🏡 index : cold-air-networks/mirror-opencl-cts.git

📄 .appveyor.yml
📄 .clang-format
📄 .gitattributes
📁 .github
📁 workflows
📄 presubmit.yml
📄 CMakeLists.txt
📄 CODE_OF_CONDUCT.md
📄 LICENSE.txt
📄 README.md
📄 check-format.sh
📄 presubmit.sh
📄 readme-spir-v-binaries.txt
📁 test_common
📄 CMakeLists.txt
📁 autotest
📄 autotest.hpp
📄 test_suite.hpp
📄 config.hpp
📁 gl
📄 gl_headers.h
📄 helpers.cpp
📄 helpers.h
📄 setup.h
📄 setup_osx.cpp
📄 setup_win32.cpp
📄 setup_x11.cpp
📁 gles
📄 gl_headers.h
📄 helpers.cpp
📄 helpers.h
📄 setup.h
📁 harness
📄 ThreadPool.cpp
📄 ThreadPool.h
📄 alloc.h
📄 clImageHelper.h
📄 cl_offline_compiler-interface.txt
📄 compat.h
📄 conversions.cpp
📄 conversions.h
📄 crc32.cpp
📄 crc32.h
📄 deviceInfo.cpp
📄 deviceInfo.h
📄 errorHelpers.cpp
📄 errorHelpers.h
📄 featureHelpers.cpp
📄 featureHelpers.h
📄 fpcontrol.h
📄 genericThread.cpp
📄 genericThread.h
📄 imageHelpers.cpp
📄 imageHelpers.h
📄 kernelHelpers.cpp
📄 kernelHelpers.h
📄 mingw_compat.c
📄 mingw_compat.h
📄 msvc9.c
📄 mt19937.cpp
📄 mt19937.h
📄 os_helpers.cpp
📄 os_helpers.h
📄 parseParameters.cpp
📄 parseParameters.h
📄 propertyHelpers.cpp
📄 propertyHelpers.h
📄 ref_counting.h
📄 rounding_mode.cpp
📄 rounding_mode.h
📄 testHarness.cpp
📄 testHarness.h
📄 test_mt19937.c
📄 threadTesting.cpp
📄 threadTesting.h
📄 typeWrappers.cpp
📄 typeWrappers.h
📁 miniz
📄 CMakeLists.txt
📄 miniz.c
📄 miniz.h
📁 test_conformance
📄 CMakeCommon.txt
📄 CMakeLists.txt
📁 SVM
📄 CMakeLists.txt
📄 common.h
📄 main.cpp
📄 test_allocate_shared_buffer.cpp
📄 test_byte_granularity.cpp
📄 test_cross_buffer_pointers.cpp
📄 test_enqueue_api.cpp
📄 test_fine_grain_memory_consistency.cpp
📄 test_fine_grain_sync_buffers.cpp
📄 test_migrate.cpp
📄 test_pointer_passing.cpp
📄 test_set_kernel_exec_info_svm_ptrs.cpp
📄 test_shared_address_space_coarse_grain.cpp
📄 test_shared_address_space_fine_grain.cpp
📄 test_shared_address_space_fine_grain_buffers.cpp
📄 test_shared_sub_buffers.cpp
📁 allocations
📄 CMakeLists.txt
📄 allocation_execute.cpp
📄 allocation_execute.h
📄 allocation_fill.cpp
📄 allocation_fill.h
📄 allocation_functions.cpp
📄 allocation_functions.h
📄 allocation_utils.cpp
📄 allocation_utils.h
📄 main.cpp
📄 testBase.h
📁 api
📄 CMakeLists.txt
📄 main.cpp
📄 negative_platform.cpp
📄 procs.h
📄 testBase.h
📄 test_api_consistency.cpp
📄 test_api_min_max.cpp
📄 test_binary.cpp
📄 test_bool.cpp
📄 test_clone_kernel.cpp
📄 test_context_destructor_callback.cpp
📄 test_create_context_from_type.cpp
📄 test_create_kernels.cpp
📄 test_device_min_data_type_align_size_alignment.cpp
📄 test_kernel_arg_changes.cpp
📄 test_kernel_arg_info.cpp
📄 test_kernel_arg_info_compatibility.cpp
📄 test_kernel_arg_multi_setup.cpp
📄 test_kernel_private_memory_size.cpp
📄 test_kernels.cpp
📄 test_mem_object_info.cpp
📄 test_mem_object_properties_queries.cpp
📄 test_mem_objects.cpp
📄 test_min_image_formats.cpp
📄 test_native_kernel.cpp
📄 test_null_buffer_arg.cpp
📄 test_pipe_properties_queries.cpp
📄 test_platform.cpp
📄 test_queries.cpp
📄 test_queue.cpp
📄 test_queue_hint.cpp
📄 test_queue_properties.cpp
📄 test_queue_properties_queries.cpp
📄 test_retain.cpp
📄 test_retain_program.cpp
📄 test_sub_group_dispatch.cpp
📄 test_zero_sized_enqueue.cpp
📁 atomics
📄 CMakeLists.txt
📄 main.cpp
📄 procs.h
📄 testBase.h
📄 test_atomics.cpp
📄 test_indexed_cases.cpp
📁 basic
📄 CMakeLists.txt
📄 main.cpp
📄 procs.h
📄 run_array
📄 run_array_image_copy
📄 run_image
📄 run_multi_read_image
📄 test_arraycopy.cpp
📄 test_arrayimagecopy.cpp
📄 test_arrayimagecopy3d.cpp
📄 test_arrayreadwrite.cpp
📄 test_astype.cpp
📄 test_async_copy.cpp
📄 test_async_copy2D.cpp
📄 test_async_copy3D.cpp
📄 test_async_copy_fence.cpp
📄 test_async_strided_copy.cpp
📄 test_barrier.cpp
📄 test_basic_parameter_types.cpp
📄 test_bufferreadwriterect.cpp
📄 test_constant.cpp
📄 test_constant_source.cpp
📄 test_createkernelsinprogram.cpp
📄 test_enqueue_map.cpp
📄 test_enqueued_local_size.cpp
📄 test_explicit_s2v.cpp
📄 test_float2int.cpp
📄 test_fpmath_float.cpp
📄 test_fpmath_float2.cpp
📄 test_fpmath_float4.cpp
📄 test_get_linear_ids.cpp
📄 test_global_linear_id.cpp
📄 test_global_work_offsets.cpp
📄 test_hiloeo.cpp
📄 test_hostptr.cpp
📄 test_if.cpp
📄 test_imagecopy.cpp
📄 test_imagecopy3d.cpp
📄 test_imagedim.cpp
📄 test_imagenpot.cpp
📄 test_imagerandomcopy.cpp
📄 test_imagereadwrite.cpp
📄 test_imagereadwrite3d.cpp
📄 test_int2float.cpp
📄 test_intmath.cpp
📄 test_kernel_call_kernel_function.cpp
📄 test_kernel_memory_alignment.cpp
📄 test_local.cpp
📄 test_local_kernel_scope.cpp
📄 test_local_linear_id.cpp
📄 test_loop.cpp
📄 test_multireadimagemultifmt.cpp
📄 test_multireadimageonefmt.cpp
📄 test_numeric_constants.cpp
📄 test_pointercast.cpp
📄 test_preprocessors.cpp
📄 test_progvar.cpp
📄 test_queue_priority.cpp
📄 test_readimage.cpp
📄 test_readimage3d.cpp
📄 test_readimage3d_fp32.cpp
📄 test_readimage3d_int16.cpp
📄 test_readimage_fp32.cpp
📄 test_readimage_int16.cpp
📄 test_rw_image_access_qualifier.cpp
📄 test_simple_image_pitch.cpp
📄 test_sizeof.cpp
📄 test_vec_type_hint.cpp
📄 test_vector_creation.cpp
📄 test_vector_swizzle.cpp
📄 test_vloadstore.cpp
📄 test_wg_barrier.cpp
📄 test_work_item_functions.cpp
📄 test_writeimage.cpp
📄 test_writeimage_fp32.cpp
📄 test_writeimage_int16.cpp
📁 buffers
📄 CMakeLists.txt
📄 array_info.cpp
📄 main.cpp
📄 procs.h
📄 test_buffer_copy.cpp
📄 test_buffer_fill.cpp
📄 test_buffer_map.cpp
📄 test_buffer_mem.cpp
📄 test_buffer_migrate.cpp
📄 test_buffer_read.cpp
📄 test_buffer_write.cpp
📄 test_image_migrate.cpp
📄 test_sub_buffers.cpp
📁 c11_atomics
📄 CMakeLists.txt
📄 common.cpp
📄 common.h
📄 host_atomics.cpp
📄 host_atomics.h
📄 main.cpp
📄 test_atomics.cpp
📁 clcpp
📄 CMakeLists.txt
📁 address_spaces
📄 CMakeLists.txt
📄 common.hpp
📄 main.cpp
📄 test_pointer_types.hpp
📄 test_storage_types.hpp
📁 api
📄 CMakeLists.txt
📄 main.cpp
📄 test_ctors.hpp
📄 test_ctors_dtors.hpp
📄 test_dtors.hpp
📄 test_spec_consts.hpp
📁 atomics
📄 CMakeLists.txt
📄 atomic_fetch.hpp
📄 main.cpp
📁 attributes
📄 CMakeLists.txt
📄 main.cpp
📄 test_ivdep.hpp
📄 test_max_size.hpp
📄 test_required_num_sub_groups.hpp
📄 common.hpp
📁 common_funcs
📄 CMakeLists.txt
📄 common_funcs.hpp
📄 main.cpp
📁 convert
📄 CMakeLists.txt
📄 convert_cast.hpp
📄 main.cpp
📁 device_queue
📄 CMakeLists.txt
📄 main.cpp
📄 test_enqueue.hpp
📄 funcs_test_utils.hpp
📁 geometric_funcs
📄 CMakeLists.txt
📄 fast_geometric_funcs.hpp
📄 geometric_funcs.hpp
📄 main.cpp
📁 images
📄 CMakeLists.txt
📄 common.hpp
📄 main.cpp
📄 test_read.hpp
📄 test_sample.hpp
📄 test_write.hpp
📁 integer_funcs
📄 24bit_funcs.hpp
📄 CMakeLists.txt
📄 bitwise_funcs.hpp
📄 common.hpp
📄 main.cpp
📄 numeric_funcs.hpp
📁 math_funcs
📄 CMakeLists.txt
📄 common.hpp
📄 comparison_funcs.hpp
📄 exponential_funcs.hpp
📄 floating_point_funcs.hpp
📄 half_math_funcs.hpp
📄 logarithmic_funcs.hpp
📄 main.cpp
📄 other_funcs.hpp
📄 power_funcs.hpp
📄 reference.hpp
📄 trigonometric_funcs.hpp
📁 pipes
📄 CMakeLists.txt
📄 main.cpp
📄 test_pipes.hpp
📁 program_scope_ctors_dtors
📄 CMakeLists.txt
📄 common.hpp
📄 main.cpp
📄 test_ctors_dtors.hpp
📁 reinterpret
📄 CMakeLists.txt
📄 as_type.hpp
📄 main.cpp
📁 relational_funcs
📄 CMakeLists.txt
📄 common.hpp
📄 comparison_funcs.hpp
📄 main.cpp
📄 select_funcs.hpp
📄 test_funcs.hpp
📁 spec_constants
📄 CMakeLists.txt
📄 common.hpp
📄 main.cpp
📄 test_spec_consts_attributes.hpp
📄 test_spec_consts_if.hpp
📄 test_spec_consts_init_vars.hpp
📄 spirv10_2016.04.27.7z
📁 subgroups
📄 CMakeLists.txt
📄 common.hpp
📄 main.cpp
📄 test_sg_all.hpp
📄 test_sg_any.hpp
📄 test_sg_broadcast.hpp
📄 test_sg_reduce.hpp
📄 test_sg_scan_exclusive.hpp
📄 test_sg_scan_inclusive.hpp
📁 synchronization
📄 CMakeLists.txt
📄 main.cpp
📁 named_barrier
📄 common.hpp
📄 test_named_barrier.hpp
📄 test_spec_example.hpp
📄 test_sub_group_barrier.hpp
📄 test_work_group_barrier.hpp
📁 utils_common
📄 errors.hpp
📄 is_vector_type.hpp
📄 kernel_helpers.hpp
📄 make_vector_type.hpp
📄 scalar_type.hpp
📄 string.hpp
📄 type_name.hpp
📄 type_supported.hpp
📄 vector_size.hpp
📁 utils_test
📄 binary.hpp
📄 compare.hpp
📁 detail
📄 base_func_type.hpp
📄 vec_helpers.hpp
📄 generate_inputs.hpp
📄 ternary.hpp
📄 unary.hpp
📁 vload_vstore
📄 CMakeLists.txt
📄 common.hpp
📄 half_utils.hpp
📄 main.cpp
📄 vload_funcs.hpp
📄 vstore_funcs.hpp
📁 workgroups
📄 CMakeLists.txt
📄 common.hpp
📄 main.cpp
📄 test_wg_all.hpp
📄 test_wg_any.hpp
📄 test_wg_broadcast.hpp
📄 test_wg_reduce.hpp
📄 test_wg_scan_exclusive.hpp
📄 test_wg_scan_inclusive.hpp
📁 workitems
📄 CMakeLists.txt
📄 main.cpp
📄 test_workitems.hpp
📁 commonfns
📄 CMakeLists.txt
📄 main.cpp
📄 procs.h
📄 test_binary_fn.cpp
📄 test_clamp.cpp
📄 test_degrees.cpp
📄 test_fmax.cpp
📄 test_fmaxf.cpp
📄 test_fmin.cpp
📄 test_fminf.cpp
📄 test_max.cpp
📄 test_maxf.cpp
📄 test_min.cpp
📄 test_minf.cpp
📄 test_mix.cpp
📄 test_radians.cpp
📄 test_sign.cpp
📄 test_smoothstep.cpp
📄 test_smoothstepf.cpp
📄 test_step.cpp
📄 test_stepf.cpp
📁 compiler
📄 CMakeLists.txt
📁 includeTestDirectory
📄 testIncludeFile.h
📄 main.cpp
📄 procs.h
📁 secondIncludeTestDirectory
📄 testIncludeFile.h
📄 testBase.h
📄 test_async_build.cpp
📄 test_build_helpers.cpp
📄 test_build_options.cpp
📄 test_compile.cpp
📄 test_compiler_defines_for_extensions.cpp
📄 test_feature_macro.cpp
📄 test_image_macro.cpp
📄 test_opencl_c_versions.cpp
📄 test_pragma_unroll.cpp
📄 test_preprocessor.cpp
📄 test_unload_platform_compiler.cpp
📄 test_unload_platform_compiler_resources.hpp
📄 write_kernel.spvasm32
📄 write_kernel.spvasm64
📁 computeinfo
📄 CMakeLists.txt
📄 conforming_version.cpp
📄 device_uuid.cpp
📄 extended_versioning.cpp
📄 main.cpp
📁 contractions
📄 CMakeLists.txt
📄 contractions.cpp
📁 conversions
📄 CMakeLists.txt
📄 Sleep.cpp
📄 Sleep.h
📄 basic_test_conversions.cpp
📄 basic_test_conversions.h
📄 fplib.cpp
📄 fplib.h
📄 run_batch
📄 test_conversions.cpp
📁 d3d10
📄 CMakeLists.txt
📄 buffer.cpp
📄 harness.cpp
📄 harness.h
📄 main.cpp
📄 misc.cpp
📄 texture2d.cpp
📄 texture3d.cpp
📁 d3d11
📄 CMakeLists.txt
📄 buffer.cpp
📄 harness.cpp
📄 harness.h
📄 main.cpp
📄 misc.cpp
📄 texture2d.cpp
📄 texture3d.cpp
📁 device_execution
📄 CMakeLists.txt
📄 device_info.cpp
📄 device_queue.cpp
📄 enqueue_block.cpp
📄 enqueue_flags.cpp
📄 enqueue_multi_queue.cpp
📄 enqueue_ndrange.cpp
📄 enqueue_profiling.cpp
📄 enqueue_wg_size.cpp
📄 execute_block.cpp
📄 host_multi_queue.cpp
📄 host_queue_order.cpp
📄 main.cpp
📄 nested_blocks.cpp
📄 procs.h
📄 utils.cpp
📄 utils.h
📁 device_partition
📄 CMakeLists.txt
📄 main.cpp
📄 procs.h
📄 testBase.h
📄 test_device_partition.cpp
📁 device_timer
📄 CMakeLists.txt
📄 main.cpp
📄 procs.h
📄 test_device_timer.cpp
📁 events
📄 CMakeLists.txt
📄 action_classes.cpp
📄 action_classes.h
📄 main.cpp
📄 procs.h
📄 testBase.h
📄 test_callbacks.cpp
📄 test_event_dependencies.cpp
📄 test_events.cpp
📄 test_userevents.cpp
📄 test_userevents_multithreaded.cpp
📄 test_waitlists.cpp
📁 extensions
📄 CMakeLists.txt
📁 cl_ext_cxx_for_opencl
📄 CMakeLists.txt
📄 cxx_for_opencl_ext.cpp
📄 cxx_for_opencl_ver.cpp
📄 main.cpp
📄 procs.h
📁 cl_khr_dx9_media_sharing
📄 CMakeLists.txt
📄 main.cpp
📄 procs.h
📄 test_create_context.cpp
📄 test_functions_api.cpp
📄 test_functions_kernel.cpp
📄 test_get_device_ids.cpp
📄 test_interop_sync.cpp
📄 test_memory_access.cpp
📄 test_other_data_types.cpp
📄 utils.cpp
📄 utils.h
📄 wrappers.cpp
📄 wrappers.h
📄 generate_spirv_offline.py
📁 generic_address_space
📄 CMakeLists.txt
📄 advanced_tests.cpp
📄 base.h
📄 basic_tests.cpp
📄 main.cpp
📄 stress_tests.cpp
📁 geometrics
📄 CMakeLists.txt
📄 main.cpp
📄 procs.h
📄 testBase.h
📄 test_geometrics.cpp
📄 test_geometrics_double.cpp
📁 gl
📄 CMakeLists.txt
📁 GLEW
📁 GL
📄 eglew.h
📄 glew.h
📄 glxew.h
📄 wglew.h
📄 glew.c
📄 common.h
📄 helpers.cpp
📄 main.cpp
📄 procs.h
📄 testBase.h
📄 test_buffers.cpp
📄 test_fence_sync.cpp
📄 test_image_methods.cpp
📄 test_images_1D.cpp
📄 test_images_1Darray.cpp
📄 test_images_2D.cpp
📄 test_images_2Darray.cpp
📄 test_images_3D.cpp
📄 test_images_depth.cpp
📄 test_images_getinfo_common.cpp
📄 test_images_multisample.cpp
📄 test_images_read_common.cpp
📄 test_images_write_common.cpp
📄 test_renderbuffer.cpp
📄 test_renderbuffer_info.cpp
📁 gles
📄 CMakeLists.txt
📄 helpers.cpp
📄 main.cpp
📄 procs.h
📄 setup_egl.cpp
📄 testBase.h
📄 test_buffers.cpp
📄 test_fence_sync.cpp
📄 test_images_2D.cpp
📄 test_images_2D_info.cpp
📄 test_images_3D.cpp
📄 test_images_3D_info.cpp
📄 test_renderbuffer.cpp
📄 test_renderbuffer_info.cpp
📁 half
📄 CMakeLists.txt
📄 Test_roundTrip.cpp
📄 Test_vLoadHalf.cpp
📄 Test_vStoreHalf.cpp
📄 cl_utils.cpp
📄 cl_utils.h
📄 main.cpp
📄 test_config.h
📄 tests.h
📁 images
📄 CMakeLists.txt
📁 clCopyImage
📄 CMakeLists.txt
📄 main.cpp
📄 test_copy_1D.cpp
📄 test_copy_1D_array.cpp
📄 test_copy_2D.cpp
📄 test_copy_2D_2D_array.cpp
📄 test_copy_2D_3D.cpp
📄 test_copy_2D_array.cpp
📄 test_copy_3D.cpp
📄 test_copy_3D_2D_array.cpp
📄 test_copy_generic.cpp
📄 test_loops.cpp
📁 clFillImage
📄 CMakeLists.txt
📄 main.cpp
📄 test_fill_1D.cpp
📄 test_fill_1D_array.cpp
📄 test_fill_2D.cpp
📄 test_fill_2D_array.cpp
📄 test_fill_3D.cpp
📄 test_fill_generic.cpp
📄 test_loops.cpp
📁 clGetInfo
📄 CMakeLists.txt
📄 main.cpp
📄 test_1D.cpp
📄 test_1D_2D_array.cpp
📄 test_2D.cpp
📄 test_3D.cpp
📄 test_loops.cpp
📁 clReadWriteImage
📄 CMakeLists.txt
📄 main.cpp
📄 test_loops.cpp
📄 test_read_1D.cpp
📄 test_read_1D_array.cpp
📄 test_read_2D.cpp
📄 test_read_2D_array.cpp
📄 test_read_3D.cpp
📄 common.cpp
📄 common.h
📁 kernel_image_methods
📄 CMakeLists.txt
📄 main.cpp
📄 test_1D.cpp
📄 test_1D_array.cpp
📄 test_2D.cpp
📄 test_2D_array.cpp
📄 test_3D.cpp
📄 test_loops.cpp
📁 kernel_read_write
📄 CMakeLists.txt
📄 main.cpp
📄 test_common.cpp
📄 test_common.h
📄 test_iterations.cpp
📄 test_loops.cpp
📄 test_read_1D.cpp
📄 test_read_1D_array.cpp
📄 test_read_2D_array.cpp
📄 test_read_3D.cpp
📄 test_write_1D.cpp
📄 test_write_1D_array.cpp
📄 test_write_2D_array.cpp
📄 test_write_3D.cpp
📄 test_write_image.cpp
📁 samplerlessReads
📄 CMakeLists.txt
📄 main.cpp
📄 test_iterations.cpp
📄 test_loops.cpp
📄 test_read_1D.cpp
📄 test_read_1D_array.cpp
📄 test_read_1D_buffer.cpp
📄 test_read_2D_array.cpp
📄 test_read_3D.cpp
📄 testBase.h
📁 integer_ops
📄 CMakeLists.txt
📄 main.cpp
📄 procs.h
📄 testBase.h
📄 test_abs.cpp
📄 test_absdiff.cpp
📄 test_add_sat.cpp
📄 test_int_basic_ops.cpp
📄 test_integers.cpp
📄 test_intmad24.cpp
📄 test_intmul24.cpp
📄 test_popcount.cpp
📄 test_sub_sat.cpp
📄 test_unary_ops.cpp
📄 test_upsample.cpp
📄 verification_and_generation_functions.cpp
📁 math_brute_force
📄 CMakeLists.txt
📄 README.txt
📄 binary_double.cpp
📄 binary_float.cpp
📄 binary_i_double.cpp
📄 binary_i_float.cpp
📄 binary_operator_double.cpp
📄 binary_operator_float.cpp
📄 binary_two_results_i_double.cpp
📄 binary_two_results_i_float.cpp
📄 function_list.cpp
📄 function_list.h
📄 i_unary_double.cpp
📄 i_unary_float.cpp
📄 macro_binary_double.cpp
📄 macro_binary_float.cpp
📄 macro_unary_double.cpp
📄 macro_unary_float.cpp
📄 mad_double.cpp
📄 mad_float.cpp
📄 main.cpp
📄 reference_math.cpp
📄 reference_math.h
📄 run_math_brute_force_in_parallel.py
📄 sleep.cpp
📄 sleep.h
📄 ternary_double.cpp
📄 ternary_float.cpp
📄 test_functions.h
📄 unary_double.cpp
📄 unary_float.cpp
📄 unary_two_results_double.cpp
📄 unary_two_results_float.cpp
📄 unary_two_results_i_double.cpp
📄 unary_two_results_i_float.cpp
📄 unary_u_double.cpp
📄 unary_u_float.cpp
📄 utility.cpp
📄 utility.h
📁 mem_host_flags
📄 CMakeLists.txt
📄 C_host_memory_block.h
📄 checker.h
📄 checker_image_mem_host_no_access.hpp
📄 checker_image_mem_host_read_only.hpp
📄 checker_image_mem_host_write_only.hpp
📄 checker_mem_host_no_access.hpp
📄 checker_mem_host_read_only.hpp
📄 checker_mem_host_write_only.hpp
📄 main.cpp
📄 mem_host_buffer.cpp
📄 mem_host_image.cpp
📄 procs.h
📄 testBase.h
📁 multiple_device_context
📄 CMakeLists.txt
📄 main.cpp
📄 procs.h
📄 testBase.h
📄 test_multiple_contexts.cpp
📄 test_multiple_devices.cpp
📁 non_uniform_work_group
📄 CMakeLists.txt
📄 TestNonUniformWorkGroup.cpp
📄 TestNonUniformWorkGroup.h
📄 main.cpp
📄 procs.h
📄 test_advanced_2d.cpp
📄 test_advanced_3d.cpp
📄 test_advanced_other.cpp
📄 test_basic.cpp
📄 tools.cpp
📄 tools.h
📄 opencl_conformance_tests_conversions.csv
📄 opencl_conformance_tests_d3d.csv
📄 opencl_conformance_tests_full.csv
📄 opencl_conformance_tests_full_binary.csv
📄 opencl_conformance_tests_full_no_math_or_conversions.csv
📄 opencl_conformance_tests_full_spirv.csv
📄 opencl_conformance_tests_math.csv
📄 opencl_conformance_tests_quick.csv
📁 pipes
📄 CMakeLists.txt
📄 kernels.h
📄 main.cpp
📄 procs.h
📄 test_pipe_info.cpp
📄 test_pipe_limits.cpp
📄 test_pipe_query_functions.cpp
📄 test_pipe_read_write.cpp
📄 test_pipe_readwrite_errors.cpp
📄 test_pipe_subgroups.cpp
📁 printf
📄 CMakeLists.txt
📄 test_printf.cpp
📄 test_printf.h
📄 util_printf.cpp
📁 profiling
📄 CMakeLists.txt
📄 copy.cpp
📄 execute.cpp
📄 execute_multipass.cpp
📄 main.cpp
📄 procs.h
📄 readArray.cpp
📄 readImage.cpp
📄 writeArray.cpp
📄 writeImage.cpp
📁 relationals
📄 CMakeLists.txt
📄 main.cpp
📄 procs.h
📄 testBase.h
📄 test_comparisons_double.cpp
📄 test_comparisons_float.cpp
📄 test_relationals.cpp
📄 test_shuffles.cpp
📄 run_conformance.py
📁 select
📄 CMakeLists.txt
📄 test_select.cpp
📄 test_select.h
📄 util_select.cpp
📁 spir
📄 CMakeLists.txt
📄 api.zip
📄 atomics.zip
📄 basic.zip
📄 binary_type.zip
📄 commonfns.zip
📄 compile_and_link.zip
📄 conversions.zip
📄 datagen.cpp
📄 datagen.h
📄 enum_values.zip
📄 exceptions.h
📄 geometrics.zip
📄 half.zip
📄 images_kernel_read_write.zip
📄 images_samplerlessRead.zip
📄 integer_ops.zip
📄 kernel_attributes.zip
📄 kernel_image_methods.zip
📄 kernelargs.cpp
📄 kernelargs.h
📄 khr.csv
📄 main.cpp
📄 math_brute_force.zip
📄 printf.zip
📄 profiling.zip
📄 relationals.zip
📄 run_build_test.cpp
📄 run_build_test.h
📄 run_services.cpp
📄 run_services.h
📄 sampler_enumeration.zip
📄 select.zip
📄 typeinfo.h
📄 vec_align.zip
📄 vec_step.zip
📁 spirv_new
📄 CMakeLists.txt
📄 README.txt
📄 assemble_spirv.py
📄 main.cpp
📄 procs.h
📁 spirv_asm
📄 atomic_dec_global.spvasm32
📄 atomic_dec_global.spvasm64
📄 atomic_inc_global.spvasm32
📄 atomic_inc_global.spvasm64
📄 branch_conditional.spvasm32
📄 branch_conditional.spvasm64
📄 branch_conditional_weighted.spvasm32
📄 branch_conditional_weighted.spvasm64
📄 branch_simple.spvasm32
📄 branch_simple.spvasm64
📄 composite_construct_int4.spvasm32
📄 composite_construct_int4.spvasm64
📄 composite_construct_struct.spvasm32
📄 composite_construct_struct.spvasm64
📄 constant_char_simple.spvasm32
📄 constant_char_simple.spvasm64
📄 constant_double_simple.spvasm32
📄 constant_double_simple.spvasm64
📄 constant_false_simple.spvasm32
📄 constant_false_simple.spvasm64
📄 constant_float_simple.spvasm32
📄 constant_float_simple.spvasm64
📄 constant_half_simple.spvasm32
📄 constant_half_simple.spvasm64
📄 constant_int3_simple.spvasm32
📄 constant_int3_simple.spvasm64
📄 constant_int4_simple.spvasm32
📄 constant_int4_simple.spvasm64
📄 constant_int_simple.spvasm32
📄 constant_int_simple.spvasm64
📄 constant_long_simple.spvasm32
📄 constant_long_simple.spvasm64
📄 constant_short_simple.spvasm32
📄 constant_short_simple.spvasm64
📄 constant_struct_int_char_simple.spvasm32
📄 constant_struct_int_char_simple.spvasm64
📄 constant_true_simple.spvasm64
📄 constant_uchar_simple.spvasm32
📄 constant_uchar_simple.spvasm64
📄 constant_uint_simple.spvasm32
📄 constant_uint_simple.spvasm64
📄 constant_ulong_simple.spvasm32
📄 constant_ulong_simple.spvasm64
📄 constant_ushort_simple.spvasm32
📄 constant_ushort_simple.spvasm64
📄 copy_char_simple.spvasm32
📄 copy_char_simple.spvasm64
📄 copy_double_simple.spvasm32
📄 copy_double_simple.spvasm64
📄 copy_float_simple.spvasm32
📄 copy_float_simple.spvasm64
📄 copy_half_simple.spvasm32
📄 copy_half_simple.spvasm64
📄 copy_int3_simple.spvasm32
📄 copy_int3_simple.spvasm64
📄 copy_int4_simple.spvasm32
📄 copy_int4_simple.spvasm64
📄 copy_int_simple.spvasm32
📄 copy_int_simple.spvasm64
📄 copy_long_simple.spvasm32
📄 copy_long_simple.spvasm64
📄 copy_short_simple.spvasm32
📄 copy_short_simple.spvasm64
📄 copy_struct_int_char_simple.spvasm32
📄 copy_struct_int_char_simple.spvasm64
📄 copy_struct_int_float_simple.spvasm32
📄 copy_struct_int_float_simple.spvasm64
📄 copy_uint_simple.spvasm64
📄 copy_ulong_simple.spvasm32
📄 copy_ulong_simple.spvasm64
📄 copy_ushort_simple.spvasm32
📄 copy_ushort_simple.spvasm64
📄 decorate_aliased.spvasm32
📄 decorate_aliased.spvasm64
📄 decorate_alignment.spvasm32
📄 decorate_alignment.spvasm64
📄 decorate_constant.spvasm32
📄 decorate_constant.spvasm64
📄 decorate_constant_fail.spvasm32
📄 decorate_constant_fail.spvasm64
📄 decorate_cpacked.spvasm32
📄 decorate_cpacked.spvasm64
📄 decorate_restrict.spvasm32
📄 decorate_restrict.spvasm64
📄 decorate_rounding_rte_double_long.spvasm32
📄 decorate_rounding_rte_double_long.spvasm64
📄 decorate_rounding_rte_float_int.spvasm32
📄 decorate_rounding_rte_float_int.spvasm64
📄 decorate_rounding_rtn_double_long.spvasm32
📄 decorate_rounding_rtn_double_long.spvasm64
📄 decorate_rounding_rtn_float_int.spvasm32
📄 decorate_rounding_rtn_float_int.spvasm64
📄 decorate_rounding_rtp_double_long.spvasm32
📄 decorate_rounding_rtp_double_long.spvasm64
📄 decorate_rounding_rtp_float_int.spvasm32
📄 decorate_rounding_rtp_float_int.spvasm64
📄 decorate_rounding_rtz_double_long.spvasm32
📄 decorate_rounding_rtz_double_long.spvasm64
📄 decorate_saturated_conversion_int.spvasm64
📄 decorate_saturated_conversion_short.spvasm32
📄 decorate_saturated_conversion_short.spvasm64
📄 decorate_saturated_conversion_uchar.spvasm32
📄 decorate_saturated_conversion_uchar.spvasm64
📄 decorate_saturated_conversion_uint.spvasm32
📄 decorate_saturated_conversion_uint.spvasm64
📄 decorate_saturated_conversion_ushort.spvasm32
📄 decorate_saturated_conversion_ushort.spvasm64
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fadd_int.spvasm32
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fadd_int.spvasm64
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fadd_uint.spvasm32
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fadd_uint.spvasm64
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fmul_int.spvasm32
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fmul_int.spvasm64
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fmul_uint.spvasm32
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fmul_uint.spvasm64
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fnegate_int.spvasm32
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fnegate_int.spvasm64
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fshiftleft_int.spvasm32
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fshiftleft_int.spvasm64
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fshiftleft_uint.spvasm32
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fshiftleft_uint.spvasm64
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fsub_int.spvasm32
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fsub_int.spvasm64
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fsub_uint.spvasm32
📄 ext_cl_khr_spirv_no_integer_wrap_decoration_fsub_uint.spvasm64
📄 fadd_double.spvasm32
📄 fadd_double.spvasm64
📄 fadd_double2.spvasm32
📄 fadd_double2.spvasm64
📄 fadd_half.spvasm64
📄 fdiv_double.spvasm32
📄 fdiv_double.spvasm64
📄 fdiv_double2.spvasm32
📄 fdiv_double2.spvasm64
📄 fdiv_float.spvasm32
📄 fdiv_float.spvasm64
📄 fdiv_float4.spvasm32
📄 fdiv_float4.spvasm64
📄 fdiv_half.spvasm32
📄 fdiv_half.spvasm64
📄 fmod_double.spvasm32
📄 fmod_double.spvasm64
📄 fmod_double2.spvasm32
📄 fmod_double2.spvasm64
📄 fmod_float.spvasm32
📄 fmod_float.spvasm64
📄 fmod_float4.spvasm32
📄 fmod_float4.spvasm64
📄 fmod_half.spvasm32
📄 fmod_half.spvasm64
📄 fmul_double.spvasm32
📄 fmul_double.spvasm64
📄 fmul_double2.spvasm32
📄 fmul_double2.spvasm64
📄 fmul_float.spvasm32
📄 frem_double.spvasm32
📄 frem_double.spvasm64
📄 frem_double2.spvasm32
📄 frem_double2.spvasm64
📄 frem_float.spvasm32
📄 frem_float.spvasm64
📄 frem_float4.spvasm32
📄 frem_float4.spvasm64
📄 frem_half.spvasm32
📄 frem_half.spvasm64
📄 fsub_double.spvasm32
📄 fsub_double.spvasm64
📄 fsub_double2.spvasm32
📄 fsub_double2.spvasm64
📄 fsub_float.spvasm32
📄 fsub_float.spvasm64
📄 fsub_float4.spvasm32
📄 fsub_float4.spvasm64
📄 fsub_half.spvasm32
📄 fsub_half.spvasm64
📄 label_simple.spvasm32
📄 label_simple.spvasm64
📄 lifetime_simple.spvasm32
📄 lifetime_simple.spvasm64
📄 linkage_export.spvasm32
📄 linkage_export.spvasm64
📄 linkage_import.spvasm32
📄 linkage_import.spvasm64
📄 loop_merge_branch_conditional_dont_unroll.spvasm32
📄 loop_merge_branch_conditional_dont_unroll.spvasm64
📄 loop_merge_branch_conditional_none.spvasm32
📄 loop_merge_branch_none.spvasm32
📄 loop_merge_branch_none.spvasm64
📄 loop_merge_branch_unroll.spvasm32
📄 loop_merge_branch_unroll.spvasm64
📄 op_function_const.spvasm32
📄 op_function_const.spvasm64
📄 op_function_inline.spvasm32
📄 op_function_inline.spvasm64
📄 op_function_noinline.spvasm32
📄 op_function_noinline.spvasm64
📄 op_function_none.spvasm32
📄 op_function_none.spvasm64
📄 op_function_pure.spvasm32
📄 op_function_pure.spvasm64
📄 op_function_pure_ptr.spvasm32
📄 op_function_pure_ptr.spvasm64
📄 op_neg_double.spvasm32
📄 op_neg_double.spvasm64
📄 op_neg_float.spvasm32
📄 op_neg_float.spvasm64
📄 op_neg_float4.spvasm32
📄 op_neg_float4.spvasm64
📄 op_neg_int.spvasm32
📄 op_neg_int.spvasm64
📄 op_neg_int4.spvasm32
📄 op_neg_int4.spvasm64
📄 op_neg_long.spvasm32
📄 op_neg_long.spvasm64
📄 op_neg_short.spvasm32
📄 op_neg_short.spvasm64
📄 op_not_int.spvasm32
📄 op_not_short.spvasm32
📄 op_not_short.spvasm64
📄 op_spec_constant_double_simple.spvasm32
📄 op_spec_constant_double_simple.spvasm64
📄 op_spec_constant_false_simple.spvasm32
📄 op_spec_constant_false_simple.spvasm64
📄 op_spec_constant_float_simple.spvasm32
📄 op_spec_constant_float_simple.spvasm64
📄 op_spec_constant_half_simple.spvasm32
📄 op_spec_constant_half_simple.spvasm64
📄 op_spec_constant_true_simple.spvasm32
📄 op_spec_constant_true_simple.spvasm64
📄 op_spec_constant_uchar_simple.spvasm32
📄 op_spec_constant_uchar_simple.spvasm64
📄 op_spec_constant_uint_simple.spvasm32
📄 op_spec_constant_uint_simple.spvasm64
📄 op_spec_constant_ulong_simple.spvasm32
📄 op_spec_constant_ulong_simple.spvasm64
📄 op_spec_constant_ushort_simple.spvasm32
📄 op_spec_constant_ushort_simple.spvasm64
📄 opaque.spvasm32
📄 opaque.spvasm64
📄 phi_2.spvasm32
📄 phi_2.spvasm64
📄 phi_3.spvasm32
📄 phi_3.spvasm64
📄 phi_4.spvasm32
📄 phi_4.spvasm64
📄 select_if_dont_flatten.spvasm32
📄 select_if_dont_flatten.spvasm64
📄 select_if_flatten.spvasm32
📄 select_switch_flatten.spvasm32
📄 select_switch_flatten.spvasm64
📄 select_switch_none.spvasm32
📄 select_switch_none.spvasm64
📄 undef_char_simple.spvasm32
📄 undef_char_simple.spvasm64
📄 undef_double_simple.spvasm32
📄 undef_double_simple.spvasm64
📄 undef_false_simple.spvasm32
📄 undef_false_simple.spvasm64
📄 undef_float_simple.spvasm32
📄 undef_float_simple.spvasm64
📄 undef_half_simple.spvasm32
📄 undef_half_simple.spvasm64
📄 undef_int3_simple.spvasm32
📄 undef_int3_simple.spvasm64
📄 undef_int4_simple.spvasm32
📄 undef_int4_simple.spvasm64
📄 undef_int_simple.spvasm32
📄 undef_int_simple.spvasm64
📄 undef_long_simple.spvasm32
📄 undef_long_simple.spvasm64
📄 undef_short_simple.spvasm32
📄 undef_short_simple.spvasm64
📄 undef_struct_int_char_simple.spvasm32
📄 undef_struct_int_char_simple.spvasm64
📄 undef_struct_int_float_simple.spvasm32
📄 undef_struct_int_float_simple.spvasm64
📄 undef_struct_struct_simple.spvasm32
📄 undef_struct_struct_simple.spvasm64
📄 undef_true_simple.spvasm32
📄 undef_ulong_simple.spvasm32
📄 undef_ulong_simple.spvasm64
📄 undef_ushort_simple.spvasm32
📄 undef_ushort_simple.spvasm64
📄 unreachable_simple.spvasm32
📄 unreachable_simple.spvasm64
📄 vector_char16_extract.spvasm32
📄 vector_char16_extract.spvasm64
📄 vector_char16_insert.spvasm32
📄 vector_char16_insert.spvasm64
📄 vector_double2_extract.spvasm32
📄 vector_double2_extract.spvasm64
📄 vector_double2_insert.spvasm32
📄 vector_double2_insert.spvasm64
📄 vector_float4_extract.spvasm32
📄 vector_float4_extract.spvasm64
📄 vector_float4_insert.spvasm32
📄 vector_float4_insert.spvasm64
📄 vector_int4_extract.spvasm32
📄 vector_int4_extract.spvasm64
📄 vector_int4_insert.spvasm32
📄 vector_int4_insert.spvasm64
📄 vector_long2_extract.spvasm32
📄 vector_long2_extract.spvasm64
📄 vector_long2_insert.spvasm32
📄 vector_long2_insert.spvasm64
📄 vector_times_scalar_double.spvasm32
📄 vector_times_scalar_double.spvasm64
📄 vector_times_scalar_float.spvasm32
📄 vector_times_scalar_float.spvasm64
📄 testBase.h
📄 test_cl_khr_spirv_no_integer_wrap_decoration.cpp
📄 test_decorate.cpp
📄 test_get_program_il.cpp
📄 test_linkage.cpp
📄 test_op_atomic.cpp
📄 test_op_branch.cpp
📄 test_op_branch_conditional.cpp
📄 test_op_composite_construct.cpp
📄 test_op_constant.cpp
📄 test_op_copy_object.cpp
📄 test_op_fmath.cpp
📄 test_op_function.cpp
📄 test_op_lifetime.cpp
📄 test_op_loop_merge.cpp
📄 test_op_negate.cpp
📄 test_op_opaque.cpp
📄 test_op_phi.cpp
📄 test_op_selection_merge.cpp
📄 test_op_spec_constant.cpp
📄 test_op_undef.cpp
📄 test_op_vector_extract.cpp
📄 test_op_vector_insert.cpp
📄 test_op_vector_times_scalar.cpp
📄 types.hpp
📁 subgroups
📄 CMakeLists.txt
📄 main.cpp
📄 procs.h
📄 subgroup_common_kernels.cpp
📄 subgroup_common_kernels.h
📄 subgroup_common_templates.h
📄 subhelpers.h
📄 test_barrier.cpp
📄 test_ifp.cpp
📄 test_queries.cpp
📄 test_subgroup.cpp
📄 test_subgroup_ballot.cpp
📄 test_subgroup_clustered_reduce.cpp
📄 test_subgroup_extended_types.cpp
📄 test_subgroup_non_uniform_arithmetic.cpp
📄 test_subgroup_non_uniform_vote.cpp
📄 test_subgroup_shuffle.cpp
📄 test_subgroup_shuffle_relative.cpp
📄 test_workitem.cpp
📄 submission_details_template.txt
📁 thread_dimensions
📄 CMakeLists.txt
📄 main.cpp
📄 procs.h
📄 test_thread_dimensions.cpp
📁 vectors
📄 CMakeLists.txt
📄 defines.h
📄 globals.cpp
📄 main.cpp
📄 procs.h
📄 structs.cpp
📄 structs.h
📄 testBase.h
📄 test_step.cpp
📄 test_vec_align.cpp
📄 type_replacer.cpp
📄 type_replacer.h
📁 workgroups
📄 CMakeLists.txt
📄 main.cpp
📄 procs.h
📄 testBase.h
📄 test_wg_all.cpp
📄 test_wg_any.cpp
📄 test_wg_broadcast.cpp
📄 test_wg_reduce.cpp
📄 test_wg_reduce_max.cpp
📄 test_wg_reduce_min.cpp
📄 test_wg_scan_exclusive_add.cpp
📄 test_wg_scan_exclusive_max.cpp
📄 test_wg_scan_exclusive_min.cpp
📄 test_wg_scan_inclusive_add.cpp
📄 test_wg_scan_inclusive_max.cpp
📄 test_wg_scan_inclusive_min.cpp 
//
// Copyright (c) 2020 The Khronos Group Inc.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
//    http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
#ifndef SUBGROUPCOMMONTEMPLATES_H
#define SUBGROUPCOMMONTEMPLATES_H

#include "typeWrappers.h"
#include <bitset>
#include "CL/cl_half.h"
#include "subhelpers.h"

#include <set>

typedef std::bitset<128> bs128;
static cl_uint4 generate_bit_mask(cl_uint subgroup_local_id,
                                  const std::string &mask_type,
                                  cl_uint max_sub_group_size)
{
    bs128 mask128;
    cl_uint4 mask;
    cl_uint pos = subgroup_local_id;
    if (mask_type == "eq") mask128.set(pos);
    if (mask_type == "le" || mask_type == "lt")
    {
        for (cl_uint i = 0; i <= pos; i++) mask128.set(i);
        if (mask_type == "lt") mask128.reset(pos);
    }
    if (mask_type == "ge" || mask_type == "gt")
    {
        for (cl_uint i = pos; i < max_sub_group_size; i++) mask128.set(i);
        if (mask_type == "gt") mask128.reset(pos);
    }

    // convert std::bitset<128> to uint4
    auto const uint_mask = bs128{ static_cast<unsigned long>(-1) };
    mask.s0 = (mask128 & uint_mask).to_ulong();
    mask128 >>= 32;
    mask.s1 = (mask128 & uint_mask).to_ulong();
    mask128 >>= 32;
    mask.s2 = (mask128 & uint_mask).to_ulong();
    mask128 >>= 32;
    mask.s3 = (mask128 & uint_mask).to_ulong();

    return mask;
}

// DESCRIPTION :
// sub_group_broadcast - each work_item registers it's own value.
// All work_items in subgroup takes one value from only one (any) work_item
// sub_group_broadcast_first - same as type 0. All work_items in
// subgroup takes only one value from only one chosen (the smallest subgroup ID)
// work_item
// sub_group_non_uniform_broadcast - same as type 0 but
// only 4 work_items from subgroup enter the code (are active)
template <typename Ty, SubgroupsBroadcastOp operation> struct BC
{
    static void gen(Ty *x, Ty *t, cl_int *m, const WorkGroupParams &test_params)
    {
        int i, ii, j, k, n;
        int ng = test_params.global_workgroup_size;
        int nw = test_params.local_workgroup_size;
        int ns = test_params.subgroup_size;
        int nj = (nw + ns - 1) / ns;
        int d = ns > 100 ? 100 : ns;
        int non_uniform_size = ng % nw;
        ng = ng / nw;
        int last_subgroup_size = 0;
        ii = 0;

        log_info("  sub_group_%s(%s)...\n", operation_names(operation),
                 TypeManager<Ty>::name());
        if (non_uniform_size)
        {
            log_info("  non uniform work group size mode ON\n");
            ng++;
        }
        for (k = 0; k < ng; ++k)
        { // for each work_group
            if (non_uniform_size && k == ng - 1)
            {
                set_last_workgroup_params(non_uniform_size, nj, ns, nw,
                                          last_subgroup_size);
            }
            for (j = 0; j < nj; ++j)
            { // for each subgroup
                ii = j * ns;
                if (last_subgroup_size && j == nj - 1)
                {
                    n = last_subgroup_size;
                }
                else
                {
                    n = ii + ns > nw ? nw - ii : ns;
                }
                int bcast_if = 0;
                int bcast_elseif = 0;
                int bcast_index = (int)(genrand_int32(gMTdata) & 0x7fffffff)
                    % (d > n ? n : d);
                // l - calculate subgroup local id from which value will be
                // broadcasted (one the same value for whole subgroup)
                if (operation != SubgroupsBroadcastOp::broadcast)
                {
                    // reduce brodcasting index in case of non_uniform and
                    // last workgroup last subgroup
                    if (last_subgroup_size && j == nj - 1
                        && last_subgroup_size < NR_OF_ACTIVE_WORK_ITEMS)
                    {
                        bcast_if = bcast_index % last_subgroup_size;
                        bcast_elseif = bcast_if;
                    }
                    else
                    {
                        bcast_if = bcast_index % NR_OF_ACTIVE_WORK_ITEMS;
                        bcast_elseif = NR_OF_ACTIVE_WORK_ITEMS
                            + bcast_index % (n - NR_OF_ACTIVE_WORK_ITEMS);
                    }
                }

                for (i = 0; i < n; ++i)
                {
                    if (operation == SubgroupsBroadcastOp::broadcast)
                    {
                        int midx = 4 * ii + 4 * i + 2;
                        m[midx] = (cl_int)bcast_index;
                    }
                    else
                    {
                        if (i < NR_OF_ACTIVE_WORK_ITEMS)
                        {
                            // index of the third
                            // element int the vector.
                            int midx = 4 * ii + 4 * i + 2;
                            // storing information about
                            // broadcasting index -
                            // earlier calculated
                            m[midx] = (cl_int)bcast_if;
                        }
                        else
                        { // index of the third
                          // element int the vector.
                            int midx = 4 * ii + 4 * i + 3;
                            m[midx] = (cl_int)bcast_elseif;
                        }
                    }

                    // calculate value for broadcasting
                    cl_ulong number = genrand_int64(gMTdata);
                    set_value(t[ii + i], number);
                }
            }
            // Now map into work group using map from device
            for (j = 0; j < nw; ++j)
            { // for each element in work_group
                // calculate index as number of subgroup
                // plus subgroup local id
                x[j] = t[j];
            }
            x += nw;
            m += 4 * nw;
        }
    }

    static int chk(Ty *x, Ty *y, Ty *mx, Ty *my, cl_int *m,
                   const WorkGroupParams &test_params)
    {
        int ii, i, j, k, l, n;
        int ng = test_params.global_workgroup_size;
        int nw = test_params.local_workgroup_size;
        int ns = test_params.subgroup_size;
        int nj = (nw + ns - 1) / ns;
        Ty tr, rr;
        int non_uniform_size = ng % nw;
        ng = ng / nw;
        int last_subgroup_size = 0;
        if (non_uniform_size) ng++;

        for (k = 0; k < ng; ++k)
        { // for each work_group
            if (non_uniform_size && k == ng - 1)
            {
                set_last_workgroup_params(non_uniform_size, nj, ns, nw,
                                          last_subgroup_size);
            }
            for (j = 0; j < nw; ++j)
            { // inside the work_group
                mx[j] = x[j]; // read host inputs for work_group
                my[j] = y[j]; // read device outputs for work_group
            }

            for (j = 0; j < nj; ++j)
            { // for each subgroup
                ii = j * ns;
                if (last_subgroup_size && j == nj - 1)
                {
                    n = last_subgroup_size;
                }
                else
                {
                    n = ii + ns > nw ? nw - ii : ns;
                }

                // Check result
                if (operation == SubgroupsBroadcastOp::broadcast_first)
                {
                    int lowest_active_id = -1;
                    for (i = 0; i < n; ++i)
                    {

                        lowest_active_id = i < NR_OF_ACTIVE_WORK_ITEMS
                            ? 0
                            : NR_OF_ACTIVE_WORK_ITEMS;
                        //  findout if broadcasted
                        //  value is the same
                        tr = mx[ii + lowest_active_id];
                        //  findout if broadcasted to all
                        rr = my[ii + i];

                        if (!compare(rr, tr))
                        {
                            log_error(
                                "ERROR: sub_group_broadcast_first(%s) "
                                "mismatch "
                                "for local id %d in sub group %d in group "
                                "%d\n",
                                TypeManager<Ty>::name(), i, j, k);
                            return TEST_FAIL;
                        }
                    }
                }
                else
                {
                    for (i = 0; i < n; ++i)
                    {
                        if (operation == SubgroupsBroadcastOp::broadcast)
                        {
                            int midx = 4 * ii + 4 * i + 2;
                            l = (int)m[midx];
                            tr = mx[ii + l];
                        }
                        else
                        {
                            if (i < NR_OF_ACTIVE_WORK_ITEMS)
                            { // take index of array where info
                              // which work_item will be
                              // broadcast its value is stored
                                int midx = 4 * ii + 4 * i + 2;
                                // take subgroup local id of
                                // this work_item
                                l = (int)m[midx];
                                // take value generated on host
                                // for this work_item
                                tr = mx[ii + l];
                            }
                            else
                            {
                                int midx = 4 * ii + 4 * i + 3;
                                l = (int)m[midx];
                                tr = mx[ii + l];
                            }
                        }
                        rr = my[ii + i]; // read device outputs for
                                         // work_item in the subgroup

                        if (!compare(rr, tr))
                        {
                            log_error("ERROR: sub_group_%s(%s) "
                                      "mismatch for local id %d in sub "
                                      "group %d in group %d - got %lu "
                                      "expected %lu\n",
                                      operation_names(operation),
                                      TypeManager<Ty>::name(), i, j, k, rr, tr);
                            return TEST_FAIL;
                        }
                    }
                }
            }
            x += nw;
            y += nw;
            m += 4 * nw;
        }
        log_info("  sub_group_%s(%s)... passed\n", operation_names(operation),
                 TypeManager<Ty>::name());
        return TEST_PASS;
    }
};

static float to_float(subgroups::cl_half x) { return cl_half_to_float(x.data); }

static subgroups::cl_half to_half(float x)
{
    subgroups::cl_half value;
    value.data = cl_half_from_float(x, CL_HALF_RTE);
    return value;
}

// for integer types
template <typename Ty> inline Ty calculate(Ty a, Ty b, ArithmeticOp operation)
{
    switch (operation)
    {
        case ArithmeticOp::add_: return a + b;
        case ArithmeticOp::max_: return a > b ? a : b;
        case ArithmeticOp::min_: return a < b ? a : b;
        case ArithmeticOp::mul_: return a * b;
        case ArithmeticOp::and_: return a & b;
        case ArithmeticOp::or_: return a | b;
        case ArithmeticOp::xor_: return a ^ b;
        case ArithmeticOp::logical_and: return a && b;
        case ArithmeticOp::logical_or: return a || b;
        case ArithmeticOp::logical_xor: return !a ^ !b;
        default: log_error("Unknown operation request"); break;
    }
    return 0;
}
// Specialize for floating points.
template <>
inline cl_double calculate(cl_double a, cl_double b, ArithmeticOp operation)
{
    switch (operation)
    {
        case ArithmeticOp::add_: {
            return a + b;
        }
        case ArithmeticOp::max_: {
            return a > b ? a : b;
        }
        case ArithmeticOp::min_: {
            return a < b ? a : b;
        }
        case ArithmeticOp::mul_: {
            return a * b;
        }
        default: log_error("Unknown operation request"); break;
    }
    return 0;
}

template <>
inline cl_float calculate(cl_float a, cl_float b, ArithmeticOp operation)
{
    switch (operation)
    {
        case ArithmeticOp::add_: {
            return a + b;
        }
        case ArithmeticOp::max_: {
            return a > b ? a : b;
        }
        case ArithmeticOp::min_: {
            return a < b ? a : b;
        }
        case ArithmeticOp::mul_: {
            return a * b;
        }
        default: log_error("Unknown operation request"); break;
    }
    return 0;
}

template <>
inline subgroups::cl_half calculate(subgroups::cl_half a, subgroups::cl_half b,
                                    ArithmeticOp operation)
{
    switch (operation)
    {
        case ArithmeticOp::add_: return to_half(to_float(a) + to_float(b));
        case ArithmeticOp::max_:
            return to_float(a) > to_float(b) || is_half_nan(b.data) ? a : b;
        case ArithmeticOp::min_:
            return to_float(a) < to_float(b) || is_half_nan(b.data) ? a : b;
        case ArithmeticOp::mul_: return to_half(to_float(a) * to_float(b));
        default: log_error("Unknown operation request"); break;
    }
    return to_half(0);
}

template <typename Ty> bool is_floating_point()
{
    return std::is_floating_point<Ty>::value
        || std::is_same<Ty, subgroups::cl_half>::value;
}

template <typename Ty, ArithmeticOp operation>
void genrand(Ty *x, Ty *t, cl_int *m, int ns, int nw, int ng)
{
    int nj = (nw + ns - 1) / ns;

    for (int k = 0; k < ng; ++k)
    {
        for (int j = 0; j < nj; ++j)
        {
            int ii = j * ns;
            int n = ii + ns > nw ? nw - ii : ns;

            for (int i = 0; i < n; ++i)
            {
                cl_ulong out_value;
                double y;
                if (operation == ArithmeticOp::mul_
                    || operation == ArithmeticOp::add_)
                {
                    // work around to avoid overflow, do not use 0 for
                    // multiplication
                    out_value = (genrand_int32(gMTdata) % 4) + 1;
                }
                else
                {
                    out_value = genrand_int64(gMTdata) % (32 * n);
                    if ((operation == ArithmeticOp::logical_and
                         || operation == ArithmeticOp::logical_or
                         || operation == ArithmeticOp::logical_xor)
                        && ((out_value >> 32) & 1) == 0)
                        out_value = 0; // increase probability of false
                }
                set_value(t[ii + i], out_value);
            }
        }

        // Now map into work group using map from device
        for (int j = 0; j < nw; ++j)
        {
            x[j] = t[j];
        }

        x += nw;
        m += 4 * nw;
    }
}

template <typename Ty, ShuffleOp operation> struct SHF
{
    static void gen(Ty *x, Ty *t, cl_int *m, const WorkGroupParams &test_params)
    {
        int i, ii, j, k, l, n, delta;
        int nw = test_params.local_workgroup_size;
        int ns = test_params.subgroup_size;
        int ng = test_params.global_workgroup_size;
        int nj = (nw + ns - 1) / ns;
        int d = ns > 100 ? 100 : ns;
        ii = 0;
        ng = ng / nw;
        log_info("  sub_group_%s(%s)...\n", operation_names(operation),
                 TypeManager<Ty>::name());
        for (k = 0; k < ng; ++k)
        { // for each work_group
            for (j = 0; j < nj; ++j)
            { // for each subgroup
                ii = j * ns;
                n = ii + ns > nw ? nw - ii : ns;
                for (i = 0; i < n; ++i)
                {
                    int midx = 4 * ii + 4 * i + 2;
                    l = (int)(genrand_int32(gMTdata) & 0x7fffffff)
                        % (d > n ? n : d);
                    switch (operation)
                    {
                        case ShuffleOp::shuffle:
                        case ShuffleOp::shuffle_xor:
                            // storing information about shuffle index
                            m[midx] = (cl_int)l;
                            break;
                        case ShuffleOp::shuffle_up:
                            delta = l; // calculate delta for shuffle up
                            if (i - delta < 0)
                            {
                                delta = i;
                            }
                            m[midx] = (cl_int)delta;
                            break;
                        case ShuffleOp::shuffle_down:
                            delta = l; // calculate delta for shuffle down
                            if (i + delta >= n)
                            {
                                delta = n - 1 - i;
                            }
                            m[midx] = (cl_int)delta;
                            break;
                        default: break;
                    }
                    cl_ulong number = genrand_int64(gMTdata);
                    set_value(t[ii + i], number);
                }
            }
            // Now map into work group using map from device
            for (j = 0; j < nw; ++j)
            { // for each element in work_group
                x[j] = t[j];
            }
            x += nw;
            m += 4 * nw;
        }
    }

    static int chk(Ty *x, Ty *y, Ty *mx, Ty *my, cl_int *m,
                   const WorkGroupParams &test_params)
    {
        int ii, i, j, k, l, n;
        int nw = test_params.local_workgroup_size;
        int ns = test_params.subgroup_size;
        int ng = test_params.global_workgroup_size;
        int nj = (nw + ns - 1) / ns;
        Ty tr, rr;
        ng = ng / nw;

        for (k = 0; k < ng; ++k)
        { // for each work_group
            for (j = 0; j < nw; ++j)
            { // inside the work_group
                mx[j] = x[j]; // read host inputs for work_group
                my[j] = y[j]; // read device outputs for work_group
            }

            for (j = 0; j < nj; ++j)
            { // for each subgroup
                ii = j * ns;
                n = ii + ns > nw ? nw - ii : ns;

                for (i = 0; i < n; ++i)
                { // inside the subgroup
                  // shuffle index storage
                    int midx = 4 * ii + 4 * i + 2;
                    l = (int)m[midx];
                    rr = my[ii + i];
                    switch (operation)
                    {
                        // shuffle basic - treat l as index
                        case ShuffleOp::shuffle: tr = mx[ii + l]; break;
                        // shuffle up - treat l as delta
                        case ShuffleOp::shuffle_up: tr = mx[ii + i - l]; break;
                        // shuffle up - treat l as delta
                        case ShuffleOp::shuffle_down:
                            tr = mx[ii + i + l];
                            break;
                        // shuffle xor - treat l as mask
                        case ShuffleOp::shuffle_xor:
                            tr = mx[ii + (i ^ l)];
                            break;
                        default: break;
                    }

                    if (!compare(rr, tr))
                    {
                        log_error("ERROR: sub_group_%s(%s) mismatch for "
                                  "local id %d in sub group %d in group %d\n",
                                  operation_names(operation),
                                  TypeManager<Ty>::name(), i, j, k);
                        return TEST_FAIL;
                    }
                }
            }
            x += nw;
            y += nw;
            m += 4 * nw;
        }
        log_info("  sub_group_%s(%s)... passed\n", operation_names(operation),
                 TypeManager<Ty>::name());
        return TEST_PASS;
    }
};

template <typename Ty, ArithmeticOp operation> struct SCEX_NU
{
    static void gen(Ty *x, Ty *t, cl_int *m, const WorkGroupParams &test_params)
    {
        int nw = test_params.local_workgroup_size;
        int ns = test_params.subgroup_size;
        int ng = test_params.global_workgroup_size;
        uint32_t work_items_mask = test_params.work_items_mask;
        ng = ng / nw;
        std::string func_name;
        work_items_mask ? func_name = "sub_group_non_uniform_scan_exclusive"
                        : func_name = "sub_group_scan_exclusive";
        log_info("  %s_%s(%s)...\n", func_name.c_str(),
                 operation_names(operation), TypeManager<Ty>::name());
        log_info("  test params: global size = %d local size = %d subgroups "
                 "size = %d work item mask = 0x%x \n",
                 test_params.global_workgroup_size, nw, ns, work_items_mask);
        genrand<Ty, operation>(x, t, m, ns, nw, ng);
    }

    static int chk(Ty *x, Ty *y, Ty *mx, Ty *my, cl_int *m,
                   const WorkGroupParams &test_params)
    {
        int ii, i, j, k, n;
        int nw = test_params.local_workgroup_size;
        int ns = test_params.subgroup_size;
        int ng = test_params.global_workgroup_size;
        uint32_t work_items_mask = test_params.work_items_mask;
        int nj = (nw + ns - 1) / ns;
        Ty tr, rr;
        ng = ng / nw;

        std::string func_name;
        work_items_mask ? func_name = "sub_group_non_uniform_scan_exclusive"
                        : func_name = "sub_group_scan_exclusive";

        uint32_t use_work_items_mask;
        // for uniform case take into consideration all workitems
        use_work_items_mask = !work_items_mask ? 0xFFFFFFFF : work_items_mask;
        for (k = 0; k < ng; ++k)
        { // for each work_group
            // Map to array indexed to array indexed by local ID and sub group
            for (j = 0; j < nw; ++j)
            { // inside the work_group
                mx[j] = x[j]; // read host inputs for work_group
                my[j] = y[j]; // read device outputs for work_group
            }
            for (j = 0; j < nj; ++j)
            {
                ii = j * ns;
                n = ii + ns > nw ? nw - ii : ns;
                std::set<int> active_work_items;
                for (i = 0; i < n; ++i)
                {
                    uint32_t check_work_item = 1 << (i % 32);
                    if (use_work_items_mask & check_work_item)
                    {
                        active_work_items.insert(i);
                    }
                }
                if (active_work_items.empty())
                {
                    log_info("  No acitve workitems in workgroup id = %d "
                             "subgroup id = %d - no calculation\n",
                             k, j);
                    continue;
                }
                else if (active_work_items.size() == 1)
                {
                    log_info("  One active workitem in workgroup id = %d "
                             "subgroup id = %d - no calculation\n",
                             k, j);
                    continue;
                }
                else
                {
                    tr = TypeManager<Ty>::identify_limits(operation);
                    int idx = 0;
                    for (const int &active_work_item : active_work_items)
                    {
                        rr = my[ii + active_work_item];
                        if (idx == 0) continue;

                        if (!compare_ordered(rr, tr))
                        {
                            log_error(
                                "ERROR: %s_%s(%s) "
                                "mismatch for local id %d in sub group %d in "
                                "group %d Expected: %d Obtained: %d\n",
                                func_name.c_str(), operation_names(operation),
                                TypeManager<Ty>::name(), i, j, k, tr, rr);
                            return TEST_FAIL;
                        }
                        tr = calculate<Ty>(tr, mx[ii + active_work_item],
                                           operation);
                        idx++;
                    }
                }
            }
            x += nw;
            y += nw;
            m += 4 * nw;
        }

        log_info("  %s_%s(%s)... passed\n", func_name.c_str(),
                 operation_names(operation), TypeManager<Ty>::name());
        return TEST_PASS;
    }
};

// Test for scan inclusive non uniform functions
template <typename Ty, ArithmeticOp operation> struct SCIN_NU
{
    static void gen(Ty *x, Ty *t, cl_int *m, const WorkGroupParams &test_params)
    {
        int nw = test_params.local_workgroup_size;
        int ns = test_params.subgroup_size;
        int ng = test_params.global_workgroup_size;
        uint32_t work_items_mask = test_params.work_items_mask;
        ng = ng / nw;
        std::string func_name;
        work_items_mask ? func_name = "sub_group_non_uniform_scan_inclusive"
                        : func_name = "sub_group_scan_inclusive";

        genrand<Ty, operation>(x, t, m, ns, nw, ng);
        log_info("  %s_%s(%s)...\n", func_name.c_str(),
                 operation_names(operation), TypeManager<Ty>::name());
        log_info("  test params: global size = %d local size = %d subgroups "
                 "size = %d work item mask = 0x%x \n",
                 test_params.global_workgroup_size, nw, ns, work_items_mask);
    }

    static int chk(Ty *x, Ty *y, Ty *mx, Ty *my, cl_int *m,
                   const WorkGroupParams &test_params)
    {
        int ii, i, j, k, n;
        int nw = test_params.local_workgroup_size;
        int ns = test_params.subgroup_size;
        int ng = test_params.global_workgroup_size;
        uint32_t work_items_mask = test_params.work_items_mask;
        int nj = (nw + ns - 1) / ns;
        Ty tr, rr;
        ng = ng / nw;

        std::string func_name;
        work_items_mask ? func_name = "sub_group_non_uniform_scan_inclusive"
                        : func_name = "sub_group_scan_inclusive";

        uint32_t use_work_items_mask;
        // for uniform case take into consideration all workitems
        use_work_items_mask = !work_items_mask ? 0xFFFFFFFF : work_items_mask;
        // std::bitset<32> mask32(use_work_items_mask);
        // for (int k) mask32.count();
        for (k = 0; k < ng; ++k)
        { // for each work_group
            // Map to array indexed to array indexed by local ID and sub group
            for (j = 0; j < nw; ++j)
            { // inside the work_group
                mx[j] = x[j]; // read host inputs for work_group
                my[j] = y[j]; // read device outputs for work_group
            }
            for (j = 0; j < nj; ++j)
            {
                ii = j * ns;
                n = ii + ns > nw ? nw - ii : ns;
                std::set<int> active_work_items;
                int catch_frist_active = -1;

                for (i = 0; i < n; ++i)
                {
                    uint32_t check_work_item = 1 << (i % 32);
                    if (use_work_items_mask & check_work_item)
                    {
                        if (catch_frist_active == -1)
                        {
                            catch_frist_active = i;
                        }
                        active_work_items.insert(i);
                    }
                }
                if (active_work_items.empty())
                {
                    log_info("  No acitve workitems in workgroup id = %d "
                             "subgroup id = %d - no calculation\n",
                             k, j);
                    continue;
                }
                else
                {
                    tr = TypeManager<Ty>::identify_limits(operation);
                    for (const int &active_work_item : active_work_items)
                    {
                        rr = my[ii + active_work_item];
                        if (active_work_items.size() == 1)
                        {
                            tr = mx[ii + catch_frist_active];
                        }
                        else
                        {
                            tr = calculate<Ty>(tr, mx[ii + active_work_item],
                                               operation);
                        }
                        if (!compare_ordered<Ty>(rr, tr))
                        {
                            log_error(
                                "ERROR: %s_%s(%s) "
                                "mismatch for local id %d in sub group %d "
                                "in "
                                "group %d Expected: %d Obtained: %d\n",
                                func_name.c_str(), operation_names(operation),
                                TypeManager<Ty>::name(), active_work_item, j, k,
                                tr, rr);
                            return TEST_FAIL;
                        }
                    }
                }
            }
            x += nw;
            y += nw;
            m += 4 * nw;
        }

        log_info("  %s_%s(%s)... passed\n", func_name.c_str(),
                 operation_names(operation), TypeManager<Ty>::name());
        return TEST_PASS;
    }
};

// Test for reduce non uniform functions
template <typename Ty, ArithmeticOp operation> struct RED_NU
{

    static void gen(Ty *x, Ty *t, cl_int *m, const WorkGroupParams &test_params)
    {
        int nw = test_params.local_workgroup_size;
        int ns = test_params.subgroup_size;
        int ng = test_params.global_workgroup_size;
        uint32_t work_items_mask = test_params.work_items_mask;
        ng = ng / nw;
        std::string func_name;

        work_items_mask ? func_name = "sub_group_non_uniform_reduce"
                        : func_name = "sub_group_reduce";
        log_info("  %s_%s(%s)...\n", func_name.c_str(),
                 operation_names(operation), TypeManager<Ty>::name());
        log_info("  test params: global size = %d local size = %d subgroups "
                 "size = %d work item mask = 0x%x \n",
                 test_params.global_workgroup_size, nw, ns, work_items_mask);
        genrand<Ty, operation>(x, t, m, ns, nw, ng);
    }

    static int chk(Ty *x, Ty *y, Ty *mx, Ty *my, cl_int *m,
                   const WorkGroupParams &test_params)
    {
        int ii, i, j, k, n;
        int nw = test_params.local_workgroup_size;
        int ns = test_params.subgroup_size;
        int ng = test_params.global_workgroup_size;
        uint32_t work_items_mask = test_params.work_items_mask;
        int nj = (nw + ns - 1) / ns;
        ng = ng / nw;
        Ty tr, rr;

        std::string func_name;
        work_items_mask ? func_name = "sub_group_non_uniform_reduce"
                        : func_name = "sub_group_reduce";

        for (k = 0; k < ng; ++k)
        {
            // Map to array indexed to array indexed by local ID and sub
            // group
            for (j = 0; j < nw; ++j)
            {
                mx[j] = x[j];
                my[j] = y[j];
            }

            uint32_t use_work_items_mask;
            use_work_items_mask =
                !work_items_mask ? 0xFFFFFFFF : work_items_mask;

            for (j = 0; j < nj; ++j)
            {
                ii = j * ns;
                n = ii + ns > nw ? nw - ii : ns;
                std::set<int> active_work_items;
                int catch_frist_active = -1;
                for (i = 0; i < n; ++i)
                {
                    uint32_t check_work_item = 1 << (i % 32);
                    if (use_work_items_mask & check_work_item)
                    {
                        if (catch_frist_active == -1)
                        {
                            catch_frist_active = i;
                            tr = mx[ii + i];
                            active_work_items.insert(i);
                            continue;
                        }
                        active_work_items.insert(i);
                        tr = calculate<Ty>(tr, mx[ii + i], operation);
                    }
                }

                if (active_work_items.empty())
                {
                    log_info("  No acitve workitems in workgroup id = %d "
                             "subgroup id = %d - no calculation\n",
                             k, j);
                    continue;
                }

                for (const int &active_work_item : active_work_items)
                {
                    rr = my[ii + active_work_item];
                    if (!compare_ordered<Ty>(rr, tr))
                    {
                        log_error("ERROR: %s_%s(%s) "
                                  "mismatch for local id %d in sub group %d in "
                                  "group %d Expected: %d Obtained: %d\n",
                                  func_name.c_str(), operation_names(operation),
                                  TypeManager<Ty>::name(), active_work_item, j,
                                  k, tr, rr);
                        return TEST_FAIL;
                    }
                }
            }
            x += nw;
            y += nw;
            m += 4 * nw;
        }

        log_info("  %s_%s(%s)... passed\n", func_name.c_str(),
                 operation_names(operation), TypeManager<Ty>::name());
        return TEST_PASS;
    }
};

#endif