This is the mail archive of the
glibc-bugs@sourceware.org
mailing list for the glibc project.
[Bug libc/24268] bench-ffsll performance degradation on glibc 2.29
- From: "fweimer at redhat dot com" <sourceware-bugzilla at sourceware dot org>
- To: glibc-bugs at sourceware dot org
- Date: Mon, 04 Mar 2019 07:50:07 +0000
- Subject: [Bug libc/24268] bench-ffsll performance degradation on glibc 2.29
- Auto-submitted: auto-generated
- References: <bug-24268-131@http.sourceware.org/bugzilla/>
https://sourceware.org/bugzilla/show_bug.cgi?id=24268
Florian Weimer <fweimer at redhat dot com> changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|WAITING |RESOLVED
Resolution|--- |INVALID
--- Comment #4 from Florian Weimer <fweimer at redhat dot com> ---
Thanks. So there is really nothing left to fix here. Increasing function
alignment in general seems rather wasteful. I suspect this is rather a cache
line conflict or something like this, and the alignment just makes the code use
a different cache line.
We could add an IFUNC that eliminates the conditional branch on modern CPUs
with BMI2. But that would be basically a defeat device because no real binaries
will call ffsll, only benchmarks, because GCC will expand ffsll calls inline.
--
You are receiving this mail because:
You are on the CC list for the bug.