Messages in this thread |  | | Date | Thu, 10 Jan 2013 08:05:00 -0500 | From | Rik van Riel <> | Subject | Re: [PATCH 4/5] x86,smp: keep spinlock delay values per hashed spinlock address |
| |
On 01/10/2013 08:01 AM, Michel Lespinasse wrote: > On Tue, Jan 8, 2013 at 2:31 PM, Rik van Riel <riel@redhat.com> wrote: >> From: Eric Dumazet <eric.dumazet@gmail.com> >> >> Eric Dumazet found a regression with the first version of the spinlock >> backoff code, in a workload where multiple spinlocks were contended, >> each having a different wait time. >> >> This patch has multiple delay values per cpu, indexed on a hash >> of the lock address, to avoid that problem. >> >> Eric Dumazet wrote: >> >> I did some tests with your patches with following configuration : >> >> tc qdisc add dev eth0 root htb r2q 1000 default 3 >> (to force a contention on qdisc lock, even with a multi queue net >> device) >> >> and 24 concurrent "netperf -t UDP_STREAM -H other_machine -- -m 128" >> >> Machine : 2 Intel(R) Xeon(R) CPU X5660 @ 2.80GHz >> (24 threads), and a fast NIC (10Gbps) >> >> Resulting in a 13 % regression (676 Mbits -> 595 Mbits) >> >> In this workload we have at least two contended spinlocks, with >> different delays. (spinlocks are not held for the same duration) >> >> It clearly defeats your assumption of a single per cpu delay being OK : >> Some cpus are spinning too long while the lock was released. >> >> We might try to use a hash on lock address, and an array of 16 different >> delays so that different spinlocks have a chance of not sharing the same >> delay. >> >> With following patch, I get 982 Mbits/s with same bench, so an increase >> of 45 % instead of a 13 % regression. > > Note that these results were with your v1 proposal. With v3 proposal, > on a slightly different machine (2 socket sandybridge) with a similar > NIC, I am not seeing the regression when not using the hash table. I > think this is because v3 got more conservative about mixed spinlock > hold times, and converges towards the shortest of the hold times in > that case.
Eric,
with just patches 1-3, can you still reproduce the regression on your system?
In other words, could we get away with dropping the complexity of patch 4, or do we still need it?
-- All rights reversed
|  |