Index: lib/Transforms/Utils/SimplifyLibCalls.cpp
===================================================================
--- lib/Transforms/Utils/SimplifyLibCalls.cpp
+++ lib/Transforms/Utils/SimplifyLibCalls.cpp
@@ -56,6 +56,36 @@
          Func == LibFunc::llabs || Func == LibFunc::strlen;
 }
 
+static bool isCallingConvCCompatible(CallInst *CI) {
+  switch(CI->getCallingConv()) {
+  case llvm::CallingConv::C:
+    return true;
+  case llvm::CallingConv::ARM_APCS:
+  case llvm::CallingConv::ARM_AAPCS:
+  case llvm::CallingConv::ARM_AAPCS_VFP: {
+
+    // The iOS ABI diverges from the standard in some cases, so for now don't
+    // try to simplify those calls.
+    if (Triple(CI->getModule()->getTargetTriple()).isiOS())
+      return false;
+
+    auto *FuncTy = CI->getFunctionType();
+
+    if (!FuncTy->getReturnType()->isPointerTy() &&
+        !FuncTy->getReturnType()->isIntegerTy() &&
+        !FuncTy->getReturnType()->isVoidTy())
+      return false;
+
+    for (auto Param : FuncTy->params()) {
+      if (!Param->isPointerTy() && !Param->isIntegerTy())
+        return false;
+    }
+    return true;
+  }
+  }
+  return false;
+}
+
 /// Return true if it only matters that the value is equal or not-equal to zero.
 static bool isOnlyUsedInZeroEqualityComparison(Value *V) {
   for (User *U : V->users()) {
@@ -1898,7 +1928,7 @@
   if (TLI->getLibFunc(*Callee, Func) && TLI->has(Func)) {
     // Make sure we never change the calling convention.
     assert((ignoreCallingConv(Func) ||
-            CI->getCallingConv() == llvm::CallingConv::C) &&
+            isCallingConvCCompatible(CI)) &&
       "Optimizing string/memory libcall would change the calling convention");
     switch (Func) {
     case LibFunc::strcat:
@@ -1965,7 +1995,7 @@
   SmallVector<OperandBundleDef, 2> OpBundles;
   CI->getOperandBundlesAsDefs(OpBundles);
   IRBuilder<> Builder(CI, /*FPMathTag=*/nullptr, OpBundles);
-  bool isCallingConvC = CI->getCallingConv() == llvm::CallingConv::C;
+  bool isCallingConvC = isCallingConvCCompatible(CI);
 
   // Command-line parameter overrides instruction attribute.
   if (EnableUnsafeFPShrink.getNumOccurrences() > 0)
@@ -2321,7 +2351,7 @@
   SmallVector<OperandBundleDef, 2> OpBundles;
   CI->getOperandBundlesAsDefs(OpBundles);
   IRBuilder<> Builder(CI, /*FPMathTag=*/nullptr, OpBundles);
-  bool isCallingConvC = CI->getCallingConv() == llvm::CallingConv::C;
+  bool isCallingConvC = isCallingConvCCompatible(CI);
 
   // First, check that this is a known library functions and that the prototype
   // is correct.
Index: test/Transforms/InstCombine/ARM/strcmp.ll
===================================================================
--- /dev/null
+++ test/Transforms/InstCombine/ARM/strcmp.ll
@@ -0,0 +1,153 @@
+; Test that the strcmp library call simplifier works correctly.
+; RUN: opt < %s -instcombine -S | FileCheck %s
+
+target datalayout = "e-m:e-p:32:32-i64:64-v128:64:128-a:0:32-n32-S64"
+
+@hello = constant [6 x i8] c"hello\00"
+@hell = constant [5 x i8] c"hell\00"
+@bell = constant [5 x i8] c"bell\00"
+@null = constant [1 x i8] zeroinitializer
+
+declare i32 @strcmp(i8*, i8*)
+
+; strcmp("", x) -> -*x
+define arm_aapcscc i32 @test1(i8* %str2) {
+; CHECK-LABEL: @test1(
+; CHECK: %strcmpload = load i8, i8* %str
+; CHECK: %1 = zext i8 %strcmpload to i32
+; CHECK: %2 = sub nsw i32 0, %1
+; CHECK: ret i32 %2
+
+  %str1 = getelementptr inbounds [1 x i8], [1 x i8]* @null, i32 0, i32 0
+  %temp1 = call arm_apcscc i32 @strcmp(i8* %str1, i8* %str2)
+  ret i32 %temp1
+
+}
+
+; strcmp(x, "") -> *x
+define arm_aapcscc i32 @test2(i8* %str1) {
+; CHECK-LABEL: @test2(
+; CHECK: %strcmpload = load i8, i8* %str
+; CHECK: %1 = zext i8 %strcmpload to i32
+; CHECK: ret i32 %1
+
+  %str2 = getelementptr inbounds [1 x i8], [1 x i8]* @null, i32 0, i32 0
+  %temp1 = call arm_aapcscc i32 @strcmp(i8* %str1, i8* %str2)
+  ret i32 %temp1
+}
+
+; strcmp(x, y)  -> cnst
+define arm_aapcscc i32 @test3() {
+; CHECK-LABEL: @test3(
+; CHECK: ret i32 -1
+
+  %str1 = getelementptr inbounds [5 x i8], [5 x i8]* @hell, i32 0, i32 0
+  %str2 = getelementptr inbounds [6 x i8], [6 x i8]* @hello, i32 0, i32 0
+  %temp1 = call arm_aapcscc i32 @strcmp(i8* %str1, i8* %str2)
+  ret i32 %temp1
+}
+
+define arm_aapcscc i32 @test4() {
+; CHECK-LABEL: @test4(
+; CHECK: ret i32 1
+
+  %str1 = getelementptr inbounds [5 x i8], [5 x i8]* @hell, i32 0, i32 0
+  %str2 = getelementptr inbounds [1 x i8], [1 x i8]* @null, i32 0, i32 0
+  %temp1 = call arm_aapcscc i32 @strcmp(i8* %str1, i8* %str2)
+  ret i32 %temp1
+}
+
+; strcmp(x, y)   -> memcmp(x, y, <known length>)
+; (This transform is rather difficult to trigger in a useful manner)
+define arm_aapcscc i32 @test5(i1 %b) {
+; CHECK-LABEL: @test5(
+; CHECK: %memcmp = call i32 @memcmp(i8* getelementptr inbounds ([6 x i8], [6 x i8]* @hello, i32 0, i32 0), i8* %str2, i32 5)
+; CHECK: ret i32 %memcmp
+
+  %str1 = getelementptr inbounds [6 x i8], [6 x i8]* @hello, i32 0, i32 0
+  %temp1 = getelementptr inbounds [5 x i8], [5 x i8]* @hell, i32 0, i32 0
+  %temp2 = getelementptr inbounds [5 x i8], [5 x i8]* @bell, i32 0, i32 0
+  %str2 = select i1 %b, i8* %temp1, i8* %temp2
+  %temp3 = call arm_aapcscc i32 @strcmp(i8* %str1, i8* %str2)
+  ret i32 %temp3
+}
+
+; strcmp(x,x)  -> 0
+define arm_aapcscc i32 @test6(i8* %str) {
+; CHECK-LABEL: @test6(
+; CHECK: ret i32 0
+
+  %temp1 = call arm_aapcscc i32 @strcmp(i8* %str, i8* %str)
+  ret i32 %temp1
+}
+
+; strcmp("", x) -> -*x
+define arm_aapcs_vfpcc i32 @test1_vfp(i8* %str2) {
+; CHECK-LABEL: @test1_vfp(
+; CHECK: %strcmpload = load i8, i8* %str
+; CHECK: %1 = zext i8 %strcmpload to i32
+; CHECK: %2 = sub nsw i32 0, %1
+; CHECK: ret i32 %2
+
+  %str1 = getelementptr inbounds [1 x i8], [1 x i8]* @null, i32 0, i32 0
+  %temp1 = call arm_aapcs_vfpcc i32 @strcmp(i8* %str1, i8* %str2)
+  ret i32 %temp1
+
+}
+
+; strcmp(x, "") -> *x
+define arm_aapcs_vfpcc i32 @test2_vfp(i8* %str1) {
+; CHECK-LABEL: @test2_vfp(
+; CHECK: %strcmpload = load i8, i8* %str
+; CHECK: %1 = zext i8 %strcmpload to i32
+; CHECK: ret i32 %1
+
+  %str2 = getelementptr inbounds [1 x i8], [1 x i8]* @null, i32 0, i32 0
+  %temp1 = call arm_aapcs_vfpcc i32 @strcmp(i8* %str1, i8* %str2)
+  ret i32 %temp1
+}
+
+; strcmp(x, y)  -> cnst
+define arm_aapcs_vfpcc i32 @test3_vfp() {
+; CHECK-LABEL: @test3_vfp(
+; CHECK: ret i32 -1
+
+  %str1 = getelementptr inbounds [5 x i8], [5 x i8]* @hell, i32 0, i32 0
+  %str2 = getelementptr inbounds [6 x i8], [6 x i8]* @hello, i32 0, i32 0
+  %temp1 = call arm_aapcs_vfpcc i32 @strcmp(i8* %str1, i8* %str2)
+  ret i32 %temp1
+}
+
+define arm_aapcs_vfpcc i32 @test4_vfp() {
+; CHECK-LABEL: @test4_vfp(
+; CHECK: ret i32 1
+
+  %str1 = getelementptr inbounds [5 x i8], [5 x i8]* @hell, i32 0, i32 0
+  %str2 = getelementptr inbounds [1 x i8], [1 x i8]* @null, i32 0, i32 0
+  %temp1 = call arm_aapcs_vfpcc i32 @strcmp(i8* %str1, i8* %str2)
+  ret i32 %temp1
+}
+
+; strcmp(x, y)   -> memcmp(x, y, <known length>)
+; (This transform is rather difficult to trigger in a useful manner)
+define arm_aapcs_vfpcc i32 @test5_vfp(i1 %b) {
+; CHECK-LABEL: @test5_vfp(
+; CHECK: %memcmp = call i32 @memcmp(i8* getelementptr inbounds ([6 x i8], [6 x i8]* @hello, i32 0, i32 0), i8* %str2, i32 5)
+; CHECK: ret i32 %memcmp
+
+  %str1 = getelementptr inbounds [6 x i8], [6 x i8]* @hello, i32 0, i32 0
+  %temp1 = getelementptr inbounds [5 x i8], [5 x i8]* @hell, i32 0, i32 0
+  %temp2 = getelementptr inbounds [5 x i8], [5 x i8]* @bell, i32 0, i32 0
+  %str2 = select i1 %b, i8* %temp1, i8* %temp2
+  %temp3 = call arm_aapcs_vfpcc i32 @strcmp(i8* %str1, i8* %str2)
+  ret i32 %temp3
+}
+
+; strcmp(x,x)  -> 0
+define arm_aapcs_vfpcc i32 @test6_vfp(i8* %str) {
+; CHECK-LABEL: @test6_vfp(
+; CHECK: ret i32 0
+
+  %temp1 = call arm_aapcs_vfpcc i32 @strcmp(i8* %str, i8* %str)
+  ret i32 %temp1
+}
Index: test/Transforms/InstCombine/ARM/strcpy.ll
===================================================================
--- /dev/null
+++ test/Transforms/InstCombine/ARM/strcpy.ll
@@ -0,0 +1,76 @@
+; Test that the strcpy library call simplifier works correctly for ARM procedure calls
+; RUN: opt < %s -instcombine -S | FileCheck %s
+;
+; This transformation requires the pointer size, as it assumes that size_t is
+; the size of a pointer.
+target datalayout = "e-m:e-p:32:32-i64:64-v128:64:128-a:0:32-n32-S64"
+
+@hello = constant [6 x i8] c"hello\00"
+@a = common global [32 x i8] zeroinitializer, align 1
+@b = common global [32 x i8] zeroinitializer, align 1
+
+declare i8* @strcpy(i8*, i8*)
+
+define arm_aapcscc void @test_simplify1() {
+; CHECK-LABEL: @test_simplify1(
+
+  %dst = getelementptr [32 x i8], [32 x i8]* @a, i32 0, i32 0
+  %src = getelementptr [6 x i8], [6 x i8]* @hello, i32 0, i32 0
+
+  call arm_aapcscc i8* @strcpy(i8* %dst, i8* %src)
+; CHECK: @llvm.memcpy.p0i8.p0i8.i32
+  ret void
+}
+
+define arm_aapcscc i8* @test_simplify2() {
+; CHECK-LABEL: @test_simplify2(
+
+  %dst = getelementptr [32 x i8], [32 x i8]* @a, i32 0, i32 0
+
+  %ret = call arm_aapcscc i8* @strcpy(i8* %dst, i8* %dst)
+; CHECK: ret i8* getelementptr inbounds ([32 x i8], [32 x i8]* @a, i32 0, i32 0)
+  ret i8* %ret
+}
+
+define arm_aapcscc i8* @test_no_simplify1() {
+; CHECK-LABEL: @test_no_simplify1(
+
+  %dst = getelementptr [32 x i8], [32 x i8]* @a, i32 0, i32 0
+  %src = getelementptr [32 x i8], [32 x i8]* @b, i32 0, i32 0
+
+  %ret = call arm_aapcscc i8* @strcpy(i8* %dst, i8* %src)
+; CHECK: call arm_aapcscc i8* @strcpy
+  ret i8* %ret
+}
+
+define arm_aapcs_vfpcc void @test_simplify1_vfp() {
+; CHECK-LABEL: @test_simplify1_vfp(
+
+  %dst = getelementptr [32 x i8], [32 x i8]* @a, i32 0, i32 0
+  %src = getelementptr [6 x i8], [6 x i8]* @hello, i32 0, i32 0
+
+  call arm_aapcs_vfpcc i8* @strcpy(i8* %dst, i8* %src)
+; CHECK: @llvm.memcpy.p0i8.p0i8.i32
+  ret void
+}
+
+define arm_aapcs_vfpcc i8* @test_simplify2_vfp() {
+; CHECK-LABEL: @test_simplify2_vfp(
+
+  %dst = getelementptr [32 x i8], [32 x i8]* @a, i32 0, i32 0
+
+  %ret = call arm_aapcs_vfpcc i8* @strcpy(i8* %dst, i8* %dst)
+; CHECK: ret i8* getelementptr inbounds ([32 x i8], [32 x i8]* @a, i32 0, i32 0)
+  ret i8* %ret
+}
+
+define arm_aapcs_vfpcc i8* @test_no_simplify1_vfp() {
+; CHECK-LABEL: @test_no_simplify1_vfp(
+
+  %dst = getelementptr [32 x i8], [32 x i8]* @a, i32 0, i32 0
+  %src = getelementptr [32 x i8], [32 x i8]* @b, i32 0, i32 0
+
+  %ret = call arm_aapcs_vfpcc i8* @strcpy(i8* %dst, i8* %src)
+; CHECK: call arm_aapcs_vfpcc i8* @strcpy
+  ret i8* %ret
+}
Index: test/Transforms/InstCombine/PR7357.ll
===================================================================
--- test/Transforms/InstCombine/PR7357.ll
+++ /dev/null
@@ -1,17 +0,0 @@
-; RUN: opt < %s "-default-data-layout=e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-n32" -instcombine -S | FileCheck %s
-@.str1 = private constant [11 x i8] c"(){};[]&|:\00", align 4
-
-; check that simplify libcalls will not replace a call with one calling
-; convention with a new call with a different calling convention.
-
-; CHECK: define arm_aapcscc i32 @foo(i32 %argc)
-; CHECK: call arm_aapcscc  i8* @strchr
-define arm_aapcscc i32 @foo(i32 %argc) nounwind {
-bb.nph:
-  %c = call arm_aapcscc  i8* @strchr(i8* getelementptr ([11 x i8], [11 x i8]* @.str1, i32 0,
-i32 0), i32 %argc) nounwind readonly
-  %p = ptrtoint i8* %c to i32
-  ret i32 %p
-}
-
-declare arm_aapcscc i8* @strchr(i8*, i32) nounwind readonly