3 ms·
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
by gpderetta 7y ago
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.
- chrismonsanto 7y 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.
- gpderetta 7y 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.
- chrismonsanto 7y agoReading 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.