MG majorant precalculation for DT #8
Conversation
Updating with main repository
valeriaRaffuzzi
requested review from
Mikolaj-A-Kowalski and
ChasingNeutrons
There was a problem hiding this comment.
Looks pretty good to me. I'm trying (and struggling) to think if there's really any circumstance that we wouldn't want this as you could in principle make it optional - probably not worth the extra code. Now we just need to do the same for CE (which I'm sure may be a bit more painful...).
| xs = ZERO | ||
| do i = 1,nMat | ||
| xs = max(xs, self % mats(i) % data(TOTAL_XS, g)) | ||
| end do | ||
| self % majorant(g) = xs |
There was a problem hiding this comment.
| xs = ZERO | |
| do i = 1,nMat | |
| xs = max(xs, self % mats(i) % data(TOTAL_XS, g)) | |
| end do | |
| self % majorant(g) = xs | |
| self % majorant(g) = maxval(self % mats(:) % data(TOTAL_XS, g) |
But please check that it really does work. My Fortran is a bit rusty.
There was a problem hiding this comment.
True Fortran magic
| xs = ZERO | ||
| do i=1,size(self % activeMats) | ||
| idx = self % activeMats(i) | ||
| xs = max(xs, self % getTotalMatXS(p, idx)) | ||
| end do |
There was a problem hiding this comment.
I think it may be a considerable problem that in the initialisation active materials are not considered.
Unless I am missing something (quite possible -> Friday evening ;-) ) the pre-calculated majorant is based on all defined materials, not only the materials which are actually included in the geometry.
I would suggest moving the majorant calculation to the activate subroutine.
There was a problem hiding this comment.
Very good point. I will do that!
|
Thanks for the development! It is really nice optimisation! I absolutely love the documentation and corrections to the typos 🦭 |
|
Done! |
Precalculating the majorant is extremely simple for MG and it buys a nice 7-8% acceleration in, for example, a 2D C5G7 problem.
The majorant here is calculated in init and saved in mgBaseNeutronDatabase, instead of adding xss at every collision.