VectorAPI VectorInsert Intrinsic

Paul Sandoz paul.sandoz at oracle.com
Mon Jun 4 20:37:57 UTC 2018


Hi Shravya,

I think we need to improve the test, but i may not be aware of certain compiler constraints you wanna ensure apply. Please consider the following as a follow on if you want to commit sooner.

Currently the with index and value are constant. I think we can take the values of the input array and insert then into a vector, which is then stored the resulting array, so the two arrays should be equal. Something like:

$abstractvectortype$<Shapes.S$bits$Bit> av = SPECIES.zero();
for (int i = 0; i < a.length; i+= SPECIES.length()) {
  for (int j = 0; j < SPECIES.length(); j++) {
    av = av.with(j, a[i + j])
  }
  av.intoArray(r, i);
}

There can be a similar test for get:

for (int i = 0; i < a.length; i+= SPECIES.length()) {
  $abstractvectortype$<Shapes.S$bits$Bit> av = SPECIES.fromArray(a, i);
  for (int j = 0; j < SPECIES.length(); j++) {
    r[i + j] = av.get(j])
  }
}

Paul.

> On Jun 4, 2018, at 11:34 AM, Rukmannagari, Shravya <shravya.rukmannagari at intel.com> wrote:
> 
> Hi All,
> 
> I would like to contribute a patch for vector insert (with) intrinsic. Could you please review the patch here:
> 
> http://cr.openjdk.java.net/~srukmannagar/VectorAPI_vectorInsert/webrev.00/
> 
> 
> 
> Thanks,
> Shravya.



More information about the panama-dev mailing list