Switched from S40 to S40.04, and it seems indeed to be 2.5 - 3% faster than S40.--lox
I thought it was only supposed to be a bug fix? When I get a chance I'll do a same-result comparison under controlled conditions. Meanwhile, I'll just install it. I did get zero credit claimed once in several hundred results processed on S-40 or S40.03, don't remember which.
Switched from S40 to S40.04, and it seems indeed to be 2.5 - 3% faster than S40.--lox
I thought it was only supposed to be a bug fix? When I get a chance I'll do a same-result comparison under controlled conditions. Meanwhile, I'll just install it. I did get zero credit claimed once in several hundred results processed on S-40 or S40.03, don't remember which.
S40.04 isn't faster than S40.03. I changed only the value of a constant.
(A multiplication was inaccurate.)
It might not be faster than S40.03, but IS FASTER THAN S40 ON MY BOX, as I wrote before. Anyway, I have not used S40.03 yet and am surprised by another (slight) improvement in comparison with S40.
And, besides all that, I was vain enough to start an "xxxx observation thread", there might not come a lot more opportunities to do so. ;)
Akosf, you might be interested in this comparison between two AMD Athlon XP3000+ machines with 1Gb RAM working the same WU: Standard app.=16598 secs......your S40.04=3708 secs. That's a 4.48x improvement!!! You have more than achieved your goal of a 4x improvement! You are truly a genius and an amazing resource for the E@H project. You must have rare insight into the inner workings of the various CPUs, mathematics, and hand coding to have achieved so much improvement. I salute you! We now have a chance to 'crunch' the data being produced by the project in a timely manner. Thanks for a great effort....Cheers, Rog.
Akosf, you might be interested in this comparison between two AMD Athlon XP3000+ machines with 1Gb RAM working the same WU: Standard app.=16598 secs......your S40.04=3708 secs. That's a 4.48x improvement!!! You have more than achieved your goal of a 4x improvement! You are truly a genius and an amazing resource for the E@H project. You must have rare insight into the inner workings of the various CPUs, mathematics, and hand coding to have achieved so much improvement. I salute you! We now have a chance to 'crunch' the data being produced by the project in a timely manner. Thanks for a great effort....Cheers, Rog.
Thanks!
edit: I'm working on S40.07 and my new cpu and mobo are coming. :-)
S40.04 Observation thread
)
I thought it was only supposed to be a bug fix? When I get a chance I'll do a same-result comparison under controlled conditions. Meanwhile, I'll just install it. I did get zero credit claimed once in several hundred results processed on S-40 or S40.03, don't remember which.
RE: RE: Switched from S40
)
S40.04 isn't faster than S40.03. I changed only the value of a constant.
(A multiplication was inaccurate.)
It might not be faster than
)
It might not be faster than S40.03, but IS FASTER THAN S40 ON MY BOX, as I wrote before. Anyway, I have not used S40.03 yet and am surprised by another (slight) improvement in comparison with S40.
And, besides all that, I was vain enough to start an "xxxx observation thread", there might not come a lot more opportunities to do so. ;)
--lox
Akosf, you might be
)
Akosf, you might be interested in this comparison between two AMD Athlon XP3000+ machines with 1Gb RAM working the same WU: Standard app.=16598 secs......your S40.04=3708 secs. That's a 4.48x improvement!!! You have more than achieved your goal of a 4x improvement! You are truly a genius and an amazing resource for the E@H project. You must have rare insight into the inner workings of the various CPUs, mathematics, and hand coding to have achieved so much improvement. I salute you! We now have a chance to 'crunch' the data being produced by the project in a timely manner. Thanks for a great effort....Cheers, Rog.
RE: Akosf, you might be
)
Thanks!
edit: I'm working on S40.07 and my new cpu and mobo are coming. :-)
RE: edit: I'm working on
)
Is this another tweak to the current app, or the 3dnow/SSE3 joint app?
RE: RE: edit: I'm working
)
S40.08 is a simple SSE code...
I will use a new letter (X) for combined 3DNow/SSE3 code.
edit: I have 0.0 credit problem again. :-) more fun :-)
edit2: I did a subtraction twice in every cycle. :-)
I found convergence in the
)
I found convergence in the power...
RE: edit: I have 0.0 credit
)
I got a 0 credit result too, good luck with finding the problem, if it's even yours. ;)
cu,
Michael
edit: 2 more 0 credit results.
25 min short wu´s ? Pentium4,
)
25 min short wu´s ?
Pentium4, 2.6Ghz needs 1403,11 sec/WU with akosf S40.04
--> http://einsteinathome.org/task/25562586
--> http://einsteinathome.org/task/24734892
really surprised :-)
thx akosf