[PATCH] D13076: [InstCombine] transform masking off of an FP sign bit into a fabs() intrinsic call (PR24886)

[Sanjay Patel via llvm-commits]

spatel created this revision.
spatel added reviewers: hfinkel, ab, sanjoy.
spatel added a subscriber: llvm-commits.

This is a partial fix for PR24886:
https://llvm.org/bugs/show_bug.cgi?id=24886

Without this IR transform, the backend (x86 at least) was producing inefficient code. 

This patch is making 2 assumptions:
1. The canonical form of a fabs() operation is, in fact, the LLVM fabs() intrinsic.
2. The high bit of an FP value is always the sign bit; as noted in the bug report, this isn't specified by the LangRef.

http://reviews.llvm.org/D13076

Files:
  lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
  test/Transforms/InstCombine/and2.ll

Index: test/Transforms/InstCombine/and2.ll
===================================================================
--- test/Transforms/InstCombine/and2.ll
+++ test/Transforms/InstCombine/and2.ll
@@ -102,3 +102,34 @@
   %add = add i64 %sub, %and
   ret i64 %add
 }
+
+define i64 @fabs_double(double %x) {
+; CHECK-LABEL: @fabs_double(
+; CHECK-NEXT: call double @llvm.fabs.f64(double %x) 
+; CHECK-NEXT: bitcast
+; CHECK-NEXT: ret
+  %bc = bitcast double %x to i64
+  %and = and i64 %bc, 9223372036854775807
+  ret i64 %and
+}
+
+define i64 @fabs_double_swap(double %x) {
+; CHECK-LABEL: @fabs_double_swap(
+; CHECK-NEXT: call double @llvm.fabs.f64(double %x) 
+; CHECK-NEXT: bitcast
+; CHECK-NEXT: ret
+  %bc = bitcast double %x to i64
+  %and = and i64 9223372036854775807, %bc
+  ret i64 %and
+}
+
+define i32 @fabs_float(float %x) {
+; CHECK-LABEL: @fabs_float(
+; CHECK-NEXT: call float @llvm.fabs.f32(float %x) 
+; CHECK-NEXT: bitcast
+; CHECK-NEXT: ret
+  %bc = bitcast float %x to i32
+  %and = and i32 %bc, 2147483647
+  ret i32 %and
+}
+
Index: lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
===================================================================
--- lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
+++ lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
@@ -1469,13 +1469,14 @@
 
 
   // fold (and (cast A), (cast B)) -> (cast (and A, B))
-  if (CastInst *Op0C = dyn_cast<CastInst>(Op0))
+  if (CastInst *Op0C = dyn_cast<CastInst>(Op0)) {
+    Value *Op0COp = Op0C->getOperand(0);
+    Type *SrcTy = Op0COp->getType();
     if (CastInst *Op1C = dyn_cast<CastInst>(Op1)) {
-      Type *SrcTy = Op0C->getOperand(0)->getType();
       if (Op0C->getOpcode() == Op1C->getOpcode() && // same cast kind ?
           SrcTy == Op1C->getOperand(0)->getType() &&
           SrcTy->isIntOrIntVectorTy()) {
-        Value *Op0COp = Op0C->getOperand(0), *Op1COp = Op1C->getOperand(0);
+        Value *Op1COp = Op1C->getOperand(0);
 
         // Only do this if the casts both really cause code to be generated.
         if (ShouldOptimizeCast(Op0C->getOpcode(), Op0COp, I.getType()) &&
@@ -1500,6 +1501,20 @@
       }
     }
 
+    // If we are masking off the sign bit of a floating-point value, convert
+    // this to the canonical fabs intrinsic call and cast back to integer.
+    // The backend should know how to optimize fabs().
+    // TODO: This transform should also apply to vectors.
+    ConstantInt *CI;
+    if (SrcTy->isFloatingPointTy() && match(Op1, m_ConstantInt(CI)) &&
+        CI->isMaxValue(true)) {
+      Module *M = I.getParent()->getParent()->getParent();
+      Function *Fabs = Intrinsic::getDeclaration(M, Intrinsic::fabs, SrcTy);
+      Value *Call = Builder->CreateCall(Fabs, Op0COp, "fabs");
+      return CastInst::CreateBitOrPointerCast(Call, I.getType());
+    }
+  }
+
   {
     Value *X = nullptr;
     bool OpsSwapped = false;


-------------- next part --------------
A non-text attachment was scrubbed...
Name: D13076.35429.patch
Type: text/x-patch
Size: 2877 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20150922/70967d96/attachment.bin>