[PATCH] D140436: [NVPTX] Add support of ptxas v12.0 to the LIT tests

Mon Jan 9 16:59:23 PST 2023

pavelkopyl added a comment.

In D140436#4026879 <https://reviews.llvm.org/D140436#4026879>, @tra wrote:

> Description could use some details. I.e. what makes sm_35 special?  I assume it's due to the fact that sm_35 is no longer supported.
>
> Instead of special-casing ptxas-12.0, perhpas a better way is to just always run ptxas with `-arch=sm_60` so it would work with ptxas from all reasonably new CUDA versions.

Thank you. I've added details to the description.

As for choosing to an actual ptxas version, yes we could do this.
But original motivation to use minimal possible version of architecture/PTX was to catch errors related to appearance of new instructions and adding appropriate checks to the codegen. For example, https://reviews.llvm.org/D123040
On the other hand, all such cases seems to have been detected and fixed for old architectures, so we could focus on current ones.

================
Comment at: llvm/test/lit.cfg.py:215
             (10, 0), (10, 1), (10, 2),
             (11, 0), (11, 1), (11, 2), (11, 3), (11, 4), (11, 5), (11, 6),
+            (12, 0),
----------------
tra wrote:
> There are also 11.7 and 11.8.
Done, thank you.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D140436/new/

https://reviews.llvm.org/D140436