search

intel / compute-runtime

Intel® Graphics Compute Runtime for oneAPI Level Zero and OpenCL™ Driver
MIT License
1.15k stars 234 forks source link

Comparing Win Neo vs Linux Neo driver and some Linux benchmarks.. #18

Closed oscarbg closed 6 years ago

oscarbg commented 6 years ago

Ok, I compiled today the driver and run and up to date clinfo tool (full report below):

extensions differences vs Windows driver are: Linux drv has (and Windows no): cl_khr_priority_hints cl_khr_throttle_hints cl_khr_create_command_queue

more info on "cl_khr_create_command_queue" extension, please? can't find info on OpenCL registry..

Win drv has (and Linux drv no): cl_khr_gl_sharing cl_khr_gl_depth_images cl_khr_gl_event cl_khr_gl_msaa_sharing

as you mention on README "delta" other driver: OpenGL sharing with MESA driver..

this means currently there is no OpenGL interop on Neo with Mesa right? are you working on it/have plans to implement it?

thanks.. Report:

Number of platforms                               1
  Platform Name                                   Intel(R) OpenCL HD Graphics
  Platform Vendor                                 Intel(R) Corporation
  Platform Version                                OpenCL 2.1 
  Platform Profile                                FULL_PROFILE
  Platform 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_subgroups cl_khr_il_program cl_khr_fp64 cl_intel_planar_yuv cl_intel_packed_yuv cl_intel_motion_estimation cl_intel_advanced_motion_estimation cl_intel_va_api_media_sharing 
  Platform Host timer resolution                  1ns
  Platform Extensions function suffix             INTEL

  Platform Name                                   Intel(R) OpenCL HD Graphics
Number of devices                                 1
  Device Name                                     Intel(R) Gen9 HD Graphics NEO
  Device Vendor                                   Intel(R) Corporation
  Device Vendor ID                                0x8086
  Device Version                                  OpenCL 2.1 NEO 
  Driver Version                                  1.0
  Device OpenCL C Version                         OpenCL C 2.0 
  Device Type                                     GPU
  Device Available                                Yes
  Device Profile                                  FULL_PROFILE
  Max compute units                               24
  Max clock frequency                             0MHz
  Device Partition                                (core)
    Max number of sub-devices                     0
    Supported partition types                     None
  Max work item dimensions                        3
  Max work item sizes                             256x256x256
  Max work group size                             256
  Compiler Available                              Yes
  Linker Available                                Yes
  Preferred work group size multiple              32
  Max sub-groups per work group                   32
  Sub-group sizes (Intel)                         8x16x32
  Preferred / native vector sizes                 
    char                                                16 / 16      
    short                                                8 / 8       
    int                                                  4 / 4       
    long                                                 1 / 1       
    half                                                 8 / 8        (cl_khr_fp16)
    float                                                1 / 1       
    double                                               1 / 1        (cl_khr_fp64)
  Half-precision Floating-point support           (cl_khr_fp16)
    Denormals                                     Yes
    Infinity and NANs                             Yes
    Round to nearest                              Yes
    Round to zero                                 Yes
    Round to infinity                             Yes
    IEEE754-2008 fused multiply-add               Yes
    Support is emulated in software               No
  Single-precision Floating-point support         (core)
    Denormals                                     Yes
    Infinity and NANs                             Yes
    Round to nearest                              Yes
    Round to zero                                 Yes
    Round to infinity                             Yes
    IEEE754-2008 fused multiply-add               Yes
    Support is emulated in software               No
    Correctly-rounded divide and sqrt operations  Yes
  Double-precision Floating-point support         (cl_khr_fp64)
    Denormals                                     Yes
    Infinity and NANs                             Yes
    Round to nearest                              Yes
    Round to zero                                 Yes
    Round to infinity                             Yes
    IEEE754-2008 fused multiply-add               Yes
    Support is emulated in software               No
  Address bits                                    64, Little-Endian
  Global memory size                              3435970560 (3.2GiB)
  Error Correction support                        No
  Max memory allocation                           1717985280 (1.6GiB)
  Unified memory for Host and Device              Yes
  Shared Virtual Memory (SVM) capabilities        (core)
    Coarse-grained buffer sharing                 Yes
    Fine-grained buffer sharing                   No
    Fine-grained system sharing                   No
    Atomics                                       No
  Minimum alignment for any data type             128 bytes
  Alignment of base address                       1024 bits (128 bytes)
  Preferred alignment for atomics                 
    SVM                                           0 bytes
    Global                                        0 bytes
    Local                                         0 bytes
  Max size for global variable                    65536 (64KiB)
  Preferred total size of global vars             1717985280 (1.6GiB)
  Global Memory cache type                        Read/Write
  Global Memory cache size                        524288 (512KiB)
  Global Memory cache line size                   64 bytes
  Image support                                   Yes
    Max number of samplers per kernel             16
    Max size for 1D images from buffer            107374080 pixels
    Max 1D or 2D image array size                 2048 images
    Base address alignment for 2D image buffers   4 bytes
    Pitch alignment for 2D image buffers          4 pixels
    Max 2D image size                             16384x16384 pixels
    Max planar YUV image size                     16384x16380 pixels
    Max 3D image size                             16384x16384x2048 pixels
    Max number of read image args                 128
    Max number of write image args                128
    Max number of read/write image args           128
  Max number of pipe args                         16
  Max active pipe reservations                    1
  Max pipe packet size                            1024
  Local memory type                               Local
  Local memory size                               65536 (64KiB)
  Max constant buffer size                        1717985280 (1.6GiB)
  Max number of constant args                     8
  Max size of kernel argument                     1024
  Queue properties (on host)                      
    Out-of-order execution                        Yes
    Profiling                                     Yes
  Queue properties (on device)                    
    Out-of-order execution                        Yes
    Profiling                                     Yes
    Preferred size                                131072 (128KiB)
    Max size                                      67108864 (64MiB)
  Max queues on device                            1
  Max events on device                            1024
  Prefer user sync for interop                    Yes
  Profiling timer resolution                      83ns
  Execution capabilities                          
    Run OpenCL kernels                            Yes
    Run native kernels                            No
    Sub-group independent forward progress        Yes
    IL version                                    SPIR-V_1.0 
    SPIR versions                                 1.2 
  printf() buffer size                            4194304 (4MiB)
  Built-in kernels                                block_motion_estimate_intel;block_advanced_motion_estimate_check_intel;block_advanced_motion_estimate_bidirectional_check_intel;
  Motion Estimation accelerator version (Intel)   2
    Device-side AVC Motion Estimation version     1
      Supports texture sampler use                Yes
      Supports preemption                         No
  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_subgroups cl_khr_il_program cl_khr_fp64 cl_intel_planar_yuv cl_intel_packed_yuv cl_intel_motion_estimation cl_intel_advanced_motion_estimation cl_intel_va_api_media_sharing
oscarbg commented 6 years ago

Some benchmarks: Platform: Intel(R) OpenCL HD Graphics half supported! Device: Intel(R) Gen9 HD Graphics NEO Driver version : 1.0 (Linux x64) Compute units : 24 Clock frequency : 0 MHz

    Half-precision compute (GFLOPS)
      half   : 748.56
      half2  : 743.20
      half4  : 741.60
      half8  : 735.49
      half16 : 729.58

    Global memory bandwidth (GBPS)
      float   : 26.32
      float2  : 25.38
      float4  : 27.56
      float8  : 28.44
      float16 : 21.64

    Single-precision compute (GFLOPS)
      float   : 436.48
      float2  : 436.28
      float4  : 436.23
      float8  : 434.85
      float16 : 429.21

    Double-precision compute (GFLOPS)
      double   : 110.00
      double2  : 109.86
      double4  : 109.64
      double8  : 109.05
      double16 : 108.10

    Transfer bandwidth (GBPS)
      enqueueWriteBuffer         : 32.54
      enqueueReadBuffer          : 12.91
      enqueueMapBuffer(for read) : 2684354.50
        memcpy from mapped ptr   : 12.79
      enqueueUnmap(after write)  : inf
        memcpy to mapped ptr     : 13.12

    Kernel launch latency : 25.37 us
BabelStream
Version: 3.3
Implementation: OpenCL
Running kernels 100 times
Precision: float
Array size: 134.2 MB (=0.1 GB)
Total size: 402.7 MB (=0.4 GB)
Using OpenCL device Intel(R) Gen9 HD Graphics NEO
Driver: 1.0
Reduction kernel config: 96 groups of size 256
Validation failed on sum. Error 0.000572205
Sum was 39.791820526123 but should be 39.7912483215332
Function    MBytes/sec  Min (sec)   Max         Average     
Copy        29190.297   0.00920     0.01023     0.00935     
Mul         28857.390   0.00930     0.01076     0.00945     
Add         28797.739   0.01398     0.01640     0.01422     
Triad       28885.098   0.01394     0.01700     0.01419     
Dot         26463.983   0.01014     0.01268     0.01038   
mixbench-ocl/read-only (v0.01-?)
Use "-h" argument to see available options
------------------------ Device specifications ------------------------
Platform:            Intel(R) OpenCL HD Graphics
Device:              Intel(R) Gen9 HD Graphics NEO/Intel(R) Corporation
Driver version:      1.0
Address bits:        64
GPU clock rate:      0 MHz
Total global mem:    3276 MB
Max allowed buffer:  1638 MB
OpenCL version:      OpenCL 2.1 NEO 
Total CUs:           24
-----------------------------------------------------------------------
Buffer size:            256MB
Workgroup size:         256
Elements per workitem:  8
Workitem fusion degree: 4
Workitem stride:        NDRange
Buffer allocation:      Device allocated
Timer:                  CL event based
Loading kernel source file...
Precompilation of kernels... [>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>]
----------------------------------------------------------------------------- CSV data -----------------------------------------------------------------------------
Experiment ID, Single Precision ops,,,,              Double precision ops,,,,              Half precision ops,,,,                Integer operations,,, 
Compute iters, Flops/byte, ex.time,  GFLOPS, GB/sec, Flops/byte, ex.time,  GFLOPS, GB/sec, Flops/byte, ex.time,  GFLOPS, GB/sec, Iops/byte, ex.time,   GIOPS, GB/sec
            0,      0.250,    6.94,    4.84,  19.34,      0.125,   14.17,    2.37,  18.95,      0.500,    6.87,    9.76,  19.53,     0.250,    6.57,    5.10,  20.41
            1,      0.750,    6.59,   15.27,  20.36,      0.375,   13.23,    7.61,  20.29,      1.500,    6.28,   32.08,  21.39,     0.750,    5.94,   16.96,  22.61
            2,      1.250,    6.00,   27.97,  22.37,      0.625,   11.71,   14.32,  22.92,      2.500,    5.53,   60.65,  24.26,     1.250,    5.38,   31.18,  24.94
            3,      1.750,    5.46,   42.98,  24.56,      0.875,   11.35,   20.69,  23.64,      3.500,    5.47,   85.81,  24.52,     1.750,    5.30,   44.31,  25.32
            4,      2.250,    5.31,   56.82,  25.25,      1.125,   11.50,   26.27,  23.35,      4.500,    5.47,  110.44,  24.54,     2.250,    5.33,   56.70,  25.20
            5,      2.750,    5.35,   69.02,  25.10,      1.375,   11.51,   32.07,  23.32,      5.500,    5.42,  136.27,  24.78,     2.750,    5.34,   69.18,  25.16
            6,      3.250,    5.33,   81.86,  25.19,      1.625,   11.62,   37.53,  23.09,      6.500,    5.40,  161.67,  24.87,     3.250,    5.31,   82.15,  25.28
            7,      3.750,    5.32,   94.64,  25.24,      1.875,   11.60,   43.38,  23.13,      7.500,    5.43,  185.34,  24.71,     3.750,    5.31,   94.71,  25.26
            8,      4.250,    5.29,  107.78,  25.36,      2.125,   11.77,   48.46,  22.80,      8.500,    5.45,  209.48,  24.64,     4.250,    5.56,  102.56,  24.13
            9,      4.750,    5.36,  119.02,  25.06,      2.375,   11.48,   55.55,  23.39,      9.500,    5.43,  234.62,  24.70,     4.750,    6.02,  105.92,  22.30
           10,      5.250,    5.32,  132.46,  25.23,      2.625,   11.58,   60.85,  23.18,     10.500,    5.44,  258.86,  24.65,     5.250,    6.49,  108.58,  20.68
           11,      5.750,    5.29,  145.84,  25.36,      2.875,   11.33,   68.12,  23.69,     11.500,    5.57,  277.26,  24.11,     5.750,    7.04,  109.66,  19.07
           12,      6.250,    5.31,  157.89,  25.26,      3.125,   11.41,   73.54,  23.53,     12.500,    5.59,  300.00,  24.00,     6.250,    7.50,  111.89,  17.90
           13,      6.750,    5.31,  170.55,  25.27,      3.375,   11.38,   79.65,  23.60,     13.500,    5.46,  331.91,  24.59,     6.750,    7.90,  114.64,  16.98
           14,      7.250,    5.43,  179.28,  24.73,      3.625,   11.67,   83.38,  23.00,     14.500,    5.89,  330.58,  22.80,     7.250,    8.39,  116.00,  16.00
           15,      7.750,    5.62,  185.18,  23.89,      3.875,   12.17,   85.45,  22.05,     15.500,    5.49,  379.18,  24.46,     7.750,    8.85,  117.52,  15.16
           16,      8.250,    5.32,  208.31,  25.25,      4.125,   12.55,   88.24,  21.39,     16.500,    5.41,  409.00,  24.79,     8.250,    9.32,  118.75,  14.39
           17,      8.750,    5.48,  214.47,  24.51,      4.375,   13.23,   88.75,  20.29,     17.500,    5.81,  404.33,  23.10,     8.750,    9.81,  119.77,  13.69
           18,      9.250,    5.53,  224.49,  24.27,      4.625,   13.88,   89.43,  19.34,     18.500,    5.47,  453.77,  24.53,     9.250,   10.27,  120.86,  13.07
           20,     10.250,    5.33,  258.17,  25.19,      5.125,   15.18,   90.60,  17.68,     20.500,    5.52,  498.43,  24.31,    10.250,   11.25,  122.25,  11.93
           22,     11.250,    5.64,  267.55,  23.78,      5.625,   16.42,   91.97,  16.35,     22.500,    5.59,  540.58,  24.03,    11.250,   12.22,  123.52,  10.98
           24,     12.250,    5.49,  299.47,  24.45,      6.125,   17.73,   92.74,  15.14,     24.500,    6.04,  544.16,  22.21,    12.250,   13.12,  125.36,  10.23
           28,     14.250,    6.11,  313.13,  21.97,      7.125,   20.35,   94.00,  13.19,     28.500,    6.60,  579.46,  20.33,    14.250,   15.00,  127.47,   8.95
           32,     16.250,    6.73,  323.90,  19.93,      8.125,   22.78,   95.74,  11.78,     32.500,    7.20,  606.20,  18.65,    16.250,   16.85,  129.47,   7.97
           40,     20.250,    8.00,  339.68,  16.77,     10.125,   27.99,   97.09,   9.59,     40.500,    8.43,  644.75,  15.92,    20.250,   20.53,  132.41,   6.54
           48,     24.250,    9.23,  352.76,  14.55,     12.125,   32.96,   98.74,   8.14,     48.500,    9.66,  673.92,  13.90,    24.250,   24.34,  133.71,   5.51
           56,     28.250,   10.44,  363.04,  12.85,     14.125,   38.12,   99.46,   7.04,     56.500,   10.98,  690.92,  12.23,    28.250,   27.56,  137.60,   4.87
           64,     32.250,   11.69,  370.32,  11.48,     16.125,   43.04,  100.56,   6.24,     64.500,   12.20,  709.70,  11.00,    32.250,   31.34,  138.10,   4.28
           80,     40.250,   14.17,  381.19,   9.47,     20.125,   52.88,  102.16,   5.08,     80.500,   14.64,  738.22,   9.17,    40.250,   38.66,  139.76,   3.47
           96,     48.250,   16.63,  389.36,   8.07,     24.125,   62.99,  102.80,   4.26,     96.500,   18.48,  700.69,   7.26,    48.250,   46.23,  140.07,   2.90
          128,     64.250,   21.55,  400.16,   6.23,     32.125,   83.37,  103.44,   3.22,    128.500,   26.36,  654.19,   5.09,    64.250,   60.20,  143.24,   2.23
          192,     96.250,   31.17,  414.44,   4.31,     48.125,  122.61,  105.37,   2.19,    192.500,   43.66,  591.74,   3.07,    96.250,   89.90,  143.70,   1.49
          256,    128.250,   40.64,  423.54,   3.30,     64.125,  162.92,  105.66,   1.65,    256.500,   47.51,  724.63,   2.83,   128.250,  119.34,  144.24,   1.12
--------------------------------------------------------------------------------------------------------------------------------------------------------------------
Geekbench 4.2.0 : http://www.geekbench.com/

System Information
  Operating System              Ubuntu Bionic Beaver (development branch) 4.16.0-041600rc1-generic x86_64
  Model                         Gigabyte Technology Co., Ltd. Z170X-Gaming 7
  Motherboard                   Gigabyte Technology Co., Ltd. Z170X-Gaming 7
  Memory                        31.3 GB 
  BIOS                          American Megatrends Inc. F22j

Processor Information
  Name                          Intel Core i5-6600K
  Topology                      1 Processor, 4 Cores
  Identifier                    GenuineIntel Family 6 Model 94 Stepping 3
  Base Frequency                3.90 GHz
  L1 Instruction Cache          32.0 KB x 2
  L1 Data Cache                 32.0 KB x 2
  L2 Cache                      256 KB x 2
  L3 Cache                      6.00 MB

OpenCL Information
  Platform Vendor               Intel(R) Corporation
  Platform Name                 Intel(R) OpenCL HD Graphics
  Device Vendor                 Intel(R) Corporation
  Device Name                   Intel(R) Gen9 HD Graphics NEO
  Maximum Frequency             0.00 Hz
  Compute Units                 24
  Device Memory                 3.20 GB

OpenCL
  Sobel                         30622         1.35 Gpixels/sec
  Histogram Equalization        29861        933.0 Mpixels/sec
  SFFT                           3760              9.37 Gflops
  Gaussian Blur                 34950        612.3 Mpixels/sec
  Face Detection                14945     4.37 Msubwindows/sec
  RAW                          128223         1.24 Gpixels/sec
  Depth of Field                36742        106.7 Mpixels/sec
  Particle Physics              14984               2368.8 FPS

Benchmark Summary
  OpenCL Score                 24427
AdamCetnerowski commented 6 years ago

On extensions in Linux, not present in Windows: NEO supports Windows and Linux using the same code base, but there is some latency in delivering the Windows closed source driver (integration, validation, signing) compared to Linux, where the source code is provided continously. These extensions will be present in the Windows shortly. New features on Windows will be present in subsequent driver releases.

On cl_khr_create_command_queue: Posting of the spec for this extension is WIP in Khronos.

On gl-sharing: The NEO driver on Windows provides sharing with the closed source OGL driver developed by VPG. On Linux, we intend to add sharing with MESA, but do not have any firm timelines as of today.

oscarbg commented 6 years ago

really thanks for info.. guess I can close..