On Thu 19 Apr 2007 16:45, Greg Ungerer pondered:
> I haven't run the dhrystone code for a while, so I don't have
> any specific numbers.

OK - no problem.

> When I last looked at this the big difference seemed to be that
> hardware vendors (well, Freescale in my example case) where using
> the fast internal RAM to run dhrystone. Running it from DRAM was
> am immediate big lose.
>
> Then on top of that was compiler.
>
> Together that meant that numbers measured under uClinux where
> significantly lower.

Yeah, I think we are in the same boat, but not quite as bad, since we have 
enough i/d cache that things don't seem to hit external SDRAM much/at all.

> Most of the work I did was with fully relocated application
> binaries - so no PIC to slow it down.

Which is why we test/measure both with flat, and fdpic. I was actually 
surprised at the low overhead of PIC for some of the tests.

With -O3 -static -fomit-frame-pointer on both, flat measured 374.69, and PIC 
measured 371.02, only a ~1% difference - which I thought was pretty good.

-Robin
_______________________________________________
uClinux-dev mailing list
[email protected]
http://mailman.uclinux.org/mailman/listinfo/uclinux-dev
This message was resent by [email protected]
To unsubscribe see:
http://mailman.uclinux.org/mailman/options/uclinux-dev

Reply via email to