1

我克隆了https://github.com/codeplaysoftware/computecpp-sdk.git并修改了 computecpp-sdk/samples/accessors/accessors.cpp文件。

我刚加了std::cout << "SYCL exception caught: " << e.get_cl_code() << '\n';

查看完全修改的代码

/***************************************************************************
 *
 *  Copyright (C) 2016 Codeplay Software Limited
 *  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
 *
 *  For your convenience, a copy of the License has been included in this
 *  repository.
 *
 *  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.
 *
 *  Codeplay's ComputeCpp SDK
 *
 *  accessor.cpp
 *
 *  Description:
 *    Sample code that illustrates how to make data available on a device
 *    using accessors in SYCL.
 *
 **************************************************************************/

#include <CL/sycl.hpp>

#include <iostream>

using namespace cl::sycl;

int main() {
  /* We define the data to be passed to the device. */
  int data = 5;

  /* The scope we create here defines the lifetime of the buffer object, in SYCL
   * the lifetime of the buffer object dictates synchronization using RAII. */
  try {
    /* We can also create a queue that uses the default selector in
     * the queue's default constructor. */
    queue myQueue;

    /* We define a buffer in order to maintain data across the host and one or
     * more devices. We construct this buffer with the address of the data
     * defined above and a range specifying a single element. */
    buffer<int, 1> buf(&data, range<1>(1));

    myQueue.submit([&](handler& cgh) {
      /* We define accessors for requiring access to a buffer on the host or on
       * a device. Accessors are are like pointers to data we can use in
       * kernels to access the data. When constructing the accessor you must
       * specify the access target and mode. SYCL also provides the
       * get_access() as a buffer member function, which only requires an
       * access mode - in this case access::mode::read_write.
       * (make_access<>() has a second template argument which defaults
       * to access::mode::global) */
      auto ptr = buf.get_access<access::mode::read_write>(cgh);

      cgh.single_task<class multiply>([=]() {
        /* We use the subscript operator of the accessor constructed above to
         * read the value, multiply it by itself and then write it back to the
         * accessor again. */
        ptr[0] = ptr[0] * ptr[0];
      });
    });

    /* queue::wait() will block until kernel execution finishes,
     * successfully or otherwise. */
    myQueue.wait();

  } catch (exception const& e) {
    std::cout << "SYCL exception caught: " << e.what() << '\n';
    std::cout << "SYCL exception caught: " << e.get_cl_code() << '\n';
    return 2;
  }

  /* We check that the result is correct. */
  if (data == 25) {
    std::cout << "Hurray! 5 * 5 is " << data << '\n';
    return 0;
  } else {
    std::cout << "Oops! Something went wrong... 5 * 5 is not " << data << "!\n";
    return 1;
  }
}

构建后,我执行了二进制文件并得到以下错误输出:

$ ./accessors 
./accessors: /usr/local/cuda-8.0/lib64/libOpenCL.so.1: no version information available (required by /usr/local/computecpp/lib/libComputeCpp.so)
./accessors: /usr/local/cuda-8.0/lib64/libOpenCL.so.1: no version information available (required by /usr/local/computecpp/lib/libComputeCpp.so)
./accessors: /usr/local/cuda-8.0/lib64/libOpenCL.so.1: no version information available (required by /usr/local/computecpp/lib/libComputeCpp.so)
SYCL exception caught: Error: [ComputeCpp:RT0101] Failed to create kernel ((Kernel Name: SYCL_class_multiply))
SYCL exception caught: -45
 SYCL Runtime closed with the following errors:
SYCL objects are still alive while the runtime is shutting down

 This probably indicates that a SYCL object was created  but not properly destroyed. 
terminate called without an active exception
Aborted (core dumped)

硬件配置如下:

$ /usr/local/computecpp/bin/computecpp_info  /usr/local/computecpp/bin/computecpp_info: /usr/local/cuda-8.0/lib64/libOpenCL.so.1: no version information available (required by /usr/local/computecpp/bin/computecpp_info) /usr/local/computecpp/bin/computecpp_info: /usr/local/cuda-8.0/lib64/libOpenCL.so.1: no version information available (required by /usr/local/computecpp/bin/computecpp_info)
********************************************************************************

ComputeCpp Info (CE 0.7.0)

********************************************************************************

Toolchain information:

GLIBC version: 2.19 GLIBCXX: 20150426 This version of libstdc++ is supported.

********************************************************************************


Device Info:

Discovered 3 devices matching:   platform    : <any>   device type : <any>

-------------------------------------------------------------------------------- Device 0:

  Device is supported                     : NO - Device does not support SPIR   CL_DEVICE_NAME                          : GeForce GTX 750 Ti   CL_DEVICE_VENDOR                        : NVIDIA Corporation  CL_DRIVER_VERSION                       : 384.111   CL_DEVICE_TYPE     : CL_DEVICE_TYPE_GPU 
-------------------------------------------------------------------------------- Device 1:

  Device is supported                     : UNTESTED - Device not tested on this OS   CL_DEVICE_NAME                          : Intel(R) HD Graphics   CL_DEVICE_VENDOR                        : Intel(R) Corporation   CL_DRIVER_VERSION                       : r5.0.63503   CL_DEVICE_TYPE                          : CL_DEVICE_TYPE_GPU 
-------------------------------------------------------------------------------- Device 2:

  Device is supported                     : YES - Tested internally by Codeplay Software Ltd.   CL_DEVICE_NAME                          : Intel(R) Core(TM) i7-4790 CPU @ 3.60GHz   CL_DEVICE_VENDOR             : Intel(R) Corporation   CL_DRIVER_VERSION                       :
1.2.0.475   CL_DEVICE_TYPE                          : CL_DEVICE_TYPE_CPU 

If you encounter problems when using any of these OpenCL devices, please consult this website for known issues: https://computecpp.codeplay.com/releases/v0.7.0/platform-support-notes

********************************************************************************

请帮助理解错误并解决相同的问题。让我知道是否需要更多信息。我想在我的 NVidia GPU 上运行这个示例代码。

4

1 回答 1

2

ComputeCpp 是开放标准 SYCL 的实现,默认输出 SPIR 指令,NVidia OpenCL 实现无法使用 SPIR 指令。相反,您需要使用 ComputeCpp 来输出 NVidia 硬件可以理解的 PTX 指令。

为此,请在使用来自 GitHub 的示例代码项目进行 cmake 调用时添加参数“-DCOMPUTECPP_BITCODE=ptx64”。

该项目中的 FindComputeCpp.cmake 文件采用此标志并提供编译器指令以输出 PTX。如果您想对自己的项目执行此操作,可以从 FindComputeCpp.cmake 文件中获取相关部分。

于 2018-06-07T14:54:29.770 回答