Mac OS X.2 client?

[Confused]

A member of our Team has the opportunity to install SB on a few OS X.2 machines, and i'm sure there's also a lot of RC5 people who might like something good for their Macs to do now that RC5 is finished.

Any chance of this? :)

Confused

[nuutti]

seventeenorbust client uses math library that is written using 386 assembler. At the moment there does not exist such library to the other platforms and I think that doing such library is very demanding task. So I guess that not in very close future (year or two). (I am not member of seventeenorbust programming team)

Yours,
Nuutti

[CyberWire]

rc5-72 just started too :/

[shifted]

In the next few weeks, i'll be attempting to optimise the non-x86 dependent version of the code so that it runs reasonably fast. Right now, according to Louie, it's about 30 times slower. We need to get it to about 50% of the speed.

After i've got it working on Solaris, MacOS 10.2 is my next target (i've got access to fewer Mac's, so it's more worth my time to do Solaris).

However, with the rate at which we are finding primes, I wonder if i'll have a port complete before the project has finished :D

[rogue]

The fastest way to get a port would be to use GMP as was done with the original client, but it has already been stated that GMP code would be 30x slower then GWs PRP code.

A slower way to get a port would be to the FFT of Glucas or Mlucas to build a client. Clients using an FFT based on one of those FFTs would be less than 3x slower (per MHz) and might even be faster on some architectures.

[shifted]

The plan was to use GMP (the client exists, it's just really slow. i was going to try to speed it up). I will investigate those other FFT algorithms.