Runtime profiling - Revision history

Robert-schade-e757@uni-paderborn.de at 11:31, 19 July 2024

2024-07-19T11:31:26Z

Daniel-schurhoff-de23@rwth-aachen.de at 08:27, 5 September 2019

2019-09-05T08:27:47Z

Jan-eitzinger-77c0@uni-erlangen.de: /* How to use gprof */

2019-06-22T05:51:10Z

How to use gprof

Jan-eitzinger-77c0@uni-erlangen.de: /* Links and more Information */

2019-06-04T12:28:46Z

Links and more Information

Jan-eitzinger-77c0@uni-erlangen.de: /* Runtime profiling for parallel applications */

2019-05-28T12:45:45Z

Runtime profiling for parallel applications

Jan-eitzinger-77c0@uni-erlangen.de: /* How to use perf */

2019-05-28T12:44:21Z

How to use perf

Jan-eitzinger-77c0@uni-erlangen.de: /* How to use gprof */

2019-05-28T12:43:31Z

How to use gprof

Jan-eitzinger-77c0@uni-erlangen.de: /* Introduction */

2019-05-28T12:41:26Z

Introduction

Jan-eitzinger-77c0@uni-erlangen.de: /* Authors */

2019-04-24T06:07:58Z

Authors

Jan-eitzinger-77c0@uni-erlangen.de at 14:19, 2 April 2019

2019-04-02T14:19:25Z

← Older revision		Revision as of 11:31, 19 July 2024
Line 1:		Line 1:
	[[Category:HPC-Developer]]		[[Category:HPC-Developer]]
		+	[[Category:Benchmarking]]<nowiki />
		+
	== Introduction ==		== Introduction ==

← Older revision		Revision as of 08:27, 5 September 2019
Line 1:		Line 1:
−		+	[[Category:HPC-Developer]]
	== Introduction ==		== Introduction ==

@@ Line 20: / Line 20: @@
 <syntaxhighlight lang="sh">
-gprof ./miniMD
+gprof ./a.out
 </syntaxhighlight>
@@ Line 26: / Line 26: @@
 <syntaxhighlight lang="sh">
-gprof ./miniMD | less
+gprof ./a.out | less
 </syntaxhighlight>

@@ Line 188: / Line 188: @@
 This was also the only tool that resolved the inlined C++ routine. There are many more collect modules available including hardware performance monitoring metrics. Please refer to the [https://software.intel.com/en-us/vtune-amplifier-help-command-line-interface Intel documentation] for more detailed information.
-== Links and more Information ==
+== Links and further information ==
 * [https://linux.die.net/man/1/gprof gprof man page]
 * [https://perf.wiki.kernel.org/index.php/Main_Page Perf Wiki pages]
 * [https://software.intel.com/en-us/vtune-amplifier-help-command-line-interface Intel Amplifier Command line Tool Documentation]

@@ Line 119: / Line 119: @@
 == Runtime profiling for parallel applications ==
-gprof as well as perf are designed to be used for single process applications. While there are possibilities to use gprof in a MPI settings (see the [https://www.lrz.de/services/compute/supermuc/tuning/gprof/ LRZ documentation]) a lightweight alternative to the more complex to use full trace tools is the command line version of Intel Amplifier. It is also using the builtin perf infrastructure in the Linux kernel.
+gprof as well as perf are designed to be used for single process applications. While there are possibilities to use gprof in a MPI setting (see the [https://www.lrz.de/services/compute/supermuc/tuning/gprof/ LRZ documentation]) a lightweight alternative to complex full trace tools is the command line version of Intel Amplifier. It is also using the builtin perf infrastructure in the Linux kernel.
 Example usage with MPI:

@@ Line 115: / Line 115: @@
 </syntaxhighlight>
-As can be seen the result is similar but not identical. Perf also reports on time spent in external entities as the kernel, mpiexec and system libraries.
+As can be seen the result is similar but not identical to gprof. Perf also reports on time spent in external entities as the kernel, mpiexec and system libraries.
 == Runtime profiling for parallel applications ==

@@ Line 6: / Line 6: @@
 == How to use gprof ==
-The first step is to compile and link the program with profiling enabled. For most compilers this is achieved by setting the '''-pg''' flag. For Intel compilers it is important to set the optimization flag afterward as the default optimization level is set to '''-O0''' when enabling profiling.
+The first step is to compile and link the program with profiling enabled. For most compilers this is done by setting the '''-pg''' flag. For Intel compilers it is important to set the optimisation flag afterward as the default optimisation level is set to '''-O0''' when profiling is enabled.
 Example for build options with Intel tool chain:
@@ Line 16: / Line 16: @@
 </syntaxhighlight>
-The generated application is executed and generates a '''gmon.out''' file containing the profiling output.
+When generated application is executed it generates a '''gmon.out''' file containing the profiling output.
 To analyse the profile the tool '''gprof''' is used with the executable as argument:

@@ Line 2: / Line 2: @@
 == Introduction ==
-The initial task in any performance analysis is to figure out in which parts of the code the runtime is spent. One wants to focus for optimisation on those regions of the code to achieve an overall speedup of the code. The tool helping to get an overview of where the time is spent is called a runtime profiler. There exist two flavours: Instrumentation based and sampling based profilers. Instrumentation based profilers insert function calls to measure the time at points in the program. Additional tasks may be performed as e.g. determining the function call stack. While it is possible to insert instrumentation calls on the binary level the common way is that the compiler adds instrumentation functions. The standard tool in this area is [https://en.wikipedia.org/wiki/Gprof gprof] and almost any compiler supports to instrument the code for gprof. Statistical sampling based profiling on the other hand are based on probing of the programs call stack triggered by operating system interrupts at regular intervals. A widespread tool for sampling based profiling is the [https://en.wikipedia.org/wiki/Perf_(Linux) perf] tool which builds on the builtin profiling infrastructure in recent Linux kernels. Both approaches have advantages and disadvantages: Instrumentation produces more accurate results but introduces more overhead and sampling has less overhead but produce less accurate results. Special care is necessary for runtime profiling of parallel (OpenMP or MPI) applications.
+The initial task in a performance analysis is to find out where the runtime is spent. You want to focus on the regions of the code that consume significant runtime.  Runtime profilers allow to measure the runtime distribution across the application code. Two profiler variants exist: Instrumentation based and sampling based. Instrumentation based profilers insert function calls to measure the time at points in the program. Additional tasks performed are e.g. determining the function call stack. While it is possible to insert instrumentation calls on the binary level usually the instrumentation calls are added during compilation. The standard tool is [https://en.wikipedia.org/wiki/Gprof gprof] and almost any compiler supports to instrument the code for gprof. Statistical sampling based profiling on the other hand is probing the program call stack at regular intervals. A widespread tool for sampling based profiling is the [https://en.wikipedia.org/wiki/Perf_(Linux) perf] tool which builds on the builtin profiling infrastructure in recent Linux kernels. Both approaches have advantages and disadvantages: Instrumentation produces more accurate results but introduces more overhead and sampling has less overhead but produce less accurate results. Special care is necessary for runtime profiling of parallel (OpenMP or MPI) applications and C++ programs (due to name mangling resolving issues).
 == How to use gprof ==

← Older revision		Revision as of 06:07, 24 April 2019
Line 192:		Line 192:
	* [https://perf.wiki.kernel.org/index.php/Main_Page Perf Wiki pages]		* [https://perf.wiki.kernel.org/index.php/Main_Page Perf Wiki pages]
	* [https://software.intel.com/en-us/vtune-amplifier-help-command-line-interface Intel Amplifier Command line Tool Documentation]		* [https://software.intel.com/en-us/vtune-amplifier-help-command-line-interface Intel Amplifier Command line Tool Documentation]
−
−	~~== Authors ==~~
−
−	~~Main author: Jan Eitzinger, [https://hpc.fau.de HPC group] at RRZE Erlangen~~

← Older revision		Revision as of 14:19, 2 April 2019
Line 187:		Line 187:

	This was also the only tool that resolved the inlined C++ routine. There are many more collect modules available including hardware performance monitoring metrics. Please refer to the [https://software.intel.com/en-us/vtune-amplifier-help-command-line-interface Intel documentation] for more detailed information.		This was also the only tool that resolved the inlined C++ routine. There are many more collect modules available including hardware performance monitoring metrics. Please refer to the [https://software.intel.com/en-us/vtune-amplifier-help-command-line-interface Intel documentation] for more detailed information.
		+
		+	== Links and more Information ==
		+	* [https://linux.die.net/man/1/gprof gprof man page]
		+	* [https://perf.wiki.kernel.org/index.php/Main_Page Perf Wiki pages]
		+	* [https://software.intel.com/en-us/vtune-amplifier-help-command-line-interface Intel Amplifier Command line Tool Documentation]
		+
		+	== Authors ==
		+
		+	Main author: Jan Eitzinger, [https://hpc.fau.de HPC group] at RRZE Erlangen