Hello Pankaj, On 01/10/18 10:31, Pankaj Bansal wrote: > This function calculates the time elaped in Naoseconds between call to > this function and BaseTime, which is passed as argument. > > This is particularly useful in detecting timeout conditions. > > Cc: Leif Lindholm <[email protected]> > Cc: Ard Biesheuvel <[email protected]> > Contributed-under: TianoCore Contribution Agreement 1.1 > Signed-off-by: Pankaj Bansal <[email protected]> > > diff --git a/ArmPkg/Library/ArmArchTimerLib/ArmArchTimerLib.c > b/ArmPkg/Library/ArmArchTimerLib/ArmArchTimerLib.c > index b81293c..0898339 100644 > --- a/ArmPkg/Library/ArmArchTimerLib/ArmArchTimerLib.c > +++ b/ArmPkg/Library/ArmArchTimerLib/ArmArchTimerLib.c > @@ -290,3 +290,39 @@ GetTimeInNanoSecond ( > > return NanoSeconds; > } > + > +/** > + Get Elapsed time in Nanoseonds w.r.t BaseTime > + > + This function calculates the time elaped in Naoseconds between call to this > + function and BaseTime, which is passed as argument. > + > + @param BaseTime BaseTime in NanoSeconds. > + > + @return The elapsed time in nanoseconds. > + > +**/ > +UINT64 > +EFIAPI > +GetElapsedTime ( > + IN UINT64 BaseTime > + ) > +{ > + UINT64 NanoSeconds; > + UINT64 Ticks; > + > + // > + // Get current Ticks. > + // > + Ticks = GetPerformanceCounter(); > + > + // > + // Convert Ticks to Nanoseconds > + // > + NanoSeconds = GetTimeInNanoSecond (Ticks); > + > + // > + // return the difference > + // > + return (NanoSeconds - BaseTime); > +} >
There are two problems with this patch set: (1) The TimerLib.h file (in the first patch) is a public library class header, for which several library instances (implementations) exist. So, introducing a new API requires adding an implementation (the same implementation, or different ones, as necessary) to *all* instances in the edk2 tree. (2) The calculation in your GetElapsedTime() function is wrong. GetTimeInNanoSecond() converts a small *difference* of ticks to time. It does not convert an absolute tick value to an absolute time. Furthermore, tick differences are less trivial to calculate than one might imagine, because (a) a performance counter may count down, and *independently*, (b) the numeric low bound of the counter range may not be zero. Earlier I proposed a new TimerTickDiffLib class (and implementation) for centralizing exactly this kind of calculation. Please see the thread at: [edk2] TimerTickDiffLib for MdePkg? http://mid.mail-archive.com/[email protected] Thanks Laszlo _______________________________________________ edk2-devel mailing list [email protected] https://lists.01.org/mailman/listinfo/edk2-devel
