From 72cce2104906dd3e3fc3820d930e7594ace3ee95 Mon Sep 17 00:00:00 2001
From: Hal Finkel <hfinkel@anl.gov>
Date: Tue, 31 Mar 2015 20:56:09 +0000
Subject: [PATCH] [PowerPC] Don't use a vector preferred memory type at -O0

Even at -O0, we fall back to SDAG when we hit intrinsics, and if the intrinsic
is a memset/memcpy/etc. we might normally use vector types. At -O0, this is
probably not a good idea (because, if there is a bug in the lowering code,
there would be no good way to turn it off). At -O0, only use scalar preferred
types.

Related to PR22754.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@233755 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/PowerPC/PPCISelLowering.cpp | 30 ++++++++++++++------------
 test/CodeGen/PowerPC/memset-nc.ll      |  9 ++++++++
 2 files changed, 25 insertions(+), 14 deletions(-)

diff --git a/lib/Target/PowerPC/PPCISelLowering.cpp b/lib/Target/PowerPC/PPCISelLowering.cpp
index 871531ece89..1c2659cbb70 100644
--- a/lib/Target/PowerPC/PPCISelLowering.cpp
+++ b/lib/Target/PowerPC/PPCISelLowering.cpp
@@ -11023,21 +11023,23 @@ EVT PPCTargetLowering::getOptimalMemOpType(uint64_t Size,
                                            bool IsMemset, bool ZeroMemset,
                                            bool MemcpyStrSrc,
                                            MachineFunction &MF) const {
-  const Function *F = MF.getFunction();
-  // When expanding a memset, require at least two QPX instructions to cover
-  // the cost of loading the value to be stored from the constant pool.
-  if (Subtarget.hasQPX() && Size >= 32 && (!IsMemset || Size >= 64) &&
-     (!SrcAlign || SrcAlign >= 32) && (!DstAlign || DstAlign >= 32) &&
-      !F->hasFnAttribute(Attribute::NoImplicitFloat)) {
-    return MVT::v4f64;
-  }
+  if (getTargetMachine().getOptLevel() != CodeGenOpt::None) {
+    const Function *F = MF.getFunction();
+    // When expanding a memset, require at least two QPX instructions to cover
+    // the cost of loading the value to be stored from the constant pool.
+    if (Subtarget.hasQPX() && Size >= 32 && (!IsMemset || Size >= 64) &&
+       (!SrcAlign || SrcAlign >= 32) && (!DstAlign || DstAlign >= 32) &&
+        !F->hasFnAttribute(Attribute::NoImplicitFloat)) {
+      return MVT::v4f64;
+    }
 
-  // We should use Altivec/VSX loads and stores when available. For unaligned
-  // addresses, unaligned VSX loads are only fast starting with the P8.
-  if (Subtarget.hasAltivec() && Size >= 16 &&
-      (((!SrcAlign || SrcAlign >= 16) && (!DstAlign || DstAlign >= 16)) ||
-       ((IsMemset && Subtarget.hasVSX()) || Subtarget.hasP8Vector())))
-    return MVT::v4i32;
+    // We should use Altivec/VSX loads and stores when available. For unaligned
+    // addresses, unaligned VSX loads are only fast starting with the P8.
+    if (Subtarget.hasAltivec() && Size >= 16 &&
+        (((!SrcAlign || SrcAlign >= 16) && (!DstAlign || DstAlign >= 16)) ||
+         ((IsMemset && Subtarget.hasVSX()) || Subtarget.hasP8Vector())))
+      return MVT::v4i32;
+  }
 
   if (Subtarget.isPPC64()) {
     return MVT::i64;
diff --git a/test/CodeGen/PowerPC/memset-nc.ll b/test/CodeGen/PowerPC/memset-nc.ll
index 46341be1c40..414a987a56f 100644
--- a/test/CodeGen/PowerPC/memset-nc.ll
+++ b/test/CodeGen/PowerPC/memset-nc.ll
@@ -1,4 +1,5 @@
 ; RUN: llc < %s | FileCheck %s
+; RUN: llc -O0 < %s | FileCheck %s -check-prefix=CHECK-O0
 target datalayout = "E-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-f128:128:128-v128:128:128-n32:64"
 target triple = "powerpc64-bgq-linux"
 
@@ -14,6 +15,10 @@ entry:
 ; CHECK: qvstfdx
 ; CHECK: qvstfdx
 ; CHECK: blr
+
+; CHECK-O0-LABEL: @test_qpx
+; CHECK-O0-NOT: qvstfdx
+; CHECK-O0: blr
 }
 
 ; Function Attrs: nounwind
@@ -31,6 +36,10 @@ entry:
 ; CHECK: stxvw4x
 ; CHECK: stxvw4x
 ; CHECK: blr
+
+; CHECK-O0-LABEL: @test_vsx
+; CHECK-O0-NOT: stxvw4x
+; CHECK-O0: blr
 }
 
 attributes #0 = { nounwind "target-cpu"="a2q" }