|
|
Log in / Subscribe / Register

Support for the TSO memory model on Arm CPUs

Support for the TSO memory model on Arm CPUs

Posted Apr 30, 2024 7:54 UTC (Tue) by epa (subscriber, #39769)
In reply to: Support for the TSO memory model on Arm CPUs by hmh
Parent article: Support for the TSO memory model on Arm CPUs

There's a moral dimension to it that other architecture extensions don't have. If a new CPU supports a new instruction, and you decide to use it to speed up your program, you are simply taking advantage of new hardware. No doubt you continue to provide an alternative, slower implementation for older CPUs. You're not being a bad person by using the new instruction -- except, perhaps, if it's patented and only available on processors from one litigious, grasping manufacturer. (Although in olden days there was sometimes a mild "naughtiness" associated with using undefined opcodes of an 8-bit processor, arising from "don't care" states in the design, which had strange but occasionally useful semantics.)

But here there is a true path revealed to us. The designers of ARM, in their wisdom, have provided a plain and humble concurrency model. The righteous programmer will take care to follow its strictures. Not for us the worldly luxury of TSO as practised by the followers of Intel and the AMD-ites. If one group of ARM developers is allowed to get lazy and become addicted to the easier memory model, that's unfair to those who have remained virtuous. Indeed, others might be led astray and start to demand support for TSO everywhere. And then where would we be?


to post comments


Copyright © 2026, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds