sr5sieve Benchmark thread
 

Athlon XP-M 2400+ (1.83GHz)

18.3 kp/s @ p = 25e9 (for 324 k's)

Pentium III 600MHz

5.1kp/s @ p=35e9

PPC G5 at 2.5 GHz:

33072 p/sec at 50e9 for 324 k

Pentium IV 3.0 GHz
11,3 kp/s @ p=30e9 for 324 k's

celeron 1.4 Ghz 512 MB
9376 P/sec p=99e9 (using 93% CPU)
324 k's

(Just testing the application, do not want to reserve a range)

AMD 64 3000+ (2GHz) - srsieve-0.3.4-mingw32-i686

21457 p/sec at 50e9 for 324 k

Carlos

[QUOTE=em99010pepe]AMD 64 3000+ (2GHz) - srsieve-0.3.4-mingw32-i686

21457 p/sec at 50e9 for 324 k[/QUOTE]
That seems a little low - can you try with the Athlon32 build as well as the P4 build?

Here you go:

AMD 64 3000+ (2GHz) 1GB

21591 p/sec at 50e9 for 324 k's - srsieve-0.3.4-mingw32-athlon
21485 p/sec at 50e9 for 324 k's - srsieve-0.3.4-mingw32-pentium4

Carlos

Oh well! Looks like clock for clock, Athlon64 is equivalent to Athlon32 for sr5sieve

[QUOTE=axn1]Oh well! Looks like clock for clock, Athlon64 is equivalent to Athlon32 for sr5sieve[/QUOTE]

The Athlon-64 speed could probably be improved by writing a suitable 64 bit mulmod in assembler and finding the right GCC options to compile it as a native 64 bit application. I don't have access to one of these machines myself, but if someone running 64 bit Linux on one wants to have a go at this I will help.

p=55e9, 324 candidates.
Srsieve 0.3.4 compiled with 'make ARCH=i686 CC="gcc -V4.1"':
Column A with 'ASSEMBLER=yes', column B with 'ASSEMBLER=no'.
Column C is sieving one candidate at a time with NewPGen 2.82.

[CODE]
A B C
Machine MHz p/sec p/sec p/sec
------------------------ ----- ----- ----- -----
P2/400 (Deschutes) 401 3470 2920 44
P3/450 (Katmai) 448 3890 3250 50
P3/600 (Coppermine) 601 5880 4740 70
P4/2.4 (Celeron 128K L2) 2880 11190 10960 100
P4/2.4 (Northwood C) 2912 12520 12200 154
-- two instances hyperthreaded 19890 18600 236
[/CODE]

It seems the assembler routines hardly benefit the P4 at all.