Hi Glen, There are no specific optimisations for the STM32F4 that I can recall.
I spent a few weeks last year making the modem and codec run fast enough for the STM32F4 by re-arranging some of the building blocks at the C code level. These changes were folded back into the source code now used on all platforms. Cheers, David On 28/07/15 09:24, glen english wrote: > Hi All > I am considering retargeting codec2 / sm1000 etc for another micro . a > '21477. > and this will include a rather strong modem. > (or I might stick with stm32F7) > > In opinion, just how much of the STMF4 code for codec2 is hard optimized > for the stm? > > Without spending alot of time looking at it, I am wondering abut the > implications for me of using a different processor. IE how much stuff is > tightly coupled to the way the stmf4 works ? > > Part of the deal with getting micros like the M4 to really fly in DSP > land (and get them even close to their pure DSP brothers) are tightly > writing to the processor and compiler architectures. > > all opinions welcome. > > next step will be actually compiling for sharc and seeing what happens. > I expect an 'average' result- as the stmf4 is good for multipurpose > things, but the SHARC compilers are not. Or rather, the sharc compiler > does not in generally utilize the very rich instruction set available to > it- only about half of the very fancy instructions of the sharc are > used, leading to lacklustre performance for GP computing. > > > glen Vk1XX > > > > > ------------------------------------------------------------------------------ > _______________________________________________ > Freetel-codec2 mailing list > [email protected] > https://lists.sourceforge.net/lists/listinfo/freetel-codec2 > ------------------------------------------------------------------------------ _______________________________________________ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2
