test-kernel-reduce-params-multi-lambda.cpp.in

//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~//
// Copyright (c) Lawrence Livermore National Security, LLC and other
// RAJA Project Developers. See top-level LICENSE and COPYRIGHT
// files for dates and other details. No copyright assignment is required
// to contribute to RAJA.
//
// SPDX-License-Identifier: (BSD-3-Clause)
//~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~//

//
// test/include headers
//
#include "RAJA_test-base.hpp"
#include "RAJA_test-camp.hpp"
#include "RAJA_test-index-types.hpp"

// for data types
#include "RAJA_test-reduce-types.hpp"
#include "RAJA_test-reduceloc-types.hpp"
#include "RAJA_test-forall-data.hpp"

//
// Header for tests in ./tests directory
//
// Note: CMake adds ./tests as an include dir for these tests.
//
#include "test-kernel-reduce-params-multi-lambda.hpp"


using Lambda0Params =  RAJA::Params<0, 2, 4, 7>;
using Lambda1Params =  RAJA::Params<1, 3, 5, 6, 7>;
using Lambda0Segs = RAJA::Segs<0, 1>;
using Lambda1Segs = RAJA::Segs<0, 1>;

//
// Exec pols for kernel region tests
//

using SequentialKernelLocExecPols =
  camp::list<

    RAJA::KernelPolicy<
      RAJA::statement::For<1, RAJA::seq_exec,  // row
        RAJA::statement::For<0, RAJA::seq_exec,  // col
          RAJA::statement::Lambda<0, Lambda0Segs, Lambda0Params>,
          RAJA::statement::Lambda<1, Lambda1Segs, Lambda1Params>
        >
      >
    >

  >;


using SequentialKernelLocForallPols =
  camp::list<
    RAJA::seq_exec
  >;

#if defined(RAJA_ENABLE_OPENMP)

using OpenMPKernelLocExecPols =
  camp::list<

    RAJA::KernelPolicy<
      RAJA::statement::For<1, RAJA::seq_exec,  // row
        RAJA::statement::For<0, RAJA::omp_parallel_for_exec,  // col
          RAJA::statement::Lambda<0, Lambda0Segs, Lambda0Params>,
          RAJA::statement::Lambda<1, Lambda1Segs, Lambda1Params>
        >
      >
    >,

    RAJA::KernelPolicy<
      RAJA::statement::For<1, RAJA::omp_parallel_for_exec,  // row
        RAJA::statement::For<0, RAJA::seq_exec,  // col
          RAJA::statement::Lambda<0, Lambda0Segs, Lambda0Params>,
          RAJA::statement::Lambda<1, Lambda1Segs, Lambda1Params>
        >
      >
    >,

    RAJA::KernelPolicy<
      RAJA::statement::Collapse<RAJA::omp_parallel_collapse_exec,
        RAJA::ArgList<1,0>,
          RAJA::statement::Lambda<0, Lambda0Segs, Lambda0Params>,
          RAJA::statement::Lambda<1, Lambda1Segs, Lambda1Params>
      >
    >

  >;

using OpenMPKernelLocForallPols =
  camp::list<
    RAJA::omp_parallel_for_exec
  >;

#endif  // RAJA_ENABLE_OPENMP

#if defined(RAJA_ENABLE_TARGET_OPENMP)

using OpenMPTargetKernelLocExecPols =
  camp::list<

    RAJA::KernelPolicy<
      RAJA::statement::For<1, RAJA::seq_exec,  // row
        RAJA::statement::For<0, RAJA::omp_target_parallel_for_exec<16>,  // col
          RAJA::statement::Lambda<0, Lambda0Segs, Lambda0Params>,
          RAJA::statement::Lambda<1, Lambda1Segs, Lambda1Params>
        >
      >
    >,

    // TODO: Works for XLC and Clang10, but only up to input size of 362x362 matrix.
    RAJA::KernelPolicy<
      RAJA::statement::Collapse<RAJA::omp_target_parallel_collapse_exec,
        RAJA::ArgList<1,0>,
          RAJA::statement::Lambda<0, Lambda0Segs, Lambda0Params>,
          RAJA::statement::Lambda<1, Lambda1Segs, Lambda1Params>
      >
    >

  >;

using OpenMPTargetKernelLocForallPols =
  camp::list<
    RAJA::omp_target_parallel_for_exec<16>
  >;

#endif  // RAJA_ENABLE_TARGET_OPENMP

#if defined(RAJA_ENABLE_CUDA)

using CudaKernelLocExecPols =
  camp::list<

     // Two full grid reductions
    RAJA::KernelPolicy<
      RAJA::statement::CudaKernel<
        RAJA::statement::For<1, RAJA::cuda_thread_x_loop,  // row
          RAJA::statement::For<0, RAJA::cuda_thread_y_loop,  // col
            RAJA::statement::Lambda<0, Lambda0Segs, Lambda0Params>,
            RAJA::statement::Lambda<1, Lambda1Segs, Lambda1Params>
          >
        >
      >
    >,

    // Two grid reductions in different directions, with seq exec mixed at different
    // levels
    RAJA::KernelPolicy<
      RAJA::statement::CudaKernel<
        RAJA::statement::For<1, RAJA::seq_exec,  // row
          RAJA::statement::For<0, RAJA::cuda_thread_x_loop,  // col
            RAJA::statement::Lambda<0, Lambda0Segs, Lambda0Params>,
            RAJA::statement::Lambda<1, Lambda1Segs, Lambda1Params>
          >
        >
      >
    >,

     RAJA::KernelPolicy<
      RAJA::statement::CudaKernel<
        RAJA::statement::For<1, RAJA::seq_exec,  // row
          RAJA::statement::For<0, RAJA::cuda_thread_y_loop,  // col
            RAJA::statement::Lambda<0, Lambda0Segs, Lambda0Params>,
            RAJA::statement::Lambda<1, Lambda1Segs, Lambda1Params>
          >
        >
      >
    >,

    RAJA::KernelPolicy<
      RAJA::statement::CudaKernel<
        RAJA::statement::For<1, RAJA::cuda_global_size_y_loop<32>,  // row
          RAJA::statement::For<0, RAJA::cuda_global_size_x_loop<32>,  // col
            RAJA::statement::Lambda<0, Lambda0Segs, Lambda0Params>,
            RAJA::statement::Lambda<1, Lambda1Segs, Lambda1Params>
          >
        >
      >
    >

  >;

using CudaKernelLocForallPols =
  camp::list<
    RAJA::cuda_exec<1024, true>
  >;

#endif  // RAJA_ENABLE_CUDA

#if defined(RAJA_ENABLE_HIP)

using HipKernelLocExecPols =
  camp::list<
    // Two full grid reductions
    RAJA::KernelPolicy<
      RAJA::statement::HipKernel<
        RAJA::statement::For<1, RAJA::hip_thread_x_loop,  // row
          RAJA::statement::For<0, RAJA::hip_thread_y_loop,  // col
            RAJA::statement::Lambda<0, Lambda0Segs, Lambda0Params>,
            RAJA::statement::Lambda<1, Lambda1Segs, Lambda1Params>
          >
        >
      >
    >,

    // Two grid reductions in different directions, with seq exec mixed at different
    // levels
    RAJA::KernelPolicy<
      RAJA::statement::HipKernel<
        RAJA::statement::For<1, RAJA::seq_exec,  // row
          RAJA::statement::For<0, RAJA::hip_thread_x_loop,  // col
            RAJA::statement::Lambda<0, Lambda0Segs, Lambda0Params>,
            RAJA::statement::Lambda<1, Lambda1Segs, Lambda1Params>
          >
        >
      >
    >,

     RAJA::KernelPolicy<
      RAJA::statement::HipKernel<
        RAJA::statement::For<1, RAJA::seq_exec,  // row
          RAJA::statement::For<0, RAJA::hip_thread_y_loop,  // col
            RAJA::statement::Lambda<0, Lambda0Segs, Lambda0Params>,
            RAJA::statement::Lambda<1, Lambda1Segs, Lambda1Params>
          >
        >
      >
    >,

    RAJA::KernelPolicy<
      RAJA::statement::HipKernel<
        RAJA::statement::For<1, RAJA::hip_global_size_y_loop<32>,  // row
          RAJA::statement::For<0, RAJA::hip_global_size_x_loop<32>,  // col
            RAJA::statement::Lambda<0, Lambda0Segs, Lambda0Params>,
            RAJA::statement::Lambda<1, Lambda1Segs, Lambda1Params>
          >
        >
      >
    >

  >;

using HipKernelLocForallPols =
  camp::list<
    RAJA::hip_exec<1024, true>
  >;

#endif  // RAJA_ENABLE_HIP

//
// Cartesian product of types used in parameterized tests
//
using @BACKEND@KernelLocTypes =
  Test< camp::cartesian_product<RawIdxTypeList,
                                ReduceDataTypeList,
                                @BACKEND@ResourceList,
                                @BACKEND@KernelLocForallPols,
                                @BACKEND@KernelLocExecPols>>::Types;

//
// Instantiate parameterized test
//
INSTANTIATE_TYPED_TEST_SUITE_P(@BACKEND@,
                               KernelParamReduceMultiLambdaTest,
                               @BACKEND@KernelLocTypes);