Open lygstate opened 2 years ago
According to the proposal:
Returns: a Temporal.Instant object representing the current system time.
I'm going to refer to my previous comment on this topic.
Is there actually a platform in existence that supports true nanosecond clock resolution for the system clock? So far I have yet to see any, but if I'm wrong then please correct me.
As such, even if the proposed Temporal object is able to store a time value in nanosecond precision does not mean that underlying hardware has sufficiently high precision timers, or that operating systems provide a way to query timers in that precision.
I think the way the proposal should be interpreted is the following: Take the current system time using the highest precision clock available, and scale it to nanoseconds. For this purpose the current port function is already more than sufficient.
According to the proposal:
Returns: a Temporal.Instant object representing the current system time.
I'm going to refer to my previous comment on this topic.
Is there actually a platform in existence that supports true nanosecond clock resolution for the system clock? So far I have yet to see any, but if I'm wrong then please correct me.
Hi, on windows, the system clock is 100ns(refer to https://docs.microsoft.com/en-us/windows/win32/api/sysinfoapi/nf-sysinfoapi-getsystemtimeasfiletime), on most posix platform, it's nanosecond, please refer to clock_gettime
As such, even if the proposed Temporal object is able to store a time value in nanosecond precision does not mean that underlying hardware has sufficiently high precision timers, or that operating systems provide a way to query timers in that precision.
I think the way the proposal should be interpreted is the following: Take the current system time using the highest precision clock available, and scale it to nanoseconds. For this purpose the current port function is already more than sufficient.
Well, in my experience the tv_nsec
member never actually contains nanosecond precision values, it's always a multiple of 1000. I'm also rather skeptical about the 100ns resolution in Windows, I would say it's also pretty likely that it's not fully utilized.
However you are correct that at least there are possibilities to query in (close to) nanosecond precision. I guess it would probably make sense to update the port function to resemble clock_gettime
, just in case, even if we can't fully utilize it yet.
According to https://tc39.es/proposal-temporal/docs/now.html#instant,
The Temporal.Now.instant() are improved the accuracy over Date.now from millisecond to nanosecond, that's need tthe
jerryscript
's port API to support nanosecond precision