Subject: Re: PC164 memory bus speeed (was: pciide performance on alpha)
To: Andrew Gillham <email@example.com>
From: Nathan J. Williams <nathanw@MIT.EDU>
Date: 11/02/1999 15:56:23
> A 'make build' of -current (same date) takes
> 1:25:00 on my Celeron 433, and 2:30 on my Alpha 500. Both with fast
> EIDE hard drives. (a 8GB 512KB buffer on the Celeron, 20GB 2MB buffer on
> the Alpha) Admittedly the Celeron has a heck of a lot more RAM (256MB
> .vs. 64MB) and I haven't tested it with only 64MB. (or only 48MB)
Remember, 'make build' is a pretty bad benchmark, because the
toolchain does wildly different amounts of work on different
platforms. For example, egcs does a lot more instruction scheduling
(the Haifa stuff) for the alpha, powerpc, and sparc targets than it
does for other processors, including the i386. The exact code built is
different as well.
> The Celeron outperforms the Alpha at dhyrstones (~800K .vs. ~540K), and
> at whetstones (250Mips .vs. 180Mips)
> The only (so far) that the Alpha wins is the DES part of rc5des. (4.9M
> .vs. 6.0M)
These are more problematic.
> If there are some optimizations (like -mcpu=21164a) that should be enabled
> I would love to hear about them. What are the options needed to specify
> the L3 cache latency?
If you're going to be fair about it, you should tweak the build
options on the i386 side for the processor as well.