[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [f-cpu] ll/sc
hi,
i don't read much the f-cpu mails these days...
but...
> Christophe Avoinne wrote:
>
> Another possibility :
>
> loopentry r3
> ...
> lload r1,[r2] --> r1 = *r2; lock LSU entry
> ...
> lstore r1,[r2],r3 --> if locked, *r2 = r1, unlock LSU entry.
> if not, jump to r3
no <- the jump destination can only be in R2.
> we jump instead of setting false a register when
> we cannot store because the lock is discarded in the LSU entry.
>
> The only interest to do so, is to reduce instructions since
> we can expect having a conditional jumping after a failed lstore.
>
> But the counterpart is that you cannot sandwitch some
> instructions between lstore and conditional jump or
> just want to jump when lstore suceeds.
The instruction format problem is much more important and makes
this approach impossible. sorry.
WHYGEE
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
*************************************************************
To unsubscribe, send an e-mail to majordomo@seul.org with
unsubscribe f-cpu in the body. http://f-cpu.seul.org/
- References:
- [f-cpu] ll/sc
- From: "Christophe Avoinne" <christophe.avoinne@laposte.net>