Skip to contentSkip to conversation

Least attached vector support

PATCH In review Yevgeny Petrilin 3 participants 3 messages May 13, 2010 – Apr 2, 2013 1 message not in the archive InfiniBand

lore
Share & tools
lore
#0 lore raw

cover Least attached vector support

Hello Roland,

Those patches where submitted a while ago, I cleaned them up a little and generated against your latest git. They allow to hw driver to choose to which EQ a CQ would be attached, considering the load on its eqs.

Thanks,
Yevgeny
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

#1 lore raw
Quoted 3 lines · Yevgeny Petrilin

Hello Roland,

Those patches where submitted a while ago, I cleaned them up a little and

generated against your latest git.

Quoted 1 line

They allow to hw driver to choose to which EQ a CQ would be attached,

considering the load on its eqs.

(replying to an e-mail from three years ago)

Hello,

Can anyone tell me what happened with this patch series ? To me this looks like a useful patch series to make the mlx4_en driver faster. However, I haven't seen anyone posting any review comments to this series and apparently this functionality isn't upstream either.

Thanks,

Bart.

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html

A message that is not in the archive (sent off-list or not yet received): [email protected]

#3 lore raw
Quoted 4 lines · Bart Van Assche

Can anyone tell me what happened with this patch series ? To me this looks like a useful patch series to make the mlx4_en driver faster. However, I haven't seen anyone posting any review comments to this series and apparently this functionality isn't upstream either.

It was never clear to me that "number of CQs attached" is a particularly good way to choose an interrupt vector. It would seem to me that we would want something that takes into account CPU affinity, possibly NUMA effects, etc. Also I don't think there were any performance numbers posted for these changes, were there?

(BTW It would be interesting to know how much improvement one could get on modern 2/4 socket systems by allocating EQ, CQ, QP, etc. memory on the "right" NUMA node -- where I don't know if "right" means on the node closest to the adapter or the node where the CPU will access the queue)
--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html