<?xml version="1.0"?>
<!DOCTYPE flagsdescription SYSTEM "http://www.spec.org/dtd/cpuflags2.dtd">
<flagsdescription>

<!-- filename to begin with "HP-Intel-Linux-Settings" -->
<filename>HP-Intel-Linux-Settings-compiler-flags</filename>

<title>SPEC CPU2006 compiler tuning Descriptions HP ProLiant Intel-based systems
applications</title>

<!-- 
  *********************************************************************************************************************
  Explanations of platform info, such as BIOS settings
  ********************************************************************************************************************* 
-->

<submit_command>
<![CDATA[

<p><b> submit= MYMASK=`printf '0x%x' \$((1&lt;&lt;\$SPECCOPYNUM))`; /usr/bin/taskset \$MYMASK $command </b></p>
<p>
       When running multiple copies of benchmarks, the SPEC config file feature 
       <b>submit</b> is sometimes used to cause individual jobs to be bound to 
       specific processors. This specific submit command is used for Linux. 
       The description of the elements of the command are:</p>
       <ul>
       <li> <b>/usr/bin/taskset [options] [mask] [pid | command [arg] ... ]</b>: <br />
           taskset is used to set or retreive the CPU affinity of a running 
           process given its PID or to launch a new COMMAND with a given CPU 
           affinity. The CPU affinity is represented as a bitmask, with the 
           lowest order bit corresponding to the first logical CPU and highest
           order bit corresponding to the last logical CPU. When the taskset 
           returns, it is guaranteed that the given program has been scheduled
           to a legal CPU. <br /><br />

           The default behaviour of taskset is to run a new command with a 
           given affinity mask: <br /><br />

                taskset [mask] [command] [arguments] </li>
       <li> <b>$MYMASK</b>: The bitmask (in hexadecimal) corresponding to a specific
           SPECCOPYNUM. For example, $MYMASK value for the first copy of a 
           rate run will be 0x00000001, for the second copy of the rate will 
           be 0x00000002 etc. Thus, the first copy of the rate run will have a
           CPU affinity of CPU0, the second copy will have the affinity CPU1 
           etc.</li>
       <li> <b>$command</b>: Program to be started, in this case, the benchmark instance 
               to be started. </li>
        </ul>

<p><b>Using numactl to bind processes and memory to cores</b></p>
         <p>For multi-copy runs or single copy runs on systems with multiple sockets, 
         it is advantageous to bind a process to a particular core.  Otherwise, the OS 
         may arbitrarily move your process from one core to another.  This can effect 
         performance.  To help, SPEC allows the use of a "submit" command where users 
         can specify a utility to use to bind processes.  We have found the utility 'numactl' 
         to be the best choice.</p>
         <p>numactl runs processes with a specific NUMA scheduling or memory placement 
         policy.  The policy is set for a command and inherited by all of its children.  
         The numactl flag "--physcpubind" specifies which core(s) to bind the process. 
         "-l" instructs numactl to keep a process memory on the local node while "-m" 
         specifies which node(s) to place a process memory.  For full details on using 
         numactl, please refer to your Linux documentation, 'man numactl'</p>


<p><b> mysubmit.pl</b></p>
<p>
       This perl script is used to ensure that for a system with N cores the first 
       N/2 benchmark copies are bound to a core that does not share its L2 cache 
       with any of the other copies. The script does this by retrieving and using 
       CPU data from /proc/cpuinfo. Note this script will only work for 6-core CPUs.</p>

       	<ul>
        <li><b>Source</b><br />
       	******************************************************************************************************<br />
       	#!/usr/bin/perl<br />
	<br />
	use strict;<br />
	use Cwd;<br />
	<br />
	# The order in which we want copies to be bound to cores<br />
	# Copies: 0, 1, 2, 3<br />
	# Cores:  0, 1, 3, 6<br />
	<br />
	my $rundir = getcwd;<br />
	<br />
	my $copynum = shift @ARGV;<br />
	<br />
	my $i;<br />
	my $j;<br />
	my $tag;<br />
	my $num;<br />
	my $core;<br />
	<br />
	my @proc;<br />
	my @cores;<br />
	<br />
	open(INPUT, "/proc/cpuinfo") or<br />
	   die "can't open /proc/cpuinfo\n";<br />
	<br />
	#open(OUTPUT, "STDOUT");<br />
	<br />
	# proc[i][0] = logical processor ID<br />
	# proc[i][1] = physical processor ID<br />
	# proc[i][2] = core ID<br />
	<br />
	$i = 0;<br />
	<br />
	while(&#60;INPUT&#62;)<br />
	{<br />
	  chop;<br />
	<br />
	  ($tag, $num) = split(/\s+:\s+/, $_);<br />
	<br />
	<br />
	  if ($tag eq "processor") {<br />
	      $proc[$i][0] = $num;<br />
	  }<br />
	<br />
	  if ($tag eq "physical id") {<br />
	      $proc[$i][1] = $num;<br />
	  }<br />
	<br />
	  if ($tag eq "core id") {<br />
	      $proc[$i][2] = $num;<br />
	      $i++;<br />
	  }<br />
	}<br />
	<br />
	$i = 0;<br />
	$j = 0;<br />
	<br />
	for $core (0, 4, 2, 1, 5, 3) {<br />
	  while ($i &lt; 24) {<br />
	     if ($proc[$i][2] == $core) {<br />
	        $cores[$j] = $proc[$i][0];<br />
	        $j++;<br />
	     }<br />
	     $i++;<br />
	  }<br />
	  $i=0;<br />
	}<br />
	<br />
	open  RUNCOMMAND, "> runcommand" or die "failed to create run file";<br />
	print RUNCOMMAND "cd $rundir\n";<br />
	print RUNCOMMAND "@ARGV\n";<br />
	close RUNCOMMAND;<br />
   	system 'taskset', '-c', $cores[$copynum], 'sh', "$rundir/runcommand";<br />
	</li></ul>


]]>
</submit_command>

<sw_environment>
<![CDATA[

<p><b> KMP_STACKSIZE=integer[B|K|M|G|T] (Linux) </b></p>
<p>
           Sets the number of bytes to allocate for each parallel thread to use as its 
           private stack. Use the optional suffix B, K, M, G, or T, to specify bytes, 
           kilobytes, megabytes, gigabytes, or terabytes. The default setting is 2M on 
           IA32 and 4M on IA64. </p>

<p><b> KMP_AFFINITY=physical,n (Linux) </b></p>
<p>
           Assigns threads to consecutive physical processors (for example, cores), 
           beginning at processor n. Specifies the static mapping of user threads to 
           physical cores, beginning at processor n. For example, if a system is configured 
           with 8 cores, and OMP_NUM_THREADS=8 and KMP_AFFINITY=physical,2 are set, then 
           thread 0 will mapped to core 2, thread 1 will be mapped to core 3, and so on in 
           a round-robin fashion.    </p>

<p><b> KMP_AFFINITY=granularity=fine,scatter </b></p>
<p>
	   The value for the environment variable KMP_AFFINITY affects how the threads from 
	   an auto-parallelized program are scheduled across processors. Specifying
	   granularity=fine selects the finest granularity level, causes each OpenMP thread 
	   to be bound to a single thread context. This ensures that there is only one thread 
	   per core on cores supporting HyperThreading Technology. Specifying scatter distributes
	   the threads as evenly as possible across the entire system. Hence a combination 
	   of these two options, will spread the threads evenly across sockets, 
	   with one thread per physical core. </p>

<p><b> OMP_NUM_THREADS=n  </b></p>
<p>
           This Environment Variable sets the maximum number of threads to use for OpenMP* 
           parallel regions to <b>n</b> if no other value is specified in the application. This 
           environment variable applies to both -openmp and -parallel (Linux) 
           or /Qopenmp and /Qparallel (Windows). Example syntax on a Linux system with 8 
           cores:<br />
           export OMP_NUM_THREADS=8<br />
           Default is the number of cores visible to the OS. 

    </p>

]]>
</sw_environment>

</flagsdescription>