5 lines of #JuliaLang is all you need to get to get the most of your hardware via #AVX512, #AppleSilicon, or other microarchitecture-specific binaries instead of plain #x86_64 / #amd64 / #x64, #aarch64 / #arm64 binaries!

… and the best part, you don’t even have to write those five lines - they’re here for you (and have been for a long time):

https://github.com/JuliaPackaging/Yggdrasil/blob/master/platforms/microarchitectures.jl

Dear Lazy web...

With the old #NixOS aarch64 build host dead and my request for access to the new one unsuccessful, I'm pondering the best path to having an aarch64 build machine again. I see three paths, in no particular order:

* Pay for an #aarch64 VPS and set it up as a build machine
* Re-build one of my Pi3's as a build machine
* Use cross compilation

None of them are ideal and there may be better ways.

Interested in your thoughts or other ideas.

I’m gonna to start aggressively migrating stuff to #aarch64 since the majority of workloads are #golang and any that are making use of specific intel version performance characteristics are already isolated to specific cpu generation and class.

Will it save money? Sure. Will it promote diversity in cpu architecture and help normalize alternative cpu architectures in the org and maybe slightly through cloud spend? Fucking maybe.

Here’s a non-ai image to express this good idea visually.

Success!!! I've migrated Immich from an x86 Debian system to Asahi Fedora on an M1 Mac Mini.
Installing Fedora was easy
Installing Docker was easy.
Installing Tailscale was super simple.
Installing and setting up CSF was easy, but only because I had notes from last time on how to make Docker and Tailscale cooperate.

Migrating the Immich database was a nightmare!! I think because my old system was a mishmash of legacy settings from 100+ releases ago... Also, my old system wasn't on the very newest release, which in itself made migration fail initially.

#immich #tailscale #asahilinux #fedora #aarch64 #csf #docker #macmini #selfhost #selfhosted

We upstreamed an SME code generator for tensor processing primitives to the LIBXSMM library. Small matrix-matrix multiplications are one of the supported primitives, and the code generation can be tested with a few commands:

```
git clone https://github.com/libxsmm/libxsmm.git
cd libxsmm; make -j BLAS=0
cd samples/xgemm; make -j
./gemm_kernel F32 F32 F32 F32 512 512 512 512 512 512 \
1 1 0 0 0 1 0 0 0 nopf nobr 0 1 10000 0
```

https://scalable.uni-jena.de/research/2025/03/31/sme-kernels.html

#aarch64 #sme #apple #m4 #hpc #ai

My first GCC commit! 🎉

Although my first C compiler was Borland Turbo C, GCC has always offered a better programming experience. My journey with GCC started in the late '90s, back when Red Hat was a rad startup, The Matrix was blowing minds, and Rage Against the Machine was blasting through boombox.

After decades of using the GCC toolchain, I now find myself contributing to its codebase. Excited to share my first commit to the GCC project, adding a few intrinsic functions to arm_acle.h.

Huge thanks to everyone on the GCC team for their support!

🔗 https://gcc.gnu.org/git/?p=gcc.git;a=commit;h=f4f7216c56fe2f67c72db5b7c4afa220725f3ed1

#gcc #compiler #opensource #aarch64 #arm