🏡 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) 2017 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.
//
#include "testBase.h"
#include "harness/conversions.h"

#define TEST_SIZE 512

#ifndef MIN
    #define MIN( _a, _b )   ((_a) < (_b) ? (_a) : (_b))
#endif
#ifndef MAX
    #define MAX( _a, _b )   ((_a) > (_b) ? (_a) : (_b))
#endif

const char *singleParamIntegerKernelSourcePattern =
"__kernel void sample_test(__global %s *sourceA, __global %s *destValues)\n"
"{\n"
"    int  tid = get_global_id(0);\n"
"    %s%s tmp = vload%s( tid, destValues );\n"
"    tmp %s= %s( vload%s( tid, sourceA ) );\n"
"    vstore%s( tmp, tid, destValues );\n"
"\n"
"}\n";

const char *singleParamSingleSizeIntegerKernelSourcePattern =
"__kernel void sample_test(__global %s *sourceA, __global %s *destValues)\n"
"{\n"
"    int  tid = get_global_id(0);\n"
"    destValues[tid] %s= %s( sourceA[tid] );\n"
"}\n";

typedef bool (*singleParamIntegerVerifyFn)( void *source, void *destination, ExplicitType vecType );
static void patchup_divide_results( void *outData, const void *inDataA, const void *inDataB, size_t count, ExplicitType vecType );
bool verify_integer_divideAssign( void *source, void *destination, ExplicitType vecType );
bool verify_integer_moduloAssign( void *source, void *destination, ExplicitType vecType );

int test_single_param_integer_kernel(cl_command_queue queue, cl_context context, const char *fnName,
                                  ExplicitType vecType, size_t vecSize, singleParamIntegerVerifyFn verifyFn,
                                     MTdata d, bool useOpKernel = false )
{
    clProgramWrapper program;
    clKernelWrapper kernel;
    clMemWrapper streams[2];
    cl_long inDataA[TEST_SIZE * 16], outData[TEST_SIZE * 16], inDataB[TEST_SIZE * 16], expected;
    int error, i;
    size_t threads[1], localThreads[1];
    char kernelSource[10240];
    char *programPtr;
    char sizeName[4];

    if (! gHasLong && strstr(get_explicit_type_name(vecType),"long"))
    {
       log_info( "WARNING: 64 bit integers are not supported on this device. Skipping %s\n", get_explicit_type_name(vecType) );
       return CL_SUCCESS;
    }

    /* Create the source */
    if( vecSize == 1 )
        sizeName[ 0 ] = 0;
    else
        sprintf( sizeName, "%d", (int)vecSize );

    if( vecSize == 1 )
        sprintf( kernelSource, singleParamSingleSizeIntegerKernelSourcePattern,
                get_explicit_type_name( vecType ), get_explicit_type_name( vecType ),
                useOpKernel ? fnName : "", useOpKernel ? "" : fnName );
    else
        sprintf( kernelSource, singleParamIntegerKernelSourcePattern,
                get_explicit_type_name( vecType ), get_explicit_type_name( vecType ),
                get_explicit_type_name( vecType ), sizeName, sizeName,
                useOpKernel ? fnName : "", useOpKernel ? "" : fnName, sizeName,
                sizeName );

    /* Create kernels */
    programPtr = kernelSource;
    if (create_single_kernel_helper(context, &program, &kernel, 1,
                                    (const char **)&programPtr, "sample_test"))
    {
        log_error("The program we attempted to compile was: \n%s\n", kernelSource);
        return -1;
    }

    /* Generate some streams */
    generate_random_data( vecType, vecSize * TEST_SIZE, d, inDataA );

    streams[0] = clCreateBuffer(
        context, CL_MEM_COPY_HOST_PTR,
        get_explicit_type_size(vecType) * vecSize * TEST_SIZE, inDataA, NULL);
    if( streams[0] == NULL )
    {
        log_error("ERROR: Creating input array A failed!\n");
        return -1;
    }

    if( useOpKernel )
    {
        // Op kernels use an r/w buffer for the second param, so we need to init it with data
        generate_random_data( vecType, vecSize * TEST_SIZE, d, inDataB );
    }
    streams[1] = clCreateBuffer(
        context, (CL_MEM_READ_WRITE | (useOpKernel ? CL_MEM_COPY_HOST_PTR : 0)),
        get_explicit_type_size(vecType) * vecSize * TEST_SIZE,
        (useOpKernel) ? &inDataB : NULL, NULL);
    if( streams[1] == NULL )
    {
        log_error("ERROR: Creating output array failed!\n");
        return -1;
    }

    /* Assign streams and execute */
    error = clSetKernelArg( kernel, 0, sizeof( streams[0] ), &streams[0] );
    test_error( error, "Unable to set indexed kernel arguments" );
    error = clSetKernelArg( kernel, 1, sizeof( streams[1] ), &streams[1] );
    test_error( error, "Unable to set indexed kernel arguments" );

    /* Run the kernel */
    threads[0] = TEST_SIZE;

    error = get_max_common_work_group_size( context, kernel, threads[0], &localThreads[0] );
    test_error( error, "Unable to get work group size to use" );

    error = clEnqueueNDRangeKernel( queue, kernel, 1, NULL, threads, localThreads, 0, NULL, NULL );
    test_error( error, "Unable to execute test kernel" );

    memset(outData, 0xFF, get_explicit_type_size( vecType ) * TEST_SIZE * vecSize );

    /* Now get the results */
    error = clEnqueueReadBuffer( queue, streams[1], CL_TRUE, 0,
                                 get_explicit_type_size( vecType ) * TEST_SIZE * vecSize,
                                 outData, 0, NULL, NULL );
    test_error( error, "Unable to read output array!" );

    // deal with division by 0 -- any answer is allowed here
    if( verifyFn == verify_integer_divideAssign || verifyFn == verify_integer_moduloAssign )
        patchup_divide_results( outData, inDataA, inDataB, TEST_SIZE * vecSize, vecType );

    /* And verify! */
    char *p = (char *)outData;
    char *in = (char *)inDataA;
    char *in2 = (char *)inDataB;
    for( i = 0; i < (int)TEST_SIZE; i++ )
    {
        for( size_t j = 0; j < vecSize; j++ )
        {
            if( useOpKernel )
                memcpy( &expected, in2, get_explicit_type_size( vecType ) );

            verifyFn( in, &expected, vecType );
            if( memcmp( &expected, p, get_explicit_type_size( vecType ) ) != 0 )
            {
                switch( get_explicit_type_size( vecType ))
                {
                    case 1:
                        if( useOpKernel )
                            log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%2.2x), got (0x%2.2x), sources (0x%2.2x, 0x%2.2x)\n",
                                      (int)i, (int)j,
                                      ((cl_uchar*)&expected)[0],
                                      *( (cl_uchar *)p ),
                                      *( (cl_uchar *)in ),
                                      *( (cl_uchar *)in2 ) );
                        else
                        log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%2.2x), got (0x%2.2x), sources (0x%2.2x)\n",
                                  (int)i, (int)j,
                                   ((cl_uchar*)&expected)[0],
                                   *( (cl_uchar *)p ),
                                   *( (cl_uchar *)in ) );
                        break;

                    case 2:
                        if( useOpKernel )
                            log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%4.4x), got (0x%4.4x), sources (0x%4.4x, 0x%4.4x)\n",
                                      (int)i, (int)j, ((cl_ushort*)&expected)[0], *( (cl_ushort *)p ),
                                      *( (cl_ushort *)in ), *( (cl_ushort *)in2 ) );
                        else
                        log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%4.4x), got (0x%4.4x), sources (0x%4.4x)\n",
                                  (int)i, (int)j, ((cl_ushort*)&expected)[0], *( (cl_ushort *)p ),
                                            *( (cl_ushort *)in ) );
                        break;

                    case 4:
                        if( useOpKernel )
                            log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%8.8x), got (0x%8.8x), sources (0x%8.8x, 0x%8.8x)\n",
                                      (int)i, (int)j, ((cl_uint*)&expected)[0], *( (cl_uint *)p ),
                                      *( (cl_uint *)in ), *( (cl_uint *)in2 ) );
                        else
                        log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%8.8x), got (0x%8.8x), sources (0x%8.8x)\n",
                                  (int)i, (int)j, ((cl_uint*)&expected)[0], *( (cl_uint *)p ),
                                            *( (cl_uint *)in ) );
                        break;

                    case 8:
                        if( useOpKernel )
                            log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%16.16llx), got (0x%16.16llx), sources (0x%16.16llx, 0x%16.16llx)\n",
                                      (int)i, (int)j, ((cl_ulong*)&expected)[0], *( (cl_ulong *)p ),
                                      *( (cl_ulong *)in ), *( (cl_ulong *)in2 ) );
                        else
                        log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%16.16llx), got (0x%16.16llx), sources (0x%16.16llx)\n",
                                  (int)i, (int)j, ((cl_ulong*)&expected)[0], *( (cl_ulong *)p ),
                                            *( (cl_ulong *)in ) );
                        break;
                }
                return -1;
            }
            p += get_explicit_type_size( vecType );
            in += get_explicit_type_size( vecType );
            in2 += get_explicit_type_size( vecType );
        }
    }

    return 0;
}

int test_single_param_integer_fn( cl_command_queue queue, cl_context context, const char *fnName, singleParamIntegerVerifyFn verifyFn, bool useOpKernel = false )
{
    ExplicitType types[] = { kChar, kUChar, kShort, kUShort, kInt, kUInt, kLong, kULong, kNumExplicitTypes };
    unsigned int vecSizes[] = { 1, 2, 3, 4, 8, 16, 0 }; // TODO 3 not tested
    unsigned int index, typeIndex;
    int retVal = 0;
    RandomSeed seed(gRandomSeed );

    for( typeIndex = 0; types[ typeIndex ] != kNumExplicitTypes; typeIndex++ )
    {
        if ((types[ typeIndex ] == kLong || types[ typeIndex ] == kULong) && !gHasLong)
            continue;

        for( index = 0; vecSizes[ index ] != 0; index++ )
        {
            if( test_single_param_integer_kernel(queue, context, fnName, types[ typeIndex ], vecSizes[ index ], verifyFn, seed, useOpKernel ) != 0 )
            {
                log_error( "   Vector %s%d FAILED\n", get_explicit_type_name( types[ typeIndex ] ), vecSizes[ index ] );
                retVal = -1;
            }
        }
    }

    return retVal;
}

bool verify_integer_clz( void *source, void *destination, ExplicitType vecType )
{
    cl_long testValue;
    int count;
    int typeBits;

    switch( vecType )
    {
        case kChar:
            testValue = *( (cl_char *)source );
            typeBits = 8 * sizeof( cl_char );
            break;
        case kUChar:
            testValue = *( (cl_uchar *)source );
            typeBits = 8 * sizeof( cl_uchar );
            break;
        case kShort:
            testValue = *( (cl_short *)source );
            typeBits = 8 * sizeof( cl_short );
            break;
        case kUShort:
            testValue = *( (cl_ushort *)source );
            typeBits = 8 * sizeof( cl_ushort );
            break;
        case kInt:
            testValue = *( (cl_int *)source );
            typeBits = 8 * sizeof( cl_int );
            break;
        case kUInt:
            testValue = *( (cl_uint *)source );
            typeBits = 8 * sizeof( cl_uint );
            break;
        case kLong:
            testValue = *( (cl_long *)source );
            typeBits = 8 * sizeof( cl_long );
            break;
        case kULong:
            // Hack for now: just treat it as a signed cl_long, since it won't matter for bitcounting
            testValue = *( (cl_ulong *)source );
            typeBits = 8 * sizeof( cl_ulong );
            break;
        default:
            // Should never happen
            return false;
    }

    count = typeBits;
    if( testValue )
    {
        testValue <<= 8 * sizeof( testValue ) - typeBits;
        for( count = 0; 0 == (testValue & CL_LONG_MIN); count++ )
            testValue <<= 1;
    }

    switch( vecType )
    {
        case kChar:
            *( (cl_char *)destination ) = count;
            break;
        case kUChar:
            *( (cl_uchar *)destination ) = count;
            break;
        case kShort:
            *( (cl_short *)destination ) = count;
            break;
        case kUShort:
            *( (cl_ushort *)destination ) = count;
            break;
        case kInt:
            *( (cl_int *)destination ) = count;
            break;
        case kUInt:
            *( (cl_uint *)destination ) = count;
            break;
        case kLong:
            *( (cl_long *)destination ) = count;
            break;
        case kULong:
            *( (cl_ulong *)destination ) = count;
            break;
        default:
            // Should never happen
            return false;
    }
    return true;
}

int test_integer_clz(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
    return test_single_param_integer_fn( queue, context, "clz", verify_integer_clz );
}


bool verify_integer_ctz( void *source, void *destination, ExplicitType vecType )
{
  cl_long testValue;
  int count;
  int typeBits;

  switch( vecType )
  {
  case kChar:
    testValue = *( (cl_char *)source );
    typeBits = 8 * sizeof( cl_char );
    break;
  case kUChar:
    testValue = *( (cl_uchar *)source );
    typeBits = 8 * sizeof( cl_uchar );
    break;
  case kShort:
    testValue = *( (cl_short *)source );
    typeBits = 8 * sizeof( cl_short );
    break;
  case kUShort:
    testValue = *( (cl_ushort *)source );
    typeBits = 8 * sizeof( cl_ushort );
    break;
  case kInt:
    testValue = *( (cl_int *)source );
    typeBits = 8 * sizeof( cl_int );
    break;
  case kUInt:
    testValue = *( (cl_uint *)source );
    typeBits = 8 * sizeof( cl_uint );
    break;
  case kLong:
    testValue = *( (cl_long *)source );
    typeBits = 8 * sizeof( cl_long );
    break;
  case kULong:
    // Hack for now: just treat it as a signed cl_long, since it won't matter for bitcounting
    testValue = *( (cl_ulong *)source );
    typeBits = 8 * sizeof( cl_ulong );
    break;
  default:
    // Should never happen
    return false;
  }

  if ( testValue == 0 )
    count = typeBits;
  else
  {
    for( count = 0; (0 == (testValue & 0x1)); count++ )
      testValue >>= 1;
  }

  switch( vecType )
  {
  case kChar:
    *( (cl_char *)destination ) = count;
    break;
  case kUChar:
    *( (cl_uchar *)destination ) = count;
    break;
  case kShort:
    *( (cl_short *)destination ) = count;
    break;
  case kUShort:
    *( (cl_ushort *)destination ) = count;
    break;
  case kInt:
    *( (cl_int *)destination ) = count;
    break;
  case kUInt:
    *( (cl_uint *)destination ) = count;
    break;
  case kLong:
    *( (cl_long *)destination ) = count;
    break;
  case kULong:
    *( (cl_ulong *)destination ) = count;
    break;
  default:
    // Should never happen
    return false;
  }
  return true;
}


int test_integer_ctz(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
  return test_single_param_integer_fn( queue, context, "ctz", verify_integer_ctz );
}

#define OP_CASE( op, sizeName, size ) \
    case sizeName: \
    {    \
        cl_##size *d = (cl_##size *)destination; \
        *d op##= *( (cl_##size *)source ); \
        break; \
    }

#define OP_CASES( op ) \
    switch( vecType ) \
    {                    \
        OP_CASE( op, kChar, char ) \
        OP_CASE( op, kUChar, uchar ) \
        OP_CASE( op, kShort, short ) \
        OP_CASE( op, kUShort, ushort ) \
        OP_CASE( op, kInt, int ) \
        OP_CASE( op, kUInt, uint ) \
        OP_CASE( op, kLong, long ) \
        OP_CASE( op, kULong, ulong ) \
        default: \
            break; \
    }

#define OP_TEST( op, opName ) \
    bool verify_integer_##opName##Assign( void *source, void *destination, ExplicitType vecType )    \
    {    \
        OP_CASES( op )    \
        return true; \
    }    \
    int test_integer_##opName##Assign(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)    \
    {    \
        return test_single_param_integer_fn( queue, context, #op, verify_integer_##opName##Assign, true ); \
    }

OP_TEST( +, add )
OP_TEST( -, subtract )
OP_TEST( *, multiply )
OP_TEST( ^, exclusiveOr )
OP_TEST( |, or )
OP_TEST( &, and )

#define OP_CASE_GUARD( op, sizeName, size ) \
    case sizeName: \
    {    \
        cl_##size *d = (cl_##size *)destination; \
        cl_##size *s = (cl_##size *)source;     \
        if( *s == 0 )                           \
            *d = -1;                            \
        else                                    \
            *d op##= *s;                        \
        break; \
    }

#define OP_CASE_GUARD_SIGNED( op, sizeName, size, MIN_VAL ) \
    case sizeName: \
    {    \
        cl_##size *d = (cl_##size *)destination; \
        cl_##size *s = (cl_##size *)source;     \
        if( *s == 0 || (*d == MIN_VAL && *s == -1))  \
            *d = -1 - MIN_VAL;                  \
        else                                    \
            *d op##= *s;                        \
        break; \
    }

#define OP_CASES_GUARD( op ) \
    switch( vecType ) \
    {                    \
        OP_CASE_GUARD_SIGNED( op, kChar, char, CL_CHAR_MIN ) \
        OP_CASE_GUARD( op, kUChar, uchar ) \
        OP_CASE_GUARD_SIGNED( op, kShort, short, CL_SHRT_MIN ) \
        OP_CASE_GUARD( op, kUShort, ushort ) \
        OP_CASE_GUARD_SIGNED( op, kInt, int, CL_INT_MIN ) \
        OP_CASE_GUARD( op, kUInt, uint ) \
        OP_CASE_GUARD_SIGNED( op, kLong, long, CL_LONG_MIN ) \
        OP_CASE_GUARD( op, kULong, ulong ) \
        default: \
            break; \
    }

#define OP_TEST_GUARD( op, opName ) \
    bool verify_integer_##opName##Assign( void *source, void *destination, ExplicitType vecType )    \
    {    \
        OP_CASES_GUARD( op )    \
        return true;            \
    }    \
    int test_integer_##opName##Assign(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)    \
    {    \
        return test_single_param_integer_fn( queue, context, #op, verify_integer_##opName##Assign, true ); \
    }

OP_TEST_GUARD( /, divide )
OP_TEST_GUARD( %, modulo )

#define PATCH_CASE( _out, _src, _dest, _count, _cl_type )           \
    {                                                               \
        const _cl_type *denom = (const _cl_type* ) _src;            \
        _cl_type *result = (_cl_type* ) _out;                       \
        for( size_t i = 0; i < _count; i++ )                        \
            if( denom[i] == 0 )                                     \
                result[i] = (_cl_type) -1;                          \
    }

#define PATCH_CASE_SIGNED( _out, _src, _dest, _count, _cl_type, _MIN_VAL )      \
    {                                                                           \
        const _cl_type *num = (const _cl_type* ) _dest;                         \
        const _cl_type *denom = (const _cl_type* ) _src;                        \
        _cl_type *result = (_cl_type* ) _out;                                   \
        for( size_t i = 0; i < _count; i++ )                                    \
            if( denom[i] == 0 || ( num[i] == _MIN_VAL && denom[i] == -1))       \
                result[i] = -1 - _MIN_VAL;                                      \
    }

static void patchup_divide_results( void *outData, const void *inDataA, const void *inDataB, size_t count, ExplicitType vecType )
{
    switch( vecType )
    {
        case kChar:
            PATCH_CASE_SIGNED( outData, inDataA, inDataB, count, cl_char, CL_CHAR_MIN )
            break;
        case kUChar:
            PATCH_CASE( outData, inDataA, inDataB, count, cl_uchar )
            break;
        case kShort:
            PATCH_CASE_SIGNED( outData, inDataA, inDataB, count, cl_short, CL_SHRT_MIN )
            break;
        case kUShort:
            PATCH_CASE( outData, inDataA, inDataB, count, cl_ushort )
            break;
        case kInt:
            PATCH_CASE_SIGNED( outData, inDataA, inDataB, count, cl_int, CL_INT_MIN )
            break;
        case kUInt:
            PATCH_CASE( outData, inDataA, inDataB, count, cl_uint )
            break;
        case kLong:
            PATCH_CASE_SIGNED( outData, inDataA, inDataB, count, cl_long, CL_LONG_MIN )
            break;
        case kULong:
            PATCH_CASE( outData, inDataA, inDataB, count, cl_ulong )
            break;
        default:
            log_error( "ERROR: internal test error -- unknown data type %d\n", vecType );
            break;
    }
}

const char *twoParamIntegerKernelSourcePattern =
"__kernel void sample_test(__global %s%s *sourceA, __global %s%s *sourceB, __global %s%s *destValues)\n"
"{\n"
"    int  tid = get_global_id(0);\n"
"    %s%s sA = %s;\n"
"    %s%s sB = %s;\n"
"    %s%s dst = %s( sA, sB );\n"
"     %s;\n"
"\n"
"}\n";

typedef bool (*twoParamIntegerVerifyFn)( void *sourceA, void *sourceB, void *destination, ExplicitType vecType );

static char * build_load_statement( char *outString, size_t vecSize, const char *name )
{
    if( vecSize != 3 )
        sprintf( outString, "%s[ tid ]", name );
    else
        sprintf( outString, "vload3( tid, %s )", name );
    return outString;
}

static char * build_store_statement( char *outString, size_t vecSize, const char *name, const char *srcName )
{
    if( vecSize != 3 )
        sprintf( outString, "%s[ tid ] = %s", name, srcName );
    else
        sprintf( outString, "vstore3( %s, tid, %s )", srcName, name );
    return outString;
}

int test_two_param_integer_kernel(cl_command_queue queue, cl_context context, const char *fnName,
                                     ExplicitType vecAType, ExplicitType vecBType, unsigned int vecSize, twoParamIntegerVerifyFn verifyFn, MTdata d )
{
    clProgramWrapper program;
    clKernelWrapper kernel;
    clMemWrapper streams[3];
    cl_long inDataA[TEST_SIZE * 16], inDataB[TEST_SIZE * 16], outData[TEST_SIZE * 16], expected;
    int error, i;
    size_t threads[1], localThreads[1];
    char kernelSource[10240];
    char *programPtr;
    char sizeName[4], paramSizeName[4];

    // embedded profiles don't support long/ulong datatypes
    if (! gHasLong && strstr(get_explicit_type_name(vecAType),"long"))
    {
       log_info( "WARNING: 64 bit integers are not supported on this device. Skipping %s\n", get_explicit_type_name(vecAType) );
       return CL_SUCCESS;
    }

    /* Create the source */
    if( vecSize == 1 )
        sizeName[ 0 ] = 0;
    else
        sprintf( sizeName, "%d", vecSize );
    if( ( vecSize == 1 ) || ( vecSize == 3 ) )
        paramSizeName[ 0 ] = 0;
        else
        sprintf( paramSizeName, "%d", vecSize );

    char sourceALoad[ 128 ], sourceBLoad[ 128 ], destStore[ 128 ];

    sprintf( kernelSource, twoParamIntegerKernelSourcePattern,
                get_explicit_type_name( vecAType ), paramSizeName,
                get_explicit_type_name( vecBType ), paramSizeName,
                get_explicit_type_name( vecAType ), paramSizeName,
                get_explicit_type_name( vecAType ), sizeName, build_load_statement( sourceALoad, (size_t)vecSize, "sourceA" ),
                get_explicit_type_name( vecBType ), sizeName, build_load_statement( sourceBLoad, (size_t)vecSize, "sourceB" ),
                get_explicit_type_name( vecAType ), sizeName,
                fnName,
                build_store_statement( destStore, (size_t)vecSize, "destValues", "dst" )
                );

    /* Create kernels */
    programPtr = kernelSource;
    if( create_single_kernel_helper( context, &program, &kernel, 1, (const char **)&programPtr, "sample_test" ) )
    {
        log_error("The program we attempted to compile was: \n%s\n", kernelSource);
        return -1;
    }

    /* Generate some streams */
    generate_random_data( vecAType, vecSize * TEST_SIZE, d, inDataA );
    generate_random_data( vecBType, vecSize * TEST_SIZE, d, inDataB );

    streams[0] = clCreateBuffer(
        context, CL_MEM_COPY_HOST_PTR,
        get_explicit_type_size(vecAType) * vecSize * TEST_SIZE, &inDataA, NULL);
    if( streams[0] == NULL )
    {
        log_error("ERROR: Creating input array A failed!\n");
        return -1;
    }
    streams[1] = clCreateBuffer(
        context, CL_MEM_COPY_HOST_PTR,
        get_explicit_type_size(vecBType) * vecSize * TEST_SIZE, &inDataB, NULL);
    if( streams[1] == NULL )
    {
        log_error("ERROR: Creating input array B failed!\n");
        return -1;
    }
    streams[2] = clCreateBuffer(
        context, CL_MEM_READ_WRITE,
        get_explicit_type_size(vecAType) * vecSize * TEST_SIZE, NULL, NULL);
    if( streams[2] == NULL )
    {
        log_error("ERROR: Creating output array failed!\n");
        return -1;
    }

    /* Assign streams and execute */
    error = clSetKernelArg( kernel, 0, sizeof( streams[0] ), &streams[0] );
    test_error( error, "Unable to set indexed kernel arguments" );
    error = clSetKernelArg( kernel, 1, sizeof( streams[1] ), &streams[1] );
    test_error( error, "Unable to set indexed kernel arguments" );
    error = clSetKernelArg( kernel, 2, sizeof( streams[2] ), &streams[2] );
    test_error( error, "Unable to set indexed kernel arguments" );

    /* Run the kernel */
    threads[0] = TEST_SIZE;

    error = get_max_common_work_group_size( context, kernel, threads[0], &localThreads[0] );
    test_error( error, "Unable to get work group size to use" );

    error = clEnqueueNDRangeKernel( queue, kernel, 1, NULL, threads, localThreads, 0, NULL, NULL );
    test_error( error, "Unable to execute test kernel" );

    memset(outData, 0xFF, get_explicit_type_size( vecAType ) * TEST_SIZE * vecSize);

    /* Now get the results */
    error = clEnqueueReadBuffer( queue, streams[2], CL_TRUE, 0,
                                 get_explicit_type_size( vecAType ) * TEST_SIZE * vecSize, outData, 0,
                                 NULL, NULL );
    test_error( error, "Unable to read output array!" );

    /* And verify! */
    char *inA = (char *)inDataA;
    char *inB = (char *)inDataB;
    char *out = (char *)outData;
    for( i = 0; i < (int)TEST_SIZE; i++ )
    {
        for( size_t j = 0; j < vecSize; j++ )
        {
            bool test = verifyFn( inA, inB, &expected, vecAType );
            if( test && ( memcmp( &expected, out, get_explicit_type_size( vecAType ) ) != 0 ) )
            {
                switch( get_explicit_type_size( vecAType ))
                {
                    case 1:
                        log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%2.2x), got (0x%2.2x), sources (0x%2.2x, 0x%2.2x), TEST_SIZE %d\n",
                                   (int)i, (int)j, ((cl_uchar*)&expected)[ 0 ], *( (cl_uchar *)out ),
                                   *( (cl_uchar *)inA ),
                                   *( (cl_uchar *)inB ) ,
                                   TEST_SIZE);
                        break;

                    case 2:
                        log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%4.4x), got (0x%4.4x), sources (0x%4.4x, 0x%4.4x), TEST_SIZE %d\n",
                                   (int)i, (int)j, ((cl_ushort*)&expected)[ 0 ], *( (cl_ushort *)out ),
                                   *( (cl_ushort *)inA ),
                                   *( (cl_ushort *)inB ),
                                   TEST_SIZE);
                        break;

                    case 4:
                        log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%8.8x), got (0x%8.8x), sources (0x%8.8x, 0x%8.8x)\n",
                                  (int)i, (int)j, ((cl_uint*)&expected)[ 0 ], *( (cl_uint *)out ),
                                            *( (cl_uint *)inA ),
                                            *( (cl_uint *)inB ) );
                        break;

                    case 8:
                        log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%16.16llx), got (0x%16.16llx), sources (0x%16.16llx, 0x%16.16llx)\n",
                                  (int)i, (int)j, ((cl_ulong*)&expected)[ 0 ], *( (cl_ulong *)out ),
                                            *( (cl_ulong *)inA ),
                                            *( (cl_ulong *)inB ) );
                        break;
                }
                return -1;
            }
            inA += get_explicit_type_size( vecAType );
            inB += get_explicit_type_size( vecBType );
            out += get_explicit_type_size( vecAType );
        }
    }

    return 0;
}

int test_two_param_integer_fn(cl_command_queue queue, cl_context context, const char *fnName, twoParamIntegerVerifyFn verifyFn)
{
    ExplicitType types[] = { kChar, kUChar, kShort, kUShort, kInt, kUInt, kLong, kULong, kNumExplicitTypes };
    unsigned int vecSizes[] = { 1, 2, 3, 4, 8, 16, 0 }; // TODO : 3 not tested
    unsigned int index, typeIndex;
    int retVal = 0;
    RandomSeed seed(gRandomSeed );

    for( typeIndex = 0; types[ typeIndex ] != kNumExplicitTypes; typeIndex++ )
    {
        if (( types[ typeIndex ] == kLong || types[ typeIndex ] == kULong) && !gHasLong)
            continue;

        for( index = 0; vecSizes[ index ] != 0; index++ )
        {
            if( test_two_param_integer_kernel(queue, context, fnName, types[ typeIndex ], types[ typeIndex ], vecSizes[ index ], verifyFn, seed ) != 0 )
            {
                log_error( "   Vector %s%d FAILED\n", get_explicit_type_name( types[ typeIndex ] ), vecSizes[ index ] );
                retVal = -1;
            }
        }
    }

    return retVal;
}

int test_two_param_unmatched_integer_fn(cl_command_queue queue, cl_context context, const char *fnName, twoParamIntegerVerifyFn verifyFn)
{
    ExplicitType types[] = { kChar, kUChar, kShort, kUShort, kInt, kUInt, kLong, kULong, kNumExplicitTypes };
    unsigned int vecSizes[] = { 1, 2, 4, 8, 16, 0 };
    unsigned int index, typeAIndex, typeBIndex;
    int retVal = 0;
    RandomSeed seed( gRandomSeed );

    for( typeAIndex = 0; types[ typeAIndex ] != kNumExplicitTypes; typeAIndex++ )
    {
        if (( types[ typeAIndex ] == kLong || types[ typeAIndex ] == kULong) && !gHasLong)
            continue;

        for( typeBIndex = 0; types[ typeBIndex ] != kNumExplicitTypes; typeBIndex++ )
        {
            if (( types[ typeBIndex ] == kLong || types[ typeBIndex ] == kULong) && !gHasLong)
                continue;

            for( index = 0; vecSizes[ index ] != 0; index++ )
            {
                if( test_two_param_integer_kernel( queue, context, fnName, types[ typeAIndex ], types[ typeBIndex ], vecSizes[ index ], verifyFn, seed ) != 0 )
                {
                    log_error( "   Vector %s%d / %s%d FAILED\n", get_explicit_type_name( types[ typeAIndex ] ), vecSizes[ index ],  get_explicit_type_name( types[ typeBIndex ] ), vecSizes[ index ] );
                    retVal = -1;
                }
            }
        }
    }

    return retVal;
}

bool verify_integer_hadd( void *sourceA, void *sourceB, void *destination, ExplicitType vecType )
{
    cl_long testValueA, testValueB, overflow;
    cl_ulong uValueA, uValueB, uOverflow;

    switch( vecType )
    {
        case kChar:
            testValueA = *( (cl_char *)sourceA );
            testValueB = *( (cl_char *)sourceB );
            *( (cl_char *)destination ) = (cl_char)( ( testValueA + testValueB ) >> 1 );
            break;
        case kUChar:
            testValueA = *( (cl_uchar *)sourceA );
            testValueB = *( (cl_uchar *)sourceB );
            *( (cl_uchar *)destination ) = (cl_uchar)( ( testValueA + testValueB ) >> 1 );
            break;
        case kShort:
            testValueA = *( (cl_short *)sourceA );
            testValueB = *( (cl_short *)sourceB );
            *( (cl_short *)destination ) = (cl_short)( ( testValueA + testValueB ) >> 1 );
            break;
        case kUShort:
            testValueA = *( (cl_ushort *)sourceA );
            testValueB = *( (cl_ushort *)sourceB );
            *( (cl_ushort *)destination ) = (cl_ushort)( ( testValueA + testValueB ) >> 1 );
            break;
        case kInt:
            testValueA = *( (cl_int *)sourceA );
            testValueB = *( (cl_int *)sourceB );
            *( (cl_int *)destination ) = (cl_int)( ( testValueA + testValueB ) >> 1 );
            break;
        case kUInt:
            testValueA = *( (cl_uint *)sourceA );
            testValueB = *( (cl_uint *)sourceB );
            *( (cl_uint *)destination ) = (cl_uint)( ( testValueA + testValueB ) >> 1 );
            break;
        case kLong:
            // The long way to avoid dropping bits
            testValueA = *( (cl_long *)sourceA );
            testValueB = *( (cl_long *)sourceB );
            overflow = ( testValueA & 0x1 ) + ( testValueB & 0x1 );
            *( (cl_long *)destination ) = ( ( testValueA >> 1 ) + ( testValueB >> 1 ) ) + ( overflow >> 1 );
            break;
        case kULong:
            // The long way to avoid dropping bits
            uValueA = *( (cl_ulong *)sourceA );
            uValueB = *( (cl_ulong *)sourceB );
            uOverflow = ( uValueA & 0x1 ) + ( uValueB & 0x1 );
            *( (cl_ulong *)destination ) = ( ( uValueA >> 1 ) + ( uValueB >> 1 ) ) + ( uOverflow >> 1 );
            break;
        default:
            // Should never happen
            return false;
    }
    return true;
}

int test_integer_hadd(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
    return test_two_param_integer_fn( queue, context, "hadd", verify_integer_hadd );
}

bool verify_integer_rhadd( void *sourceA, void *sourceB, void *destination, ExplicitType vecType )
{
    cl_long testValueA, testValueB, overflow;
    cl_ulong uValueA, uValueB, uOverflow;

    switch( vecType )
    {
        case kChar:
            testValueA = *( (cl_char *)sourceA );
            testValueB = *( (cl_char *)sourceB );
            *( (cl_char *)destination ) = (cl_char)( ( testValueA + testValueB + 1 ) >> 1 );
            break;
        case kUChar:
            testValueA = *( (cl_uchar *)sourceA );
            testValueB = *( (cl_uchar *)sourceB );
            *( (cl_uchar *)destination ) = (cl_uchar)( ( testValueA + testValueB + 1 ) >> 1 );
            break;
        case kShort:
            testValueA = *( (cl_short *)sourceA );
            testValueB = *( (cl_short *)sourceB );
            *( (cl_short *)destination ) = (cl_short)( ( testValueA + testValueB + 1 ) >> 1 );
            break;
        case kUShort:
            testValueA = *( (cl_ushort *)sourceA );
            testValueB = *( (cl_ushort *)sourceB );
            *( (cl_ushort *)destination ) = (cl_ushort)( ( testValueA + testValueB + 1 ) >> 1 );
            break;
        case kInt:
            testValueA = *( (cl_int *)sourceA );
            testValueB = *( (cl_int *)sourceB );
            *( (cl_int *)destination ) = (cl_int)( ( testValueA + testValueB + 1 ) >> 1 );
            break;
        case kUInt:
            testValueA = *( (cl_uint *)sourceA );
            testValueB = *( (cl_uint *)sourceB );
            *( (cl_uint *)destination ) = (cl_uint)( ( testValueA + testValueB + 1 ) >> 1 );
            break;
        case kLong:
            // The long way to avoid dropping bits
            testValueA = *( (cl_long *)sourceA );
            testValueB = *( (cl_long *)sourceB );
            overflow = ( testValueA | testValueB ) & 0x1;
            *( (cl_long *)destination ) = ( ( testValueA >> 1 ) + ( testValueB >> 1 ) ) + overflow;
            break;
        case kULong:
            // The long way to avoid dropping bits
            uValueA = *( (cl_ulong *)sourceA );
            uValueB = *( (cl_ulong *)sourceB );
            uOverflow = ( uValueA | uValueB ) & 0x1;
            *( (cl_ulong *)destination ) = ( ( uValueA >> 1 ) + ( uValueB >> 1 ) ) + uOverflow;
            break;
        default:
            // Should never happen
            return false;
    }
    return true;
}

int test_integer_rhadd(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
    return test_two_param_integer_fn( queue, context, "rhadd", verify_integer_rhadd );
}

#define MIN_CASE( type, const ) \
    case const : \
    {            \
        cl_##type valueA = *( (cl_##type *)sourceA ); \
        cl_##type valueB = *( (cl_##type *)sourceB ); \
        *( (cl_##type *)destination ) = (cl_##type)( valueB < valueA ? valueB : valueA ); \
        break; \
    }

bool verify_integer_min( void *sourceA, void *sourceB, void *destination, ExplicitType vecType )
{
    switch( vecType )
    {
        MIN_CASE( char, kChar )
        MIN_CASE( uchar, kUChar )
        MIN_CASE( short, kShort )
        MIN_CASE( ushort, kUShort )
        MIN_CASE( int, kInt )
        MIN_CASE( uint, kUInt )
        MIN_CASE( long, kLong )
        MIN_CASE( ulong, kULong )
        default:
            // Should never happen
            return false;
    }
    return true;
}

int test_integer_min(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
    return test_two_param_integer_fn( queue, context, "min", verify_integer_min);
}

#define MAX_CASE( type, const ) \
    case const : \
    {            \
        cl_##type valueA = *( (cl_##type *)sourceA ); \
        cl_##type valueB = *( (cl_##type *)sourceB ); \
        *( (cl_##type *)destination ) = (cl_##type)( valueA < valueB ? valueB : valueA ); \
        break; \
    }

bool verify_integer_max( void *sourceA, void *sourceB, void *destination, ExplicitType vecType )
{
    switch( vecType )
    {
            MAX_CASE( char, kChar )
            MAX_CASE( uchar, kUChar )
            MAX_CASE( short, kShort )
            MAX_CASE( ushort, kUShort )
            MAX_CASE( int, kInt )
            MAX_CASE( uint, kUInt )
            MAX_CASE( long, kLong )
            MAX_CASE( ulong, kULong )
        default:
            // Should never happen
            return false;
    }
    return true;
}

int test_integer_max(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
    return test_two_param_integer_fn( queue, context, "max", verify_integer_max );
}


void multiply_unsigned_64_by_64( cl_ulong sourceA, cl_ulong sourceB, cl_ulong &destLow, cl_ulong &destHi )
{
    cl_ulong lowA, lowB;
    cl_ulong highA, highB;

    // Split up the values
    lowA = sourceA & 0xffffffff;
    highA = sourceA >> 32;
    lowB = sourceB & 0xffffffff;
    highB = sourceB >> 32;

    // Note that, with this split, our multiplication becomes:
    //     ( a * b )
    // = ( ( aHI << 32 + aLO ) * ( bHI << 32 + bLO ) ) >> 64
    // = ( ( aHI << 32 * bHI << 32 ) + ( aHI << 32 * bLO ) + ( aLO * bHI << 32 ) + ( aLO * bLO ) ) >> 64
    // = ( ( aHI * bHI << 64 ) + ( aHI * bLO << 32 ) + ( aLO * bHI << 32 ) + ( aLO * bLO ) ) >> 64
    // = ( aHI * bHI ) + ( aHI * bLO >> 32 ) + ( aLO * bHI >> 32 ) + ( aLO * bLO >> 64 )

    // Now, since each value is 32 bits, the max size of any multiplication is:
    // ( 2 ^ 32 - 1 ) * ( 2 ^ 32 - 1 ) = 2^64 - 4^32 + 1 = 2^64 - 2^33 + 1, which fits within 64 bits
    // Which means we can do each component within a 64-bit integer as necessary (each component above marked as AB1 - AB4)
    cl_ulong aHibHi = highA * highB;
    cl_ulong aHibLo = highA * lowB;
    cl_ulong aLobHi = lowA * highB;
    cl_ulong aLobLo = lowA * lowB;

    // Assemble terms.
    //  We note that in certain cases, sums of products cannot overflow:
    //
    //      The maximum product of two N-bit unsigned numbers is
    //
    //          (2**N-1)^2 = 2**2N - 2**(N+1) + 1
    //
    //      We note that we can add the maximum N-bit number to the 2N-bit product twice without overflow:
    //
    //          (2**N-1)^2 + 2*(2**N-1) = 2**2N - 2**(N+1) + 1 + 2**(N+1) - 2 = 2**2N - 1
    //
    //  If we breakdown the product of two numbers a,b into high and low halves of partial products as follows:
    //
    //                                              a.hi                a.lo
    // x                                            b.hi                b.lo
    //===============================================================================
    //  (b.hi*a.hi).hi      (b.hi*a.hi).lo
    //                      (b.lo*a.hi).hi      (b.lo*a.hi).lo
    //                      (b.hi*a.lo).hi      (b.hi*a.lo).lo
    // +                                        (b.lo*a.lo).hi      (b.lo*a.lo).lo
    //===============================================================================
    //
    // The (b.lo*a.lo).lo term cannot cause a carry, so we can ignore them for now.  We also know from above, that we can add (b.lo*a.lo).hi
    // and (b.hi*a.lo).lo to the 2N bit term [(b.lo*a.hi).hi + (b.lo*a.hi).lo] without overflow.  That takes care of all of the terms
    // on the right half that might carry.  Do that now.
    //
    cl_ulong aLobLoHi = aLobLo >> 32;
    cl_ulong aLobHiLo = aLobHi & 0xFFFFFFFFULL;
    aHibLo += aLobLoHi + aLobHiLo;

    // That leaves us with these terms:
    //
    //                                              a.hi                a.lo
    // x                                            b.hi                b.lo
    //===============================================================================
    //  (b.hi*a.hi).hi      (b.hi*a.hi).lo
    //                      (b.hi*a.lo).hi
    //                    [ (b.lo*a.hi).hi + (b.lo*a.hi).lo + other ]
    // +                                                                (b.lo*a.lo).lo
    //===============================================================================

    // All of the overflow potential from the right half has now been accumulated into the [ (b.lo*a.hi).hi + (b.lo*a.hi).lo ] 2N bit term.
    // We can safely separate into high and low parts. Per our rule above, we know we can accumulate the high part of that and (b.hi*a.lo).hi
    // into the 2N bit term (b.lo*a.hi) without carry.  The low part can be pieced together with (b.lo*a.lo).lo, to give the final low result

    destHi = aHibHi + (aHibLo >> 32 ) + (aLobHi >> 32);             // Cant overflow
    destLow = (aHibLo << 32) | ( aLobLo & 0xFFFFFFFFULL );
}

void multiply_signed_64_by_64( cl_long sourceA, cl_long sourceB, cl_ulong &destLow, cl_long &destHi )
{
    // Find sign of result
    cl_long aSign = sourceA >> 63;
    cl_long bSign = sourceB >> 63;
    cl_long resultSign = aSign ^ bSign;

    // take absolute values of the argument
    sourceA = (sourceA ^ aSign) - aSign;
    sourceB = (sourceB ^ bSign) - bSign;

    cl_ulong hi;
    multiply_unsigned_64_by_64( (cl_ulong) sourceA, (cl_ulong) sourceB, destLow, hi );

    // Fix the sign
    if( resultSign )
    {
        destLow ^= resultSign;
        hi  ^= resultSign;
        destLow -= resultSign;

        //carry if necessary
        if( 0 == destLow )
            hi -= resultSign;
    }

    destHi = (cl_long) hi;
}

bool verify_integer_mul_hi( void *sourceA, void *sourceB, void *destination, ExplicitType vecType )
{
    cl_long testValueA, testValueB, highSigned;
    cl_ulong highUnsigned, lowHalf;

    switch( vecType )
    {
        case kChar:
            testValueA = *( (cl_char *)sourceA );
            testValueB = *( (cl_char *)sourceB );
            *( (cl_char *)destination ) = (cl_char)( ( testValueA * testValueB ) >> 8 );
            break;
        case kUChar:
            testValueA = *( (cl_uchar *)sourceA );
            testValueB = *( (cl_uchar *)sourceB );
            *( (cl_uchar *)destination ) = (cl_uchar)( ( testValueA * testValueB ) >> 8 );
            break;
        case kShort:
            testValueA = *( (cl_short *)sourceA );
            testValueB = *( (cl_short *)sourceB );
            *( (cl_short *)destination ) = (cl_short)( ( testValueA * testValueB ) >> 16 );
            break;
        case kUShort:
            testValueA = *( (cl_ushort *)sourceA );
            testValueB = *( (cl_ushort *)sourceB );
            *( (cl_ushort *)destination ) = (cl_ushort)( ( testValueA * testValueB ) >> 16 );
            break;
        case kInt:
            testValueA = *( (cl_int *)sourceA );
            testValueB = *( (cl_int *)sourceB );
            *( (cl_int *)destination ) = (cl_int)( ( testValueA * testValueB ) >> 32 );
            break;
        case kUInt:
            testValueA = *( (cl_uint *)sourceA );
            testValueB = *( (cl_uint *)sourceB );
            *( (cl_uint *)destination ) = (cl_uint)( ( testValueA * testValueB ) >> 32 );
            break;
        case kLong:
            testValueA = *( (cl_long *)sourceA );
            testValueB = *( (cl_long *)sourceB );

            multiply_signed_64_by_64( testValueA, testValueB, lowHalf, highSigned );
            *( (cl_long *)destination ) = highSigned;
            break;
        case kULong:
            testValueA = *( (cl_ulong *)sourceA );
            testValueB = *( (cl_ulong *)sourceB );

            multiply_unsigned_64_by_64( testValueA, testValueB, lowHalf, highUnsigned );
            *( (cl_ulong *)destination ) = highUnsigned;
            break;
        default:
            // Should never happen
            return false;
    }
    return true;
}

int test_integer_mul_hi(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
    return test_two_param_integer_fn( queue, context, "mul_hi", verify_integer_mul_hi );
}

bool verify_integer_rotate( void *sourceA, void *sourceB, void *destination, ExplicitType vecType )
{
    cl_ulong testValueA;
    char numBits;

    switch( vecType )
    {
        case kChar:
        case kUChar:
            testValueA = *( (cl_uchar *)sourceA );
            numBits = *( (cl_uchar *)sourceB );
            numBits &= 7;
            if ( numBits == 0 )
                *( (cl_uchar *)destination ) =  (cl_uchar)testValueA;
            else
                *( (cl_uchar *)destination ) = (cl_uchar)( ( testValueA << numBits ) | ( testValueA >> ( 8 - numBits ) ) );
            break;
        case kShort:
        case kUShort:
            testValueA = *( (cl_ushort *)sourceA );
            numBits = *( (cl_ushort *)sourceB );
            numBits &= 15;
            if ( numBits == 0 )
                *( (cl_ushort *)destination ) =  (cl_ushort)testValueA;
            else
                *( (cl_ushort *)destination ) = (cl_ushort)( ( testValueA << numBits ) | ( testValueA >> ( 16 - numBits ) ) );
            break;
        case kInt:
        case kUInt:
            testValueA = *( (cl_uint *)sourceA );
            numBits = *( (cl_uint *)sourceB );
            numBits &= 31;
            if ( numBits == 0 )
                *( (cl_uint *)destination ) =  (cl_uint) testValueA;
            else
                *( (cl_uint *)destination ) = (cl_uint)( ( testValueA << numBits ) | ( testValueA >> ( 32 - numBits ) ) );
            break;
        case kLong:
        case kULong:
            testValueA = *( (cl_ulong *)sourceA );
            numBits = *( (cl_ulong *)sourceB );
            numBits &= 63;
            if ( numBits == 0 )
                *( (cl_ulong *)destination ) =  (cl_ulong)testValueA;
            else
                *( (cl_ulong *)destination ) = (cl_ulong)( ( testValueA << numBits ) | ( testValueA >> ( 64 - numBits ) ) );
            break;
        default:
            // Should never happen
            log_error( "Unknown type encountered in verify_integer_rotate. Test failed. Aborting...\n" );
            abort();
            return false;
    }
    return true;
}

int test_integer_rotate(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
    return test_two_param_integer_fn( queue, context, "rotate", verify_integer_rotate );
}

const char *threeParamIntegerKernelSourcePattern =
"__kernel void sample_test(__global %s%s *sourceA, __global %s%s *sourceB, __global %s%s *sourceC, __global %s%s *destValues)\n"
"{\n"
"    int  tid = get_global_id(0);\n"
"    %s%s sA = %s;\n"
"    %s%s sB = %s;\n"
"    %s%s sC = %s;\n"
"    %s%s dst = %s( sA, sB, sC );\n"
"     %s;\n"
"\n"
"}\n";

typedef bool (*threeParamIntegerVerifyFn)( void *sourceA, void *sourceB, void *sourceC, void *destination,
                                            ExplicitType vecAType, ExplicitType vecBType, ExplicitType vecCType, ExplicitType destType );

int test_three_param_integer_kernel(cl_command_queue queue, cl_context context, const char *fnName,
                                  ExplicitType vecAType, ExplicitType vecBType, ExplicitType vecCType, ExplicitType destType,
                                    unsigned int vecSize, threeParamIntegerVerifyFn verifyFn, MTdata d )
{
    clProgramWrapper program;
    clKernelWrapper kernel;
    clMemWrapper streams[4];
    cl_long inDataA[TEST_SIZE * 16], inDataB[TEST_SIZE * 16], inDataC[TEST_SIZE * 16], outData[TEST_SIZE * 16], expected;
    int error, i;
    size_t threads[1], localThreads[1];
    char kernelSource[10240];
    char *programPtr;
    char sizeName[4], paramSizeName[4];

    if (! gHasLong && strstr(get_explicit_type_name(vecAType),"long"))
    {
        log_info( "WARNING: 64 bit integers are not supported on this device. Skipping %s\n", get_explicit_type_name(vecAType) );
        return CL_SUCCESS;
    }


    /* Create the source */
    if( vecSize == 1 )
        sizeName[ 0 ] = 0;
    else
        sprintf( sizeName, "%d", vecSize );
    if( ( vecSize == 1 ) || ( vecSize == 3 ) )
        paramSizeName[ 0 ] = 0;
        else
        sprintf( paramSizeName, "%d", vecSize );

    char sourceALoad[ 128 ], sourceBLoad[ 128 ], sourceCLoad[ 128 ], destStore[ 128 ];

    sprintf( kernelSource, threeParamIntegerKernelSourcePattern,
            get_explicit_type_name( vecAType ), paramSizeName,
            get_explicit_type_name( vecBType ), paramSizeName,
            get_explicit_type_name( vecCType ), paramSizeName,
            get_explicit_type_name( destType ), paramSizeName,
            get_explicit_type_name( vecAType ), sizeName, build_load_statement( sourceALoad, (size_t)vecSize, "sourceA" ),
            get_explicit_type_name( vecBType ), sizeName, build_load_statement( sourceBLoad, (size_t)vecSize, "sourceB" ),
            get_explicit_type_name( vecCType ), sizeName, build_load_statement( sourceCLoad, (size_t)vecSize, "sourceC" ),
            get_explicit_type_name( destType ), sizeName,
            fnName,
            build_store_statement( destStore, (size_t)vecSize, "destValues", "dst" )
            );

    /* Create kernels */
    programPtr = kernelSource;
    if( create_single_kernel_helper( context, &program, &kernel, 1, (const char **)&programPtr, "sample_test" ) )
    {
    log_error("The program we attempted to compile was: \n%s\n", kernelSource);
        return -1;
    }

    /* Generate some streams */
    generate_random_data( vecAType, vecSize * TEST_SIZE, d, inDataA );
    generate_random_data( vecBType, vecSize * TEST_SIZE, d, inDataB );
    generate_random_data( vecCType, vecSize * TEST_SIZE, d, inDataC );

    streams[0] = clCreateBuffer(
        context, CL_MEM_COPY_HOST_PTR,
        get_explicit_type_size(vecAType) * vecSize * TEST_SIZE, &inDataA, NULL);
    if( streams[0] == NULL )
    {
        log_error("ERROR: Creating input array A failed!\n");
        return -1;
    }
    streams[1] = clCreateBuffer(
        context, CL_MEM_COPY_HOST_PTR,
        get_explicit_type_size(vecBType) * vecSize * TEST_SIZE, &inDataB, NULL);
    if( streams[1] == NULL )
    {
        log_error("ERROR: Creating input array B failed!\n");
        return -1;
    }
    streams[2] = clCreateBuffer(
        context, CL_MEM_COPY_HOST_PTR,
        get_explicit_type_size(vecCType) * vecSize * TEST_SIZE, &inDataC, NULL);
    if( streams[2] == NULL )
    {
        log_error("ERROR: Creating input array C failed!\n");
        return -1;
    }
    streams[3] = clCreateBuffer(
        context, CL_MEM_READ_WRITE,
        get_explicit_type_size(destType) * vecSize * TEST_SIZE, NULL, NULL);
    if( streams[3] == NULL )
    {
        log_error("ERROR: Creating output array failed!\n");
        return -1;
    }

    /* Assign streams and execute */
    error = clSetKernelArg( kernel, 0, sizeof( streams[0] ), &streams[0] );
    test_error( error, "Unable to set indexed kernel arguments" );
    error = clSetKernelArg( kernel, 1, sizeof( streams[1] ), &streams[1] );
    test_error( error, "Unable to set indexed kernel arguments" );
    error = clSetKernelArg( kernel, 2, sizeof( streams[2] ), &streams[2] );
    test_error( error, "Unable to set indexed kernel arguments" );
    error = clSetKernelArg( kernel, 3, sizeof( streams[3] ), &streams[3] );
    test_error( error, "Unable to set indexed kernel arguments" );

    /* Run the kernel */
    threads[0] = TEST_SIZE;

    error = get_max_common_work_group_size( context, kernel, threads[0], &localThreads[0] );
    test_error( error, "Unable to get work group size to use" );

    error = clEnqueueNDRangeKernel( queue, kernel, 1, NULL, threads, localThreads, 0, NULL, NULL );
    test_error( error, "Unable to execute test kernel" );

    memset(outData, 0xFF, get_explicit_type_size( destType ) * TEST_SIZE * vecSize);

    /* Now get the results */
    error = clEnqueueReadBuffer( queue, streams[3], CL_TRUE, 0, get_explicit_type_size( destType ) * TEST_SIZE * vecSize, outData, 0, NULL, NULL );
    test_error( error, "Unable to read output array!" );

    /* And verify! */
    char *inA = (char *)inDataA;
    char *inB = (char *)inDataB;
    char *inC = (char *)inDataC;
    char *out = (char *)outData;
    for( i = 0; i < (int)TEST_SIZE; i++ )
    {
        for( size_t j = 0; j < vecSize; j++ )
        {
            bool test = verifyFn( inA, inB, inC, &expected, vecAType, vecBType, vecCType, destType );
            if( test && ( memcmp( &expected, out, get_explicit_type_size( destType ) ) != 0 ) )
            {
                switch( get_explicit_type_size( vecAType ))
                {
                    case 1:
                        log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%2.2x), got (0x%2.2x), sources (0x%2.2x, 0x%2.2x, 0x%2.2x)\n",
                                  (int)i, (int)j, ((cl_uchar*)&expected)[ 0 ], *( (cl_uchar *)out ),
                                            *( (cl_uchar *)inA ),
                                            *( (cl_uchar *)inB ),
                                            *( (cl_uchar *)inC ) );
                        break;

                    case 2:
                        log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%4.4x), got (0x%4.4x), sources (0x%4.4x, 0x%4.4x, 0x%4.4x)\n",
                                  (int)i, (int)j, ((cl_ushort*)&expected)[ 0 ], *( (cl_ushort *)out ),
                                            *( (cl_ushort *)inA ),
                                            *( (cl_ushort *)inB ),
                                            *( (cl_ushort *)inC ) );
                        break;

                    case 4:
                        log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%8.8x), got (0x%8.8x), sources (0x%8.8x, 0x%8.8x, 0x%8.8x)\n",
                                  (int)i, (int)j, ((cl_uint*)&expected)[ 0 ], *( (cl_uint *)out ),
                                            *( (cl_uint *)inA ),
                                            *( (cl_uint *)inB ),
                                            *( (cl_uint *)inC ) );
                        break;

                    case 8:
                        log_error( "ERROR: Data sample %d:%d does not validate! Expected (0x%16.16llx), got (0x%16.16llx), sources (0x%16.16llx, 0x%16.16llx, 0x%16.16llx)\n",
                                  (int)i, (int)j, ((cl_ulong*)&expected)[ 0 ], *( (cl_ulong *)out ),
                                            *( (cl_ulong *)inA ),
                                            *( (cl_ulong *)inB ),
                                            *( (cl_ulong *)inC ) );
                        break;
                }
                return -1;
            }
            inA += get_explicit_type_size( vecAType );
            inB += get_explicit_type_size( vecBType );
            inC += get_explicit_type_size( vecCType );
            out += get_explicit_type_size( destType );
        }
    }

    return 0;
}

int test_three_param_integer_fn(cl_command_queue queue, cl_context context, const char *fnName, threeParamIntegerVerifyFn verifyFn)
{
    ExplicitType types[] = { kChar, kUChar, kShort, kUShort, kInt, kUInt, kLong, kULong, kNumExplicitTypes };
    unsigned int vecSizes[] = { 1, 2, 3, 4, 8, 16, 0 };
    unsigned int index, typeAIndex;
    int retVal = 0;
    RandomSeed seed(gRandomSeed);

    for( typeAIndex = 0; types[ typeAIndex ] != kNumExplicitTypes; typeAIndex++ )
    {
        if ((types[ typeAIndex ] == kLong || types[ typeAIndex] == kULong) && !gHasLong)
            continue;

        for( index = 0; vecSizes[ index ] != 0; index++ )
        {
            if( test_three_param_integer_kernel(queue, context, fnName, types[ typeAIndex ], types[ typeAIndex ], types[ typeAIndex ], types[ typeAIndex ], vecSizes[ index ], verifyFn, seed ) != 0 )
            {
                log_error( "   Vector %s%d,%s%d,%s%d FAILED\n", get_explicit_type_name( types[ typeAIndex ] ), vecSizes[ index ],
                                                            get_explicit_type_name( types[ typeAIndex ] ), vecSizes[ index ] ,
                                                            get_explicit_type_name( types[ typeAIndex ] ), vecSizes[ index ] );
                retVal = -1;
            }
        }
    }

    return retVal;
}

bool verify_integer_clamp( void *sourceA, void *sourceB, void *sourceC, void *destination,
                        ExplicitType vecAType, ExplicitType vecBType, ExplicitType vecCType, ExplicitType destType )
{
    if( vecAType == kULong || vecAType == kUInt || vecAType == kUShort || vecAType == kUChar )
    {
        cl_ulong valueA, valueB, valueC;

        switch( vecAType )
        {
            case kULong:
                valueA = ((cl_ulong*) sourceA)[0];
                valueB = ((cl_ulong*) sourceB)[0];
                valueC = ((cl_ulong*) sourceC)[0];
                break;
            case kUInt:
                valueA = ((cl_uint*) sourceA)[0];
                valueB = ((cl_uint*) sourceB)[0];
                valueC = ((cl_uint*) sourceC)[0];
                break;
            case kUShort:
                valueA = ((cl_ushort*) sourceA)[0];
                valueB = ((cl_ushort*) sourceB)[0];
                valueC = ((cl_ushort*) sourceC)[0];
                break;
            case kUChar:
                valueA = ((cl_uchar*) sourceA)[0];
                valueB = ((cl_uchar*) sourceB)[0];
                valueC = ((cl_uchar*) sourceC)[0];
                break;
            default:
                //error -- should never get here
                abort();
                break;
        }


        if(valueB > valueC) {
            return false; // results are undefined : let expected alone.
        }

        switch( vecAType )
        {
            case kULong:
                ((cl_ulong*) destination)[0] = MAX(MIN(valueA, valueC), valueB);
                break;
            case kUInt:
                ((cl_uint*) destination)[0] = (cl_uint)
                    (MAX(MIN(valueA, valueC), valueB));
                break;
            case kUShort:
                ((cl_ushort*) destination)[0] = (cl_ushort)
                    (MAX(MIN(valueA, valueC), valueB));
                break;
            case kUChar:
                ((cl_uchar*) destination)[0] = (cl_uchar)
                    (MAX(MIN(valueA, valueC), valueB));
                break;
            default:
                //error -- should never get here
                abort();
                break;
        }




    }
    else
    {
        cl_long valueA, valueB, valueC;


        switch( vecAType )
        {
            case kLong:
                valueA = ((cl_long*) sourceA)[0];
                valueB = ((cl_long*) sourceB)[0];
                valueC = ((cl_long*) sourceC)[0];
                break;
            case kInt:
                valueA = ((cl_int*) sourceA)[0];
                valueB = ((cl_int*) sourceB)[0];
                valueC = ((cl_int*) sourceC)[0];
                break;
            case kShort:
                valueA = ((cl_short*) sourceA)[0];
                valueB = ((cl_short*) sourceB)[0];
                valueC = ((cl_short*) sourceC)[0];
                break;
            case kChar:
                valueA = ((cl_char*) sourceA)[0];
                valueB = ((cl_char*) sourceB)[0];
                valueC = ((cl_char*) sourceC)[0];
                break;
            default:
                //error -- should never get here
                abort();
                break;
        }

        if(valueB > valueC) {
            return false; // undefined behavior : leave "expected" alone
        }

        switch( vecAType )
        {
            case kLong:
                ((cl_long*) destination)[0] = MAX(MIN(valueA, valueC), valueB);
                break;
            case kInt:
                ((cl_int*) destination)[0] = (cl_int)
                    (MAX(MIN(valueA, valueC), valueB));
                break;
            case kShort:
                ((cl_short*) destination)[0] = (cl_short)
                    (MAX(MIN(valueA, valueC), valueB));
                break;
            case kChar:
                ((cl_char*) destination)[0] = (cl_char)
                    (MAX(MIN(valueA, valueC), valueB));
                break;
            default:
                //error -- should never get here
                abort();
                break;
        }

    }
    return true;
}

int test_integer_clamp(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
    return test_three_param_integer_fn( queue, context, "clamp", verify_integer_clamp );
}

bool verify_integer_mad_sat( void *sourceA, void *sourceB, void *sourceC, void *destination,
                        ExplicitType vecAType, ExplicitType vecBType, ExplicitType vecCType, ExplicitType destType )
{
    if( vecAType == kULong || vecAType == kUInt || vecAType == kUShort || vecAType == kUChar )
    {
        cl_ulong valueA, valueB, valueC;

        switch( vecAType )
        {
            case kULong:
                valueA = ((cl_ulong*) sourceA)[0];
                valueB = ((cl_ulong*) sourceB)[0];
                valueC = ((cl_ulong*) sourceC)[0];
                break;
            case kUInt:
                valueA = ((cl_uint*) sourceA)[0];
                valueB = ((cl_uint*) sourceB)[0];
                valueC = ((cl_uint*) sourceC)[0];
                break;
            case kUShort:
                valueA = ((cl_ushort*) sourceA)[0];
                valueB = ((cl_ushort*) sourceB)[0];
                valueC = ((cl_ushort*) sourceC)[0];
                break;
            case kUChar:
                valueA = ((cl_uchar*) sourceA)[0];
                valueB = ((cl_uchar*) sourceB)[0];
                valueC = ((cl_uchar*) sourceC)[0];
                break;
            default:
                //error -- should never get here
                abort();
                break;
        }

        cl_ulong multHi, multLo;
        multiply_unsigned_64_by_64( valueA, valueB, multLo, multHi );

        multLo += valueC;
        multHi += multLo < valueC;  // carry if overflow
        if( multHi )
            multLo = 0xFFFFFFFFFFFFFFFFULL;

        switch( vecAType )
        {
            case kULong:
                ((cl_ulong*) destination)[0] = multLo;
                break;
            case kUInt:
                ((cl_uint*) destination)[0] = (cl_uint) MIN( multLo, (cl_ulong) CL_UINT_MAX );
                break;
            case kUShort:
                ((cl_ushort*) destination)[0] = (cl_ushort) MIN( multLo, (cl_ulong) CL_USHRT_MAX );
                break;
            case kUChar:
                ((cl_uchar*) destination)[0] = (cl_uchar) MIN( multLo, (cl_ulong) CL_UCHAR_MAX );
                break;
            default:
                //error -- should never get here
                abort();
                break;
        }
    }
    else
    {
        cl_long valueA, valueB, valueC;

        switch( vecAType )
        {
            case kLong:
                valueA = ((cl_long*) sourceA)[0];
                valueB = ((cl_long*) sourceB)[0];
                valueC = ((cl_long*) sourceC)[0];
                break;
            case kInt:
                valueA = ((cl_int*) sourceA)[0];
                valueB = ((cl_int*) sourceB)[0];
                valueC = ((cl_int*) sourceC)[0];
                break;
            case kShort:
                valueA = ((cl_short*) sourceA)[0];
                valueB = ((cl_short*) sourceB)[0];
                valueC = ((cl_short*) sourceC)[0];
                break;
            case kChar:
                valueA = ((cl_char*) sourceA)[0];
                valueB = ((cl_char*) sourceB)[0];
                valueC = ((cl_char*) sourceC)[0];
                break;
            default:
                //error -- should never get here
                abort();
                break;
        }

        cl_long multHi;
        cl_ulong multLo;
        multiply_signed_64_by_64( valueA, valueB, multLo, multHi );

        cl_ulong sum = multLo + valueC;
        // carry if overflow
        if( valueC >= 0 )
        {
            if( multLo > sum )
            {
                multHi++;
                if( CL_LONG_MIN == multHi )
                {
                    multHi = CL_LONG_MAX;
                    sum = CL_ULONG_MAX;
                }
            }
        }
        else
        {
            if( multLo < sum )
            {
                multHi--;
                if( CL_LONG_MAX == multHi )
                {
                    multHi = CL_LONG_MIN;
                    sum = 0;
                }
            }
        }

        // saturate
        if( multHi > 0 )
            sum = CL_LONG_MAX;
        else if( multHi < -1 )
            sum = CL_LONG_MIN;
        cl_long result = (cl_long) sum;

        switch( vecAType )
        {
            case kLong:
                ((cl_long*) destination)[0] = result;
                break;
            case kInt:
                result = MIN( result, (cl_long) CL_INT_MAX );
                result = MAX( result, (cl_long) CL_INT_MIN );
                ((cl_int*) destination)[0] = (cl_int) result;
                break;
            case kShort:
                result = MIN( result, (cl_long) CL_SHRT_MAX );
                result = MAX( result, (cl_long) CL_SHRT_MIN );
                ((cl_short*) destination)[0] = (cl_short) result;
                break;
            case kChar:
                result = MIN( result, (cl_long) CL_CHAR_MAX );
                result = MAX( result, (cl_long) CL_CHAR_MIN );
                ((cl_char*) destination)[0] = (cl_char) result;
                break;
            default:
                //error -- should never get here
                abort();
                break;
        }
    }
    return true;
}

int test_integer_mad_sat(cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
    return test_three_param_integer_fn( queue, context, "mad_sat", verify_integer_mad_sat );
}

bool verify_integer_mad_hi( void *sourceA, void *sourceB, void *sourceC, void *destination,
                            ExplicitType vecAType, ExplicitType vecBType, ExplicitType vecCType, ExplicitType destType )
{
    if( vecAType == kULong || vecAType == kUInt || vecAType == kUShort || vecAType == kUChar )
    {
        cl_ulong valueA, valueB, valueC;

        switch( vecAType )
        {
            case kULong:
                valueA = ((cl_ulong*) sourceA)[0];
                valueB = ((cl_ulong*) sourceB)[0];
                valueC = ((cl_ulong*) sourceC)[0];
                break;
            case kUInt:
                valueA = ((cl_uint*) sourceA)[0];
                valueB = ((cl_uint*) sourceB)[0];
                valueC = ((cl_uint*) sourceC)[0];
                break;
            case kUShort:
                valueA = ((cl_ushort*) sourceA)[0];
                valueB = ((cl_ushort*) sourceB)[0];
                valueC = ((cl_ushort*) sourceC)[0];
                break;
            case kUChar:
                valueA = ((cl_uchar*) sourceA)[0];
                valueB = ((cl_uchar*) sourceB)[0];
                valueC = ((cl_uchar*) sourceC)[0];
                break;
            default:
                //error -- should never get here
                abort();
                break;
        }

        cl_ulong multHi, multLo;
        multiply_unsigned_64_by_64( valueA, valueB, multLo, multHi );

        switch( vecAType )
        {
            case kULong:
                ((cl_ulong*) destination)[0] = multHi + valueC;
                break;
            case kUInt:
                ((cl_uint*) destination)[0] = (cl_uint) (( multLo >> 32) + valueC );
                break;
            case kUShort:
                ((cl_ushort*) destination)[0] = (cl_ushort) (( multLo >> 16) + valueC );
                break;
            case kUChar:
                ((cl_uchar*) destination)[0] = (cl_uchar) (( multLo >> 8) + valueC );
                break;
            default:
                //error -- should never get here
                abort();
                break;
        }
    }
    else
    {
        cl_long valueA, valueB, valueC;

        switch( vecAType )
        {
            case kLong:
                valueA = ((cl_long*) sourceA)[0];
                valueB = ((cl_long*) sourceB)[0];
                valueC = ((cl_long*) sourceC)[0];
                break;
            case kInt:
                valueA = ((cl_int*) sourceA)[0];
                valueB = ((cl_int*) sourceB)[0];
                valueC = ((cl_int*) sourceC)[0];
                break;
            case kShort:
                valueA = ((cl_short*) sourceA)[0];
                valueB = ((cl_short*) sourceB)[0];
                valueC = ((cl_short*) sourceC)[0];
                break;
            case kChar:
                valueA = ((cl_char*) sourceA)[0];
                valueB = ((cl_char*) sourceB)[0];
                valueC = ((cl_char*) sourceC)[0];
                break;
            default:
                //error -- should never get here
                abort();
                break;
        }

        cl_long multHi;
        cl_ulong multLo;
        multiply_signed_64_by_64( valueA, valueB, multLo, multHi );

        switch( vecAType )
        {
            case kLong:
                ((cl_long*) destination)[0] = multHi + valueC;
                break;
            case kInt:
                ((cl_int*) destination)[0] = (cl_int) ((multLo >> 32) + valueC);
                break;
            case kShort:
                ((cl_short*) destination)[0] = (cl_int) ((multLo >> 16) + valueC);
                break;
            case kChar:
                ((cl_char*) destination)[0] = (cl_char) (cl_int) ((multLo >> 8) + valueC);
                break;
            default:
                //error -- should never get here
                abort();
                break;
        }
    }
    return true;
}

int test_integer_mad_hi( cl_device_id deviceID, cl_context context, cl_command_queue queue, int num_elements)
{
    return test_three_param_integer_fn( queue, context, "mad_hi", verify_integer_mad_hi );
}