Messages in this thread | | | Date | Mon, 27 May 2013 17:57:11 +0200 | From | Manfred Spraul <> | Subject | Re: [PATCH 0/4] ipc/sem.c: Bug fixes, regression fixes, v3 |
| |
Hi Davidlohr,
On 05/26/2013 10:50 PM, Davidlohr Bueso wrote: > > In lack of getting my swingbench DSS environment back, I ran these > changes against the semop-multi program on my laptop. For 256 threads, > with Manfred's patchset the ops/sec suffers around -7.3%.
Could you also check the performance of only patch#1? I fear that it might be slower than all 4 together.
With regards to semop-multi: Is this the tool? http://marc.info/?l=linux-kernel&m=136208613626892&q=p3
I think the logic is the wrong: Locking a semaphore is substraction, unlocking adding. Thus multiple tasks can run in parallel - and the task switch code is never triggered. Could you double check that the number of context switches matches the output?
I usually use this tool: http://marc.info/?l=linux-kernel&m=125038376609750
-- Manfred
| |