Applying the Profile-Guided Optimization (PGO) to Fluid Animate

Fluid animate is one of a class of algorithms for calculating fluid flow. There is one implementation using Intel® Cilk Plus technology. This sample uses the same serial implementation to demonstrate the use of Intel Profile Guided Optimization (PGO) without any code change. The PGO of Intel® C++ Compiler improves application performance by reorganizing code layout to reduce instruction-cache problems and branch mispredictions. With the collected application runtime information Intel C++ Compiler is able to be more selective and specific in optimizing the application. This may also help with application size introduced by aggressive inlining. You can find more in-depth explanation in this article Guide to Profile-guided Optimization of Computational Fluid Dynamics with Intel® Compiler.

This code originally written as part of the Princeton Parsec benchmark suite by Richard O. Lee and later modified by Christian Bienia and Christian Fensch.

• System Requirements
• Hardware:
  • Any Intel® processor like 2nd Generation Intel Core™ i3, i5, or i7 processors and Intel Xeon® E3 or E5 processor family
  For Microsoft* Windows*:
  • Microsoft Visual Studio 2010*, 2012*, or 2013* Professional Edition or above
  • Intel® Parallel Studio XE 2015 Composer Edition for C++ Windows*
  For Linux*:
  • GNU* GCC 4.5 or newer
  • Intel® Parallel Studio XE 2015 Composer Edition for C++ Linux*
  For OS X*:
  • OS X 10.9 or above
  • Xcode* 5.0 or above
  • Intel® Parallel Studio XE 2015 Composer Edition for C++ OS X OR
    Intel® Integrated Native Developer Experience 2015 Build Edition for OS X (Intel® INDE 2015 Build Edition for OS X)

Performance Gain with PGO

Build Instructions:

• For Microsoft Visual Studio 2010, 2012 or 2013 users:
• Open Visual Studio 2010, 2012 or 2013 and load the solution .sln file
  Choose a configuration (for best performance, choose a release configuration):

  • Intel-debug and Intel-release: uses Intel C++ compiler
  • VSC-debug and VSC-release: uses Microsoft Visual C++* compiler
  Run application to get a baseline run time for later use. Use "Intel-Release" configuration for performance testing

  • Project Properties -> C/C++ -> Preprocessor -> Preprocessor Definitions: add PERF_NUM
  Build with Intel C++ compiler PGO:

  • In the Solution Explorer window right click on FluidAnimate project name, select "Intel Compiler XE 15.0 > Profile Guided Optimization". It brings up "Profile Guided Optimization Dialog.
  • Without changing any default setting, click [Run] button. This step will start the following:
    • Build the FluidAnimate with /Qprof-gen option
    • Run the application to collect the profiling data
    • Rebuild the application with /Qprof-use option
    • At the end of this step the PGO optimized FluidAnimate binary is created at "PGOSample\Intel-Release\FluidAnimate.exe"
  Run PGO optimized binary to get the performance number: press Ctrl+F5 to run the application.

• For Windows Command Line users:
• For Intel C++ Compiler:

  • Open the appropriate Intel C++ compiler command prompt from Start menu
  • Build and run the application normally without PGO:
    build
build run -o 0
  • Build with PGO:
    build pgo
    This command will complete the followings:
    • Build FluidAnimate with /Qprof-gen
    • Run the application to collect profile data
    • Rebuild FluidAnimate with /Qprof-use and generate PGO optimized binary "release\FuildAnimatePGO.exe"
  [Optional]For Visual C++ Compiler (only linear/scalar will run):

  • Open the appropriate Microsoft Visual Studio command prompt from the Start menu and navigate to project folder
  • To compile: Build.bat [perf_num]
    • [perf_num]: collect performance numbers (will run example 5 times and average time taken)
  • To run: Build.bat run

• For Linux* or OS X* users:
• From a terminal window, navigate to the project folder
  Using Intel® C++ compiler:

  • Set the environment: source <icc-install-dir>/bin/compilervars.sh ia32 or intel-64
  • To compile with PGO: run buildPGO.sh shell script ". ./BuildPGO.sh"
    This step completes the followings:
    • Generate an instrumented binary
    • Run the instrucmented binary to collect profile data
    • Build the application again using the profile data and generates PGO optimized binary intel-release/FluidAnimatePGO
    To run the PGO optimized binary:
    . ./intel-release/FluidAnimatePGO
  [Optional] Using gcc (only linear/scalar will run):

  • To compile: make gcc [perf_num=1]
    • [perf_num=1]: collect performance numbers (will run example 5 times and average time taken)
  • To run: make run

• For OS X* users using Intel® Integrated Native Developer Experience Build Edition for OS X* (Intel® INDE Build Edition for OS X* ):
• From a terminal window, navigate to the project folder
  Using Intel® C++ compiler:

  • Set the environment: source <icc-install-dir>/bin/compilervars.sh ia32 or intel-64
  • To compile with PGO: run buildPGO_inde.sh shell script ". ./BuildPGO_inde.sh"
    This step completes the followings:
    • Generate an instrumented binary
    • Run the instrucmented binary to collect profile data
    • Build the application again using the profile data and generates PGO optimized binary intel-release/FluidAnimatePGO
    To run the PGO optimized binary:
    . ./intel-release/FluidAnimatePGO
  [Optional] Using gcc (only linear/scalar will run):

  • To compile: make gcc [perf_num=1]
    • [perf_num=1]: collect performance numbers (will run example 5 times and average time taken)
  • To run: make run