Re: 9.2beta1, parallel queries, ReleasePredicateLocks, CheckForSerializableConflictIn in the oprofile

On Wed, Jun 6, 2012 at 2:53 PM, Sergey Koposov <koposov(at)ast(dot)cam(dot)ac(dot)uk> wrote:
> On Wed, 6 Jun 2012, Ants Aasma wrote:
>
>> On Wed, Jun 6, 2012 at 2:27 PM, Sergey Koposov <koposov(at)ast(dot)cam(dot)ac(dot)uk>
>> wrote:
>>>
>>> I've quickly tested your lockfree-getbuffer.patch patch with the test
>>> case
>>> you provided and I barely see any improvement (2% at max)
>>> https://docs.google.com/open?id=0B7koR68V2nM1QVBxWGpZdW4wd0U
>>> tested with 24 core (48 ht cores, Xeon E7- 4807).
>>> Although the tps vs number of threads looks weird....
>>
>>
>> Was this the range scan on the test table? (sorry about the error in
>> the query, the x should really be id) In that case the results look
>> really suspicious.
>
>
> Yes, my fault partially, because without much thought I've put "value"
> instead of "x" in the script. Now after replacing it by "id" the tps are
> much smaller.
>
> Here is the tps vs nthreads I did test up to 10 threads on my 24 cpu system
> (I disabled HT though):
> https://docs.google.com/open?id=0B7koR68V2nM1Nk9OcWNJOTRrYVE
>
> Your patch clearly improve the situation (the peak tps is ~ 10% higher), but
> the general picture is the same: flattening of tps vs nthreads.

I think this is the expected result. In the single user case the
spinklock never spins and only has to make the cpu-locking cache
instructions once. can we see results @24 threads?

merlin