[hashcash] Re: Speed problem with 1.03 on Mac G4
- From: Jonathan Morton <chromi@xxxxxxxxxxxxxxxxxxxxx>
- To: hashcash@xxxxxxxxxxxxx
- Date: Tue, 24 Aug 2004 07:19:26 +0100
I saw this too on my AMD machine, but doing "make 386" instead of
"make
x86-mmx" fixed it. Maybe the AMD does not have the MMX or the
compiler
does not do it right for that architecture?
No, I wrote the core using an Athlon-XP, so it should work on any
machine which supports MMX. It looks like there's a bug (which would
show up on any MMX machine), which I should track down and fix.
Here we are - a patch against 1.03pr4 which fixes the MMX cores. It
seems to be a compiler bug to do with pointer aliasing, which for some
reason got masked during my usual testing but was exposed by the
integration process.
This patch also adds a "generic" target to the Makefile, to allow easy
compiling on platforms other than x86 and PPC. Additionally, the
benchmark is rejigged to a larger workload, since it is no longer used
in production and a more accurate result is desirable.
Just for kicks, I got someone with a G5 to benchmark for me. His
Xserve (I believe all Xserves are twin 2GHz jobs) gets over 9 million
attempts per second using a single thread, using either of the
double-pipe Altivec cores. If two threads were used to fill up the
CPUs, that translates to nearly 20 20-bit tokens a second, or an easy
24-bit token every second.
--------------------------------------------------------------
from: Jonathan "Chromatix" Morton
mail: chromi@xxxxxxxxxxxxxxxxxxxxx
website: http://www.chromatix.uklinux.net/
tagline: The key to knowledge is not to rely on people to teach you it.
Other related posts:
