Glibc 2.29 Is Offering Up Some Nice Performance Improvements

[phoronix]

Phoronix: Glibc 2.29 Is Offering Up Some Nice Performance Improvements

Glibc 2.29 was released a few days back and like most GNU C Library releases -- particularly in recent times -- does offer up more CPU performance optimizations... Some early benchmarks done this weekend do show some nice performance improvements in select workloads at least out of our initial benchmarking.

http://www.phoronix.com/vr.php?view=27491

[ksec]

Would be nice if there were some benchmarks comparing Glibc vs muslc.

[jabl]

Would be nice with some analysis of profiling data so we can have an idea of what specific improvements caused which changes.

[pegasus]

Would be nice if these changes land in rhel 8. Btw Michael, beta is already out, care to add it to your regular tests?

[S.Pam]

I'm surprised glibc affected performance enough in the first place to see these improvements

[mlau]

I'm surprised glibc affected performance enough in the first place to see these improvements

There were quite notable performance improvements to floating points functions (i.e. libm), driven by the ARM folks, which also benefit more recent x86.

[FireBurn]

I'm running Glibc 2.29 on Gentoo... just needed to unmask it

[S.Pam]

I'm running Glibc 2.29 on Gentoo... just needed to unmask it

Any issues?

[carewolf]

I'm surprised glibc affected performance enough in the first place to see these improvements

It is very important for performance. One of the biggest factors are the malloc/free implementation and the string/memory routines, but they have been optimized so much through the years there is little left to improve.

[Michael]

Would be nice if these changes land in rhel 8. Btw Michael, beta is already out, care to add it to your regular tests?

Have the ISO but unfortunately no repo access unless having a RHEL subscriber account.