Closed a-corni closed 3 months ago
test_perf_sim.pdf Conclusion: no major changes on the computation time. I think that the RampWaveform is the worst case scenario for this condition on the max_step
Should I add the "false positive" simulation in the tests ? In tests/test_simresults.py
? Basically just checking that the final state is not equal to the ground state.
Should I add the "false positive" simulation in the tests ? In
tests/test_simresults.py
? Basically just checking that the final state is not equal to the ground state.
Hahaha you read my mind
Should I add the "false positive" simulation in the tests ? In
tests/test_simresults.py
? Basically just checking that the final state is not equal to the ground state.Hahaha you read my mind
Currently, the max_step is defined automotically from the Sequence if the user don't provide it. This definition was determined in this PR three years ago https://github.com/pasqal-io/Pulser/pull/225 However, it fails if a user provides one Pulse composed of a waveform (Composite) that would be equal to 0 during a long period and then has a short variation before going back to 0. This is a typical issue of qutip (and adams solver) as can be seen in the issues related to the previous PR and on this forum https://groups.google.com/g/qutip/c/slz4MCnkolw I think it's best to define the max_step from the shortest variation of the waveforms that compose a Pulse.
example of a code that fails:
To be discussed: With the new implementation, this will surely increase the number of time steps for any InterpolatedWaveform, CustomWaveform, BlackmanWaveform. I think it will make the computation more robust, which is what we want for this package.