On Wednesday 31 January 2007 07:43, Larry Finger wrote:
> Michael,
>
> I think the patch below is needed. On my 4311 system, including this change
> improves throughput at
> least by a factor of 5 at 1 Mb/s. It also matches the old specifications
> where txctl1 was multiplied
> by 0x10.
>
> Larry
>
>
> Index: wireless-dev/drivers/net/wireless/d80211/bcm43xx/bcm43xx_lo.c
> ===================================================================
> --- wireless-dev.orig/drivers/net/wireless/d80211/bcm43xx/bcm43xx_lo.c
> +++ wireless-dev/drivers/net/wireless/d80211/bcm43xx/bcm43xx_lo.c
> @@ -401,7 +401,7 @@ static void lo_measure_txctl_values(stru
> & 0xFFF0) | txctl2);
> tmp0 = lo_measure_feedthrough(dev, lna, pga,
> trsw_rx);
> if (tmp0 < minimum) {
> - lo->txctl1 = txctl1;
> + lo->txctl1 = txctl1 << 4;
> lo->txctl2 = txctl2;
> minimum = tmp0;
> }
I don't think this is correct, because the txctl1 value is bits 0-3
in the lo->txctl1 variable. So this would mean you get a better TX by using
a txctl1 value of zero. Your TX power improvement may come from other
bugs that are now hidden by the txctl1==0 thing.
Specs might also be wrong here. Afaik the broadcom driver indeed stores the
txctl1 value in bits 4-7. We don't do this. So we need to be careful
with changes like the above.
Anyway, all the txctl related code will be completely rewritten really
soon, as there are other bigger bugs than that. ;)
The txctl stuff currently is really confusing, as we are using the
name "txctl" for a few different things. But I think Joseph is doing
good progress at actually cleaning it up in v4 specs by assigning
more meaningful names to the variables.
--
Greetings Michael.
_______________________________________________
Bcm43xx-dev mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/bcm43xx-dev
