InsertBatch seems unnecessarily fussy about parameters

[ringerc]

benchmarks/src/main/java/org/postgresql/benchmark/statement/InsertBatch.java

Line 70 in 0534647

} else if (!(p2multi == 128 || p1nrows == 1024)) {

It seems to only run if multiinsert is 128 or nrows is 1024.

Perhaps something closer to

    } else if ((p1nrows % p2multi) != 0) {
      System.err.println("p1nrows must be evenly divisible by p2multi");
      System.exit(-1);
    }

was intended?

[vlsi]

That is there to save time on measurements.

Otherwise there are too many permutations of the arguments with little additional value.

[ringerc]

That makes sense. I wonder if there's a better way to do that, perhaps at the jmh level? Right now, if you want a run like

java -jar target/benchmarks.jar \
    '^org.postgresql.benchmark.statement.InsertBatch.insertBatch$' \
    -bm Throughput \
    -p p1nrows=100,1000,10000,100000 -p p2multi=1

you can't presently do it.

For now I've patched it in my copy.

I definitely suggest an informative error like "permutation skipped" or "invalid parameter combination"

[vlsi]

See https://bugs.openjdk.java.net/browse/CODETOOLS-7901296

[vlsi]

I think system property can be added to "run all the permutations".

[ringerc]

Makes sense. I think the main thing is to give it a meaningful message and useful comment (issue link, explanation)

[ringerc]

It's easy enough to patch if you know what and why