[PATCH] D34723: [AMDGPU] truncate left shift

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added inline comments.


================
Comment at: lib/Target/AMDGPU/AMDGPUISelLowering.cpp:2761-2762
 
+SDValue AMDGPUTargetLowering::performTruncCombine(SDNode *N,
+                                                  DAGCombinerInfo &DCI) const {
+  EVT VT = N->getValueType(0);
----------------
rampitec wrote:
> rampitec wrote:
> > arsenm wrote:
> > > This exact combine is already done in DAGCombiner:
> > > 
> > >   // trunc (shl x, K) -> shl (trunc x), K => K < VT.getScalarSizeInBits()
> > > 
> > > Why isn't it triggering? Is the other combine you added missing an AddToWorklist or something?
> > Ah. That one only works with constants.
> Do you want me to transfer it there?
Yes, it would be better if it was all in one place


https://reviews.llvm.org/D34723