Two things worth reading in this context: - Tony Hoare's book on CSP (Communicating Sequential Processes); this was the origin of both the transputer and the foundation for the XMOS design. - Dave May's description of the/his Xmos architecture design, download from the xmos site; he is one kool kat; superbright and a gentleman of a character.
while its true davide may is the man, and his transputer vision has inspired many,alas it seems the actual xmos/xcore executives as rather limited in their view, as in fixating on USB2 audio applications only,.and pricing the actual 16 core dev boards into outer space ;o( for a time, way back, i tried to convince the xmos people at the trade shows to finally start providing self contained
while it's true david may is the man, and his transputer vision has inspired many, alas it seems the actual xmos/xcore executives today, (as i understand it, dave has no real input there today!) are rather limited in their view , as in, fixating on high end USB2 audio AVB applications only. and pricing the actual 16 core dev (multi mike etc) boards into outer space ;o( for a time, way back, i tried to convince the xmos people at the trade shows to finally start providing self contained
Nice video. How is its DSP capability? Is it as powerful as NXP DSP56K Symphony Processor? I am looking for a chip for my convolver that has to carry out 16 low latency convolutions on 8 channel 24 bit / 96khz audio signal.
+valleynomad Thank you. To be honest I don't think I'm qualified to answer your question. I had a quick look at the XMOS because it was an interesting approach back then (and still is). I have a bunch of projects where it probably is a good fit, but they are all still in a very early stage and will remain there for quite some time. I therefore have no practical experience on demanding real-world applications with this hardware and giving advice would be unfair. That said, I found the XMOS forums to be an excellent resource for those kinds of questions. I also came across this which might allow you to take a first guess on whether this is the hardware you're looking for: www.xmos.com/download/private/AN01011%3A-DSP-performance-on-XS1-L-devices%281.0.1rc1%29.pdf www.xmos.com/download/private/lib_dsp-[userguide]%282.0.0rc2%29.pdf www.xmos.com/download/private/AN00209%3A-xCORE-200-DSP-Elements-Library%282.0.0rc2%29.pdf I hope that helps at least a little. All the best and good luck with the project!
Create video dude. I surely will continue to watch the others. I had to make the decition to use a FPGA or a uC. When i found xMOS I get exited, cause i can program my solution with C and some sugar on it. Endly i conclude i should use a FPGA (real-time visual algorithms). It wasn't a very informed decition, it had to be take quikly. I got no time to truly understand xCORE technology. Greetings.
Hello and thank you for your comment! I am currently a bit busy with the spectrum analyzer series, but I want to keep the regular videos on other electronics topics going as well. Do you have something in particular that you would like to see? On the FPGA vs. microcontroller issue that you mentioned: I have a little bit of background in vision algorithms and from what I can tell, using a normal uC will not be sufficient for most applications. It is true that the XMOS seems to bring some of the "FPGA goodness" without actually having to go the FPGA way. Upon closer inspection however, this might be good enough for some vision-based applications, but I don't think that this can replace FPGAs in general. (I hope my video didn't imply something different). When it comes to vision, some of the newer algorithms massively exploit parallelism, and on the XMOS, the degree of parallelism that you can achieve is still much less than what you can do on an FPGA, presuming that your algorithm can be sliced nicely in the first place. I see some potential for the algorithms that work well in a pipeline, but again nothing you could not do on an FPGA. Also with the chip on the XMOS StartKit, I found it to be a bit of a hack to move a lot of data from the chip to a computer for further processing -- this is one of the major drawbacks of the kit in my opinion. Thank you again and best regards Matt