On 2011-05-29 18:44, Avi Kivity wrote:
> On 05/29/2011 07:41 PM, Jan Kiszka wrote:
>> >
>> > void qemu_cpu_kick(void *_env)
>> > {
>> > CPUState *env = _env;
>> >
>> > qemu_cond_broadcast(env->halt_cond);
>> > if (!env->thread_kicked) {
>> > qemu_cpu_kick_thread(env);
>> > env->thread_kicked = true;
>> > }
>> > }
>> >
>> > Seems to have redundancies - we're both signalling a condition
>> variable
>> > and sending a signal.
>> >
>>
>> The target may block on the halt condition or run in guest mode. I don't
>> think we can (and should) try to find out which wakeup call is
>> sufficient.
>
> What we could do is make both waiters wait for the same event.
That may only work if we are able to switch halt_cond to a signal-based
mechanism. As usual, a portability issue. I don't think there is much to
win /wrt performance (cond_broadcast is fast). So the question is if
code complexity could actually benefit from this.
Jan
Attachment:
signature.asc
Description: OpenPGP digital signature
