RC5 & P4 = lousy Mkeys/s?

[John]

I remember putting RC5 on a P4 system a while back and it was not too happy about it. Did DC fix the P4 bug?

 

[Poof]

It wasn't a bug as far as I know. There just wasn't a core compiled to take advantage of the P4. I remember some discussions here last year of people considering working on some P4 compile things but then didn't hear anymore about it.

 

[John]

I guess RC5 is a good reason to stick with AMD on the PC side of things.

 

[Assimilator1]

Yeah ,P4's do nicely in SETI though

 

[ViRGE]

What little improvements that can be done with the P4 have for the most part been done. Its design is simply not very good when it comes to RC5, as it doesn't have the right hardware rotate instuctions natively.

 

[ElFenix]

it has the rotate instructions its just they are very high latency compared to amd and intel's 6th gen design. takes quite a few clock cycles to do whereas the 6th gen parts did it in one or 2.

 

[SpaceWalker]

ElFenix,

The P4 is the only Intel IA-32 CPU that does the rotate instruction software, unlike the other CPU's that do it in hardware. It's this software rotate emulation that causes the P4 to suck on RC5. As per DNET here

 

[GeoffS]

In my experience,

(RC5 + P4 = lousy Mkeys/s) = TRUE

Geoff

 

[BlitzRommel]

Yeah, I have a P4 1.6GHz Compaq computer that only crunches about 2.5MKey/s -- I had a PIII 933 that crunched a bit more than that. Bump against P4's for RC5.

 

[networkman]

P4 + RC5 = SUCK!