Intel® SDK for OpenCL™ Applications

OpenCL™ Device Fission 助力 CPU 性能

下载 PDF

概要

Device Fission 是 OpenCL™ 规范的一种特性,可为 OpenCL 编程人员提供更强的能力和控制,以更好地管理哪些计算单元运行 OpenCL 命令。 从根本上讲,Device Fission 支持将设备再次划分为一个或多个子设备,如果使用得当,这可以提供出色的性能优势,尤其是在运行 CPU 时。

面向 OpenCL™ 应用的英特尔® 软件开发套件是专为基于英特尔® 架构的平台上的 OpenCL 应用提供的一个全面的软件开发环境。 该软件开发套件支持开发人员在使用 Windows* 和 Linux* 操作系统的英特尔® CPU 上开发 OpenCL 应用并以其为目标。

  • Linux*
  • Microsoft Windows* (XP, Vista, 7)
  • OpenCL™ Code Builder
  • Intel® SDK for OpenCL™ Applications
  • OpenCL*
  • Parallel Computing
  • Using SPIR for fun and profit with Intel® OpenCL™ Code Builder

    This short tutorial provides a brief introduction to Khronos SPIR. It touches on the differences between a SPIR binary and an Intel proprietary Intermediate Binary, demonstrates ways to create SPIR binaries using tools shipped with Intel® SDK for OpenCL™ Applications , and explains how to use SPIR binaries in your OpenCL program.
  • Professors
  • Students
  • Microsoft Windows* (XP, Vista, 7)
  • Microsoft Windows* 8.x
  • Windows*
  • C/C++
  • Intermediate
  • Intel® Media Server Studio Essentials Edition
  • Intel® Media Server Studio Professional Edition
  • OpenCL™ Code Builder
  • Intel® SDK for OpenCL™ Applications
  • OpenCL*
  • SPIR
  • SPIR-V
  • Development Tools
  • Graphics
  • Intel® Core™ Processor
  • Optimization
  • Parallel Computing
  • License Agreement: 

    Work-Group Size Considerations for Intel® Xeon Phi™ Coprocessors

    The recommended work-group size for kernels is multiple of 16, which equals the SIMD width for the float and int data type. The automatic vectorization module packs the work-items into SIMD packets of 16 items (for double as well), and processed the rest (“tail”) of the work-group in a scalar way. In other words, a work-group with the size of 2*SIMD_WIDTH executes faster than, the one with the size of 2* SIMD_WIDTH-1.

    Subscribe to Intel® SDK for OpenCL™ Applications