Glibc Enables A Per-Thread Cache For Malloc - Big Performance Win

[phoronix]

Phoronix: Glibc Enables A Per-Thread Cache For Malloc - Big Performance Win

Glibc has added a per-thread cache to malloc and enabled it by default...

http://www.phoronix.com/scan.php?page=news_item&px=glibc-malloc-thread-cache

[chuckula]

I think this email has some more detailed information: https://sourceware.org/ml/libc-alpha.../msg00452.html

[vladpetric]

About effing time!

TCMalloc (thread caching malloc) has been doing that for more than a decade

http://goog-perftools.sourceforge.net/doc/tcmalloc.html

[cj.wijtmans]

not much info about this news. how many nanoseconds did it save?

[starshipeleven]

not much info about this news. how many nanoseconds did it save?

See the email linked in first post, it's around 20% decrease in malloc times.

I know too little about this subject to translate that into nanoseconds saved.

[Zan Lynx]

I am pretty sure glibc has tried similar things in the past. It isn't as if it had no support for threads. It just wasn't very good.

[pal666]

Just a note: golang has a lockless per-thread allocation cache since its introduction in 2009.

to have malloc cache you have to provide malloc. otherwise your cache will not affect code calling malloc directly like glibc itself

[DrYak]

Genuine question :
- Which developer in his right mind would be calling malloc in a performance-critical part of the code ?

Are there that many *non-synthetic benchmarks* use cases where faster malloc really impacts performance of the software ?

[Hi-Angel]

Are there that many *non-synthetic benchmarks* use cases where faster malloc really impacts performance of the software ?

I imagine Firefox might benefit. AFAIK it still have disabled GPU acceleration by default, and so is pretty CPU-heavy.

[coder]

Genuine question :
- Which developer in his right mind would be calling malloc in a performance-critical part of the code ?

Well, pretty much anything not dealing only with small strings and fixed-sized arrays. So, that's basically most real-world software.

Most interesting data structures require dynamic allocation.