Fellow ET folks,
As many of you know, I am now working on rewriting the Illinois group's
staggered vector potental GRMHD code for insertion into the latest ET code
as an open-source thorn called IllinoisGRMHD. The motivations for rewriting
the existing thorn are threefold:
1) Make the existing code easier to read and extend, to quickly and
efficiently do new science
2) Speed up the code
3) Make the code smaller and easier to insert into ET.
There are 3 tasks involved in getting our code ready for insertion into ET,
and I am happy to announce that the first, most difficult task is complete,
as well as Task 2A. Task 2B is underway:
Task 1) Rewrite the guts of our old GRMHD thorn, including the, Ai->Bi
routine, PPM, flux evaluation, and RHS computation routines. To speed up
the turnaround time in getting this code ready and quickly try new
techniques for optimization, I developed it as a standalone piece of
software, allocating memory for gridfunctions filled with random but
physical data (think: 10^{-2} noise atop Minkowski background), and then
going through the full reconstruction process in the new code & measuring
the speed. The code "guts" I described now take up only about 1300 lines of
code, where before they took up several times that.
Task 2A) Embed the code developed in Task 1 into a new thorn called
IllinoisGRMHD in our years-old version of ET, adding to it our primitives
solver (already open sourced) and boundary condition routines. (We will be
abandoning our old boundary condition routines once we move to the latest
ET, as ET apparently has a much nicer infrastructure for BCs.)
Task 2B) I will then write a very simple initial data thorn that sets all
input gridfunctions to have random but physical values at all gridpoints.
Using these initial data, I will then perform Cowling evolutions and
compare output from our current code with the new code. If after N
timesteps the results are identical to roundoff error, I will announce that
Task 2 is complete, and publicly release the IllinoisGRMHD thorn as an
open-source piece of software.
Task 3) Work with the ET community to embed our new open-sourced thorn into
the latest ET. I plan to work with Philipp Moesta on this task, as he has
already ported our open-sourced AMR prolongation/restriction operators into
a relatively recent version of Carpet. We will work together to embed our
work into the *latest* ET.
-= Timeline =-
The initial data thorn in Task 2B was written this past weekend, and I am
now working on comparing results from our old GRMHD thorn and the new
IllinoisGRMHD thorn, using these initial data as input for an evolution.
Differences beyond roundoff error in the evolution are regarded as bugs,
and I am deep into backtracing differences between the evolution codes at
this moment. I hope to complete this task within a couple weeks, and once
it is complete I will publicly release the thorn as an open-source module
and begin Task 3.
Please feel free to ask questions or share any feedback/suggestions.
-Zach
