![]() |
compiled msieve on PS3
hi all,
i just want to announce i successfully compiled msieve on the playstation3 (on yellow-dog linux) using a single PPE i got a 272 bits number factored in 37 mins (its not canned heat but its ok) for compiling you have to fix the file dd.h (because no ppc_intrinsics.h on PS3) starting at: #if defined(__GNUC__) && (defined(__powerpc__) || defined(__ppc64__)) //#include <ppc_intrinsics.h> /* * fmsub - Floating Multiply-Subract (Double-Precision) * * (a * c - b) double precision */ static inline double __fmsub (double a, double c, double b) __attribute__((always_inline)); static inline double __fmsub (double a, double c, double b) { double result; __asm__ ("fmsub %0, %1, %2, %3" /* outputs: */ : "=f" (result) /* inputs: */ : "f" (a), "f" (c), "f" (b)); return result; } regards |
[QUOTE=bigmic;102428]
i just want to announce i successfully compiled msieve on the playstation3 (on yellow-dog linux) using a single PPE i got a 272 bits number factored in 37 mins (its not canned heat but its ok) [/QUOTE] Very cool. Just to be pedantic, the compiled binary would not be using the special Cell processing engines everybody is drooling over, but rather the 64-bit PowerPC core. Once in a while I think about what it would take to make sieving SIMD-friendly, but any of the ideas I get would require major code transformations to even test. jasonp |
haha :-)
hey - we got the toy for 3 days now in ancient europe :-) so first tests with the spe's are made - i also don't expect fast progress but some stuff is already working with spe support (coded some RC4 code running on SPE's :-) cu |
| All times are UTC. The time now is 01:30. |
Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.