[ARM64] Prefer generation of bzero on Darwin only

[bsmith]

Summary

This patch modifies the behavior of the ARM64 backend to only prefer generating bzero for zero memsets when the target is Darwin, since not all target libraries will have this function. This matches the behavior of x86 (almost, x86 is MacOSX >= 10.6 however for ARM64 I believe that just checking for Darwin should be sufficient), no other backends use bzero.

[t.p.northover]

Hi Bradley,

This looks fine to me.

Cheers.

Tim.

lib/Target/ARM64/ARM64Subtarget.cpp

 }
 
 /// This function returns the name of a function which has an interface
 /// like the non-standard bzero function, if such a function exists on
 /// the current subtarget and it is considered prefereable over
 /// memset with zero passed as the second argument. Otherwise it
 /// returns null.
 const char *ARM64Subtarget::getBZeroEntry() const {
-  // At the moment, always prefer bzero.
+  // Prefer bzero on Darwin only.
+  if(isTargetDarwin())
     return "bzero";
+
+  return nullptr;
 }
 
 void ARM64Subtarget::overrideSchedPolicy(MachineSchedPolicy &Policy,
                                          MachineInstr *begin, MachineInstr *end,
                                          unsigned NumRegionInstrs) const {
   // LNT run (at least on Cyclone) showed reasonably significant gains for
   // bi-directional scheduling. 253.perlbmk.
   Policy.OnlyTopDown = false;
   Policy.OnlyBottomUp = false;
 }

test/CodeGen/ARM64/memset-to-bzero.ll

-; RUN: llc %s -march arm64 -o - | FileCheck %s
+; RUN: llc %s -mtriple=arm64-apple-darwin -o - | \
+; RUN:   FileCheck --check-prefix=CHECK-DARWIN --check-prefix=CHECK %s
+; RUN: llc %s -mtriple=arm64-linux-gnu -o - | \
+; RUN:   FileCheck --check-prefix=CHECK-LINUX --check-prefix=CHECK %s
 ; <rdar://problem/14199482> ARM64: Calls to bzero() replaced with calls to memset()
 
 ; CHECK: @fct1
 ; For small size (<= 256), we do not change memset to bzero.
 ; CHECK: memset
 define void @fct1(i8* nocapture %ptr) {
 entry:
   tail call void @llvm.memset.p0i8.i64(i8* %ptr, i8 0, i64 256, i32 1, i1 false)
   ret void
 }
 
 declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i32, i1)
 
 ; CHECK: @fct2
 ; When the size is bigger than 256, change into bzero.
-; CHECK: bzero
+; CHECK-DARWIN: bzero
+; CHECK-LINUX: memset
 define void @fct2(i8* nocapture %ptr) {
 entry:
   tail call void @llvm.memset.p0i8.i64(i8* %ptr, i8 0, i64 257, i32 1, i1 false)
   ret void
 }
 
 ; CHECK: @fct3
 ; For unknown size, change to bzero.
-; CHECK: bzero
+; CHECK-DARWIN: bzero
+; CHECK-LINUX: memset
 define void @fct3(i8* nocapture %ptr, i32 %unknown) {
 entry:
   %conv = sext i32 %unknown to i64
   tail call void @llvm.memset.p0i8.i64(i8* %ptr, i8 0, i64 %conv, i32 1, i1 false)
   ret void
 }
 
 ; CHECK: @fct4
 ; Size <= 256, no change.
 ; CHECK: memset
 define void @fct4(i8* %ptr) {
 entry:
   %tmp = tail call i64 @llvm.objectsize.i64(i8* %ptr, i1 false)
   %call = tail call i8* @__memset_chk(i8* %ptr, i32 0, i64 256, i64 %tmp)
   ret void
 }
 
 declare i8* @__memset_chk(i8*, i32, i64, i64)
 
 declare i64 @llvm.objectsize.i64(i8*, i1)
 
 ; CHECK: @fct5
 ; Size > 256, change.
-; CHECK: bzero
+; CHECK-DARWIN: bzero
+; CHECK-LINUX: memset
 define void @fct5(i8* %ptr) {
 entry:
   %tmp = tail call i64 @llvm.objectsize.i64(i8* %ptr, i1 false)
   %call = tail call i8* @__memset_chk(i8* %ptr, i32 0, i64 257, i64 %tmp)
   ret void
 }
 
 ; CHECK: @fct6
 ; Size = unknown, change.
-; CHECK: bzero
+; CHECK-DARWIN: bzero
+; CHECK-LINUX: memset
 define void @fct6(i8* %ptr, i32 %unknown) {
 entry:
   %conv = sext i32 %unknown to i64
   %tmp = tail call i64 @llvm.objectsize.i64(i8* %ptr, i1 false)
   %call = tail call i8* @__memset_chk(i8* %ptr, i32 0, i64 %conv, i64 %tmp)
   ret void
 }