[PATCH] D88783: [InstCombine] matchFunnelShift - fold or(shl(a,x),lshr(b,sub(bw,x))) -> fshl(a,b,x) iff x < bw

Fri Oct 9 08:48:34 PDT 2020

spatel added a comment.

In D88783#2321425 <https://reviews.llvm.org/D88783#2321425>, @lebedev.ri wrote:

> In D88783#2321414 <https://reviews.llvm.org/D88783#2321414>, @RKSimon wrote:
>
>> Just so I don't unnecessarily waste time on this - are your main/blocking concerns that backend legalization/lowering is still poor or is it more fundamental than that?
>
> I'm mainly interested in a better spelled-out motivation for the knownbits restriction.

Agree - the default backend expansion seems fine for most targets. If the riscv case is an outlier that can be fixed, then it's probably ok to proceed here (but we should file a bug to raise awareness).

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D88783/new/

https://reviews.llvm.org/D88783