jemalloc performance tunning for Embedded Platform

[amol pise]

Dear Jason,

Thank you very much for your quick response.
As you suggested I changed the chunk size  i.e. LG_CHUNK_DEFAULT value
from 22 to 20 ,  But still the speed performance on MIPS target with
jemalloc not improved.

My current configuration is as below:
{{{
enable_autogen='1'
enable_debug='0'
enable_dss='0'
enable_dynamic_page_shift='0'
enable_fill='0'
enable_lazy_lock='0'
enable_prof='0'
enable_stats='0'
enable_swap='0'
enable_sysv='0'
enable_tcache='1'
enable_tiny='1'
enable_tls='1'
enable_xmalloc='0'
}}}

I also come across the following thread
https://bugzilla.mozilla.org/show_bug.cgi?id=446096
"Integrate upstream jemalloc optimizations", here you suggested to Remove the
red-black trees which improves the allocation speed. It means adopting
the fix in the patch improves the performance.

Please let me know if my understanding is wrong.

Thank You,
Amol Pise

On 5/17/12, Jason Evans <jasone at canonware.com> wrote:
> On May 16, 2012, at 12:41 AM, amol pise wrote:
>> I am planning to use jemalloc library for the embedded platform (ARM and
>> MIPS).
>> I wanted to know the areas where we can tunned its performance for
>> Embedded platforms ?
>>
>> Your suggestions and inputs are very important for me to go ahead.
>>
>> I realy appriciated the help in this regards.
>>
>> Waiting for the reply.
>
> You probably want to disable every optional feature at configuration time.
> Depending on the virtual memory system for your embedded platform, you may
> want to reduce the chunk size.  I can't think of any other general
> guidelines that apply to all embedded use cases.
>
> Jason