Re: [f-cpu] Instruction census

[Yann Guidon <whygee@f-cpu.org>]

hi,

nico wrote:

> On Mon, 13 Jan 2003 00:07:19 +0100
> Michael Riepe <michael@stud.uni-hannover.de> wrote:
>
> > The vast majority of instructions are load/store, add/sub/loadaddr[i]
> >    
> It's even worse ! There is much more load than store !
isn't it the usual case ?
MR's numbers look ok.

> Could you calculate the distribution of distance between the load and the
> effective use of the data ?
in "classical code" it's immediate.
i guess they haven't yet "teached" GCC to heavily prefetch things.

> Because load
> latency is udge(2-3 on L1, until ~200 from the main memory). Store is
> not a such problem if we use store buffer and write allocate.
>  
To quote MR : "life's a bitch".

YG

*************************************************************
To unsubscribe, send an e-mail to majordomo@seul.org with
unsubscribe f-cpu       in the body. http://f-cpu.seul.org/