Using the Xeon Phi - Revision history

Admin at 12:52, 22 June 2020

2020-06-22T12:52:01Z

← Older revision		Revision as of 14:52, 22 June 2020
Line 79:		Line 79:

	<pre>		<pre>
	job submit /scheduler:fi--didemrchnb /numnodes:1 /singlenode:false /jobtemplate:Phi /workdir:\\fi--~~didef2~~\Tmp\Wes\Phi /stdout:out.txt /stderr:err.txt run.bat		job submit /scheduler:fi--didemrchnb /numnodes:1 /singlenode:false /jobtemplate:Phi /workdir:\\fi--didef3\Tmp\Wes\Phi /stdout:out.txt /stderr:err.txt run.bat
	</pre>		</pre>

Line 89:		Line 89:

	<pre>		<pre>
	\\fi--didef2\Tmp\Wes\Phi>set MIC_LD_LIBRARY_PATH=\\fi--~~didef2~~\Tmp\Wes\Phi\lib		\\fi--didef2\Tmp\Wes\Phi>set MIC_LD_LIBRARY_PATH=\\fi--didef3\Tmp\Wes\Phi\lib

	\\fi--didef2\Tmp\Wes\Phi>\\fi--~~didef2~~\Tmp\Wes\Phi\PhiTest.exe		\\fi--didef2\Tmp\Wes\Phi>\\fi--didef3\Tmp\Wes\Phi\PhiTest.exe
	Offload thread 112		Offload thread 112
	Offload thread 43		Offload thread 43

Admin at 12:51, 22 June 2020

2020-06-22T12:51:27Z

← Older revision		Revision as of 14:51, 22 June 2020
Line 4:		Line 4:
	=== Prerequesites ===		=== Prerequesites ===
	* You'll need Visual Studio, and the Intel C++ Parallel Compiler - probably most recent editions.		* You'll need Visual Studio, and the Intel C++ Parallel Compiler - probably most recent editions.
	* You'll also need the Phi copressor software - https://software.intel.com/en-us/articles/intel-manycore-platform-software-stack-mpss . According to the docs, you need both the Coprocessor, and the Essentials installer. (Possibly the latest local copy will be at \\fi--~~didef2~~\Software\HPC\PhiDrivers)		* You'll also need the Phi copressor software - https://software.intel.com/en-us/articles/intel-manycore-platform-software-stack-mpss . According to the docs, you need both the Coprocessor, and the Essentials installer. (Possibly the latest local copy will be at \\fi--didef3\Software\HPC\PhiDrivers)

	=== Setup the Project ===		=== Setup the Project ===
Line 63:		Line 63:
	=== Prepare a cluster job ===		=== Prepare a cluster job ===

	* Decide where to run the job as normal. I'm going for T:\Wes\Phi (which is \\fi--~~didef2~~\Tmp\Wes\Phi) in this example.		* Decide where to run the job as normal. I'm going for T:\Wes\Phi (which is \\fi--didef3\Tmp\Wes\Phi) in this example.
	* Copy the executable there, which will be in Release\x64 in your project folder. Mine is called PhiTest.exe.		* Copy the executable there, which will be in Release\x64 in your project folder. Mine is called PhiTest.exe.
	* Find a folder something like: C:\Program Files (x86)\IntelSWTools\parallel_studio_xe_2016.1.051\compilers_and_libraries_2016\windows\redist\intel64_win\compiler		* Find a folder something like: C:\Program Files (x86)\IntelSWTools\parallel_studio_xe_2016.1.051\compilers_and_libraries_2016\windows\redist\intel64_win\compiler

Admin at 16:43, 11 December 2015

2015-12-11T16:43:09Z

← Older revision		Revision as of 18:43, 11 December 2015
Line 2:		Line 2:

	== Using Visual Studio with Intel C++ Parallel Compiler ==		== Using Visual Studio with Intel C++ Parallel Compiler ==
			=== Prerequesites ===
			* You'll need Visual Studio, and the Intel C++ Parallel Compiler - probably most recent editions.
			* You'll also need the Phi copressor software - https://software.intel.com/en-us/articles/intel-manycore-platform-software-stack-mpss . According to the docs, you need both the Coprocessor, and the Essentials installer. (Possibly the latest local copy will be at \\fi--didef2\Software\HPC\PhiDrivers)

	=== Setup the Project ===		=== Setup the Project ===

Admin at 15:52, 11 December 2015

2015-12-11T15:52:10Z

← Older revision		Revision as of 17:52, 11 December 2015
Line 68:		Line 68:
	I'll call this run.bat, and put it in T:\Wes\Phi. I'll assume we'll have the working directory set, so...		I'll call this run.bat, and put it in T:\Wes\Phi. I'll assume we'll have the working directory set, so...
	<pre>		<pre>
	set ~~MIC_LD_LIBRARY~~=\\fi--didef2\Tmp\Wes\Phi\lib		set MIC_LD_LIBRARY_PATH=\\fi--didef2\Tmp\Wes\Phi\lib
	PhiTest.exe		PhiTest.exe
	</pre>		</pre>

Admin at 15:42, 11 December 2015

2015-12-11T15:42:38Z

← Older revision		Revision as of 17:42, 11 December 2015
Line 1:		Line 1:
	The Xeon Phi is essentially linux ~~installed~~ on a chip, on a card, inside some of our nodes. It's '''extremely''' multi-core, so offers itself for algorithms that have reasonably simple but very parallel sections. Here is the experience I've collected so far in using it.		The Xeon Phi is essentially linux on a chip, on a special card, inside some of our HPC nodes. It's '''extremely''' multi-core, so offers itself for algorithms that have reasonably simple but very parallel sections. Here is the experience I've collected so far in using it.

	== Using Visual Studio with Intel C++ Parallel Compiler ==		== Using Visual Studio with Intel C++ Parallel Compiler ==

Admin at 15:39, 11 December 2015

2015-12-11T15:39:50Z

← Older revision		Revision as of 17:39, 11 December 2015
Line 13:		Line 13:
	** And check the Target Device below is also set to '''Intel MIC Architecture'''.		** And check the Target Device below is also set to '''Intel MIC Architecture'''.
	** C/C++, Language [Intel C++], OpenMP Support: '''Generate Parallel Code'''		** C/C++, Language [Intel C++], OpenMP Support: '''Generate Parallel Code'''
	** C/C++, Language, Runtime Library: '''Multi-threaded (/MT)''' - ~~for all the good it does. We~~'ll copy DLLs later.		** C/C++, Language, Runtime Library: '''Multi-threaded (/MT)''' - we'll copy more DLLs across later.

	=== Some Code ===		=== Some Code ===

Admin at 15:38, 11 December 2015

2015-12-11T15:38:40Z

← Older revision		Revision as of 17:38, 11 December 2015
Line 11:		Line 11:
	** (The above is a workaround - the compiler assumes everyone has fortran installed and cries if we don't.)		** (The above is a workaround - the compiler assumes everyone has fortran installed and cries if we don't.)
	** C/C++, Code Generation [Intel C++], Enable OpenMP Offloading Compilation... - choose '''Intel MIC Architecture'''		** C/C++, Code Generation [Intel C++], Enable OpenMP Offloading Compilation... - choose '''Intel MIC Architecture'''
			** And check the Target Device below is also set to '''Intel MIC Architecture'''.
	** C/C++, Language [Intel C++], OpenMP Support: '''Generate Parallel Code'''		** C/C++, Language [Intel C++], OpenMP Support: '''Generate Parallel Code'''
	** C/C++, Language, Runtime Library: '''Multi-threaded (/MT)''' - for all the good it does. We'll copy DLLs later.		** C/C++, Language, Runtime Library: '''Multi-threaded (/MT)''' - for all the good it does. We'll copy DLLs later.

Admin at 15:34, 11 December 2015

2015-12-11T15:34:03Z

← Older revision		Revision as of 17:34, 11 December 2015
Line 25:		Line 25:
	int main(int argc, char *argv[]) {		int main(int argc, char *argv[]) {

	int ~~th_id~~;		int thread_id;

			// Here's an OpenMP loop using main processor

			#pragma omp parallel private (thread_id)
			{
			thread_id = omp_get_thread_num();
			printf("Local thread %d\n", thread_id);
			#pragma omp barrier
			#pragma omp single
			printf("There are %d local threads\n", omp_get_num_threads());
			}

			// Insert an extra offload pragma to do OpenMP on the Phi.

	#pragma offload target(mic)		#pragma offload target(mic)
	#pragma omp parallel private (~~th_id~~)		#pragma omp parallel private (thread_id)
	{		{
	~~th_id~~ = omp_get_thread_num();		thread_id = omp_get_thread_num();
	printf("Offload thread %d\n", ~~th_id~~);		printf("Offload thread %d\n", thread_id);
	#pragma omp barrier		#pragma omp barrier
	#pragma omp single		#pragma omp single
	printf("There are %d threads\n", omp_get_num_threads());		printf("There are %d offload threads\n", omp_get_num_threads());
	}		}
	return EXIT_SUCCESS;		return EXIT_SUCCESS;
	}		}

	</pre>		</pre>

Line 73:		Line 87:

	\\fi--didef2\Tmp\Wes\Phi>\\fi--didef2\Tmp\Wes\Phi\PhiTest.exe		\\fi--didef2\Tmp\Wes\Phi>\\fi--didef2\Tmp\Wes\Phi\PhiTest.exe
	~~Hello World from~~ thread ~~194~~		Offload thread 112
	~~Hello World from~~ thread ~~182~~		Offload thread 43
	~~Hello World from~~ thread ~~114~~		Offload thread 117
	~~Hello World from~~ thread ~~176~~		....
			Offload thread 45
			There are 240 offload threads

			Local thread 12
			Local thread 5
			Local thread 9
	....		....
	~~Hello World from thread 18~~		There are 16 local threads
	There are ~~240~~ threads
	</pre>		</pre>

	That's a lot of threads.		That's a lot of threads. Note that my code did the local bit first, but the output has come out in reverse. There may be interleaving issues with stdout, so in real code, do it better!

Admin at 15:29, 11 December 2015

2015-12-11T15:29:39Z

← Older revision		Revision as of 17:29, 11 December 2015
Line 5:		Line 5:

	* Create a new 64-bit project.		* Create a new 64-bit project.
			* Choose to use the Intel C++ Compiler from the Projects menu.
	* Get into Release, x64 mode with the menus at the top.		* Get into Release, x64 mode with the menus at the top.
	* Choose the Intel Compiler, from the Projects menu.
	* In Project Preferences:-		* In Project Preferences:-
	** Linker, General [Intel C++], Additional Options for MIC Offload Linker, add '''-no-fortlib'''		** Linker, General [Intel C++], Additional Options for MIC Offload Linker, add '''-no-fortlib'''
Line 15:		Line 15:

	=== Some Code ===		=== Some Code ===

			In this example. I've setup a project called PhiTest, and a single file, main.cpp.

	<pre>		<pre>
Line 23:		Line 25:
	int main(int argc, char *argv[]) {		int main(int argc, char *argv[]) {

	int th_id~~, nthreads~~;		int th_id;

	#pragma offload target(mic)		#pragma offload target(mic)

Admin at 15:27, 11 December 2015

2015-12-11T15:27:20Z

← Older revision		Revision as of 17:27, 11 December 2015
Line 9:		Line 9:
	* In Project Preferences:-		* In Project Preferences:-
	** Linker, General [Intel C++], Additional Options for MIC Offload Linker, add '''-no-fortlib'''		** Linker, General [Intel C++], Additional Options for MIC Offload Linker, add '''-no-fortlib'''
			** (The above is a workaround - the compiler assumes everyone has fortran installed and cries if we don't.)
	** C/C++, Code Generation [Intel C++], Enable OpenMP Offloading Compilation... - choose '''Intel MIC Architecture'''		** C/C++, Code Generation [Intel C++], Enable OpenMP Offloading Compilation... - choose '''Intel MIC Architecture'''
	** C/C++, Language [Intel C++], OpenMP Support: '''Generate Parallel Code'''		** C/C++, Language [Intel C++], OpenMP Support: '''Generate Parallel Code'''
Line 48:		Line 49:
	** Make a folder called '''lib''' in your test folder (ie, T:\Wes\Phi\Lib), and copy all the files you just found into it, including the "locale" folder. If you like command-line copying, then something like <pre>xcopy . /e T:\Wes\Phi\Lib</pre>		** Make a folder called '''lib''' in your test folder (ie, T:\Wes\Phi\Lib), and copy all the files you just found into it, including the "locale" folder. If you like command-line copying, then something like <pre>xcopy . /e T:\Wes\Phi\Lib</pre>
	==== A batch file to run the job ====		==== A batch file to run the job ====
	I'll call this run.bat, and put it in T:\Wes\Phi.		I'll call this run.bat, and put it in T:\Wes\Phi. I'll assume we'll have the working directory set, so...
	<pre>		<pre>
	set MIC_LD_LIBRARY=\\fi--didef2\Tmp\Wes\Phi\lib		set MIC_LD_LIBRARY=\\fi--didef2\Tmp\Wes\Phi\lib
	~~\\fi--didef2\Tmp\Wes\Phi\~~PhiTest.exe		PhiTest.exe
	</pre>		</pre>

Line 64:		Line 65:
	=== And the result ===		=== And the result ===

	In out.txt, ~~we get~~...		In my out.txt, I have...

	<pre>		<pre>
	\\fi--didef2\Tmp\Wes\~~Phi2~~>set MIC_LD_LIBRARY_PATH=\\fi--didef2\Tmp\Wes\Phi\lib		\\fi--didef2\Tmp\Wes\Phi>set MIC_LD_LIBRARY_PATH=\\fi--didef2\Tmp\Wes\Phi\lib

	\\fi--didef2\Tmp\Wes\~~Phi2~~>\\fi--didef2\Tmp\Wes\~~Phi2~~\PhiTest.exe		\\fi--didef2\Tmp\Wes\Phi>\\fi--didef2\Tmp\Wes\Phi\PhiTest.exe
	Hello World from thread 194		Hello World from thread 194
	Hello World from thread 182		Hello World from thread 182
	Hello World from thread 114		Hello World from thread 114
	Hello World from thread 176		Hello World from thread 176
	~~Hello World from thread 42~~
	~~Hello World from thread 37~~
	~~Hello World from thread 231~~
	~~Hello World from thread 74~~
	~~Hello World from thread 39~~
	~~Hello World from thread 134~~
	....		....
	Hello World from thread 18		Hello World from thread 18