dpnp icon indicating copy to clipboard operation
dpnp copied to clipboard

ENH: kernels for `random.vonmisses`; part 2

Open samir-nasibli opened this issue 4 years ago • 6 comments

Description

Enable computations on devices [CPU/GPU].

Tests

  • DPNP own:
tests/test_random.py::TestDistributionsVonmises::test_moments[large_kappa] PASSED
tests/test_random.py::TestDistributionsVonmises::test_moments[small_kappa] PASSED
tests/test_random.py::TestDistributionsVonmises::test_invalid_args PASSED
tests/test_random.py::TestDistributionsVonmises::test_seed[large_kappa] FAILED
tests/test_random.py::TestDistributionsVonmises::test_seed[small_kappa] FAILED
  • + numpy external

TODO

tests/test_random.py::TestDistributionsVonmises::test_seed failed on both devices. Bug.

samir-nasibli avatar Apr 14 '21 18:04 samir-nasibli

@samir-nasibli Is this PR ready to review or still in development stage?

shssf avatar Jun 22 '21 17:06 shssf

@samir-nasibli Is this PR ready to review or still in development stage?

I will update this PR or move some part of this changes to another PR with closing this.

samir-nasibli avatar Jun 22 '21 20:06 samir-nasibli

@LukichevaPolina The use of extra memory with the amount of data is not a good practice in optimization. We must avoid this cases. We have to remove the possibilities for a potential race condition in the algorithm.

samir-nasibli avatar Oct 07 '21 13:10 samir-nasibli

The use of extra memory with the amount of data is not a good practice in optimization. We must avoid this cases. - ideas? We have to remove the possibilities for a potential race condition in the algorithm. - done

densmirn avatar Oct 07 '21 13:10 densmirn

The use of extra memory with the amount of data is not a good practice in optimization. We must avoid this cases. - ideas? We have to remove the possibilities for a potential race condition in the algorithm. - done

  1. Any optimization with the use of additional memory can actually degrade (depending on the input data) and underestimate all the benefits from parallelism. Allocation/Deallocation/Working with additional memory is expensive.

  2. We have to remove the possibilities for a potential race condition in the algorithm. - done

using extra mem is brute force approach.

  1. - ideas?

We need to investigate it.

  1. I also recommend to use perf tests and profiler tools during optimization. Comparative analysis is important in such work.

samir-nasibli avatar Oct 07 '21 14:10 samir-nasibli

Stale PR?

oleksandr-pavlyk avatar May 06 '22 16:05 oleksandr-pavlyk