CI: Add a run of FastRNN benchmarks in default executor/fuser configuration.

[ZolotukhinM]

Stack from ghstack:

• Benchmarks: tweak PE config settings. #45349 Benchmarks: tweak PE config settings.
• CI: Add a run of FastRNN benchmarks in default executor/fuser configuration. #45348 CI: Add a run of FastRNN benchmarks in default executor/fuser configuration.
• Benchmarks: add 'default' options for fuser and executor. #45347 Benchmarks: add 'default' options for fuser and executor.

Differential Revision: D23935520

[Krovatkin]

[wconstab]

i'm fine with this, especially in the spirit of moving fast for PE work, but i also wonder if this approach (of running multiple configs on the same commit) makes sense long term, vs having the approach that each commit we measure in CI has a particular setting?

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (gh/ZolotukhinM/349/base@59014b4). Click here to learn what that means.
The diff coverage is n/a.

@@                    Coverage Diff                     @@
##             gh/ZolotukhinM/349/base   #45348   +/-   ##
==========================================================
  Coverage                           ?   68.06%           
==========================================================
  Files                              ?      393           
  Lines                              ?    50987           
  Branches                           ?        0           
==========================================================
  Hits                               ?    34706           
  Misses                             ?    16281           
  Partials                           ?        0           

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 59014b4...68c3a44. Read the comment docs.

[ZolotukhinM]

i'm fine with this, especially in the spirit of moving fast for PE work, but i also wonder if this approach (of running multiple configs on the same commit) makes sense long term, vs having the approach that each commit we measure in CI has a particular setting?

I have a similar concern but slightly swaying towards checking potentially redundant tests. Imho it's better than accidentally not testing what's important. In the long run we hopefully would remove the LE/OF configuration, and keep only the default, which would be PE/TE.

[facebook-github-bot]

@ZolotukhinM merged this pull request in a07d829.