From 5518a1355b8b09bf92419b65ea4e4854734b0ebc Mon Sep 17 00:00:00 2001
From: Shuxin Yang <shuxin.llvm@gmail.com>
Date: Sun, 9 Dec 2012 03:12:46 +0000
Subject: - Re-enable population count loop idiom recognization - fix a bug
 which cause sigfault. - add two testing cases which was causing crash

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@169687 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/X86/X86ISelLowering.cpp | 11 +++++++++++
 lib/Target/X86/X86ISelLowering.h   |  8 ++++++++
 lib/Target/X86/X86TargetMachine.h  |  2 +-
 3 files changed, 20 insertions(+), 1 deletion(-)

(limited to 'lib/Target')

diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index af772ed..84e5677 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -17679,6 +17679,17 @@ FindInConvertTable(const X86TypeConversionCostTblEntry *Tbl, unsigned len,
   return -1;
 }
 
+ScalarTargetTransformInfo::PopcntHwSupport
+X86ScalarTargetTransformImpl::getPopcntHwSupport(unsigned TyWidth) const {
+  assert(isPowerOf2_32(TyWidth) && "Ty width must be power of 2");
+  const X86Subtarget &ST = TLI->getTargetMachine().getSubtarget<X86Subtarget>();
+
+  // TODO: Currently the __builtin_popcount() implementation using SSE3
+  //   instructions is inefficient. Once the problem is fixed, we should
+  //   call ST.hasSSE3() instead of ST.hasSSE4().
+  return ST.hasSSE41() ? Fast : None;
+}
+
 unsigned
 X86VectorTargetTransformInfo::getArithmeticInstrCost(unsigned Opcode,
                                                      Type *Ty) const {
diff --git a/lib/Target/X86/X86ISelLowering.h b/lib/Target/X86/X86ISelLowering.h
index 1fdac34..1042fe1 100644
--- a/lib/Target/X86/X86ISelLowering.h
+++ b/lib/Target/X86/X86ISelLowering.h
@@ -932,6 +932,14 @@ namespace llvm {
                              const TargetLibraryInfo *libInfo);
   }
 
+  class X86ScalarTargetTransformImpl : public ScalarTargetTransformImpl {
+  public:
+    explicit X86ScalarTargetTransformImpl(const TargetLowering *TL) :
+      ScalarTargetTransformImpl(TL) {};
+
+    virtual PopcntHwSupport getPopcntHwSupport(unsigned TyWidth) const;
+  };
+
   class X86VectorTargetTransformInfo : public VectorTargetTransformImpl {
   public:
     explicit X86VectorTargetTransformInfo(const TargetLowering *TL) :
diff --git a/lib/Target/X86/X86TargetMachine.h b/lib/Target/X86/X86TargetMachine.h
index a3de042..792f721 100644
--- a/lib/Target/X86/X86TargetMachine.h
+++ b/lib/Target/X86/X86TargetMachine.h
@@ -118,7 +118,7 @@ class X86_64TargetMachine : public X86TargetMachine {
   X86SelectionDAGInfo TSInfo;
   X86TargetLowering TLInfo;
   X86JITInfo        JITInfo;
-  ScalarTargetTransformImpl STTI;
+  X86ScalarTargetTransformImpl STTI;
   X86VectorTargetTransformInfo VTTI;
 public:
   X86_64TargetMachine(const Target &T, StringRef TT,
-- 
cgit v1.1