Very slow simulation and high memory use for modified FRQI circuit with qudits (4 ququarts + 1 qubit)

[Keno-00]

Hello, thank you for creating MQT.Qudits!

I am experimenting with a modified FRQI-style encoding that uses qudits. The circuit encodes a 16×16 grayscale image using four ququarts for the pixel index and one qubit for the intensity register. The construction uses several controlled RY gates for the amplitude encoding step.

My issue is that the simulation becomes extremely slow and consumes large amounts of RAM. A 16×16 image already pushes my system near out-of-memory conditions when using tnsim. The misim backend appears incompatible with controlled RY gates, and the fakebackend simulation takes around 23 hours to complete a single run with 1 million shots. I would like to understand if this behaviour is expected for the current simulator architecture or if there are recommended techniques to manage memory or restructure the circuit to keep the simulation tractable.

System information:
• Windows 11
• AMD Ryzen 5 6600H (Nitro V15), 16 GB DDR5 RAM
• Python 3.13.7
• mqt.qudits 0.4.0

Any guidance on how to optimise this or confirm the expected limits of the available backends would be greatly appreciated.

[KevinMTO]

Hey I contacted you on Linkedin, we can have a chat and you can show me exactly what's the problem :)

[KevinMTO]

I will look into this as soon as possible and give a solution.
Understanding also the application would help drive the debugging :)

[Keno-00]

Thank you. I have sent you a message on LinkedIn. Please check it out when you have the time.