Ok, thanks!
On March 18, 2021 1:01:41 PM UTC, Ron Economos <[email protected]> wrote: >There are no libvolk optimizations for the LDPC encoder. I don't even >think it's possible. The encoder is just a giant XOR, but the XOR >operands are not from contiguous locations in memory. > >Ron > >On 3/18/21 05:42, Adrian Musceac wrote: >> Hi Ron, >> Tangential to the subject only, do you know if LDPC has optimizations > >> for ARM archs in libvolk? Arm64 mainly. >> >> Thanks, >> Adrian >> >> On March 18, 2021 11:54:42 AM UTC, Ron Economos <[email protected]> >wrote: >> >> I haven't tried it, but it may be possible on a Raspberry Pi 4. I >> wouldn't bother with anything less than a RPi4. I have >successfully run >> a modified DVB-T2 flow graph on a Beagleboard-X15, but only at 5 >MHz >> channel width (5.714285 Msps). >> >> I've attached the flow graph. It uses short frames and 5/6 code >rate to >> limit the amount of LDPC processing. It also sends a lot of dummy >> symbols, again to limit the amount of processing. However, it >still >> attains 19.057285 Mbps at 5 MHz channel width. If you can get it >to run >> at 8 MHz channel width, the bitrate will be 30.491656 Mbps. >> >> It's also tuned for UHD v3.13.0.3-rc1. If you're using a later >version >> of UHD, change send_frame_size=65536 to send_frame_size=8192 in >the UHD >> Sink block. Using large USB frames helps to significantly lower >> processor load, but this feature has been removed from later >versions of >> UHD. >> >> Ron >> >> On 3/18/21 04:17, Ali G. Dezfuli wrote: >> >> Hi all, Is there any experience in transmitting DVB-T2 with >> gnuradio and raspberry pi (any version)? I just wonder if it >> can work real-time with about 9 MHz sampling frequency of 8 >> MHz DVB-T2. Thank you! BR, AGD >>
