[Bug 281765] llvm - ThinLTO never uses more than a single thread by default

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=281765

--- Comment #12 from Mark Millard <marklmi26-fbsd@yahoo.com> ---
(In reply to Daniel Engberg from comment #11)

For devel/llvm-devel there was:

QUOTE
author  Brooks Davis <brooks@FreeBSD.org>       2024-11-14 22:35:53 +0000
committer       Brooks Davis <brooks@FreeBSD.org>       2024-11-14 22:35:53
+0000

commit  daa99286ef8104d1df6c22e35d02b081cab2ed0f (patch)
tree    c0fb26926fe04858909faabb85fcf708882fc123 /devel/llvm-devel
parent  6a2b178e527439794d5efba5bcf01561f4073121 (diff)

download        ports-daa99286ef8104d1df6c22e35d02b081cab2ed0f.tar.gz
ports-daa99286ef8104d1df6c22e35d02b081cab2ed0f.zip

devel/llvm-devel: new snapshot, 32-bit improvements

Apply a patch based on one posted by Mark Millard in the comments of
https://reviews.freebsd.org/D46239 which defaults lld to using a single
thread on 32-bit systems.
END QUOTE

This does not address the parallelism issue for 64 bit(+)
processes but keeps the 32-bit contexts from getting in
the way of having such parallelism for 64+ bits, at least
for devel/llvm-devel for now.

I do not know if any other devel/llvm* will get such or if
it might be upstreamed. With such in place to deal with
avoiding process size limitation issues, RAM+SWAP can be
part of what allows managing whatever parallelism rule is
picked. (There will still be folks building on small
aarch64 boards with only 4 cores.)

-- 
You are receiving this mail because:
You are the assignee for the bug.