[bmenrigh]

I recently started using Microsoft's mimalloc (via an LD_PRELOAD) to better use huge (1 GB) pages in a memory intensive program. The performance gains are significant (around 20%). It feels rather strange using an open source MS library for performance on my Linux system.

There needs to be more competition in the malloc space. Between various huge page sizes and transparent huge pages, there are a lot of gains to be had over what you get from a default GNU libc.

[skavi]

We evaluated a few allocators for some of our Linux apps and found (modern) tcmalloc to consistently win in time and space. Our applications are primarily written in Rust and the allocators were linked in statically (except for glibc). Unfortunately I didn't capture much context on the allocation patterns. I think in general the apps allocate and deallocate at a higher rate than most Rust apps (or more than I'd like at least).

Our results from July 2025:

rows are <allocator>: <RSS>, <time spent for allocator operations>

  app1:
  glibc: 215,580 KB, 133 ms
  mimalloc 2.1.7: 144,092 KB, 91 ms
  mimalloc 2.2.4: 173,240 KB, 280 ms
  tcmalloc: 138,496 KB, 96 ms
  jemalloc: 147,408 KB, 92 ms

  app2, bench1
  glibc: 1,165,000 KB, 1.4 s
  mimalloc 2.1.7: 1,072,000 KB, 5.1 s
  mimalloc 2.2.4:
  tcmalloc: 1,023,000 KB, 530 ms

  app2, bench2
  glibc: 1,190,224 KB, 1.5 s
  mimalloc 2.1.7: 1,128,328 KB, 5.3 s
  mimalloc 2.2.4: 1,657,600 KB, 3.7 s
  tcmalloc: 1,045,968 KB, 640 ms
  jemalloc: 1,210,000 KB, 1.1 s

  app3
  glibc: 284,616 KB, 440 ms
  mimalloc 2.1.7: 246,216 KB, 250 ms
  mimalloc 2.2.4: 325,184 KB, 290 ms
  tcmalloc: 178,688 KB, 200 ms
  jemalloc: 264,688 KB, 230 ms

tcmalloc was from github.com/google/tcmalloc/tree/24b3f29.

i don't recall which jemalloc was tested.

[hedora]

I’m surprised (unless they replaced the core tcmalloc algorithm but kept the name).

tcmalloc (thread caching malloc) assumes memory allocations have good thread locality. This is often a double win (less false sharing of cache lines, and most allocations hit thread-local data structures in the allocator).

Multithreaded async systems destroy that locality, so it constantly has to run through the exception case: A allocated a buffer, went async, the request wakes up on thread B, which frees the buffer, and has to synchronize with A to give it back.

Are you using async rust, or sync rust?

[skavi]

modern tcmalloc uses per CPU caches via rseq [0]. We use async rust with multithreaded tokio executors (sometimes multiple in the same application). so relatively high thread counts.

[0]: https://github.com/google/tcmalloc/blob/master/docs/design.m...

[usrnm]

How do you control which CPU your task resumes on? If you don't then it's still the same problem described above, no?

[skavi]

on the OS scheduler side, i'd imagine there's some stickiness that keeps tasks from jumping wildly between cores. like i'd expect migration to be modelled as a non zero cost. complete speculation though.

tokio scheduler side, the executor is thread per core and work stealing of in progress tasks shouldn't be happening too much.

for all thread pool threads or threads unaffiliated with the executor, see earlier speculation on OS scheduler behavior.

[packetlost]

Correct. The Linux scheduler has been NUMA aware + sticky for awhile (which is more or less what this reduces to in common scenarios).

[jhalstead]

> I’m surprised (unless they replaced the core tcmalloc algorithm but kept the name).

Indeed, it's not the old gperftools version.

Blog: https://abseil.io/blog/20200212-tcmalloc

History / Diffs: https://google.github.io/tcmalloc/gperftools.html

[skavi]

also:

1. tcmalloc is actually the only allocator I tested which was not using thread local caches. even glibc malloc has tcache.

2. async executors typically shouldn’t have tasks jumping willy nilly between threads. i see the issue u describe more often with the use of thread pools (like rayon or tokio’s spawn_blocking). i’d argue that the use of thread pools isn’t necessarily an inherent feature of async executors. certainly tokio relies on its threadpool for fs operations, but io-uring (for example) makes that mostly unnecessary.

[ComputerGuru]

That’s a considerable regression for mimalloc between 2.1 and 2.2 – did you track it down or report it upstream?

Edit: I see mimalloc v3 is out – I missed that! That probably moots this discussion altogether.

[skavi]

nope.

[codexon]

This is similar to what I experienced when I tested mimalloc many years ago. If it was faster, it wasn't faster by much, and had pretty bad worst cases.

[pjmlp]

If you go into Dr Dobbs, The C/C++ User's Journal and BYTE digital archives, there will be ads of companies whose product was basically special cased memory allocator.

Even toolchains like Turbo Pascal for MS-DOS, had an API to customise the memory allocator.

The one size fits all was never a solution.

[adgjlsfhk1]

One of the best parts about GC languages is they tend to have much more efficient allocation/freeing because the cost is much more lumped together so it shows up better in a profile.

[pjmlp]

Agreed, however there is also a reason why the best ones also pack multiple GC algorithms, like in Java and .NET, because one approach doesn't fit all workloads.

[nevdka]

Then there’s perl, which doesn’t free at all.

[hedora]

Perl frees memory. It uses refcounting, so you need to break heap cycles or it will leak.

(99% of the time, I find this less problematic than Java’s approach, fwiw).

[wredcoll]

Unless this has changed recently, perl doesn't free memory to the kernel, only within its own process/vm.

[cermicelli]

Freedom is overrated... :P

[NooneAtAll3]

doesn't java also?

I heard that was a common complaint for minecraft

[adgjlsfhk1]

Minecraft for somewhat silly reasons was largely stuck using Java8 for ~a decade longer than it should have which meant that it was using some fairly outdated GC algorithms.

[NooneAtAll3]

"silly reasons" being Java breaking backwards compatibility

decade seems a usual timescale for that, considering f.e. python 2->3

[kbolino]

So much software was stuck on Java 8 and for so long that some of the better GC algorithms got backported to it.

[xxs]

What do you mean - if Java returns memory to the OS? Which one - Java heap of the malloc/free by the JVM?

[cogman10]

Java is pretty greedy with the memory it claims. Especially historically it was pretty hard to get the JVM to release memory back to the OS.

To an outsider, that looks like the JVM heap just steadily growing, which is easy to mistake for a memory leak.

[CyberDildonics]

Any extra throughput is far overshadowed by trying to control pauses and too much heap allocations happening because too much gets put on the heap. For anything interactive the options are usually fighting the gc or avoiding gc.

[bluGill]

When it works. Many programs in GC language end up fighting the GC by allocating a large buffer and managing it by hand anyway because when performance counts you can't have allocation time in there at all. (you see this in C all the time as well)

[cogman10]

That's generally a bad idea. Not always, but generally.

It was a better idea when Java had the old mark and sweep collector. However, with the generational collectors (which are all Java collectors now. except for epsilon) it's more problematic. Reusing buffers and objects in those buffers will pretty much guarantees that buffer ends up in oldgen. That means to clear it out, the VM has to do more expensive collections.

The actual allocation time for most of Java's collectors is almost 0, it's a capacity check and a pointer bump in most circumstances. Giving the JVM more memory will generally solve issues with memory pressure and GC times. That's (generally) a better solution to performance problems vs doing the large buffer.

Now, that said, there certainly have been times where allocation pressure is a major problem and removing the allocation is the solution. In particular, I've found boxing to often be a major cause of performance problems.

[drob518]

If your workload is very regular, you can still do better with an arena allocator. Within the arena, it uses the same pointer-bump allocation as Java normally uses, but then you can free the whole area back to the start by resetting the pointer to its initial value. If you use the arena for servicing a single request, for instance, you then reset as soon as you're done with the request, setting you up with a totally free area for the next request. That's more efficient than a GC. But it also requires your algorithm to fall into that pattern where you KNOW that you can and should throw everything from the request away. If you can't guarantee that, then modern collectors are pretty magical and tunable.

[CyberDildonics]

If people didn't need to do it, they wouldn't generally do it. Not always, but generally.

[cogman10]

People do stuff they shouldn't all the time.

For example, some code I had to clean up pretty early on in my career was a dev, for unknown reasons, reinventing the `ArrayList` and then using that invention as a set (doing deduplication by iterating over the elements and checking for duplicates). It was done in the name of performance, but it was never a slow part of the code. I replaced the whole thing with a `HashSet` and saved ~300 loc as a result.

This individual did that sort of stuff all over the code base.

[CyberDildonics]

Reinventing data structures poorly is very common.

Heap allocation in java is something trivial happens constantly. People typically do funky stuff with memory allocation because they have to, because the GC is causing pauses.

People avoid system allocators in C++ too, they just don't have to do it because of uncontrollable pauses.

[m463]

I remember in the early days of web services, using the apache portable runtime, specifically memory pools.

If you got a web request, you could allocate a memory pool for it, then you would do all your memory allocations from that pool. And when your web request ended - either cleanly or with a hundred different kinds of errors, you could just free the entire pool.

it was nice and made an impression on me.

I think the lowly malloc probably has lots of interesting ways of growing and changing.

[Sesse__]

This is called “an arena” more generally, and it is in wide use across many forms of servers, compilers, and others.

[jra_samba]

Look into talloc, used inside Samba (and other FLOSS projects like sssd). Exactly this.

[pocksuppet]

In many cases you can also do better than using malloc e.g. if you know you need a huge page, map a huge page directly with mmap

Yes, if you want to use huge pages with arbitrary alloc/free, then use a third-party malloc. If your alloc/free patterns are not arbitrary, you can do even better. We treat malloc as a magic black box but it's actually not very good.

[IshKebab]

I feel like the real thing that needs to change is we need a more expressive allocation interface than just malloc/realloc. I'm sure that memory allocators could do a significantly better job if they had more information about what the program was intending to do.

[liuliu]

There are, look no further than jemalloc API surface itself:

https://jemalloc.net/jemalloc.3.html

One thing to call out: sdallocx integrates well with C++'s sized delete semantics: https://isocpp.org/files/papers/n3778.html

[hedora]

You can also play tricks with inlining and constant propagation in C (especially on the malloc path, where the ground-truth allocation size is usually statically known).

[Dylan16807]

I think some operating system improvements could get people motivated to use huge pages a lot better. In particular make them less fragile on linux and make them not need admin rights on windows. The biggest factor causing problems there is that neither OS can swap a 2MB page. So someone needs to care enough to fix that.

[anthk]

I used mimalloc to run zenlisp under OpenBSD as it would clash with the paranoid malloc of base.

[jeffbee]

Just out of curiosity are you getting 1GB huge pages on Xeon or some other platform? I always thought this class of page is the hardest to exploit, considering that the machine only has, if I recall correctly, one TLB slot for those.

[bmenrigh]

Modern x86_64 has supported multiple page sizes for a long time. I'm on commodity Zen 5 hardware (9900X) with 128 GiB of RAM. Linux will still use a base page size of 4kb but also supports both 2 MiB and 1 GiB huge pages. You can pass something like `default_hugepagesz=2M hugepagesz=1G hugepages=16` to your kernel on boot to use 2 MiB pages but reserve 16 1 GiB pages for later use.

The nice thing about mimalloc is that there are a ton of configurable knobs available via env vars. I'm able to hand those 16 1 GiB pages to the program at launch via `MIMALLOC_RESERVE_HUGE_OS_PAGES=16`.

EDIT: after re-reading your comment a few times, I apologize if you already knew this (which it sounds like you did).

[jeffbee]

Right but on Intel the 1G page size has historically been the odd one. For example Skylake-X has 1536 L2 shared TLB entries for either 4K or 2M pages, but it only has 16 entries that can be used for 1G pages. It wasn't unified until Cascade Lake. But Skylake-like Xeon is still incredibly common in the cloud so it's hard to target the later ones.

[Dylan16807]

So for any process that's using less than 16GB, it's a significant performance boost. And most processes using more RAM, but not splitting accesses across more than 16 zones in rapid succession, will also see a performance boost.

My old Intel CPU only has 4 slots for 1GB pages, and that was enough to get me about a 20% performance boost on Factorio. (I think a couple percent might have been allocator change but the boost from forcing huge pages was very significant)

[jeffbee]

That strikes me as a common hugepages win. People never believe you, though, when you say you can make their thing 20% faster for free.

[menaerus]

Then it should be pretty easy to display that 20% "faster for free", no? But as always the devil is in the details. I experimented a lot with huge pages, and although in theory you should see the performance boost, the workloads I have been using to test this hypothesis did not end up with anything statistically significant/measurable. So, my conclusion was ... it depends.

[tosti]

    > commodity
    > zen 5
    > 128GiB

Are you from the future?

[Dylan16807]

I'm not sure what point you're trying to make.

In the middle of last year, a 9900X was around $350 and 128GB of memory was also around $350. That's very easily "commodity" range.

[tosti]

Damn. I feel old and must've missed that boat. Several other boats too, I guess.

Here I was thinking 16GiB is pretty good. I get to compile LibreOffice in an afternoon. QtWebEngine overnight.

Doesn't 128GiB make rowhammer much more feasible? You'd have 32GiB per DIMM.

Oh well

[Dylan16807]

Two 64GiB DIMMs would be the more likely setup. The current CPUs strongly prefer having only one stick of DDR5 per channel.

The effectiveness of rowhammer depends on how well the manufacturer implemented target row refresh. But the internal ECC on DDR5 should help defend against it somewhat.

Personally I've been in the 24-32GiB range since 2013, and that's despite the fact that I'm still on DDR3.

[sylware]

If there is so much performance difference among generic allocators, it means you need semantic optimized allocators (unless performance is actually not that much important in the end).

[Cloudef]

You are not wrong and this is indeed what zig is trying to push by making all std functions that allocate take a allocator parameter.

[codexon]

Agreed mostly. Going from standard library to something like jemalloc or tcmalloc will give you around 5-10% wins which can be significant, but the difference between those generic allocators seem small. I just made a slab allocator recently for a custom data type and got speedups of 100% over malloc.

[sylware]

Here you go.

[codexon]

I've been using jemalloc for over 10 years and don't really see a need for it to be updated. It always holds up in benchmarks against any new flavor of the month malloc that comes out.

Last time I checked mimalloc which was admittedly a while ago, probably 5 years, it was noticebly worse and I saw a lot of people on their github issues agreeing with me so I just never looked at it again.

[adgjlsfhk1]

Mimalloc v3 has just come out (about a month ago) and is a significant improvement over both v2 and v1 (what you likely last tested)