Hello, I've left your original email intact below for comparison,
but I can tell you that on my single processor Celeron 500
(an HP Pavilion 8560c with non-stock video and sound)
does the same check upon bootup, and finds the pII_mmx
to be the fastest, I believe in the 1200mb/sec range.
On Sun, 07 Nov 1999, Jason Costomiris wrote:
> I'm running a Dual Celeron setup (300A's on BP6 oc'd to 477). I noticed
> during boot the following:
>
> Nov 3 20:28:49 anthrax kernel: raid5: measuring checksumming speed
> Nov 3 20:28:49 anthrax kernel: raid5: MMX detected, trying high-speed MMX checksum
>routines
> Nov 3 20:28:49 anthrax kernel: pII_mmx : 1069.467 MB/sec
> Nov 3 20:28:49 anthrax kernel: p5_mmx : 1115.949 MB/sec
> Nov 3 20:28:49 anthrax kernel: 8regs : 819.531 MB/sec
> Nov 3 20:28:49 anthrax kernel: 32regs : 481.965 MB/sec
> Nov 3 20:28:49 anthrax kernel: using fastest function: p5_mmx (1115.949 MB/sec)
>
> I see that the p5_mmx instructions are actually faster than the pII_mmx
> instructions. I've also read that this condition is present in the P-III
> CPUs. Would the P-III patches help in this case? I doubt it, but I
> suppose anything's possible.
>
> The above came from the stock 2.2.12-20smp kernel. I'm currently running
> 2.2.13 without the RAID patches (they didn't seem to get along w/ the
> new IDE patches that add support for the HPT366 onboard the BP6). Would the
> P-III patches and compiling w/P-III optimization add performance, or would
> that just be a waste of cycles?
--
To unsubscribe: mail [EMAIL PROTECTED] with "unsubscribe"
as the Subject.
