Refresh mlx-swift fork and embedded MLX - #6

Merged

a-github-name merged 18 commits into

mainfrom

codex/mlx-swift-0.32.1-upstream-refresh

Aug 14, 2026

Commits on Jul 7, 2026

Add fp8 conversion utilities (#435)
lucasnewman
authored
7a8d05a
build iOS in CI (#440)
davidkoski
authored
de3342c

Commits on Jul 13, 2026

Add global scale support to quantized layers (#426)
aleroot
authored
09051ed

Commits on Aug 4, 2026

Fix FinalizerCaptureState leak in MLXArray(rawPointer:_:dtype:finalizer:) (#448)

agerjura
and
Fred Jura
authored
9a09a03

Commits on Aug 10, 2026

Derive the jit-source lists in update-mlx.sh instead of hand-listing them (#445)
GoodOlClint
authored
da31870

Commits on Aug 14, 2026

build: support Swift 6.0 without dropping CUDA
a-github-name
committed
fdb1a7c
fix: retain Linux MLX fast symbols
a-github-name
committed
a1ea6fd
Refresh MLX 0.32.1 runtime bridge
a-github-name
committed
c2dbb6f
Make async MLX streams executor-safe
a-github-name
committed
d4f707f
Preserve actor isolation for async MLX streams
a-github-name
committed
802bb53
Pin quantized custom Metal kernel support
a-github-name
committed
08e9ba7
Keep CPU and GPU streams executor-safe
a-github-name
committed
d353dcc
Avoid unavailable GPU companion streams
a-github-name
committed
0ae3580
Ship bit-exact NVFP4 staging rewrite
a-github-name
committed
f92684f
Regenerate MLX 0.32.1 runtime sources
a-github-name
committed
2878bab
Format refreshed fast-kernel tests
a-github-name
committed
b28ba70
Pin QMV dispatch correction
a-github-name
committed
f8536fe
Pin Linux matmul compatibility fix
a-github-name
committed
3e6df6d