On Fri, 10 Sep 2021 08:32:02 GMT, Kartik Ohri <github.com+27751938+amcap1...@openjdk.org> wrote:
> Hi all! > > Please review this PR to add a benchmark comparing the performance of > Arrays.mismatch intrinsic in the JDK with that of the Vector API. Kindly > refer to this [thread] on panama-dev regarding some initial discussion about > this benchmark. I have attached the [results] of the full benchmark run along > with the [assembly] output of a shorter run I had done while analysing the > results. The benchmarks were run against the latest build of panama-vector > available from builds.shipilev.net. > > Also, I have not added the copyright header to this file yet as I am an > individual contributor (OCA signed) and do not know what to put there. > > [thread]: > https://mail.openjdk.java.net/pipermail/panama-dev/2021-September/014839.html > [results]: https://github.com/openjdk/jdk/files/7142452/results.csv > [assembly]: https://github.com/openjdk/jdk/files/7142362/benchmarks.txt > > Regards, > Kartik Benchmark looks good (assuming license is added). test/micro/org/openjdk/bench/jdk/incubator/vector/ArrayMismatchBenchmark.java line 52: > 50: static final VectorSpecies<Long> LONG_SPECIES_PREFERRED = > LongVector.SPECIES_PREFERRED; > 51: > 52: static final Random random = new Random(); We could use the recently added `RandomGenerator` instead in the spirit of encouraging the use of new and preferred APIs: - remove the static field - replace `FLOAT_SPECIES` with `DOUBLE_SPECIES` - in `setup` create an instance `RandomGenerator rg = RandomGenerator.getDefault()` - remove `createRandomFloats` and use `rg.doubles(...).toArray()` ------------- PR: https://git.openjdk.java.net/jdk/pull/5459