There is no difference in the application versions. However, the set of the workunits changed (previously "G2", now "V2"). These were timed on the CPU to run about the same time, however on the GPU they seem to behave very differently. For now I will adjust for it by doubling the credit, we should certainly factor in that runtime difference in future setups.
This is weird - apparently in all other cases credit was granted properly when fixing the validate errors, but jut in this case it wasn't. This certainly need some manual work.
There is no difference in the application versions. However, the set of the workunits changed (previously "G2", now "V2"). These were timed on the CPU to run about the same time, however on the GPU they seem to behave very differently. For now I will adjust for it by doubling the credit, we should certainly factor in that runtime difference in future setups.
There is indeed a part of the computation which efficiency is actually in opposite directions on CPU and GPU (with larger input size on the CPU it gets more efficient, while on the GPU it gets less efficient). There's, however, not much we can do about it during a "sub-run". "V2" will be finished next week, though, we will continue (on the GPUs) with an extension of "G2", which shouldn't have this problem.
I'm running 1 task on a 1050
)
I'm running 1 task on a 1050 and I do also see a few invalid results again. Zero invalid results before.
The runtimes have about doubled.
There is no difference in the
)
There is no difference in the application versions. However, the set of the workunits changed (previously "G2", now "V2"). These were timed on the CPU to run about the same time, however on the GPU they seem to behave very differently. For now I will adjust for it by doubling the credit, we should certainly factor in that runtime difference in future setups.
BM
Here is a good example of one
)
Here is a good example of one that failed on multiple computers.
https://einsteinathome.org/workunit/428854168It was a very productive day,
)
It was a very productive day, I got a bunch ,ore validate errors.
That was indeed a problem in
)
That was indeed a problem in validation. Should be fixed now. "Validate errors" should have been corrected and credit granted.
BM
Betreger wrote:Here is a good
)
This is weird - apparently in all other cases credit was granted properly when fixing the validate errors, but jut in this case it wasn't. This certainly need some manual work.
BM
Fixed.
)
Fixed.
BM
It seems to be so, now we can
)
That seems to be so, LOL now we can wait for the next bug to surface.
Bernd Machenschalk
)
There is indeed a part of the computation which efficiency is actually in opposite directions on CPU and GPU (with larger input size on the CPU it gets more efficient, while on the GPU it gets less efficient). There's, however, not much we can do about it during a "sub-run". "V2" will be finished next week, though, we will continue (on the GPUs) with an extension of "G2", which shouldn't have this problem.
BM
Are there no longer any CPU
)
Are there no longer any CPU based GW tasks to run? I haven't received any for over a week. I am however surprised that no-one else has said anything.