On Wed, Jun 24, 2026 at 03:08:24PM +0200, Jan Stary wrote: > Dear all, > > the resolution of CLOCK_PROCESS_CPUTIME_ID, > as reported by clock_getres(), is 10000000 ns; > this seems to be the case on all archs I have. > > This resolution seems to be coarser than > the actual cpu time reported (in a process that does nothing else), > such as 7233653 ns on this machine here. > > FWIW, each of the OSes below (+Debian) returns 1 ns.
Yes this is wrong the CLOCK_PROCESS_CPUTIME_ID returns the tu_runtime which uses nanouptime() (see tuagg_add_runtime). This should be changed to report 1ns like the other clocks. -- :wq Claudio
