What's New? Intel® SDK for OpenCL™ Applications 2016, R2

New Features in the Linux* version

  • Beta release of Source and Assembly level GPU Kernel Debugging on Linux*
    • Seamless debugging experience that enables you to debug OpenCL™ Kernels running on Intel© Processor Graphics with real-time inspection of variables and registers across the hardware execution units threads
    • GDB* Support
    • Supported only on 6th Generation Intel© Core™ Processors with Intel© Iris™ Graphics, Iris™ Pro Graphics and HD Graphics
  • New Kernel Development Framework plugin for Eclipse* on Linux*
    • Assembly view of generated Intel© Processor Graphics assembly code with cross mapping support (from assembly code to OpenCL code)
    • Kernel analysis report with compute matrices and deep level analysis views
    • Host C++ code generator - generates full OpenCL application host C++ code for a given session
    • Lightweight "Run" option to execute the kernel and get immediate results. You can later run a full analysis on the kernel execution session
    • Output validation - assign reference to an output buffer or image object and get immediate results if the reference does or does not match the output variable
    • Command line interface - enables you to perform most of the tasks that are available through the IDE, including: capturing a kernel execution from an OpenCL application, executing a kernel and viewing its results, viewing the kernel's performance analysis results, and more
    • Advanced buffer and image viewer with pixel overlay presentation, options for showing image as buffer, compare mode and more

New Features in the Windows* version

  • OpenCL™ 2.1 and SPIR-V* support
    • OpenCL 2.1 development environment with the experimental CPU-only runtime for OpenCL 2.1
    • SPIR-V generation support with Intel© Code Builder for OpenCL™ offline compiler and Kernel Development Framework including textual representing of SPIR-V binaries
  • New features in Kernel Development Framework
    • Host C++ code generator - generates full OpenCL™ application host C++ code for a given session
    • Lightweight "Run" option to execute the kernel and get immediate results. You can later run a full analysis on the kernel execution session
    • Output validation - assign reference to an output buffer or image object and get immediate results if the reference does or does not match the output variable
    • Command line interface - enables you to perform most of the tasks that are available through the IDE, including: capture a kernel execution from an OpenCL application, execute a kernel and view its results, view the kernel's performance analysis results, and more
    • Advanced buffers and images viewer with pixel overlay presentation, show image as buffer options, compare mode and more
  • Beta2 release of Source and Assembly level GPU Kernel Debugging on Windows* OSes including bug fixes and improvements
  • Latency analysis on 6th Generation Intel© Core™ Processors with Intel© Iris™ Graphics, Iris™ Pro Graphics and HD Graphics
For more complete information about compiler optimizations, see our Optimization Notice.