Guidance regarding reporting a bug probably related with VectorAPI

[Paul Sandoz]

Hi Aurelian,

This is the right place. The alternative is to log a report via bugreport.java.com <http://bugreport.java.com/>.

I cannot reproduce locally on my Intel-based MacBook (6-Core Intel Core i7) using the latest tip of the JDK, or JDK 21 and JDK 20.

What version of the JDK are you using?
If you can please share the hs_err files on this thread or via a gist.

Thanks,
Paul.

> On Nov 30, 2023, at 11:07 AM, Aurelian Tutuianu <padreati at yahoo.com> wrote:
> 
> Hello everyone,
> 
> I am a Java veteran user, but a completely disoriented bug reporter. While working on building a tensor library leveraging VectorAPI and concurrency, I went into what it looks like a bug. 
> 
> The scenario is simple, I simulate two strided arrays and compute dot product between them. For vectorized computation I take values from the strided array using fromArray in it's simple form if the stride is 1 or the complex form with indexes if we have non unit stride. Since I know from the beginning if a strided vector is unitary or not, I tried to take out the condition outside the loop. That causes problems (it looks like from time to time it goes to the wrong branch and calls the other function. 
> 
> I have built a small reproducible example which you can find it here: https://github.com/padreati/rapaio/blob/master/rapaio-experiment/src/bugs/VectorTest.java. I have also hs_err files if needed. It has the same behavior on OpenJdk, Corretto, Zulu, others not tested. I use Linux Ubuntu, i7 gen11, 8 cores.
> 
> It is possible to guide me into how to report that, if I may? Or if somebody can reproduce also? Or perhaps I made a mistake that I am not aware of and I might learn from that?
> 
> I am astonished of your work and it is the first time I take enough courage to bother you with issues I have. 
> Best regards, Aurelian Tutuianu
> 
> 
>