I saw the juliacon22 video from Christine Flood about Julia GC. I’m glad to see that the GC is finally getting some attention and not just the compiler. It’s long overdue. Let’s just address the elephant in the room: The Julia GC has a memory fragmentation problem. There I said it. I brought this p…

Yes but in those languages you know that you’re responsible for memory management. In fact in those languages they want to manage all of that explicitely for full control including allocation/deallocation strategies like only use static memory in C for embedded systems or arena allocators in C++ etc…

Note that we are actively collecting GC benchmarks at GitHub - JuliaCI/GCBenchmarks . If you have good problem cases for GC aren’t yet represented in the test cases we would love PRs.

When I have some time I’ll try to go back to some old crappy code and extract just the code needed to blow up the memory :slight_smile:

[image] Steven_Sagaert: Let’s just address the elephant in the room: The Julia GC has a memory fragmentation problem. Thanks for bringing it up, but I think it’s a non-issue, or I mean with an alternative allocator a small issue. It’s time to revisit mimalloc (I looked into linking alternati…

[image] gbaraldi: Switching to a compacting GC would be awesome As I said, I’m not sure, I didn’t yet read the Ruby … Deep Dive, I guess it does help most far small objects, because of cache issues, not (huge) arrays of small, i.e. matrices Julia is most optimized for. At least I would first …

In a classic result, Robson showed that all such allocators can suffer from catastrophic memory fragmentation [30]. This increase in memory consumption can be as high as the log of the ratio between the largest and smallest object sizes allo- cated. For example, for an application that allocate…

I guess we could run the GCBenchmarks with different mallocs, not sure how easy/hard that would be, but maybe there is some possible benefit there

[image] Steven_Sagaert: julia heap would need more than the specified limit it exits with an OutOfMemory exception. So only julia exits cleanly and you host server never crashes. I’ m sure sysadmins will appreciate that :slightly_smiling_face: systemd-oomd does something like this at a system…

[image] Palli: but in 64-bit it’s at least close to free Imagine your heap has say 10MB chunks each of which has 100kB of randomly located integer arrays… Let’s say you’ve got 1000 of these arrays, so you have 100MB allocated, but in 10GB of virtual address… The OS knows nothing about the fac…

Nice. I didn’t know that. Guess I’m too old school :wink: Although I deploy on Linux I develop on Windows and there you can still get into the situation where you get to 100% RAM taken and then the julia GC frantically trying to free memory and very large (GBs) fluctuations in julia memory but neve…

Julia GC, heap fragmentation, out of memory, push!/append!

Internals & Design

dlakelan August 10, 2022, 1:46pm 23

Incidentally, it’s very likely this issue bit me recently…

Topic		Replies	Views
Improved allocation design, with 4-byte pointers, and sometimes 5-byte in effect Internals & Design	10	925	December 12, 2024
Julia vs Microsoft's (Rust-inspired) Verona language and e.g. snmalloc allocator Offtopic	13	3689	May 9, 2020
What exactly is "allocation" in Julia? Performance question , memory-allocation	45	6593	November 4, 2022
Multithreaded memory usage General Usage	19	1951	July 14, 2023
Julia position in the Debian Benchmark Game can be improved, and categorization of some Julia there is unfair Performance	29	2204	December 12, 2024

Julia GC, heap fragmentation, out of memory, push!/append!

Related topics