[FFmpeg-devel] [PATCH] aarch64: Use cntvct_el0 as timer register on Android

[Martin Storsjö]

On Fri, 7 Jun 2024, Rémi Denis-Courmont wrote:

> Le 7 juin 2024 12:12:45 GMT+03:00, "Martin Storsjö" <martin at martin.st> a écrit :
>> The default timer register pmccntr_el0 usually requires enabling
>> access with e.g. a kernel module.
>> ---
>> cntvct_el0 has significantly better resolution than
>> av_gettime_relative (while the unscaled nanosecond output of
>> clock_gettime is much higher resolution).
>>
>> In one tested case, the cntvct_el0 timer has a frequency of 25 MHz
>> (readable via the register cntfrq_el0).
>> ---
>> libavutil/aarch64/timer.h | 9 +++++++++
>> 1 file changed, 9 insertions(+)
>>
>> diff --git a/libavutil/aarch64/timer.h b/libavutil/aarch64/timer.h
>> index fadc9568f8..966f17081a 100644
>> --- a/libavutil/aarch64/timer.h
>> +++ b/libavutil/aarch64/timer.h
>> @@ -33,7 +33,16 @@ static inline uint64_t read_time(void)
>>     uint64_t cycle_counter;
>>     __asm__ volatile(
>>         "isb                   \t\n"
>> +#if defined(__ANDROID__)
>> +        // cntvct_el0 has lower resolution than pmccntr_el0, but is usually
>> +        // accessible from user space by default.
>> +        "mrs %0, cntvct_el0        "
>> +#else
>> +        // pmccntr_el0 has higher resolution, but is usually not accessible
>> +        // from user space by default (but access can be enabled with a custom
>> +        // kernel module).
>>         "mrs %0, pmccntr_el0       "
>> +#endif
>
> It feels a little newbie-hostile choice to pick something that's broken 
> by default but only on non-Android Linux.

Actually, on aarch64 Windows (both native Windows, and Linux binaries in 
WSL), pmccntr_el0 is accessible in user mode (and this is a documented 
feature, not an accident). Hence the "usually" in the comments.

But yes, the default is a bit hostile in that sense, as it's usually not 
usable as such on Linux.

This patch is at least a small baby step in a different direction; more 
configurability probably would be good, but I didn't want to take on 
fixing that here. But for those actually working on assembly, pmccntr_el0 
still is the best and usually worth the hassle - so we want to get rid of 
it entirely.

> Is there at least a runtime warning?

Since ac40c3bb07781e72f3eb1e30ea450019cc1f6302 we do have a runtime test 
in checkasm, if --bench is passed, giving the user a somewhat generic 
message, rather than crashing. (For other code using START/STOP_TIMER, 
there's no runtime check/warning though.)

// Martin