[llvm-commits] [llvm] r75845 - in /llvm/trunk: lib/Transforms/Scalar/InstructionCombining.cpp test/Transforms/InstCombine/fold-bin-operand.ll

Eli Friedman eli.friedman at gmail.com
Wed Jul 15 15:13:49 PDT 2009


Author: efriedma
Date: Wed Jul 15 17:13:34 2009
New Revision: 75845

URL: http://llvm.org/viewvc/llvm-project?rev=75845&view=rev
Log:
Don't restrict the set of instructions where we try to constant-fold the 
operands; it's possible to end up with a constant-foldable operand to 
most instructions, even those which can't trap.


Added:
    llvm/trunk/test/Transforms/InstCombine/fold-bin-operand.ll
Modified:
    llvm/trunk/lib/Transforms/Scalar/InstructionCombining.cpp

Modified: llvm/trunk/lib/Transforms/Scalar/InstructionCombining.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Scalar/InstructionCombining.cpp?rev=75845&r1=75844&r2=75845&view=diff

==============================================================================
--- llvm/trunk/lib/Transforms/Scalar/InstructionCombining.cpp (original)
+++ llvm/trunk/lib/Transforms/Scalar/InstructionCombining.cpp Wed Jul 15 17:13:34 2009
@@ -13021,9 +13021,7 @@
       continue;
     }
 
-    if (TD &&
-        (I->getType()->getTypeID() == Type::VoidTyID ||
-         I->isTrapping())) {
+    if (TD) {
       // See if we can constant fold its operands.
       for (User::op_iterator i = I->op_begin(), e = I->op_end(); i != e; ++i)
         if (ConstantExpr *CE = dyn_cast<ConstantExpr>(i))

Added: llvm/trunk/test/Transforms/InstCombine/fold-bin-operand.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/InstCombine/fold-bin-operand.ll?rev=75845&view=auto

==============================================================================
--- llvm/trunk/test/Transforms/InstCombine/fold-bin-operand.ll (added)
+++ llvm/trunk/test/Transforms/InstCombine/fold-bin-operand.ll Wed Jul 15 17:13:34 2009
@@ -0,0 +1,13 @@
+; RUN: llvm-as < %s | opt -instcombine | llvm-dis | not grep icmp
+
+define i1 @f(i1 %x) {
+	%b = and i1 %x, icmp eq (i8* inttoptr (i32 1 to i8*), i8* inttoptr (i32 2 to i8*))
+	ret i1 %b
+}
+
+; FIXME: This doesn't fold at the moment!
+; define i32 @f(i32 %x) {
+;	%b = add i32 %x, zext (i1 icmp eq (i8* inttoptr (i32 1000000 to i8*), i8* inttoptr (i32 2000000 to i8*)) to i32)
+;	ret i32 %b
+;}
+





More information about the llvm-commits mailing list