On 4/20/22 12:30, Eli Zaretskii wrote:
I see the time samples change in jumps of 15 msec.
Could you give the first part of the output? I would like to see what the the samples are jumping from and to, and how often they jump.
Something like the following is what I'd hope to see from the first lines of the output of 'gllib/test-gettime-res x'. What are you seeing?
gettime_res returned 15625000 ns time = 1650496481.515625000 time = 1650496481.531250000 time = 1650496481.546875000 time = 1650496481.562500000 time = 1650496481.578125000 time = 1650496481.593750000 time = 1650496481.609375000 time = 1650496481.625000000 time = 1650496481.640625000 time = 1650496481.656250000
Which is expected on MS-Windows, given the scheduler time tick, but what does that have to do with the system's time resolution?
The resolution of Elisp's (time-convert nil t) is determined by the smallest nonzero gap between timestamps that are returned by C's current_timespec. This is the system time resolution as far as Elisp is concerned, because Elisp cannot return the current time at a finer resolution than what current_timespec gives it. This resolution is not necessarily the same as the time resolution of the motherboard clock, OS high-res timestamp, file timestamps, etc.
And how is the 0.625 msec number reported by the program obtained from those samples?
Use the largest resolution R ns consistent with the samples, such that 1000000000 is an integer multiple of R so that timestamp computations are exact.
