Hi Bin, On Tue, 7 Jul 2020 at 01:09, Bin Meng <bmeng...@gmail.com> wrote: > > Hi Simon, > > On Sat, Jul 4, 2020 at 12:38 AM Simon Glass <s...@chromium.org> wrote: > > > > The current get_timer_us() uses 64-bit arithmetic. When implementing > > microsecond-level timeouts, 32-bits is plenty. Add a new function to > > support this. > > > > Signed-off-by: Simon Glass <s...@chromium.org> > > --- > > > > (no changes since v1) > > > > include/time.h | 11 +++++++++++ > > lib/time.c | 5 +++++ > > 2 files changed, 16 insertions(+) > > > > diff --git a/include/time.h b/include/time.h > > index e99f9c8012..434e63b075 100644 > > --- a/include/time.h > > +++ b/include/time.h > > @@ -17,6 +17,17 @@ unsigned long get_timer(unsigned long base); > > unsigned long timer_get_us(void); > > uint64_t get_timer_us(uint64_t base); > > > > +/** > > + * get_timer_us_long() - Get the number of elapsed microseconds > > + * > > + * This uses 32-bit arithmetic on 32-bit machines, which is enough to > > handle > > + * delays of over an hour. > > + * > > + *@base: Base time to consider > > + *@return elapsed time since @base > > + */ > > +unsigned long get_timer_us_long(unsigned long base); > > The function name does not clear indicates this is 32-bit value > because unsigned long is still 64-bit when building for 64-bit U-Boot.
Yes that's right. My purpose is to use the natural long time. Shall I update the commit message? Regards, Simon
