[PATCH] D86364: [ValueTracking] Interpret GEPs as a series of adds multiplied by the related scaling factor

Tue Oct 20 11:32:51 PDT 2020

spatel added a comment.

In D86364#2340420 <https://reviews.llvm.org/D86364#2340420>, @qcolombet wrote:

> Hi @nikic, @spatel, @lebedev.ri, @RKSimon, @efriedma,
>
> Following the conversation in D87342 <https://reviews.llvm.org/D87342> , this patch has been rebased and all the previous feedbacks have been addressed.
>
> We're left with the compile time issue that @nikic brought up.
>
> Any idea on how to tackle that?

I have not stepped through this in detail, so not sure if it helps:

1. It's been a couple of months, so we should confirm that the time impact is still the same.
2. It was suggested earlier to trigger the analysis from a different/dedicated pass rather than instcombine. Is that not feasible?
3. We're using the more expensive analysis on all GEPs. Could we use the existing analysis as the common case and limit the more expensive some way (for example, only GEPs with N or more operands, inbounds, etc)?

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D86364/new/

https://reviews.llvm.org/D86364