[llvm-commits] [llvm] r55982 - /llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp

[Evan Cheng]

Actually my patch isn't quite the same. :-) Why is the check only  
needed for ConstantInt? Isn't it needed for all cases. Also, why do  
you return "false" when the function is supposed to return an unsigned?

I think the check should be here:
   MVT::SimpleValueType VT = TLI.getValueType(V- 
 >getType()).getSimpleVT();
   if (!TLI.isTypeLegal(VT))
     return 0;

What do you think?

Evan

On Sep 8, 2008, at 11:32 PM, Owen Anderson wrote:

> Author: resistor
> Date: Tue Sep  9 01:32:02 2008
> New Revision: 55982
>
> URL: http://llvm.org/viewvc/llvm-project?rev=55982&view=rev
> Log:
> Check for type legality before materializing integer constants in  
> fast isel.  With this change,
> all of MultiSource/Applications passes on Darwin/X86 under FastISel.
>
> 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=55982&r1=55981&r2=55982&view=diff
>
> = 
> = 
> = 
> = 
> = 
> = 
> = 
> = 
> ======================================================================
> --- llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp (original)
> +++ llvm/trunk/lib/CodeGen/SelectionDAG/FastISel.cpp Tue Sep  9  
> 01:32:02 2008
> @@ -39,6 +39,8 @@
>                                        MBB->getParent()- 
> >getConstantPool());
>     // Don't cache constant materializations.  To do so would require
>     // tracking what uses they dominate.
> +    if (!TLI.isTypeLegal(VT))
> +      return false;
>     Reg = FastEmit_i(VT, VT, ISD::Constant, CI->getZExtValue());
>   } else if (isa<GlobalValue>(V)) {
>     return TargetMaterializeConstant(dyn_cast<Constant>(V),
>
>
> _______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits