Public bug reported: [Description] Update glibc to include this patch for checking lock availability before acquisition to reduce cache line bouncing. This significantly improves trylock throughput on multi-core systems under heavy contention.
This has been accepted upstream glibc-2.43+ and backported to glibc-2.28: https://sourceware.org/bugzilla/show_bug.cgi?id=33704 https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=2ca35051530486b0695b007814ddf3db73d603a9 [Hardware Information] Architecture: Intel / AMD (x86_64) Platform(s): Emerald Rapids Granite Rapids Sierra Forest Date HW is expected at Canonical: Component(s): x86 [Software Information] Target Version: 24.04 Target Kernel: glibc-2.39 Commit IDs: 2ca35051530486b0695b007814ddf3db73d603a9 nptl: Optimize trylock for high cache contention workloads (BZ #33704) External Links: [Business Justification] Provides performance improvement for all x86 processors, especially higher core count ones like Xeon 6. [Testing guidance] Use glibc benchmarks: bench-pthread-mutex-trylock-throughput, bench-pthread-mutex-trylock-recursive-throughput [External ID] ** Affects: intel Importance: Undecided Status: New ** Affects: ubuntu Importance: Undecided Status: New ** Tags: 24.04 emr gnr srf x86-64 ** Information type changed from Private to Public ** Description changed: [Description] - Update glibc to include this patch for checking lock availability before acquisition to reduce cache line bouncing. This significantly improves trylock throughput on multi-core systems under heavy contention. - - This has been accepted upstream glibc-2.43+ and backported to glibc-2.28: https://sourceware.org/bugzilla/show_bug.cgi?id=33704 - https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=2ca35051530486b0695b007814ddf3db73d603a9 + Update glibc to include this patch for checking lock availability before acquisition to reduce cache line bouncing. This significantly improves trylock throughput on multi-core systems under heavy contention. + + This has been accepted upstream glibc-2.43+ and backported to glibc-2.28: https://sourceware.org/bugzilla/show_bug.cgi?id=33704 + https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=2ca35051530486b0695b007814ddf3db73d603a9 [Hardware Information] - Architecture: - Intel / AMD (x86_64) - Platform(s): - Emerald Rapids - Granite Rapids - Sierra Forest - Date HW is expected at Canonical: - - Component(s): - x86 + Architecture: + Intel / AMD (x86_64) + Platform(s): + Emerald Rapids + Granite Rapids + Sierra Forest + Date HW is expected at Canonical: + + Component(s): + x86 [Software Information] - Target Version: - 24.04 - Target Kernel: - glibc-2.28 - Commit IDs: - 2ca35051530486b0695b007814ddf3db73d603a9 nptl: Optimize trylock for high cache contention workloads (BZ #33704) - External Links: - + Target Version: + 24.04 + Target Kernel: + glibc-2.39 + Commit IDs: + 2ca35051530486b0695b007814ddf3db73d603a9 nptl: Optimize trylock for high cache contention workloads (BZ #33704) + External Links: [Business Justification] - Provides performance improvement for all x86 processors, especially higher core count ones like Xeon 6. + Provides performance improvement for all x86 processors, especially higher core count ones like Xeon 6. [Testing guidance] - Use glibc benchmarks: bench-pthread-mutex-trylock-throughput, bench-pthread-mutex-trylock-recursive-throughput + Use glibc benchmarks: bench-pthread-mutex-trylock-throughput, bench-pthread-mutex-trylock-recursive-throughput [External ID] -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/2138255 Title: [GNR] glibc: NPTL: Optimize trylock for high cache contention workloads To manage notifications about this bug go to: https://bugs.launchpad.net/intel/+bug/2138255/+subscriptions -- ubuntu-bugs mailing list [email protected] https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
