[PATCH v3] Improve the performance of --num-threads -d 31

["Zhou, Wenjian/周文剑"]

Hi,

I did the performance test.
v2 + Minoru's patch and v3 almost have the same performance in -d 31.

-- 
Thanks
Zhou

On 03/07/2016 04:39 PM, Minfei Huang wrote:
> On 03/07/16 at 08:16am, Minoru Usui wrote:
>> Hi, Minfei
>>>>
>>>> Hi, Wenjian.
>>>>
>>>> Since your patch applied Minoru Usui's draft patch, I have vierified it.
>>>> And it works well under 4T memory.
>>>>
>>>> applied makedumpfile with option -d 31
>>>> real    3m48.681s
>>>>
>>>> applied makedumpfile with option -d 31 --num-threads 32
>>>> real    3m3.533s
>>>>
>>>> original makedumpfile with option -d 31
>>>> real    3m57.163s
>>>>
>>>> Now with or without option --num-thread, it takes almost same time.
>>>
>>> Thank you for your test on big memory machine.
>>> I am relieved to see your test results.
>>
>> I'm sorry, I may be misunderstood.
>> Is above result is Zhou's v3 patch or v2 + my patch?
>
> I have applied Zhou's v3 patch.
>
>>
>> If above result is Zhou's v3 patch, could you test v2 + my patch?
>
> Your patch fails to be applied based on Zhou's v2 patch. There is a rej
> file to be generated during applying.
>
> Thanks
> Minfei
>
>