PATCH: SelectionDAG: Use MVT::i32 for vector indices

Tom Stellard tom at stellard.net
Mon Aug 5 15:43:29 PDT 2013


On Mon, Aug 05, 2013 at 12:19:44PM -0700, Owen Anderson wrote:
> 
> On Aug 5, 2013, at 11:36 AM, Tom Stellard <tom at stellard.net> wrote:
> 
> > On Mon, Aug 05, 2013 at 10:26:05AM -0700, Tom Stellard wrote:
> >> On Mon, Aug 05, 2013 at 10:18:37AM -0700, Owen Anderson wrote:
> >>> 
> >>> On Aug 5, 2013, at 9:51 AM, Tom Stellard <tom at stellard.net> wrote:
> >>> 
> >>>> On Thu, Aug 01, 2013 at 07:34:30AM -0700, Tom Stellard wrote:
> >>>>> On Wed, Jul 31, 2013 at 10:09:03PM -0700, Owen Anderson wrote:
> >>>>>> 
> >>>>>> On Jul 31, 2013, at 7:25 PM, Tom Stellard <tom at stellard.net> wrote:
> >>>>>> 
> >>>>>>>> 
> >>>>>>>> The first version of this patch that I wrote:
> >>>>>>>> http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20130715/181799.html
> >>>>>>>> added TargetLowering::getVectorIdxTy() which targets could use to specify
> >>>>>>>> whatever type the wanted.  What do you think of using the version one patch
> >>>>>>>> instead?
> >>>>>>>> 
> >>>>>>> 
> >>>>>>> Ping.
> >>>>>> 
> >>>>>> FWIW, I prefer the approach using a target-specified type.
> >>>>>> 
> >>>>> 
> >>>>> Thanks Owen,
> >>>>> 
> >>>>> Do you any comments on my original patch?
> >>>>> 
> >>>>> http://lists.cs.uiuc.edu/pipermail/llvm-commits/Week-of-Mon-20130715/181799.html
> >>>>> 
> >>> 
> >>> 
> >>>> -  SDValue Idx = ZExtPromotedInteger(N->getOperand(2));
> >>>> +  SDValue Idx = DAG.getZExtOrTrunc(N->getOperand(2), SDLoc(N), MVT::i32);
> >>> Should this be a parameterized type rather than i32?
> >>> 
> >> 
> >> Yes it should, I will fix this.
> >> 
> > 
> > Hi Owen,
> > 
> > Here is an updated patch, which fixes the problem you pointed out plus
> > fixes the index size for a few more EXTRACT_VECTOR_ELT creations which
> > were added since I first wrote this patch.
> > 
> > -Tom
> > <0001-TargetLowering-Add-getVectorIdxTy-function-v2.patch>
> 
> LGTM.
> 

Thanks, Owen. Committed as r187748.

-Tom



More information about the llvm-commits mailing list