You are not logged in.
Hi
i was wondring how much is the processing time for tranmitting through a relay node.for me its taking great deal of time as from results.
pls sugest me some ways in which i can reduce this time.
I tried to work on rxPacket.i make som changes as i receive it and than jst send it.nocopy payload etc.bt m not able to achieve the desired result(reduce time).
i also tried to complete above task in rxphygood handler not going to upper layes but it gave error on generating linker script all zeros.
Thanks
Offline
You can't implement a zero-copy scheme in the old reference design. The Tx and Rx packet buffers are distinct dual-port RAMs inside the PHY and can only be accessed via the OPB. Zero-copy payload processing is possible in the new reference design, where the packet buffers live in a single PLB_BRAM core.
Offline
if i jst wan to forward the pckt that i receive frm air in OFDM_TXRX_MIMO_RXPKTBUFFER
why cant i jst use the (src) as OFDM_TXRX_MIMO_RXPKTBUFFER_BASEADDR+6*sizeof(int) in the call below
memcpy(OFDM_TXRX_MIMO_TXPKTBUFFER_BASEADDR,(src),(len))
in warpmac_sendofdm.
Thanx.
Offline
You can, but that memcpy is going to take a very long time. It gets compiled as a series of load/store instructions, copying every 32-bit word via the OPB through the PPC's register file. At one point, I measured the OPB as requiring 18 OPB_CLK cycles per 32-bit copy to/from the PPC. So to copy a 1500 byte packet from OPB to PPC back to OPB would require 250+ usec, an eternity on MAC timescales. This is why we re-architected the hardware design with a unified packet buffer on the PLB.
Offline