undefined | Better HN

0 pointsgpderetta6y ago0 comments

No, to implement x86 atomic semantics is the guarantee that a single cache line can be held in exlusive mode for a minimum lenght of time.

As forward progress is a pretty basic requirements, in practice even LL/SC platforms in practice do that, but is instead of having a single instruction with guaranteed forward progress you have to use a few special (but sometimes underspecified) sequences of instructions between the ll/sc pairs.

0 comments

chrismonsanto6y ago

> in practice

FWIW RISC-V guarantees forward progress for reasonable uses:

> We mandate that LR/SC sequences of bounded length (16 consecutive static instructions) will eventually succeed, provided they contain only base ISA instructions other than loads, stores, and taken branches.

gpderettaOP6y ago

[sorry for the late reply]

what happens if those 16 instructions touch 16 different cache lines? I'm not an hardware expert (and even less on coherency protocols), but I think it would be extremely hard to make sure livelocking is avoided in all cases, short of having some extremely expensive and heavy handed global 'bus' lock fallback.

chrismonsanto6y ago

Reading and writing memory are excluded from the guarantee, aside from the LR/SC instructions that bookend a transaction. Inside the transaction you're basically limited to register-register ALU and aborting branches.

j / k navigate · click thread line to collapse

0 comments

chrismonsanto6y ago

> in practice

FWIW RISC-V guarantees forward progress for reasonable uses:

gpderettaOP6y ago

[sorry for the late reply]

chrismonsanto6y ago

j / k navigate · click thread line to collapse