From 818ff34bc0841edda10951b7b043076b6e7159ef Mon Sep 17 00:00:00 2001
From: Chris Lattner <sabre@nondot.org>
Date: Sat, 23 Jan 2010 18:49:30 +0000
Subject: [PATCH] implement a simple instcombine xform that has been in the
 readme forever.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@94318 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Target/README.txt                         | 20 ++----------
 .../InstCombine/InstCombineShifts.cpp         | 32 ++++++++++++++++---
 test/Transforms/InstCombine/intrinsics.ll     | 15 +++++++++
 3 files changed, 44 insertions(+), 23 deletions(-)

diff --git a/lib/Target/README.txt b/lib/Target/README.txt
index 6c6290a70a1..c68ae76a257 100644
--- a/lib/Target/README.txt
+++ b/lib/Target/README.txt
@@ -237,24 +237,6 @@ define void @test(i32* %P) {
 
 //===---------------------------------------------------------------------===//
 
-dag/inst combine "clz(x)>>5 -> x==0" for 32-bit x.
-
-Compile:
-
-int bar(int x)
-{
-  int t = __builtin_clz(x);
-  return -(t>>5);
-}
-
-to:
-
-_bar:   addic r3,r3,-1
-        subfe r3,r3,r3
-        blr
-
-//===---------------------------------------------------------------------===//
-
 quantum_sigma_x in 462.libquantum contains the following loop:
 
       for(i=0; i<reg->size; i++)
@@ -294,6 +276,8 @@ unsigned long reverse(unsigned v) {
 
 //===---------------------------------------------------------------------===//
 
+[LOOP RECOGNITION]
+
 These idioms should be recognized as popcount (see PR1488):
 
 unsigned countbits_slow(unsigned v) {
diff --git a/lib/Transforms/InstCombine/InstCombineShifts.cpp b/lib/Transforms/InstCombine/InstCombineShifts.cpp
index 321c91d57d2..9dd60dd4a25 100644
--- a/lib/Transforms/InstCombine/InstCombineShifts.cpp
+++ b/lib/Transforms/InstCombine/InstCombineShifts.cpp
@@ -12,6 +12,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "InstCombine.h"
+#include "llvm/IntrinsicInst.h"
 #include "llvm/Support/PatternMatch.h"
 using namespace llvm;
 using namespace PatternMatch;
@@ -69,10 +70,9 @@ Instruction *InstCombiner::FoldShiftByConstant(Value *Op0, ConstantInt *Op1,
   if (Op1->uge(TypeBits)) {
     if (I.getOpcode() != Instruction::AShr)
       return ReplaceInstUsesWith(I, Constant::getNullValue(Op0->getType()));
-    else {
-      I.setOperand(1, ConstantInt::get(I.getType(), TypeBits-1));
-      return &I;
-    }
+    // ashr i32 X, 32 --> ashr i32 X, 31
+    I.setOperand(1, ConstantInt::get(I.getType(), TypeBits-1));
+    return &I;
   }
   
   // ((X*C1) << C2) == (X * (C1 << C2))
@@ -387,7 +387,29 @@ Instruction *InstCombiner::visitShl(BinaryOperator &I) {
 }
 
 Instruction *InstCombiner::visitLShr(BinaryOperator &I) {
-  return commonShiftTransforms(I);
+  if (Instruction *R = commonShiftTransforms(I))
+    return R;
+  
+  Value *Op0 = I.getOperand(0), *Op1 = I.getOperand(1);
+  
+  if (ConstantInt *Op1C = dyn_cast<ConstantInt>(Op1))
+    if (IntrinsicInst *II = dyn_cast<IntrinsicInst>(Op0)) {
+      // ctlz.i32(x)>>5  --> zext(x == 0)
+      // cttz.i32(x)>>5  --> zext(x == 0)
+      // ctpop.i32(x)>>5 --> zext(x == -1)
+      if ((II->getIntrinsicID() == Intrinsic::ctlz ||
+           II->getIntrinsicID() == Intrinsic::cttz ||
+           II->getIntrinsicID() == Intrinsic::ctpop) &&
+          (1ULL << Op1C->getZExtValue()) ==
+            Op0->getType()->getScalarSizeInBits()) {
+        bool isCtPop = II->getIntrinsicID() == Intrinsic::ctpop;
+        Constant *RHS = ConstantInt::getSigned(Op0->getType(), isCtPop ? -1 : 0);
+        Value *Cmp = Builder->CreateICmpEQ(II->getOperand(1), RHS);
+        return new ZExtInst(Cmp, II->getType());
+      }
+    }
+  
+  return 0;
 }
 
 Instruction *InstCombiner::visitAShr(BinaryOperator &I) {
diff --git a/test/Transforms/InstCombine/intrinsics.ll b/test/Transforms/InstCombine/intrinsics.ll
index c63475c2bd7..08dcfa731a9 100644
--- a/test/Transforms/InstCombine/intrinsics.ll
+++ b/test/Transforms/InstCombine/intrinsics.ll
@@ -144,3 +144,18 @@ entry:
 ; CHECK-NEXT: %pop.cmp = icmp eq i32 %b, 0
 ; CHECK-NEXT: volatile store i1 %pop.cmp, i1* %c
 }
+
+
+define i32 @cttz_simplify1(i32 %x) nounwind readnone ssp {
+  %tmp1 = tail call i32 @llvm.ctlz.i32(i32 %x)    ; <i32> [#uses=1]
+  %shr3 = lshr i32 %tmp1, 5                       ; <i32> [#uses=1]
+  ret i32 %shr3
+  
+; CHECK: @cttz_simplify1
+; CHECK: icmp eq i32 %x, 0
+; CHECK-NEXT: zext i1 
+; CHECK-NEXT: ret i32
+}
+
+declare i32 @llvm.ctlz.i32(i32) nounwind readnone
+