Date: Mon, 07 Oct 2024 09:00:58 +0000 From: bugzilla-noreply@freebsd.org To: toolchain@FreeBSD.org Subject: [Bug 281765] llvm - ThinLTO never uses more than a single thread by default Message-ID: <bug-281765-29464-1lo0VDG0vB@https.bugs.freebsd.org/bugzilla/> In-Reply-To: <bug-281765-29464@https.bugs.freebsd.org/bugzilla/> References: <bug-281765-29464@https.bugs.freebsd.org/bugzilla/>
next in thread | previous in thread | raw e-mail | index | archive | help
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=3D281765 David Chisnall <theraven@FreeBSD.org> changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |theraven@FreeBSD.org --- Comment #3 from David Chisnall <theraven@FreeBSD.org> --- (In reply to Dimitry Andric from comment #2) > The reason for this was that lld was using way too many threads on many-c= ore machines: nowadays just creating "hw.ncpu" threads is no longer a good = default. I don't believe this is true (unless RAM is not properly scaled with core count). The problem is build systems that try to parallelise link jobs.=20 Building LLVM with PARALLEL_LINK_JOBS=3D1, the upstream default remains the= right one. Building FreeBSD, it is not.=20=20 The fix for this in the bmake system is probably to wrap lld in a shell scr= ipt that acquires a lock, invokes lld, and then releases the lock and exits with lld's exit code. --=20 You are receiving this mail because: You are the assignee for the bug.=
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?bug-281765-29464-1lo0VDG0vB>