jemallocator
jemallocator copied to clipboard
gnzlbg
Rust allocator using jemalloc as a backend
We should check that `-C debug-info=N` is appropriately forwarded to `-gN` when building `jemalloc` in jemalloc-sys. Debug symbols can have a big impact on binary size.
See https://github.com/jemalloc/jemalloc/issues/1320
See: https://github.com/jemalloc/jemalloc/issues/1274
Background threads fail in musl with: ``` : Error in dlsym(RTLD_NEXT, "pthread_create") ```
@davidtgoldblatt mentioned on gitter: > there's an optimization that Rust can do that C can't btw > > Since rust always passes in the alignment, it can support more size...
The tests fail on `sparc64-unknown-linux-gnu` due to a `libstd` initialization failure: https://github.com/rust-lang/rust/issues/55204
The library fails to build for: * arm-linux-androideabi * armv7-linux-androideabi
The library fails to build for: * arm-unknown-linux-musleabi * armv7-unknown-linux-musleabihf
Switching to this crate with `#[global_allocator]` apparently caused Servo to consume 7x more virtual memory than with `alloc_jemalloc`: https://github.com/servo/servo/issues/19010. This seems surprising since the same jemalloc code is used ultimately,...
But currently isn't, and AFAIK there isn't a way to annotate FFI functions with these attributes. This might explain why the benchmark for `alloc_excess` when the result is not used...