Fix periodic boundary bug in #3199 #3243
Conversation
|
I had setup |
|
Maybe |
|
In that case, it seems that we need to change the code in EDIT - crossed in the air with Weiqun. |
| if (p.pos(idim) > rhi[idim]) { | ||
| p.pos(idim) = rhi[idim]; | ||
| if (p.pos(idim) >= rhi[idim]) { | ||
| p.pos(idim) = std::nextafter(p.pos(idim), rlo[idim]); |
There was a problem hiding this comment.
Yes, this looks like the right thing to do.
|
Hi, I think we have a similar test in pyAMReX that breaks with GNU 10.1 now. |
|
There seem to be also a series of regressions in WarpX: |
There was a flaw in amrex::Geometry::computeRoundoffDomain. If probhi is close to zero, std::nextafter towards negative infinity will be a very small number (e.g., 1.e-300). So the function will be able to find the lowest value of roundoff_hi that's outside the domain within a reasonable number of iterations. In the new implementation, we use bisection to find the lowest value of roundoff_lo that's inside the domain and the highest value fo roundoff_hi that's inside the domain, up to a tolerance. X-ref: - #3243 - #3199
PR AMReX-Codes#3199 broke the Nyx regression tests. The issue is we require the particle positions to be in [plo, phi). This won't happen if rhi is exactly equal to phi. The fix is to make sure rhi is always slightly inside the domain. EDIT - fixed this in a different way. Now, when applying periodic boundaries, the particle positions will be set to be strictly less than rhi.
There was a flaw in amrex::Geometry::computeRoundoffDomain. If probhi is close to zero, std::nextafter towards negative infinity will be a very small number (e.g., 1.e-300). So the function will be able to find the lowest value of roundoff_hi that's outside the domain within a reasonable number of iterations. In the new implementation, we use bisection to find the lowest value of roundoff_lo that's inside the domain and the highest value fo roundoff_hi that's inside the domain, up to a tolerance. X-ref: - AMReX-Codes#3243 - AMReX-Codes#3199
* Rework handling of roundoff domain extent (AMReX-Codes#3199) * Fix periodic boundary bug in AMReX-Codes#3199 (AMReX-Codes#3243) * Fix Roundoff Domain (AMReX-Codes#3247) * Add Tests/RoundoffDomain (AMReX-Codes#3248) * Roundoff errors in computeRoundoffDomain (AMReX-Codes#3255)
PR #3199 broke the Nyx regression tests. The issue is we require the particle positions to be in [plo, phi). This won't happen if rhi is exactly equal to phi. The fix is to make sure rhi is always slightly inside the domain.
EDIT - fixed this in a different way. Now, when applying periodic boundaries, the particle positions will be set to be strictly less than rhi.
The proposed changes: