@never_released@mastodon.social
2026-09-07 05:07 UTC
> We find that Apple silicon
TSO mode preserves Arm weak-memory optimizations, typically yielding execution times within 3% of Arm mode across
modern applications and classic benchmarks.
Replies (1)
-
@never_released@mastodon.social 2026-09-07 05:07
> We argue that these issues are addressable in hardware; in fact, we find that the M4 already resolves the SIMD slowdown. Moreover, the slowdowns can be mitigated with simple software changes. With these software changes, the harmonic mean of per-application TSO slowdowns across our entire benchmark suite decreases from 4.2% to 1.9%.