Synth using m4/m7

#1

Just out of curiosity, are there any commercial synths (or FX) using the ARM cortex M4/M7 ?

#2

Waldorf Rocket - STM32L151
Waldorf Streichfett - ???
some Mutable Instruments modules - STM32xxxx
LXR - STM32F4
Hoxton Owl - STM32F4
Teenage Engineering Pocket Operator - EFM32

Older manufacturers seem to stay with their development/investment in ASICs.

#3

Also PreenFM2 - STM32F405

#4

I haven't seen yet people use the M7, which claims to be 50% faster, quite a bump! Any idea if the M7 may make its way into the axoloti in the near future? I've seen several demo patches hitting the cpu limit on the axoloti this is why I'm a bit hesitant to get one...

#5

Axoloti with an M7 is not around the corner. The M7 architecture relies on caching and some small 64bit wide memory regions to get the speed increase, which is not so easy to benefit from in general.

#6

Thanks for replying! Their marketing team made me believe otherwise... oh well, that explains the lack of adoption of the M7. :smile:

Is there a way to estimate how long it would take axoloti's to process a 1000x1000 matrix where computations involving the immediate 8 neighbors of each entry? This would need to be generated at each sample of each block.

Computations involve multiplication and summing. I'm interested in physical modeling and how feasible it would be to develop my own objects using the axoloti frame work.

#7

So 1000x1000x9x48000 multiplies and adds per second, that's 432 Giga-Multiply-Accumulates per second??? That is high-end FPGA or GPU territory. Axoloti Core is far from that.

I do have dreams about Axoloti targeting FPGA, or GPU, but really, don't hold your breath...

The art of audio dsp programming is finding a good balance between accuracy and computational load. I suggest doing non-realtime simulations first, and then try to reduce it drastically with a minimum impact on the result.

I believe Axoloti is capable of one dimensional digital waveguides, but in 2D it would get extremely limited.