Question: Will anyone be able to speed up this code even more? Opportunity to coauthor a paper. A new largescale sparse DAE and stiff ODE Solver

A robust largescale index-1 DAE (and stiff ODE) solver has been developed in Maple.

PS: These codes run only with the Java version of Maple. Codes are saved as *.mws for the ease of typing. One can download the files and save them as *.mw as well.
Details about our approach are provided in a paper just submitted. 
See arxiv at
https://arxiv.org/abs/2212.02630

We encourage everyone to test these codes and report bugs. All the examples can be run with Intel's Pardiso (provided users have access to libraries) by calling DAESolverP.txt and by calling "IMPDAEP" instead of "IMPDAE". This is useful for large-scale problems. The symbolic capability and ListTools search capability of Maple are very good and can be used for developing optimization solvers as well.

I would like to know if CPU time/memory usage can be reduced significantly. In particular,  for examples 5 and 6. Some ways to contribute include

(1) Running the code in evalhf or compiled form. This may be hard.
(2) Providing options to run other parallel open-source linear solvers (eg., MUMPS).

(3) Other examples that show the use of the developed solver. We are able to solve > 100,000 DAEs.

(4) Helping in converting the code to Maple 14 or earlier (by doing sparse LU Decomposition. Just using LinearSolve will slow down the code).

Please avoid ~,*, etc (shortcuts) unless it improves the speed of calculation.
Thanks 
Dr. Venkat Subramanian

Download examples_1_3.mws

Download example_4.mws

Download example_5.mws

Download example_6.mws

mw format files are given below
Download examples_1_3.mw

Download example_4.mw

Download example_5.mw

Download example_6.mw

DAESolver.txt

DAESolverP.txt