CUDA C to OpenCL 代码转换器

Edison

http://www.multiscalelab.org/swan

Swan: A simple tool for porting CUDA kernels to OpenCL

Download latest version

For the impatient: download the latest version of Swan here (by downloading this file, you are accepting the GPL2 license http://www.gnu.org/licenses/gpl-2.0.txt). To build, untar and follow the instructions in the README. the requirements are:

    * A modern(ish) Linux OS
    * CUDA toolkit and/or OpenCL runtime and a matching GPU
    * Perl (with the package Text::Balanced)
    * gcc/g++ 3.x or later

What is it?

Swan is a small tool that aids the reversible conversion of existing CUDA codebases to OpenCL. It does several useful things:

    * Translates CUDA kernel source-code to OpenCL.
    * Provides a common API that abstracts both CUDA and OpenCL runtimes.
    *

      Preserves the convenience of the CUDA <<< grid, block >>> kernel launch syntax by generating C source-code for kernel entry-point functions.

Why might you want it?

Possible uses include:

    * Evaluating OpenCL performance of an existing CUDA code.
    * Maintaining a dual-target OpenCL and CUDA code.
    * Reducing dependence on NVCC when compiling host code.
    * Support multiple CUDA compute capabilities in a single binary

Limitations

It's not a drop-in replacement for nvcc. Host code needs to have all kernel invocations and CUDA API calls re-written.

Swan does not support a few things. In particular:

    * CUDA C++ templating in kernel code.
    * OpenCL Images/Samplers (analogous to Textures).
    * Multiple device management in a single process.
    * Compiling kernels for the CPU.
    * CUDA device-emulation mode.

Furthermore, it's a work in progress. It works for our code but no promises it will for yours.

Use

The essential steps for converting a CUDA code:

    * Separate CUDA kernels into separate source files.
    * Compile the kernel source using swan. For a CUDA target:

swan --cuda kernel.kh kernel.cu

or for OpenCL

swan --opencl kernel.kh kernel.cu

    *

      The output file kernel.kh contains:
          o a data block which holds the compiled kernel source code (PTX or IL).
          o autogenerated source-code functions for running each of the kernels
    *

      #include "kernel.kh" into the host source.
    * Re-write all kernel invocations to use the corresponding entry-point functions. For example, the kernel with formal declaration:

__global__ void vec_add( float *in, float *out, int N );

will have an entry point defined as:

void k_vec_add( int3 grid, int3 block, int shmem, float *in, float *out, int N );

where grid, block and shmem correspond to the first three arguments of the <<< >>> syntax.

Entry points are named systematically. For any kernel kernel(), there will be two entry points: k_kernel() and k_kernel_async(). The former is a synchronous call (blocks until kernel completes execution).

    *

      Re-write all of the CUDA API calls with the Swan analogues (see swan_api.h)
    *

      Compile and link code against the appropriate Swan library (libswan_ocl or libswan_cuda).

Examples can be found in the package.

Contact

For queries, problems and suggestions, contact Matt @ M.J.Harvey (at) imperial (dot) ic (dot) ac (dot) uk.

Publications

    *

      Experiences porting from CUDA to OpenCL Presentation at the Daresbury Machine Evaluation Workshop, 2009

License

Swan is licensed under the GPL version 2

什么？

这个工具NV应该很不高兴看到吧

disruptor

对于nv来说，opencl就是cudac的一个马甲

什么？

请问粗黑体如何得出的结论？

jocover

CUDA和clock函数他怎么在opencl上实现