On 25 April 2017 at 23:32, Lionel Landwerlin <lionel.g.landwerlin@xxxxxxxxx> wrote: > From: Robert Bragg <robert@xxxxxxxxxxxxx> > > There were a couple of problems with both of these tests that could lead > to false negatives addressed by this patch. > > 1) The upper limit for the number of iterations missed a +1 to consider > that there might be a sample immediately available at the start of the > loop. > > 2) The tests didn't consider that a duration measured in terms of > (end-start) ticks could be +- 1 tick since we don't know the > fractional part of the tick counts. Our threshold for stime being < > one tick could have a false negative for any real stime between 1 to > 10 milliseconds depending on luck. > > The tests now both run for a lot longer (1000 x tick duration, or > typically 10 seconds each) so that a single tick represents a much > smaller proportion of the total duration (0.1%) and the stime thresholds > are now set at 1% of the total duration. > > Signed-off-by: Robert Bragg <robert@xxxxxxxxxxxxx> I did r-b this in the past, so: Reviewed-by: Matthew Auld <matthew.auld@xxxxxxxxx> _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
