On Fri, Sep 23, 2011 at 06:38:14PM +0200, Julian Andres Klode wrote:
> Allow TX buffers to be allocated only in the lower 75% of the pool
> to avoid a completely filled buffer preventing the driver from
> processing responses.
Looking a second time at it, it makes more sense to reserve the
lower 25% for received things and have the upper 75% shared. This
keeps 16 events free for events and responses, and we can thus
often find the response slot in O(1) instead of O(2):
0=request => [0 is not free] [1 is free]
17=request => [0 is free]
I also reduces the complexity a bit. Something like:
int i = (category == NVEC_MSG_TX) ? (NVEC_POOL_SIZE / 4) : 0;
for (; i < NVEC_POOL_SIZE; i++) {
if (atomic_xchg(&nvec->msg_pool[i].used, 1) == 0) {
dev_vdbg(nvec->dev, "INFO: Alloc %i\n", i);
return &nvec->msg_pool[i];
}
}
I might be more readable if I invert he condition and use a continue
though.
--
Julian Andres Klode - Debian Developer, Ubuntu Member
See http://wiki.debian.org/JulianAndresKlode and http://jak-linux.org/.
_______________________________________________
devel mailing list
devel@xxxxxxxxxxxxxxxxxxxxxx
http://driverdev.linuxdriverproject.org/mailman/listinfo/devel
