[PATCH] D107225: [Polly][Isl] Move to the new-polly-generator branch version of isl-noexceptions.h. NFCI

Thu Aug 12 04:17:18 PDT 2021

Meinersbur added a comment.

In D107225#2940080 <https://reviews.llvm.org/D107225#2940080>, @patacca wrote:

> In D107225#2939965 <https://reviews.llvm.org/D107225#2939965>, @Meinersbur wrote:
>
>> I don't see when you are using `isl::size::release()` and when `IslAssertNoError`. We talked about a follow-up patch for the assertions, but what is the concept for this patch?
>> Is it somehow possible to void the `static_cast<unsigned>` from `isl::size`? Part of the solution might be to make `islAssertNoError` return the unsigned result, as we talked about yesterday.
>
> My idea was to rename and maybe reimplement the `IslAssertNoError` in a different patch because this one is getting pretty big and has already a lot of changes.
> I know that in the last meeting we agreed to include both `IslAssertNoError` and `isl::size::release` in this patch but I am now thinking of landing the patch withouth the `IslAssertNoError` and opening a new one in which the assertion and the casting to `unsigned` is better handled. Is it ok?

Yes, that would be great.

> I've run the llvm-test-suite and I also got a single failing test case: `MultiSource/Applications/lambda-0.1.3/lambda.test`
> I am still unsure if it is related to this patch or not but I am investigating

It is not related to this patch, it failed before as well. See http://meinersbur.de:8011/#/builders/76/builds/337
(I added `-mllvm -polly-process-unprofitable` to the builder, then the failures showed up; It wasn't testing `-mllvm -polly-process-unprofitable` before that).

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D107225/new/

https://reviews.llvm.org/D107225