Does the cortex have any sort of onboard gpu or chip capable of parallel computing?

I ask because when implementing an advanced kalman filter, I found it took too long to calculate it’s results, even after using dedicated matrix math functions that are built in instead of while loops. My question is if there is any sort of parallel computing chip that is onboard that I can use to do matrix math quicker with?

what language are you using? I ask because RobotC is an interpreter I believe, or at least one level of abstraction from the core. Whereas, PROS and Convex are direct compiles. Should be faster. I don’t use PROS or Convex but believe this to be true.

I don’t believe so. Technically, there are 2 processors, but the master processor is not there to run user code. VEXU teams use external microcontrollers for this generally.

You can’t get a simple kalman filter to run realtime in RobotC, let alone an extended kalman filter. I haven’t tried doing it in PROS, though. You will have better luck getting rid of the VM.

I don’t imagine the jump to pros will make enough of a difference. You really need a couple orders of magnitude improvement not 2x ish.

What performance are you seeing (ie. how long does it take to run) ?
Perhaps post the code and we can make suggestions for optimization.

I thinking of more than the core PROS speedup, though I worded it badly. When you have C and C++ support, you can cross-compile math libraries like Eigen that use lazy evaluation for speedup.

That’s fair. Real implementation of matrix operations are going to be a lot more efficient than naive stuff in ROBOTC.

What was your performance when you tried to implement one in ROBOTC? For comparison?

It took around 18 ms to compute on the PC emulator, I never tested on a real cortex after seeing that time.

I beg to differ with you on the performance of PROS over RobotC. I converted the Whetstone Floating point Benchmark to run under RobotC and PROS and PROS was nearly 6 times faster. https://en.wikipedia.org/wiki/Whetstone_(benchmark)

Make sure you are doing any of your math in seperate tasks in either language and it will probably multithread the 2nd, 3rd,… tasks.

We are not disagreeing with you. We know PROS is faster.