Hi everyone,

Just wanted to update you that we've found the root cause of the slowdown and we're testing a fix now.

I hope to have 1.05 published today that vastly improves runtimes across all configurations.

For the next batch of workunits, I intend on increasing the credits to 10000 to make up for lost work due to this bug. I won't keep it at 10000 forever, but I hope this is suitable.

I appreciate your patience everyone. The light at the end of the tunnel is here! :)

For reference on runtime improvements:

1.04 runtime on config 53, 1 million iterations: ~1600 seconds
1.05 runtime on config 53, 1 million iterations: ~112 seconds

I'm testing the dynamically changing the iteration count based on the config file, since they seem to have varying runtimes if we use the same iteration count across them.

I suggest crediting the 10,000 for the finished work that took so long and also the ones that went Invalid becoming of this issue. This would be more fair as we already did the work. Crediting more on new WUs only rewards those who stopped running and then come back when things are great for them which is unfair to those who still ran WUs through this issue.

For what it's worth, there are only 88 tasks that returned past the deadline that ran for an extraordinary amount of time.

I'm not sure how I can give credit to tasks that provided no output. I'm not sure BOINC supports manually manipulating the data of the task to do so.

I can assign a credit value to those 88 tasks, but I'm not sure what else you'd like me to do given the system's constraints.

3, 4, 5 days of runtime and 5000 points...yay!

Aborted them and will dedicate my CPU to something else until it gets fix properly.

Thanks for your hard work fixing these bugs though.

3, 4, 5 days of runtime and 5000 points...yay!

Aborted them and will dedicate my CPU to something else until it gets fix properly.

Thanks for your hard work fixing these bugs though.

It's already fixed. Download the new 1.05 tasks.