3 ms·
The PRUs are a waste of perfectly good silicon until there is a C compiler for them. I haven't seen any project outside of TI's demos that uses these in any ad
by tmuir 12y ago
The PRUs are a waste of perfectly good silicon until there is a C compiler for them. I haven't seen any project outside of TI's demos that uses these in any advanced fashion.
It sounds like an excellent idea. Completely separate processors on the same memory bus that allow higher IO speeds than the main processor. But since it can only be coded in assembly, Altera SOCs are much more attractive. Instead of cortex M4s, there is an FPGA on the same chip. Instead of making the interprocessor feature a crippled afterthought, the FPGA/ARM interface is central to the system, and the bus can be made as wide as practical. Since you can create the precise peripheral to suit your application on the FPGA, its kind of a blank slate.
Edit: As it turns out, there is now a C compiler for the PRU, 3 years after the hardware was released.
http://processors.wiki.ti.com/index.php/PRU-ICSS_Getting_Started_Guide http://processors.wiki.ti.com/index.php/PRU-ICSS_Getting_Sta...
- joezydeco 12y agoThis has always kind of been TI's thing. I worked with a group that wrote H.264 codecs running on the IVA of the old OMAP2/3 and Hollywood chips. The IVA was straightforward DSP but the "bridge" method of getting that to talk to the CPU was just...awful.
- rasz_pl 12y agoWhen I read nonsense like that I immediately think you meant C++, or java, or python. Embedded is hard, If you cant handle assembler you wont be able to handle verilog either.
- tmuir 12y agoExcept that there is no way to scale assembly. You write every instruction. There are no companies that offer prewritten assembly for different common functions. Toga vendors offer all sorts of ip. Understanding assembly and choosing to develop with assembly are two different things. I see absolutely no correlation between learning assembly and learning verilog. They are about as orthogonal as you can get.