Live Q&A with Laurent Le Faucheur - DSP/ML computing libraries for IoT
Laurent Le Faucheur - Watch Now - Duration: 25:09
I also have a question about many sequential shifts and MAC. The slide on the processors calculates the most optimal times, yes? Back when I worked with processors that could reorder instructions, it seemed that we had to be careful using the raw instruction times, because the processor could skip a load (if I remember correctly) by chaining operations, and in the other direction cache misses were critical to minimize.
A potential reference, that I've only just skimmed through:
https://blog.stuffedcow.net/2013/05/measuring-rob-capacity/
(Aside: It'd be nice to have slide numbers to reference questions to :)
On the CSD slide, the final line would read A x 76 = -A<<2 + A<<4 + A<<6, correct? I assume the final line would be A x 0.594 = (-A<<2 + A<<4 + A<<6) >> 7?
CSD looks fascinating.
11:34:33 From Emanuele Ziglioli : Helium and Neon, will they be forever separate universes? 11:38:11 From Thomas Wagner : Is there any situation where implementing algorithms yourself rather than using CMSIS might still provide value? In other words are there domains where CMSIS might not be the most helpful? 11:38:12 From Emanuele Ziglioli : Thank you, didn't know you could compile CMSIS for Neon, cheers 11:42:51 From John Phillips : I applaud ARM's avoidance of memory-bound limitations in their development of the CMSIS-DSP libraries. It is an impressive effort to place these libraries on small sleepy Cortex M series devices and 16-bits also. How many MFCC coefficientss can be accomodated ? 11:45:39 From Emanuele Ziglioli : Microcontrollers with CMSIS are present in smartphones as "Contex Hub" or "Sensor Hub" (never really understood the difference). I've always been curious but always found hard just to get started. Root access on Android was needed for example. Vendors like Qualcomm do things differently, with their own embedded fixed point DSP. Do you know anything about software tools for Sensor Hubs? 11:57:41 From Emanuele Ziglioli : Merci beaucoup!
Yes, the renormalization ends with an arithmetic right shift.
Cortex-M have a short pipeline with internal bypass, the compiler takes care of instruction reordering when needed.