MS Windows Beta Test App 4.24 available
3 Aug 2006 15:49:12 UTC
Topic 13403
(moderation:
A new Beta Test App is available on our usual Beta Test Page.
It features a codebase similar to the Linux 4.17 Beta Test App, with hand-coded optimized assembler calculation from Akos (and me).
See how much this speeds up things in general.
I haven't seen a single invalid result from the Linux Apps with the same analysis code, so I'm pretty sure this App won't cause any problems. Please report invalid results ASAP.
Happy crunching!
BM
BM
MS Windows Beta Test App 4.24 available
)
Would really be courious to know how the 4.24 compares to the 4.17 on the same machine. If I read you results right, there's still a big difference. Is this consistent with wall-clock time or is there a problem with reporting the CPU-time?
BM
BM
RE: Validate error: result
)
Thanks for the report.
Apparently the search finished successfully, but the uploaded result file wasn't compressed (as it should be) and the file looks truncated. The file, however, is much longer than it should be (for a short WU).
The other results this machine has returned, even from the same charge of WUs (h1_0324.5_S5R1) look ok and are valid so far. Any idea what has gone wrong? Did you temporary ran short on disk space?
BM
BM
RE: Just installed 4.24,
)
It probably is. All "work" (tasks) that has once been assigned to 4.02 will be finished with that App. New tasks should get assigned to the 4.24.
BM
BM
RE: This should get changed
)
Give me some time to take a closer look at the results from these Apps. I think the Apps will be made official ones some time at the end of next week (or shortly after the weekend), if nothing suspicious shows up. We will then adjust the credit shortly after that, too, to stay conformant to the other projects.
BM
BM
RE: Hi everyone, first of
)
Well, I promised to work on that...
That's good news.
Thanks!
BM
BM
RE: Well, see fo yourself.
)
I found "infinity" numbers in the result files the machine sent back, i.e. in the ones already marked as validate error and in a few your host sent back after them, too. However recent results from this machine (l1_1014.0_S5R1__285_S5R1a_0, id=38625736) look fairly ok. Looks like a (temporary) overheated CPU to me.
BM
BM
RE: Makes you wonder about
)
How many of them have you met? How many compilers have you written on your own? If I would have to write a compiler for this worse-of-all x86 architecture, I would really be happy if this would give me anything working (and conforming to the standards) at all.
I've not seen any speedup with the current code on the Intel Mac where we use Apples's modified gcc, but actually a slight slowdown. Seems that this compiler knows the CPU better than we.
BM
BM
Bruce has made the Beta Apps
)
Bruce has made the Beta Apps 4.17 and 4.24 official. You can now remove the app_info.xml and restart the client to switch back to the official, auto-update path.
Thanks a lot for participating in this Beta Test!
BM
BM
RE: RE: Makes you wonder
)
Sorry for having been that harsh. But I don't like to talk bad about people which I don't know, or to judge their work when I don't know the problems they're facing at it. And I'm especially sensitive when these people are of my kind (programmers).
BM
BM
RE: Could you get your
)
I doubt that. It didn't help much with earlier code, and it refuses to compile the current one "cannot allocate registers for asm statement". I'll plan to look into that later this week again.
Edit: I expect a bit more from the SSE2 stuff I'm coding right now, but I may also be wrong there. Thanks to Akos, we have now reached a level of optimization where the effects of any change (on the speed) are hard to predict - just try and see. And it gets very different for different CPUs. Luckily all the Intel Macs have (more or less) the same.
BM
BM