I have attached the valgrind results for a 0 second simulation and a 100 second simulation. I have also attached the backendDAEinfo logs.

I am using v1.9.4-dev.beta2.120+g23cee7b at the moment. I am aware that the amount of memory leaked is very little, but the 4.3 MB was after 10 seconds of simulation time. This can become much greater when the simulation time is extended.

I have made the alterations to the post to show the Valgrind screenshot.

When simulating any model in OpenModelica, the RAM usage increases throughout the lifetime of the simulation. The problem is that if a model simulates for too long a time, all the RAM may be used causing the machine to crash.

I used Valgrind to identify the memory leak issues. It quickly becomes clear that there are some severe memory leaks present in the code if a large model is being simulated. I have uploaded the Valgrind summary of a model that I ran for 10 seconds simulation times, indicating that memory leaks are present in the code. I believe this may be due to the fact that some variables are not deallocated (freed) correctly.

Is anyone aware of the QSS solvers (http://sourceforge.net/projects/qssengine/) by Joaquın Fernandez & Ernesto Kofman? They have a complete range of QSS solvers as well as 'n MicroModelica framework to utilise these solvers. Has there been any interaction between the development team at OpenModelica with the team working on MicroModelica?

The solver is an Linearly Explicit QSS solver that we are working on. We were planning on implementing this solver ourselves. I believe this solver would be an adaptation of the existing QSS with a second order approach. If we could collaborate on working on the existing QSS solver or creating and implenting a new solver, that would be much appreciated.

I have been working on a new QSS solver for ODE systems. The issue that I am confronted with is editing and packaging the code in order for OpenModelica to utilise the code to solve ODE systems. I am currently not sure if there is any standard guide on how to add a solver. Also, does the solver need to be written in C or C++?

Hi

The above mentioned branch "https://openmodelica.org/svn/OpenModelica/branches/cuda_codegen" has been closed. Is there any alternative/updated branch making use of GPU capabilities?

Regards
Anton