Re: dcache-ac6-D - dcache threading

Date: Mon Jun 05 2000 - 07:08:23 EST

Andrey Savochkin writes:
> On Mon, Jun 05, 2000 at 06:37:43PM +0900, wrote:
> >
> > Following values are that you requested:
> > lock aquire 280ns/call
> > movzwl(inw) 1580ns/call or 750ns/loop
> Seems to be too long, even if assume a synchronous PCI transaction (it's
> about 22 PCI cycles on 33MHz PCI). I don't know how to explain it.

Please be careful to interpret the values. The measurement is based
on statistic sampling and super scalar execution can distort the

For example, serializing instruction such as cli/sti/in/out must flush
previous outstanding execution, which takes longer time than one
instruction execution.
In a pipelined superscaler CPU, it is impossible to define execution
time of any single instruction, actually.

But if you keep this caution, these values will help you a lot.

Computer Systems Laboratory, Fujitsu Labs.

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to Please read the FAQ at

This archive was generated by hypermail 2b29 : Wed Jun 07 2000 - 21:00:21 EST