[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Re: [f-cpu] Smooth Register backup issues...

>> -SRB could be slow : SBR add muxes on the read port of the register bank
>> which are on the critical data path
> Instead of shadow, I should rather have it called a 4032-bit wide 1 word
> deep FIFO. If we shorten the critical data path, we win again. I have
> to look how COMPLEXITY is affected. If weget rid of 4032 MUX ports,
> it's worth looking at the story.
>> Linux is jiffies are about 1kHz. We should try to evaluate network
>> things
>> (10 us for a paquet ?) but most of the IO are done by DMA.
> AFAIK, it's 100 Hz unless you install a low latency kernel, which
> suggests 512 Hz during configure for compile.

1khz on the 2.6
100Hz on the 2.4 and before
512Hz on the patched Red Hat kernel 2.4

> Before we talk about winning 1% performance, we should avoid 100%
> performance
> loss (i.e. get the CPU work at all).


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