Changing the default userspace memory allocator for a distribution is less a technical choice than you think: lists.fedoraproject.org/archives/list/
Conversation
Replying to
jemalloc, tcmalloc, hardened_malloc, mimalloc, etc. are as general purpose as glibc malloc. glibc malloc isn't good at any niche. It's not a modern/competitive allocator design. It's not simply choosing different tradeoffs than modern allocators. It's an obsolete design overall.
1
9
Almost everyone who caring mostly about memory usage and performance would be happier with jemalloc with suitable defaults for libc inclusion (see: FreeBSD, Android). Essentially no one would replace jemalloc with glibc malloc if it was the other way around with jemalloc in libc.
1
2
FreeBSD uses jemalloc as their default libc allocator. Android used it for a long time tuned for low memory usage rather than parallelism until recent versions replaced it with a more security focused allocator (Scudo) outside low memory devices. What's less general purpose?
I can agree that glibc's approach as a whole doesn't have much to do with technical reasons but rather political ones and nepotism. I see what's happening as misinformation being spread about other projects, and not just ours but also jemalloc. Is this really a fight you want?
1
2
Signal boosting this stuff elsewhere has escalated it further and now I feel the need to respond to Red Hat employees spreading misinformation.
I'm not going to limit myself to debunking misinformation but also calling out corruption and unethical behavior in your projects.
1
Show replies