What's New

Intel® Advisor 2019 Beta

  • Flow graph analyzer:
    • New rapid visual prototyping environment
    • Interactively build, validate, and visualize algorithms
  • Enhanced roofline analysis
    • Preview Feature: Integer OP/S-based Roofline – Roofline chart for codes with integer data types. To enable this functionality set the environment variable ADVIXE_EXPERIMENTAL=int_roofline
    • Ability to adjust the Roofline to a custom number of threads to see the practical performance limits for given a application.
    • Export the Roofline chart to HTML from command line without configuring the graphical user interface in cluster environment.
  • MacOS* user interface
    • View and analyze data collected on Linux* or Windows*

Intel® Advisor 2018 Update 2

  • Memory and Dependencies analyses
    • Decreasing overhead for Memory Access Patterns and Dependencies analyses via limiting loop call count and analysis duration in project properties.
    • New metric Memory footprint for the first instance of the loop (calculated in bytes)
    • Ability to select loops in command line by source file and line: advixe-cl --mark-up-loops --select main.cpp:12,other.cpp:198
  • Recommendations
    • New style for recommendations pane and parameters or peel/remainder recommendation specific to certain user code
    • New recommendation to use non-temporal store (NTS) instructions to improve memory bound application performance.
  • Usability
    • Font size can be now customized in Options menu. This may help to adjust GUI appearance in SSH X-forwarding sessions.
    • Filter state is made persistent on re-opening results
  • Roofline
    • Roofline benchmarks can be synchronized for multi-rank MPI applications with the --benchmark-sync option.
    • Visualizing performance limits for each loop on Roofline chart
    • Ability to compare several Roofline results on the same chart
    • "Filter-in" arbitrary subset of dots on Roofline chart for simplifying complex charts analysis and saving only selected dots to bitmaps.
  • New OS support
    • Support for Fedora* 27
    • Support for Ubuntu* 17.10
    • Microsoft* Windows* 10 version 1709

Intel® Advisor 2018 Update 1

  • Selective profiling for Roofline, FLOPS and Trip Counts collections to decrease analysis scope and decrease overhead.
    • Select only loops of interest in GUI/CLI
    • Markup region of interest in source code via ITT "pause/resume" APIs
  • Improved finalization speed: Perform static instruction mix analysis on demand.
  • Usability improvements:
    • Run the Roofline analysis in command line with a single command
    • Improved UI responsiveness
    • Introduced progress bar for issues detection
    • Single HTML report generated from Python to share data easily in one file
    • Controls in Project Properties for suppressing warnings on missing debug info from system modules
    • Hints to configure search paths if symbols are not resolved
  • Fixed some issues with the Microsoft Visual Studio* 2017 integration
  • Bug fixes

Intel® Advisor 2018


  • General availability of Cache-aware Roofline performance characterization model, that graphically represents main performance limiting factors for the whole application and individual loops and functions.

  • Preview feature: Hierarchical Roofline: to enable, please set the environment variable ADVIXE_EXPERIMENTAL=roofline_ex before launching the Intel® Advisor (data re-collection may be required).

Deep characterization:

  • Total FLOPS metric in Top Down view that shows FLOPS distribution across call tree and allows to aggregate data on different hierarchy levels.

  • Dynamic Instruction Mix metric in Code Analytics.

  • Call Count metric for functions in addition to Trip Counts data.

New recommendations:

  • Force scalar remainder for loops with low mask utilization on Intel® Advanced Vector Extensions 512 (Intel® AVX-512).

  • Extended "Gather recommendation" with "Constant (non-unit)" pattern.

  • "Possible inefficient conflict-detection instructions present" recommendation.

  • Reporting names of serialized or scalar functions preventing vectorization.

  • "Vectorize call(s) to virtual method" recommendation.

Data reporting:

  • Joined Survey and Refinement command line report: advixe-cl -R joined

  • MKL breakdown in Summary and MKL filter in Survey report.

  • Divided read/write access in Memory Access Patterns report.

Usability improvements:

  • Toolbar was redesigned to use single Start button.

  • Ability to re-finalize Advisor Survey result in order to apply updated binary and source search paths.

  • New design for Recommendations tab with improved navigation.

  • Support for cross-OS analysis to all license types. Download installation packages for additional operating systems from registrationcenter.intel.com.

  • Filter by module for Survey, FLOPS and Trip Counts collections, that helps to decrease overhead and narrow down results for big applications.

Support for new processors and operating systems:

  • Added support for Intel® Xeon® Scalable processors

  • Added support for Microsoft Visual Studio* 2017

  • Added support for Microsoft Windows Server* 2016

  • Added support for Fedora* 26

  • Added support for Ubuntu* 17.04

  • Added support for SUSE* Linux Enterprise Server* 12 SP2

  • Added support for Debian* 9

Preview features:

  • Flow Graph Analyzer for analyzing applications using Intel® TBB Flow Graph (available as a separate download).
  • Cache simulator providing cache line utilization, number of cache misses, memory loads and stores for arbitrary cache size. To enable, set the environment variable ADVIXE_EXPERIMENTAL=cachesim before launching Intel® Advisor (data re-collection may be required).
  • Python* profiling for Survey collection and preliminary support for FLOPS, TripCounts and Roofline collections.
  • Experimental support for accessing Advisor data via Python* API. See examples in the {install_dir}/pythonapi/examples directory. Please note that the API is subject to change in the future releases.
Optimization Notice: 
For more complete information about compiler optimizations, see our Optimization Notice.