Re: [Qemu-devel] [PATCH 00/12] trace: [tcg] Allow tracing guest events in TCG-generated code

[Lluís Vilanova]

Peter Maydell writes:

> On 7 February 2014 14:49, Richard Henderson <address@hidden> wrote:
>> On 02/06/2014 11:26 AM, Lluís Vilanova wrote:
>>>> > At which point, option 2 would be the best bet, I think.
>>> Right. For the 3rd option I was also thinking about having per-vCPU tracing
>>> states (in the case of guest events), so that you can trace separate events
>>> depending on the vCPU.
>>> 
>>> Which reminds me, I should add a vCPU pointer to the "guest_vmem" event, 
>>> since
>>> otherwise you cannot differentiate accesses among different vCPUs.
>> 
>> The TB cache is shared, so you won't be able to do option 3 for different
>> events for different cpus.  You could only do option 1 for that

> I have a vague plan in the back of my mind that we should make the
> TB cache be per-CPU, as a step towards better handling of multithreaded
> linux-user guest binaries and also heterogenous multiple CPU support...

What I implemented is something in between. I have multiple TB caches, and each
CPU selects one of them according to what tcg tracing events they have enabled
(i.e., if two CPUs have the same events, they use the same TB cache).

Although I can see how having one private TB per core could improve TB cache
hits when running multiple applications.


Lluis

-- 
 "And it's much the same thing with knowledge, for whenever you learn
 something new, the whole world becomes that much richer."
 -- The Princess of Pure Reason, as told by Norton Juster in The Phantom
 Tollbooth