[ale] BogoMips uselsess?

Michael Hirsch hirsch at mathcs.emory.edu
Tue May 11 00:46:10 EDT 1999


Clay Lawrence writes:
> I've found make -j n where n=RAM/8 works fairly well and keeps this from
> happening. Where RAM=32 n is 4 but for people like me where RAM=512  is
> 64 :-)

Hmm.  I recently ran tests on my dual PPro with 128Meg.  According to
this rule n = 16 should be optimal, but I found 5 or 6 was best. 

I imagine the proper formula will depend on the ratio between
processor and hard disk speed.  Also, my tests were on compiling the
kernel which is a little different from compiling normal code, though
I don't know how much difference that will make.

--Michael






More information about the Ale mailing list