search

hughperkins / tf-coriander

OpenCL 1.2 implementation for Tensorflow
Apache License 2.0
791 stars 90 forks source link

clCreateCommandQueue always return `-6` for dual GPUs #74

Closed ghost closed 6 years ago

ghost commented 6 years ago

First of all the tensorflow included seems outdated. I had updated the hash of dependencies to continue the compilation. Compiling is good.

Tested on Ubuntu 16.04

uname -a:

Linux karl-1230v3 4.10.0-42-generic #46~16.04.1-Ubuntu SMP Mon Dec 4 15:57:59 UTC 2017 x86_64 x86_64 x86_64 GNU/Linux

CPU: Intel Xeon E3-1230 v3, 4 Cores, 8 Threads RAM: 16G DDR3 1600MHz

tf-coriander: https://github.com/hughperkins/tf-coriander/commit/9fd0a7502c3361fc1de6b58ed02a5b22772a5f45

Simple test

karl@karl-1230v3:~/github.com/hughperkins/tf-coriander$ python3 tensorflow/stream_executor/cl/test/test_simple.py
OpenCL platform: AMD Accelerated Parallel Processing
OpenCL device: Tahiti
terminate called after throwing an instance of 'std::runtime_error'
  what():  Error creating OpenCL command queue, OpenCL errorcode: -6

Aborted (core dumped)

CLINFO:

Number of platforms                               1
  Platform Name                                   AMD Accelerated Parallel Processing
  Platform Vendor                                 Advanced Micro Devices, Inc.
  Platform Version                                OpenCL 2.1 AMD-APP (2527.3)
  Platform Profile                                FULL_PROFILE
  Platform Extensions                             cl_khr_icd cl_amd_event_callback cl_amd_offline_devices 
  Platform Host timer resolution                  <printPlatformInfo:5: get CL_PLATFORM_HOST_TIMER_RESOLUTION : error -30>
  Platform Extensions function suffix             AMD

  Platform Name                                   AMD Accelerated Parallel Processing
Number of devices                                 2
  Device Name                                     Tahiti
  Device Vendor                                   Advanced Micro Devices, Inc.
  Device Vendor ID                                0x1002
  Device Version                                  OpenCL 1.2 AMD-APP (2527.3)
  Driver Version                                  2527.3
  Device OpenCL C Version                         OpenCL C 1.2 
  Device Type                                     GPU
  Device Profile                                  FULL_PROFILE
  Device Board Name (AMD)                         AMD Radeon HD 7900 Series
  Device Topology (AMD)                           PCI-E, 03:00.0
  Max compute units                               16
  SIMD per compute unit (AMD)                     4
  SIMD width (AMD)                                16
  SIMD instruction width (AMD)                    1
  Max clock frequency                             1000MHz
  Graphics IP (AMD)                               6.0
  Device Partition                                (core)
    Max number of sub-devices                     16
    Supported partition types                     none specified
  Max work item dimensions                        3
  Max work item sizes                             1024x1024x1024
  Max work group size                             256
  Preferred work group size multiple              64
  Wavefront width (AMD)                           64
  Preferred / native vector sizes                 
    char                                                 4 / 4       
    short                                                2 / 2       
    int                                                  1 / 1       
    long                                                 1 / 1       
    half                                                 1 / 1        (n/a)
    float                                                1 / 1       
    double                                               1 / 1        (cl_khr_fp64)
  Half-precision Floating-point support           (n/a)
  Single-precision Floating-point support         (core)
    Denormals                                     No
    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
    Correctly-rounded divide and sqrt operations  No
  Address bits                                    32, Little-Endian
  Global memory size                              2009325568 (1.871GiB)
  Global free memory (AMD)                        <printDeviceInfo:68: get number of CL_DEVICE_GLOBAL_FREE_MEMORY_AMD : error -33>
  Global memory channels (AMD)                    12
  Global memory banks per channel (AMD)           16
  Global memory bank width (AMD)                  256 bytes
  Error Correction support                        No
  Max memory allocation                           1613091430 (1.502GiB)
  Unified memory for Host and Device              No
  Minimum alignment for any data type             128 bytes
  Alignment of base address                       2048 bits (256 bytes)
  Global Memory cache type                        Read/Write
  Global Memory cache size                        16384
  Global Memory cache line                        64 bytes
  Image support                                   Yes
    Max number of samplers per kernel             16
    Max size for 1D images from buffer            134217728 pixels
    Max 1D or 2D image array size                 2048 images
    Base address alignment for 2D image buffers   256 bytes
    Pitch alignment for 2D image buffers          256 bytes
    Max 2D image size                             16384x16384 pixels
    Max 3D image size                             2048x2048x2048 pixels
    Max number of read image args                 128
    Max number of write image args                8
  Local memory type                               Local
  Local memory size                               32768 (32KiB)
  Local memory syze per CU (AMD)                  65536 (64KiB)
  Local memory banks (AMD)                        32
  Max constant buffer size                        65536 (64KiB)
  Max number of constant args                     8
  Max size of kernel argument                     1024
  Queue properties                                
    Out-of-order execution                        No
    Profiling                                     Yes
  Prefer user sync for interop                    Yes
  Profiling timer resolution                      1ns
  Profiling timer offset since Epoch (AMD)        1513400371708032628ns (Sat Dec 16 12:59:31 2017)
  Execution capabilities                          
    Run OpenCL kernels                            Yes
    Run native kernels                            No
    Thread trace supported (AMD)                  No
    SPIR versions                                 1.2
  printf() buffer size                            4194304 (4MiB)
  Built-in kernels                                
  Device Available                                Yes
  Compiler Available                              Yes
  Linker Available                                Yes
  Device Extensions                               cl_khr_fp64 cl_amd_fp64 cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_int64_base_atomics cl_khr_int64_extended_atomics cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_gl_sharing cl_amd_device_attribute_query cl_amd_vec3 cl_amd_printf cl_amd_media_ops cl_amd_media_ops2 cl_amd_popcnt cl_khr_image2d_from_buffer cl_khr_spir cl_khr_gl_event 

  Device Name                                     Tahiti
  Device Vendor                                   Advanced Micro Devices, Inc.
  Device Vendor ID                                0x1002
  Device Version                                  OpenCL 1.2 AMD-APP (2527.3)
  Driver Version                                  2527.3
  Device OpenCL C Version                         OpenCL C 1.2 
  Device Type                                     GPU
  Device Profile                                  FULL_PROFILE
  Device Board Name (AMD)                         AMD Radeon HD 7900 Series
  Device Topology (AMD)                           PCI-E, 04:00.0
  Max compute units                               16
  SIMD per compute unit (AMD)                     4
  SIMD width (AMD)                                16
  SIMD instruction width (AMD)                    1
  Max clock frequency                             1000MHz
  Graphics IP (AMD)                               6.0
  Device Partition                                (core)
    Max number of sub-devices                     16
    Supported partition types                     none specified
  Max work item dimensions                        3
  Max work item sizes                             1024x1024x1024
  Max work group size                             256
  Preferred work group size multiple              64
  Wavefront width (AMD)                           64
  Preferred / native vector sizes                 
    char                                                 4 / 4       
    short                                                2 / 2       
    int                                                  1 / 1       
    long                                                 1 / 1       
    half                                                 1 / 1        (n/a)
    float                                                1 / 1       
    double                                               1 / 1        (cl_khr_fp64)
  Half-precision Floating-point support           (n/a)
  Single-precision Floating-point support         (core)
    Denormals                                     No
    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
    Correctly-rounded divide and sqrt operations  No
  Address bits                                    32, Little-Endian
  Global memory size                              3099906048 (2.887GiB)
  Global free memory (AMD)                        <printDeviceInfo:68: get number of CL_DEVICE_GLOBAL_FREE_MEMORY_AMD : error -33>
  Global memory channels (AMD)                    12
  Global memory banks per channel (AMD)           16
  Global memory bank width (AMD)                  256 bytes
  Error Correction support                        No
  Max memory allocation                           2420617216 (2.254GiB)
  Unified memory for Host and Device              No
  Minimum alignment for any data type             128 bytes
  Alignment of base address                       2048 bits (256 bytes)
  Global Memory cache type                        Read/Write
  Global Memory cache size                        16384
  Global Memory cache line                        64 bytes
  Image support                                   Yes
    Max number of samplers per kernel             16
    Max size for 1D images from buffer            134217728 pixels
    Max 1D or 2D image array size                 2048 images
    Base address alignment for 2D image buffers   256 bytes
    Pitch alignment for 2D image buffers          256 bytes
    Max 2D image size                             16384x16384 pixels
    Max 3D image size                             2048x2048x2048 pixels
    Max number of read image args                 128
    Max number of write image args                8
  Local memory type                               Local
  Local memory size                               32768 (32KiB)
  Local memory syze per CU (AMD)                  65536 (64KiB)
  Local memory banks (AMD)                        32
  Max constant buffer size                        65536 (64KiB)
  Max number of constant args                     8
  Max size of kernel argument                     1024
  Queue properties                                
    Out-of-order execution                        No
    Profiling                                     Yes
  Prefer user sync for interop                    Yes
  Profiling timer resolution                      1ns
  Profiling timer offset since Epoch (AMD)        1513400371708032628ns (Sat Dec 16 12:59:31 2017)
  Execution capabilities                          
    Run OpenCL kernels                            Yes
    Run native kernels                            No
    Thread trace supported (AMD)                  No
    SPIR versions                                 1.2
  printf() buffer size                            4194304 (4MiB)
  Built-in kernels                                
  Device Available                                Yes
  Compiler Available                              Yes
  Linker Available                                Yes
  Device Extensions                               cl_khr_fp64 cl_amd_fp64 cl_khr_global_int32_base_atomics cl_khr_global_int32_extended_atomics cl_khr_local_int32_base_atomics cl_khr_local_int32_extended_atomics cl_khr_int64_base_atomics cl_khr_int64_extended_atomics cl_khr_3d_image_writes cl_khr_byte_addressable_store cl_khr_gl_sharing cl_amd_device_attribute_query cl_amd_vec3 cl_amd_printf cl_amd_media_ops cl_amd_media_ops2 cl_amd_popcnt cl_khr_image2d_from_buffer cl_khr_spir cl_khr_gl_event 

NULL platform behavior
  clGetPlatformInfo(NULL, CL_PLATFORM_NAME, ...)  No platform
  clGetDeviceIDs(NULL, CL_DEVICE_TYPE_ALL, ...)   No platform
  clCreateContext(NULL, ...) [default]            No platform
  clCreateContext(NULL, ...) [other]              Success [AMD]
  clCreateContextFromType(NULL, CL_DEVICE_TYPE_CPU)  No devices found in platform
  clCreateContextFromType(NULL, CL_DEVICE_TYPE_GPU)  Success (2)
    Platform Name                                 AMD Accelerated Parallel Processing
    Device Name                                   Tahiti
    Device Name                                   Tahiti
  clCreateContextFromType(NULL, CL_DEVICE_TYPE_ACCELERATOR)  No devices found in platform
  clCreateContextFromType(NULL, CL_DEVICE_TYPE_CUSTOM)  No devices found in platform
  clCreateContextFromType(NULL, CL_DEVICE_TYPE_ALL)  Success (2)
    Platform Name                                 AMD Accelerated Parallel Processing
    Device Name                                   Tahiti
    Device Name                                   Tahiti
ghost commented 6 years ago

Also, I had tried to run without GPU, seems working for simple test running on CPU.

ghost commented 6 years ago

According to https://streamhpc.com/blog/2013-04-28/opencl-error-codes/, -6 seems mean CL_OUT_OF_HOST_MEMORY. So that means the host with 16G RAM is insufficient for a simple test? Or 6GB GDDR ram insufficient?

ghost commented 6 years ago

I just tested with the OPENCL demo from http://laanwj.github.io/2016/05/06/opencl-ubuntu1604.html And I find out the reason is maybe your code just create the clContext using the device NULL. If yiou provide the device id with NULL, the function clCreateCommandQueue will always return -6.

As my test results, this is definitely a potential problem for AMD graphics cards when there are dual GPUs.

Some other users complain about the AMD driver implementation: https://community.amd.com/thread/220038

Also, I will provide my test code in a moment.

wileyyugioh commented 6 years ago

This error is not only for dual-gpus, but was also happening for my single HD 7770. A solution I that worked for me was https://github.com/fireice-uk/xmr-stak-amd/issues/69#issuecomment-318248745

#!/bin/bash
export GPU_FORCE_64BIT_PTR=1
export GPU_USE_SYNC_OBJECTS=1
export GPU_MAX_ALLOC_PERCENT=100
export GPU_SINGLE_ALLOC_PERCENT=100
export GPU_MAX_HEAP_SIZE=100