Teach InstCombine to hoist FABS and FNEG through FPTRUNC instructions…

…. The application of these operations commutes with the truncation, so we should prefer to do them in the smallest size we can, to save register space, use smaller constant pool entries, etc. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@172117 91177308-0d34-0410-b5e6-96231b3b80d8
Boscop · Jan 10, 2013 · e9d4eba · e9d4eba
1 parent 244b7a4
commit e9d4eba
Show file tree

Hide file tree

Showing 2 changed files with 46 additions and 1 deletion.
diff --git a/lib/Transforms/InstCombine/InstCombineCasts.cpp b/lib/Transforms/InstCombine/InstCombineCasts.cpp
@@ -1204,8 +1204,34 @@ Instruction *InstCombiner::visitFPTrunc(FPTruncInst &CI) {
       }
       break;  
     }
+
+    // (fptrunc (fneg x)) -> (fneg (fptrunc x))
+    if (BinaryOperator::isFNeg(OpI)) {
+      Value *InnerTrunc = Builder->CreateFPTrunc(OpI->getOperand(1),
+                                                 CI.getType());
+      return BinaryOperator::CreateFNeg(InnerTrunc);
+    }
   }
-
+
+  IntrinsicInst *II = dyn_cast<IntrinsicInst>(CI.getOperand(0));
+  if (II) {
+    switch (II->getIntrinsicID()) {
+      default: break;
+      case Intrinsic::fabs: {
+        // (fptrunc (fabs x)) -> (fabs (fptrunc x))
+        Value *InnerTrunc = Builder->CreateFPTrunc(II->getArgOperand(0),
+                                                   CI.getType());
+        Type *IntrinsicType[] = { CI.getType() };
+        Function *Overload =
+          Intrinsic::getDeclaration(CI.getParent()->getParent()->getParent(),
+                                    II->getIntrinsicID(), IntrinsicType);
+
+        Value *Args[] = { InnerTrunc };
+        return CallInst::Create(Overload, Args, II->getName());
+      }
+    }
+  }
+
   // Fold (fptrunc (sqrt (fpext x))) -> (sqrtf x)
   CallInst *Call = dyn_cast<CallInst>(CI.getOperand(0));
   if (Call && Call->getCalledFunction() && TLI->has(LibFunc::sqrtf) &&

diff --git a/test/Transforms/InstCombine/fpcast.ll b/test/Transforms/InstCombine/fpcast.ll
@@ -13,3 +13,22 @@ define i8 @test2() {
 ; CHECK: ret i8 -1
 }
 
+; CHECK: test3
+define half @test3(float %a) {
+; CHECK: fptrunc
+; CHECK: llvm.fabs.f16
+  %b = call float @llvm.fabs.f32(float %a)
+  %c = fptrunc float %b to half
+  ret half %c
+}
+
+; CHECK: test4
+define half @test4(float %a) {
+; CHECK: fptrunc
+; CHECK: fsub
+  %b = fsub float -0.0, %a
+  %c = fptrunc float %b to half
+  ret half %c
+}
+
+declare float @llvm.fabs.f32(float) nounwind readonly