Until #1166 / #1404 / #1480 / #1505 / #1506 / #1509 are resolved, we'll allocate all available memory for each aCompCor_cosine_filter thread so that such nodes will necessarily run one-threaded and won't overrun memory as long as the total allocation is sufficient
Project sub-parts and estimates
Enable throttling ①
Set throttling on aCompCor_cosine_filter ①
Test throttled nodes ⑤
Timeline notes
No response
Timesink notes
No response
Frustration notes
No response
Design decisions and notes
[ ] Throttle all cosine_filters or just aCompCor?
[ ] Possibly also throttle calc_ants_warp ― it's been consistently problematic with overrunning its memory allocation (#1308)
User interface changes
This node is using way more than its allocated memory again, this time with the
fmriprep-options
preconfig on some HCP data.Until #1166 / #1404 / #1480 / #1505 / #1506 / #1509 are resolved, we'll allocate all available memory for each
aCompCor_cosine_filter
thread so that such nodes will necessarily run one-threaded and won't overrun memory as long as the total allocation is sufficientProject sub-parts and estimates
aCompCor_cosine_filter
①Timeline notes
No response
Timesink notes
No response
Frustration notes
No response
Design decisions and notes
cosine_filter
s or justaCompCor
?calc_ants_warp
― it's been consistently problematic with overrunning its memory allocation (#1308)cal_DVARS
? (#1301)