Re: [Qemu-devel] [PATCH v2 0/6] Reduce lock contention on TCG hot-path

[Alex Bennée]

Emilio G. Cota <address@hidden> writes:

> On Tue, Jul 05, 2016 at 17:18:10 +0100, Alex Bennée wrote:
>> Well this is the first re-spin of the series posted last week. I've
>> added a bunch of additional patches to be more aggressive with
>> avoiding bouncing locks but to be honest the numbers don't seem to
>> make it worth it.
>
> How many threads are you using? With just a few threads I wouldn't
> expect a measurable difference.

My setup is 8 "cores" which pigz expands to use but I know you have
beefier machines on hand ;-)

>
>> I think the first 3 patches are ready to take if the TCG maintainers
>> want to:
>>
>>     tcg: Ensure safe tb_jmp_cache lookup out of 'tb_lock'
>>     tcg: set up tb->page_addr before insertion
>>     tcg: cpu-exec: remove tb_lock from the hot-path
>
> I think it would be simpler to use tb_lock_recursive and
> tb_lock_reset, as pointed out in v1 of this series.

I didn't realise people were suggesting asymmetric lock taking/reseting.
It seems ugly IMHO.

>
> Thanks,
>
>               Emilio


--
Alex Bennée