On 11 January 2015 at 11:58, Bill Fischofer <[email protected]>
wrote:

> The patch wasn't intended as a bug fix, but rather is intended to be a
> more refined implementation of the buffer allocator that can permit better
> performance scalability testing as we move into 2015.  It does have the
> effect, however, of fixing a couple of reported bugs that were the result
> of the previous implementation.
>
> I can post a repackaged version of this as v2 if that will ease getting it
> reviewed for inclusion in ODP 0.8, however, it just means asking reviewers
> to keep track of and apply multiple patches instead of one patch.
>

I think it is worth it, as Ciprian found out with the big buffer patch we
had impacted our ability to use bisect to narrow down an issue.


>
>
>
> On Sun, Jan 11, 2015 at 1:56 AM, Shmulik Ladkani <
> [email protected]> wrote:
>
>> On Sat, 10 Jan 2015 20:51:28 -0600 Bill Fischofer <
>> [email protected]> wrote:
>> > Because in the presence of local buffer caching it is not clear whether
>> > lockless or lock-based allocation will scale better, this patch adds
>> > compile-time support for selecting which sychronization mechanism to
>> use.
>> > By default lockless allocation is used. To enable lock-based allocation
>> > change the USE_BUFFER_POOL_LOCKS define to 1.
>>
>> Thanks.
>>
>> May I suggest splitting this as a patch set?
>> Introducing USE_BUFFER_POOL_LOCKS seems independent of the actual bug
>> fix.
>>
>> Regards,
>> Shmulik
>>
>
>
> _______________________________________________
> lng-odp mailing list
> [email protected]
> http://lists.linaro.org/mailman/listinfo/lng-odp
>
>


-- 
*Mike Holmes*
Linaro  Sr Technical Manager
LNG - ODP
_______________________________________________
lng-odp mailing list
[email protected]
http://lists.linaro.org/mailman/listinfo/lng-odp

Reply via email to