Loading ...
Sorry, an error occurred while loading the content.

Re: SL on Raspberry Pi?

Expand Messages
  • vbifyz
    ... This is why I mentioned DosBox. It would serve as the x86 emulation layer. Another x86 emulation alternative I know about is QEMU. I realize the
    Message 1 of 5 , Apr 30, 2013
    • 0 Attachment
      --- In SpectrumLabUsers@yahoogroups.com, wolf_dl4yhf <dl4yhf@...> wrote:
      >
      > Utterly impossible, because SL is written in Borland C++ Builder (using
      > Borlands VCL) - no chance to produce ARM code with it.
      > The ARM target would have to emulate x86 instructions.

      This is why I mentioned DosBox. It would serve as the x86 emulation layer. Another x86 emulation alternative I know about is QEMU.
      I realize the performance will be poor, but it may be ok for some applications.

      73, Mike
    • wolf_dl4yhf
      ... Ok yes, a CPU emulator would be ok for less CPU-intense applications, but not for mainly number-crunching software like SL. Also the target CPU definitely
      Message 2 of 5 , Apr 30, 2013
      • 0 Attachment
        Am 30.04.2013 23:38, schrieb vbifyz:
         



        --- In SpectrumLabUsers@yahoogroups.com, wolf_dl4yhf <dl4yhf@...> wrote:
        >
        > Utterly impossible, because SL is written in Borland C++ Builder (using
        > Borlands VCL) - no chance to produce ARM code with it.
        > The ARM target would have to emulate x86 instructions.

        This is why I mentioned DosBox. It would serve as the x86 emulation layer. Another x86 emulation alternative I know about is QEMU.
        I realize the performance will be poor, but it may be ok for some applications.


        Ok yes, a CPU emulator would be ok for less CPU-intense applications, but not for mainly number-crunching software like SL.
        Also the target CPU definitely needs a floating point unit (in hardware, not emulated) which adds to the complexity.
        Re-writing the DSP stuff from floating point to integer just to run it on a small ARM would be overkill. There are some nice optimized fixed-point (integer) FFT algorithms which look pretty neat, but all the rest (GPS-timestamp-based processing, filters implemented in the time domain, etc) calls for a CPU with a bit more 'kick'. For my applications, the BeagleBone would be the first choice for an embedded system running some of the DSP functions, for example as a 'remote frontend'. But I wouldn't even call that a 'future plan' right now, due to the ever present lack of time.

        All the best,
          Wolf .

      Your message has been successfully submitted and would be delivered to recipients shortly.