[CCPPETMR/SIRF] Objective function value wrongly displayed as zero when using STIR with openmp (#157)

evgueni-ovtchinnikov

unread,

Mar 22, 2018, 10:07:12 AM3/22/18

to CCPPETMR/SIRF, Subscribed

if STIR is built with openmp, steepest_ascent demo prints the objective function values as zeros

at the same time, optimization function scipy.optimize.fminbound seem to get correct objective function values: exactly the same values of the argument x and changes in the image after an iteration are computed as without openmp - but objective function values are reported to be 0!

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub, or mute the thread.

Kris Thielemans

unread,

May 10, 2018, 7:51:38 AM5/10/18

to CCPPETMR/SIRF, Subscribed

confirmed that this is still the case. This must be a STIR bug.

Kris Thielemans

unread,

May 10, 2018, 10:21:23 AM5/10/18

to CCPPETMR/SIRF, Subscribed

fixed on STIR master at UCL/STIR@cc7c164.

We therefore need to bump the default_STIR_TAG in the superbuild

Kris Thielemans

unread,

May 11, 2018, 11:23:02 AM5/11/18

to CCPPETMR/SIRF, Subscribed

Closed #157.

Reply all

Reply to author

Forward