Hi Laura

Probably a silly question, but do you have "Enable Large Tx Frames"
enabled as a parameter option on the 10Ge core? The ~8000 data word
limit only applies when this option is chosen, otherwise it is a _lot_
smaller. When buffer overflows happen, bad things occur, the 10Ge core
does not recover, no packets are sent, it must be reset.

Regards
Andrew

On Wed, 2012-05-30 at 10:54 +0200, Henno Kriel wrote:
> Hi Laura
> 
> 
> The 64 x 64b FIFO are switched in behind the 8KB FIFO, so this is
> actually fine.
> 
> 
> HK
> 
> On Wed, May 30, 2012 at 8:29 AM, Henno Kriel <[email protected]> wrote:
>         Hi Laura 
>         
>         I had a look at the xco coregen file for TX Fifo when the
>         "Enable Large Tx Frames" option is used (8K Packets). The size
>         of the FIFO seems to be only 64 deep!?
>         
>         I will investigate further...
>         
>         HK
>         
>         
>         On Wed, May 30, 2012 at 1:45 AM, Laura Vertatschitsch
>         <[email protected]> wrote:
>                 Hey guys,
>                 
>                 
>                 I've got several designs where I consistently clock
>                 512 64-bit samples into the ten gbe0 (v2 block) at a
>                 rate of ~14MHz.  The fpga clock is running at 200.
>                  It's my understanding that I still should have some
>                 room to get faster data off.  I went up by a factor of
>                 2 (the only change in my design) to clock the data,
>                 still only 512 samples, at a new rate of ~28MHz.  Not
>                 a single packet is sent, and the tx_afull and
>                 tx_overrun flags are consistently high.  I was
>                 surprised by this, as I am only creating 4096Bytes in
>                 the frame and I know ~8000B is the limit.  I've
>                 simulated in Simulink to be sure that the tx_valid and
>                 eof flags are being set when I want.  
>                 
>                 
>                 Could there be something else going on here?  Could
>                 the tx_overrun flag also be set in the event that the
>                 buffer cannot clear out the data before the next data
>                 point in?
>                 
>                 
>                 --Laura Vertatschitsch
>         
>         
>         
>         
>         -- 
>         Henno Kriel
>         
>         DSP Engineer
>         Digital Back End
>         meerKAT
>         
>         SKA South Africa
>         Third Floor
>         The Park
>         Park Road (off Alexandra Road)
>         Pinelands
>         7405
>         Western Cape
>         South Africa
>         
>         Latitude: -33.94329 (South); Longitude: 18.48945 (East).
>         
>         (p) +27 (0)21 506 7300
>         (p) +27 (0)21 506 7365 (direct)
>         (f) +27 (0)21 506 7375
>         (m) +27 (0)84 504 5050
> 
> 
> 
> 
> -- 
> Henno Kriel
> 
> DSP Engineer
> Digital Back End
> meerKAT
> 
> SKA South Africa
> Third Floor
> The Park
> Park Road (off Alexandra Road)
> Pinelands
> 7405
> Western Cape
> South Africa
> 
> Latitude: -33.94329 (South); Longitude: 18.48945 (East).
> 
> (p) +27 (0)21 506 7300
> (p) +27 (0)21 506 7365 (direct)
> (f) +27 (0)21 506 7375
> (m) +27 (0)84 504 5050
> 



Reply via email to