OpenCL 1.1 规格正式发布

Edison

http://www.khronos.org/registry/cl/
Khronos OpenCL API Registry

The OpenCL API registry contains specifications of the core API; specifications of Khronos- and vendor-approved OpenCL extensions; header files corresponding to the specifications; and other related documentation.

OpenCL Core API Specification, Headers, and Documentation

The current version of OpenCL is OpenCL 1.1.

• OpenCL 1.1 Specification(revision 33, June 11, 2010).
• OpenCL 1.1 C++ Bindings Specification(revision 3, June 11, 2010).
• OpenCL 1.1 Online Manual Pages.
• All of the following headers should be present in a directory CL/ (or OpenCL/ on MacOS X). The single header file opencl.hincludes all of the other headers as appropriate for the target platform, and simply including opencl.h should be all that applications need to do.
  • opencl.h- OpenCL 1.1 Single Header File for Applications.
  • cl_platform.h- OpenCL 1.1 Platform-Dependent Macros.
  • cl.h- OpenCL 1.1 Core API Header File.
  • cl_ext.h- OpenCL 1.1 Extensions Header File.
  • cl_d3d10.h- OpenCL 1.1 Khronos OpenCL/Direct3D 10 Extensions Header File.
  • cl_gl.h- OpenCL 1.1 Khronos OpenCL/OpenGL Extensions Header File.
  • cl_gl_ext.h- OpenCL 1.1 Vendor OpenCL/OpenGL Extensions Header File.
  • Extension templatefor writing an OpenCL extension specification. Extensions in the registry (listed below) follow the structure of this document, which describes the purpose of each section in an extension specification.

temp444

PC上还是DirectX为王。

alleny

问题是现在哪里去找opencl的应用？自己写还是等3年？
fish2fish 发表于 2010-6-15 19:07

    自己写吧，写完了就去卖钱

fish2fish

问题是现在哪里去找opencl的应用？自己写还是等3年？

gz_easy

不知AMD Stream SDK何时更新支持OCL 1.1。

alleny

一个人玩的寂寞，两个人玩的才算是竞争，AMD写驱动能力貌似没有NV强，NV号称自己是软件公司。

餐具

不错～～～有了标准，两家就可以开始竞争了

jocover

258.19驱动支持opencl 1.1了

alleny

感觉NV在openCL上又领先了一步。驱家新闻http://news.mydrivers.com/1/167/167120.htm

alleny

回复 5# Edison


    OpenCL 1.1高处不胜寒，一般人想玩也买不起，呵呵。一般只能眼馋。什么时候fermi中端卡出了，再尝鲜了。lz不知道现在open source的好的项目也没有什么可一推荐的？

Edison

Well，很大部分的特性是为 Fermi 度身定制的，同时也做了一些必要的下限设置来满足 Cypress 的支持。

pp70145

能写成哪些显卡能支持那就更好了。。。

细节不是平民该懂的

限量极

幼儿园没毕业。。看不懂

Edison

Appendix E – Changes

E.1  Summary of changes from OpenCL 1.0

The following features are added to the OpenCL platform layer and runtime (sections 4 and 5):

  Following queries to table 4.3
o  CL_DEVICE_NATIVE_VECTOR_WIDTH_{CHAR | SHORT | INT | LONG | FLOAT |
DOUBLE | HALF}
o  CL_DEVICE_HOST_UNIFIED_MEMORY
o  CL_DEVICE_OPENCL_C_VERSION

  CL_CONTEXT_NUM_DEVICES to the list of queries specified to clGetContextInfo.

  Optional image formats: CL_Rx, CL_RGx and CL_RGBx.

  Support for sub-buffer objects – ability to create a buffer object that refers to a specific
region in another buffer object using clCreateSubBuffer.

  clEnqueueReadBufferRect, clEnqueueWriteBufferRect and
clEnqueueCopyBufferRect APIs to read from, write to and copy a rectangular region of
a buffer object respectively.  

  clSetMemObjectDestructorCallback API to allow a user to register a callback function
that will be called when the memory object is deleted and its resources freed.

  Options that control the OpenCL C version used when building Pro-Agram executable.  
These are described in section 5.6.3.5.

  CL_KERNEL_PREFERRED_WORK_GROUP_SIZE_MULTIPLE to the list of queries
specified to clGetKernelWorkGroupInfo.

  Support for user events.  User events allow applications to enqueue commands that wait
on a user event to finish before the command is executed by the device.  Following new
APIs are added - clCreateUserEvent and clSetUserEventStatus.

  clSetEventCallback API to register a callback function for a specific command
execution status.

The following modifications are made to the OpenCL platform layer and runtime (sections 4 and
5):


o  CL_DEVICE_MAX_PARAMETER_SIZE from 256 to 1024 bytes
o  CL_DEVICE_LOCAL_MEM_SIZE from 16 KB to 32 KB.

  The global_work_offset argument in clEnqueueNDRangeKernel can be a non-NULL
value.

  All API calls except clSetKernelArg are thread-safe.

The following features are added to the OpenCL C programming language (section 6):

  3-component vector data types.

  New built-in functions
o  get_global_offset work-item function defined in section 6.11.1.
o  minmag, maxmag math functions defined in section 6.11.2.
o  clamp integer function defined in section 6.11.3.   
o  (vector, scalar) variant of integer functions min and max in section 6.11.3.
o  async_work_group_strided_copy defined in section 6.11.10.
o  vec_step, shuffle and shuffle2 defined in section 6.11.12.

  cl_khr_byte_addressable_store extension is a core feature.

  cl_khr_global_int32_base_atomics, cl_khr_global_int32_extended_atomics,
cl_khr_local_int32_base_atomics and cl_khr_local_int32_extended_atomics
extensions are core features.  The built-in atomic function names are changed to use the
atomic_ prefix instead of atom_.

  Macros CL_VERSION_1_0 and CL_VERSION_1_1.

The following features in OpenCL 1.0 are deprecated:

  The clSetCommandQueueProperty API is no longer supported in OpenCL 1.1.
  The __ROUNDING_MODE__ macro is no longer supported in OpenCL C 1.1.

The following new extensions are added to section 9:

  cl_khr_gl_event – Creating a CL event object from a GL sync object.
  cl_khr_d3d10_sharing – Sharing memory objects with Direct3D 10.

The following modifications are made to the OpenCL ES Profile described in section 10:

  64-bit integer support is optional.