[SyneRBI/SIRF] test PET__PYTHON_ALGEBRA is very slow (Issue #1013)

Kris Thielemans

unread,

Oct 18, 2021, 5:37:39 PM10/18/21

to SyneRBI/SIRF, Subscribed

It takes 120s on my machine. Would be worth finding out which bit is so slow (or is it everything?)

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub, or unsubscribe.
Triage notifications on the go with GitHub Mobile for iOS or Android.

Evgueni Ovtchinnikov

unread,

Oct 19, 2021, 6:00:31 AM10/19/21

to SyneRBI/SIRF, Subscribed

@KrisThielemans out of 39 unit tests, 26 work with two AcquisitionData objects (stored in file in 13 tests) currently of size 357 x 126 x 344 = 15 473 808 - with my_forward_projection.hs it takes 20 sec on my laptop

Kris Thielemans

unread,

Oct 19, 2021, 11:51:58 AM10/19/21

to SyneRBI/SIRF, Subscribed

I guess the AcquisitionData tests still take a lot of time (hopefully only the InFile ones). It certainly makes sense to do it with smaller data, as it's just a waste of resources otherwise.

Evgueni Ovtchinnikov

unread,

Oct 19, 2021, 12:07:18 PM10/19/21

to SyneRBI/SIRF, Subscribed

PET AcquisitionData can be created from the scanner name - is it possible to create a small enough one for testing? My guess is that max_ring_diff can help - what about span and view_mash_factor?

Edoardo Pasca

unread,

Oct 23, 2021, 1:06:32 PM10/23/21

to SyneRBI/SIRF, Subscribed

To change to smaller dataset it is sufficient to replace the self.image1 and self.image2 in all the test classes.

https://github.com/SyneRBI/SIRF/blob/master/src/xSTIR/pSTIR/tests/test_algebra.py

Evgueni Ovtchinnikov

unread,

Dec 9, 2021, 12:19:07 PM12/9/21

to SyneRBI/SIRF, Subscribed

Closed #1013 via bb1c437.

Evgueni Ovtchinnikov

unread,

Dec 9, 2021, 12:20:28 PM12/9/21

to SyneRBI/SIRF, Subscribed

@KrisThielemans now takes 6 sec

Reply all

Reply to author

Forward