> Hi Richard,
>
> I think your test procedure is flawed. I upped the
> SIZE parameter in the makefile to 63kbytes in order
> to simulate the situtation where the network buffer is
> not already in the 1st level cache, which I am guessing
> is the typical case. Your routines did very badly, ie
> they were twice as slow as any of the other versions.
> This is on a dual PPro 180 with 256kbytes 2nd level cache
> per processor. It has separate 8k I and D caches for the
> 1st level. I compiled with -O2.
>
> --
> Erik Corry erik@arbat.com Ceterum censeo, Microsoftem esse delendam!
>
What does this have to do with the real world when the typical
MTU is 1500 bytes??
Further, with -O6 -funroll-loops, the comparison is still:
Counting C loops for 3 seconds
Counting ASM loops for 3 seconds
C routine : 3388
AS routine : 6471
Change : 1.91 times faster
AS clocks : 76035 clocks/byte : 1.24
C clocks : 145993 clocks/byte : 2.38
Cheers,
Dick Johnson
***** FILE SYSTEM WAS MODIFIED *****
Penguin : Linux version 2.1.127 on an i586 machine (66.15 BogoMips).
Warning : It's hard to remain at the trailing edge of technology.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/