Update reducer types allowed in parallel reduction.

[pkestene]

Trying to fix #220

[NaderAlAwar]

The changes look good. Would it be possible to add a test that confirms that this is working fine? I think you can extend this test

pykokkos/tests/test_parallelreduce.py

Line 79 in 01f920f

def test_squaresum_types(series_max, dtype):

by simply parameterizing it on the other data types you added.

[pkestene]

The changes look good. Would it be possible to add a test that confirms that this is working fine? I think you can extend this test

pykokkos/tests/test_parallelreduce.py

Line 79 in 01f920f

def test_squaresum_types(series_max, dtype):

by simply parameterizing it on the other data types you added.

Sure, I'll update the test.

But just to clarify beforehand:
in

pykokkos/tests/test_parallelreduce.py

Line 58 in 01f920f

def squaresum(self, i: float, acc: pk.Acc[pk.double]):

the iterate variable is a float, but should really be an integral type. I was even surprised the test builds. I think, kokkos should check that internally, no ?

[NaderAlAwar]

the iterate variable is a float, but should really be an integral type. I was even surprised the test builds. I think, kokkos should check that internally, no ?

Looking at the generated C++, it seems that PyKokkos ignores the type supplied by the user and generates an int32_t. In Kokkos, it is possible to set the index type through a RangePolicy template argument, but we do not support that yet. I think the test should be changed to an int iterate variable for now.