|
From: Tom H. <to...@co...> - 2007-12-21 15:47:16
|
In message <e2e...@ma...>
Bart Van Assche <bar...@gm...> wrote:
> On Dec 21, 2007 10:04 AM, Tom Hughes <to...@co...> wrote:
>>
>> Surely if it was changed before VG_TRACK(start_client_code)() then it
>> would have been fine for it to cache it?
>>
>> Did you mean that it was changed after that tracking call but before
>> the code in the handler was executed?
>
> What happens is the following:
> * Core changes VG_(running_tid).
> * Core notifies tool about client memory accesses.
> * Core calls VG_TRACK(start_client_code)().
>
> This is why it is not sufficient to cache VG_(running_tid) upon
> VG_TRACK(start_client_code)() notifications.
Ah right. I guess the question there is "which thread are those
accesses made on behalf off?".
Tom
--
Tom Hughes (to...@co...)
http://www.compton.nu/
|