Hi Sylvain,

You are absolutely correct.  We are looking at splitting the
axi_packet_gate into a buffer and a packet gate.  That should effectively
reduce the address FIFO to a reasonable size and reduce the resource usage
back to reasonable levels.

Regards,
Michael

On Thu, Oct 4, 2018 at 4:10 AM Sylvain Munaut <246...@gmail.com> wrote:

> Hi Michael,
>
> > Thank you for letting us know.  That is quite a jump in utilization.  We
> will take a look when we have some time.
> >
> > We did recently make recent bug fixes in axi_packet_gate.  It looks like
> that, and possibly other changes, inadvertently increased resource
> utilization more than expected for B200.  Our priority is always first to
> make sure the FPGA images successfully build with the bug fixes so we can
> get those bug fixes out to users as soon as possible.  We do try to
> minimize resource utilization, but that is a secondary priority.
>
> AFAICT, it's all due to the new "address FIFO". But that FIFO seems to
> be setup to be unreasonably large.
>
> I mean it can store as many addresses as data words ... while AFAICT
> it only has to store one address per packet in the buffer, so reducing
> it to say 32 entries should be enough. This way 6 bits can fit in 1
> CLB using 32x2 distrams instead of using block rams.
>
> Cheers,
>
>     Sylvain
>
_______________________________________________
USRP-users mailing list
USRP-users@lists.ettus.com
http://lists.ettus.com/mailman/listinfo/usrp-users_lists.ettus.com

Reply via email to