jcstress results weirdness

[Aleksey Shipilev]

On 16.04.2015 12:14, Andrew Haley wrote:
> I'm seeing this:
> 
>                       Test                          Cycles      Results
>    org.openjdk.jcstress.tests.seqcst.volatiles
>       L1_L1_L1_L1_L1_S1__Test                  > 10^-2147483648 PASSED
>       L1_L1_L1_L1_S1_L1__Test                  > 10^-2147483648 PASSED
>       L1_L1_L1_L1_S1_S1__Test                  > 10^-2147483648 PASSED
>       L1_L1_L1_L1_S1__S1__Test                 > 10^7           PASSED
>       L1_L1_L1_L1_S1__Test                     > 10^-2147483648 PASSED
>       L1_L1_L1_S1_L1_L1__Test                  > 10^-2147483648 PASSED
>       L1_L1_L1_S1_L1_S1__Test                  > 10^-2147483648 PASSED
> 
> Is this perhaps arithmetic overflows generating the results?

Thanks, I think that's underflow when the "cycles" are close to zero.
(It's interesting to follow up why it is so). Anyhow, the printing is
fixed:
  https://bugs.openjdk.java.net/browse/CODETOOLS-7901391
  http://hg.openjdk.java.net/code-tools/jcstress/rev/2c43cdee5309

-Aleksey