Re: [PATCH 3/3] cpu-timers, icount: new modules

[Claudio Fontana <cfontana@suse.de>]

On 7/31/20 12:09 AM, Paolo Bonzini wrote:
> On 30/07/20 18:33, Claudio Fontana wrote:
>> One problem I noticed is that qemu_clock_get_ns is util/qemu-timer.c,
>> which is tools _and_ softmmu, while I tried to extract the
>> softmmu-only timer code in softmmu/cpu-timers.c,
> 
> Not all of it, only the VIRTUAL clock which is
> 
>         if (use_icount) {
>             return cpu_get_icount();
>         } else {
>             return cpu_get_clock();
>         }
> 
> and would be changed to something like
> 
> 	return cpu_get_virtual_clock();
> 
> In turn cpu_get_virtual_clock() is
> 
> 	return (accel_ops->cpu_get_virtual clock ?: cpu_get_clock)();
> 
> in the emulators, plus a stub that replaces stubs/cpu-get-icount.c and
> is just
> 
> 	return get_clock_realtime();
> 
> as in stubs/cpu-get-clock.c.
> 
> Paolo
> 
> 

works, hooking up cpu_get_ticks() also works.

qemu_start_warp_timer seems to make sense only for tcg/icount and not for qtest, which directly sets and warps the clock, is that right?

Would you want a start_warp_timer cpu accel interface that is actually useful only for tcg, to avoid if (icount_enabled()) { qemu_start_warp_timer()? }

Thanks,

Claudio