[llvm] r175853 - Fix isa<> check which could never be true.

Pete Cooper peter_cooper at apple.com
Thu Feb 21 17:50:38 PST 2013


Author: pete
Date: Thu Feb 21 19:50:38 2013
New Revision: 175853

URL: http://llvm.org/viewvc/llvm-project?rev=175853&view=rev
Log:
Fix isa<> check which could never be true.

It was incorrectly checking a Function* being an IntrinsicInst* which
isn't possible.  It should always have been checking the CallInst* instead.

Added test case for x86 which ensures we only get one constant load.
It was 2 before this change.

rdar://problem/13267920

Added:
    llvm/trunk/test/CodeGen/X86/fast-isel-constant.ll
Modified:
    llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp

Modified: llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp?rev=175853&r1=175852&r2=175853&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp Thu Feb 21 19:50:38 2013
@@ -705,7 +705,7 @@ bool FastISel::SelectCall(const User *I)
   // all the values which have already been materialized,
   // appear after the call. It also makes sense to skip intrinsics
   // since they tend to be inlined.
-  if (!isa<IntrinsicInst>(F))
+  if (!isa<IntrinsicInst>(Call))
     flushLocalValueMap();
 
   // An arbitrary call. Bail.

Added: llvm/trunk/test/CodeGen/X86/fast-isel-constant.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/X86/fast-isel-constant.ll?rev=175853&view=auto
==============================================================================
--- llvm/trunk/test/CodeGen/X86/fast-isel-constant.ll (added)
+++ llvm/trunk/test/CodeGen/X86/fast-isel-constant.ll Thu Feb 21 19:50:38 2013
@@ -0,0 +1,32 @@
+; RUN: llc < %s -march=x86-64 -O0 | FileCheck %s
+; Make sure fast-isel doesn't reset the materialised constant map
+; across an intrinsic call.
+
+; CHECK: movl	$100000
+; CHECK-NOT: movl	$100000
+define i1 @test1(i32 %v1, i32 %v2, i32* %X) nounwind {
+entry:
+  %a = shl i32 100000, %v1
+  %t = call {i32, i1} @llvm.sadd.with.overflow.i32(i32 %a, i32 %v2)
+  %ext = extractvalue {i32, i1} %t, 0
+  %sum = shl i32 100000, %ext
+  %obit = extractvalue {i32, i1} %t, 1
+  br i1 %obit, label %overflow, label %normal
+
+normal:
+  store i32 %sum, i32* %X
+  br label %overflow
+
+overflow:
+  ret i1 false
+  
+; X32: test4:
+; X32: addl
+; X32-NEXT: jo
+
+; X64:        test4:
+; X64:          addl	%e[[A1:si|dx]], %e[[A0:di|cx]]
+; X64-NEXT:	jo
+}
+
+declare {i32, i1} @llvm.sadd.with.overflow.i32(i32, i32)





More information about the llvm-commits mailing list