @llvm.lifetime.end.p0i8(i64 10, i8* nonnull %x)
+  call void @llvm.lifetime.start.p0i8(i64 10, i8* nonnull %y)
+  call void @llvm.memmove.p0i8.p0i8.i32(i8* %ptr, i8* %ptr1, i32 1, i1 true)
+  call void @llvm.lifetime.end.p0i8(i64 10, i8* nonnull %y)
+  ret i32 0
+}
+
 ; Check an alloca with potentially out of range GEP to ensure it gets a tag and
 ; check.
 define i32 @test_potentially_out_of_range(i32* %a) sanitize_hwaddress {
@@ -117,6 +235,23 @@
   ret i32 0
 }
 
+define i32 @test_potentially_out_of_range2(i8* %a) sanitize_hwaddress {
+entry:
+  ; CHECK-LABEL: @test_potentially_out_of_range2
+  ; NOSAFETY: call {{.*}}__hwasan_generate_tag
+  ; NOSAFETY: call {{.*}}__hwasan_memmove
+  ; SAFETY-NOT: call {{.*}}__hwasan_generate_tag
+  ; SAFETY: call {{.*}}__hwasan_memmove
+  ; NOSTACK-NOT: call {{.*}}__hwasan_generate_tag
+  ; NOSTACK: call {{.*}}__hwasan_memmove
+  %buf.sroa.0 = alloca [10 x i8], align 4
+  %ptr = getelementptr [10 x i8], [10 x i8]* %buf.sroa.0, i32 0, i32 9
+  %x = bitcast [10 x i8]* %buf.sroa.0 to i8*
+  call void @llvm.lifetime.start.p0i8(i64 10, i8* nonnull %x)
+  call void @llvm.memmove.p0i8.p0i8.i32(i8* %ptr, i8* %a, i32 1, i1 true)
+  call void @llvm.lifetime.end.p0i8(i64 10, i8* nonnull %x)
+  ret i32 0
+}
 ; Check an alloca with potentially out of range GEP to ensure it gets a tag and
 ; check.
 define i32 @test_unclear(i32* %a) sanitize_hwaddress {
@@ -179,6 +314,10 @@
 ; Function Attrs: argmemonly mustprogress nofree nosync nounwind willreturn
 declare void @llvm.lifetime.end.p0i8(i64 immarg, i8* nocapture)
 
+declare void @llvm.memset.p0i8.i32(i8*, i8, i32, i1)
+declare void @llvm.memcpy.p0i8.p0i8.i32(i8*, i8*, i32, i1)
+declare void @llvm.memmove.p0i8.p0i8.i32(i8*, i8*, i32, i1)
+
 declare i1 @cond()
 declare void @use(i8* nocapture)
 declare i32 @getoffset()