[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Rep:[f-cpu] about the scheduler...



hi,

--- cedric <cedric.bail@free.fr> wrote:
> > Question : do we discard the instructions that
> > write to register zero ? It can save a few cycles,
> > but on the other hands it might become too complex
> > to check.
> 
> I think that's not usefull, because you have a nop,
> so if
> you use it it's only for scheduling reason. Too
> complex
> and not usefull.

i mean do the instruction, but don't set the register
write queue and the x-bar write queue.
it looks easy to me, it can be easily done in the
decoder.

(the scheduler cycle runs after the decoder cycle)
the only thing the scheduler dous during decode
is check for posible bypasses.

it may not look very helpful for fast instructions but
we don't heve only two write slots per instruction
so it seems a pity to spoil them.

when long instructions use both slots to save the
result, than that almost automaticaly results
in a stall further down the line, becouse there
are no free slots. therfore it looks very inportent to
me to not use a free write slot if we write the result
to register 0 anyway.

__________________
(small) simulator update, i got (unconditional)
JMP(/CALL/RET) working :-)

jaap.



__________________________________________________
Do You Yahoo!?
Yahoo! Health - Feel better, live better
http://health.yahoo.com
*************************************************************
To unsubscribe, send an e-mail to majordomo@seul.org with
unsubscribe f-cpu       in the body. http://f-cpu.seul.org/