rmuir commented on PR #12311:
URL: https://github.com/apache/lucene/pull/12311#issuecomment-1559035945
couple of random TODOs i had to look into, we can think about:
* disable panama provider unless user has at least 128-bit vectors? Nobody
has benched the case where SPECIES_PREFERRED is 64-bits: that's the case where
vectorization isnt enabled at all for some reason (VM, unsupported platform). I
am suspicious it will do any good at all for these algorithms. we can try to
test it with QEMU if we really care.
* detect C1-only ("client mode") and don't enable panama provider then
either. We already have this issue in our own test suite. This one is even
worse because java side "thinks it has vectors" but compiler does something
excruciating slow.
* try to organize the benchmarks in a better way, than being at
github.com/rmuir? I don't really like touching this code without verifying
that. In my experience before, its very easy to change one small thing and make
it all go very slow (e.g. stuff starts getting bounds-checked or something)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
