Things move pretty fast here. I woke up this morning to find out that there's a new stable version (S5S0007) and a bunch of different test versions. My current work unit has about 2 hours or so to go using the S5S0003 patch, so I'm going to let it finish out before applying a new patch. The total estimated completion time on this is roughly 14.75 hours, and it's about 86% complete right now. I've got the latest stable patch, but I'll check back after my current unit finishes to see where we're at and then decide which patch to apply. Thanks for all your work akosf. These new work units are so long, I'm glad to see that you can make some optimizations to reduce the crunching time.
Same for this S5T0303result - "Checked, but no consensus yet" - now "Invalid". And, in both cases, the comparison was against a "standard app". Will try S5T0304 next.
Quote:
This result using S5T0302 is another 10+% faster BUT it is listed as "Checked, but no consensus yet" - now "Invalid"
Quote:
This is my first result using S5T0301 - an improvement of 6 or 7% over S5T0003 (validated OK).
Quote:
This is my first full result with the S5T0003 patch. And this result was a hybrid - about 10% standard app/90% patched. It appears the patch is 10 to 12% faster than the standard app on my 2.4 GHz P4 w/Windows XP Pro.
Edit: This is my "Results for computer" (for comparison).
This result is a Hybird of SST0003 and SST0304 showing initial
A savings on this WU of almost 2 hrs from stock app...40,000 instead of 47,000..am going to let this run again on the SST0304 completely before I try another...
using Pent4 2.8GHz SSE3 no HT....thanks Akos.you have that way of keeping this interesting!!!
Things move pretty fast here.
)
Things move pretty fast here. I woke up this morning to find out that there's a new stable version (S5S0007) and a bunch of different test versions. My current work unit has about 2 hours or so to go using the S5S0003 patch, so I'm going to let it finish out before applying a new patch. The total estimated completion time on this is roughly 14.75 hours, and it's about 86% complete right now. I've got the latest stable patch, but I'll check back after my current unit finishes to see where we're at and then decide which patch to apply. Thanks for all your work akosf. These new work units are so long, I'm glad to see that you can make some optimizations to reduce the crunching time.
S5T0710.dat - eliminated
)
S5T0710.dat
- eliminated double jumps
- reduced amount of FPU macro ops
- removed double loads on general purpose registers
- better SSE register usage
- reduced memory and integer register usage
- optimized branch structure
- faster FPU comparisons
- SSE3 truncation
- reordered instruction sequence
CPU: SSE3 compatible
RE: S5T0710.dat This
)
This leads to an HTML-File. ;)
cu,
Michael
[edit]Seems to be fixed now. :-)[/edit]
Just left-click, do not save
)
Just left-click, do not save target as:)
Welcome To Team China!
deleted - Yin Gang was faster
)
deleted - Yin Gang was faster
Another valid result for
)
Another valid result for S5T0709 http://einsteinathome.org/task/35066904 . My machine : CELERON D .Compared with stock app
gain aprox. 30% .Nice work :)
S5T0710 SSE3 / 2107,61 sec
)
S5T0710 SSE3 / 2107,61 sec / 35,1min / result
S5T0710 SSE3 / 2184,97 sec / 36,4min / result
very fast but the results are invalid :-/
This result using S5T0709 -
)
This result using S5T0709 - "Valid" and a 35+% improvement over S5T0301
This result using S5T0304 - still "Pending"
This result is a Hybird of
)
This result is a Hybird of SST0003 and SST0304 showing initial
A savings on this WU of almost 2 hrs from stock app...40,000 instead of 47,000..am going to let this run again on the SST0304 completely before I try another...
using Pent4 2.8GHz SSE3 no HT....thanks Akos.you have that way of keeping this interesting!!!
Ok just woke up. Quick
)
Ok just woke up. Quick question.
Is S5S0007 SSE3 Only, or can be used by SSE machines?
Come Join us at Hawaiian Beach Bums