Tuesday, February 12, 2019

OpenCL 2.1 NEO

Here my test of integrated OpenCL kernels. I compile CapsBasic-beignet and it shows me that I have new NEO driver that replace beignet.

Number of available platforms: 2
Platform names:
    [0] Intel(R) OpenCL HD Graphics [Selected]
    [1] Intel Gen OCL Driver
Number of devices available for each type:
    CL_DEVICE_TYPE_CPU: 0
    CL_DEVICE_TYPE_GPU: 1
    CL_DEVICE_TYPE_ACCELERATOR: 0

*** Detailed information for each device ***

CL_DEVICE_TYPE_GPU[0]
    CL_DEVICE_NAME: Intel(R) Gen9 HD Graphics NEO
    CL_DEVICE_AVAILABLE: 1
    CL_DEVICE_VENDOR: Intel(R) Corporation
    CL_DEVICE_PROFILE: FULL_PROFILE
    CL_DEVICE_VERSION: OpenCL 2.1 NEO
    CL_DRIVER_VERSION: 19.03.12192
    CL_DEVICE_OPENCL_C_VERSION: OpenCL C 2.0
    CL_DEVICE_MAX_COMPUTE_UNITS: 23
    CL_DEVICE_MAX_CLOCK_FREQUENCY: 1100
    CL_DEVICE_MAX_WORK_GROUP_SIZE: 256
    CL_DEVICE_ADDRESS_BITS: 64
    CL_DEVICE_MEM_BASE_ADDR_ALIGN: 1024
    CL_DEVICE_MAX_MEM_ALLOC_SIZE: 1595310080
    CL_DEVICE_GLOBAL_MEM_SIZE: 3190620160
    CL_DEVICE_MAX_CONSTANT_BUFFER_SIZE: 1595310080
    CL_DEVICE_GLOBAL_MEM_CACHE_SIZE: 524288
    CL_DEVICE_GLOBAL_MEM_CACHELINE_SIZE: 64
    CL_DEVICE_LOCAL_MEM_SIZE: 65536
    CL_DEVICE_PROFILING_TIMER_RESOLUTION: 83
    CL_DEVICE_IMAGE_SUPPORT: 1
    CL_DEVICE_ERROR_CORRECTION_SUPPORT: 0
    CL_DEVICE_HOST_UNIFIED_MEMORY: 1
    CL_DEVICE_EXTENSIONS: cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_fp16 cl_khr_depth_images cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_icd cl_khr_image2d_from_buffer cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_intel_subgroups cl_intel_required_subgroup_size cl_intel_subgroups_short cl_khr_spir cl_intel_accelerator cl_intel_media_block_io cl_intel_driver_diagnostics cl_intel_device_side_avc_motion_estimation cl_khr_priority_hints cl_khr_throttle_hints cl_khr_create_command_queue cl_khr_fp64 cl_khr_subgroups cl_khr_il_program cl_intel_spirv_device_side_avc_motion_estimation cl_intel_spirv_media_block_io cl_intel_spirv_subgroups cl_khr_mipmap_image cl_khr_mipmap_image_writes cl_intel_planar_yuv cl_intel_packed_yuv cl_intel_motion_estimation cl_intel_advanced_motion_estimation cl_intel_va_api_media_sharing
    CL_DEVICE_PREFERRED_VECTOR_WIDTH_INT: 4
    CL_DEVICE_PREFERRED_VECTOR_WIDTH_LONG: 1
    CL_DEVICE_PREFERRED_VECTOR_WIDTH_FLOAT: 1
    CL_DEVICE_PREFERRED_VECTOR_WIDTH_DOUBLE: 1
    CL_DEVICE_NATIVE_VECTOR_WIDTH_INT: 4
    CL_DEVICE_NATIVE_VECTOR_WIDTH_LONG: 1
    CL_DEVICE_NATIVE_VECTOR_WIDTH_FLOAT: 1
    CL_DEVICE_NATIVE_VECTOR_WIDTH_DOUBLE: 1

Then I tested Intel(R) OpenCL HD Graphics with fractal-gen-opencl but results for existing implementation is not optimistic because it cannot perform
10000000 iterations to use on each pixel to determine its set membership of mandelbrot fractal. I was waiting a day but it still calculate pixels. 1000000 iterations works very fast ~8sec to create 768x786 fractal.

mandelbrot

 
burning-ship


debug magazine archive

  71 jounals still available on issuu with great story of netlabels time.  debug_mag Publisher Publications - Issuu