submit=echo 'pbind -b...' > dobmk; sh dobmk
When running multiple copies of benchmarks, the SPEC config file feature submit is sometimes used to
cause individual jobs to be bound to specific processors. If so, the specific command may be found in the config file; here
is a brief guide to understanding that command:
One or more of the following settings may have been applied to the testbed. If so, the "Platform Notes" section of the report will say so; and you can read below to find out more about what these settings mean.
autoup=<n>
When the file system flush daemon fsflush runs, it writes to disk all modified file buffers that are more
than n seconds old.
lpg_alloc_prefer=<n>
0 = the OS may allocate remote pages if the size requested is readily available in a remote locality group (default)
1 = Set lgroup page allocation to strongly prefer local pages.
maxusers=<n>
To increase the number of user processes derived by the system
ncsize=<n>
Defines the number of entries in the directory name look-up cache (DNLC). This parameter is used by UFS, NFS, and
ZFS to cache elements of path names that have been resolved.
rlim_fd_cur=<n>
Defines the soft limit on file descriptors that a single process can have open.
tune_t_fsflushr=<n>
Controls the number of seconds between runs of the file system flush daemon, fsflush.
zfs:zfs_arc_min=<n>
Controls the minimum amount of memory used in bytes by ZFS for caching of file system buffers.
zfs:zfs_arc_max=<n>
Controls the maximum amount of memory used in bytes by ZFS for caching of file system buffers.
ZFS RAIDZ
In mirrored storage pool configuration, ZFS provides a RAID-Z configuration with either single or double parity fault tolerance. Single-parity RAID-Z is similar to RAID-5. Double-parity RAID-Z is similar to RAID-6.
OMP_DYNAMIC=<TRUE|FALSE>
Enables (TRUE) or disables (FALSE) dynamic adjustment of the number of threads available for execution of parallel
regions. The default is TRUE.
MTEXCLUSIVE=Y
By default, libmtmalloc allocates 2*NCPUS buckets from which allocations occur. Threads share buckets based on their thread ID. If MTEXCLUSIVE is invoked, then 4*NCPUS buckets are used. Threads with thread id less than 2*NCPUS receive an exclusive bucket and thus do not need to use locks. Allocation perfor- mance for these buckets may be dramatically increased. One enabled MTEXCLUSIVE can not be dis- abled. This feature can be enabled by setting the MTMALLOC_OPTION MTEXCLUSIVE to "Y" or "y" or any- thing beginning with "y". Alternatively it can be enabled by a call to mallocctl(3MALLOC).
OMP_NESTED=<TRUE|FALSE>
Enables or disables nested parallelism. Value is either TRUE or FALSE. The default is FALSE.
OMP_NUM_THREADS=<n>
If programs have been compiled with -xautopar, this environment variable can be set to the number of
processors that programs should use.
PARALLEL=<n>
If programs have been compiled with -xautopar, this environment variable can be set to the number of
processors that programs should use.
STACKSIZE=<n>
Set the size of the stack (temporary storage area) for each slave thread of a multithreaded program.
SUNW_MP_PROCBIND=<n>
This environment variable can be used to bind the LWPs (lightweight processes) managed by the microtasking library,
libmtsk, to processors. Performance can be enhanced with processor binding, but performance degradation will occur if
multiple LWPs are bound to the same processor.
The value for SUNW_MP_PROCBIND can be:
Integers in the above denote the "logical" processor IDs to which the LWPs are to be bound. Logical processor IDs are
consecutive integers that start with 0, and may or may not be identical to the actual processsor IDs. If n processors are
available online, then their logical processor IDs are 0, 1, ..., n-1.
By default, LWPs are not bound to processors. It is left up to the operating system, Solaris, to schedule LWPs onto processors.
If the value "TRUE" is used, the operating system will bind processes to processors, starting with processor 0.
If the value "SCATTER" is used, then the threads will be bound to virtual processors that are far apart.
SUNW_MP_THR_IDLE=SPIN
Controls the end-of-task status of each helper thread executing the parallel part of a program. You can set the value
to spin, sleep ns, or sleep nms. The default is SPIN -- the thread spins (or busy-waits) after completing a parallel task
until a new parallel task arrives. You can set the value to be one of the following: SLEEP( times), SLEEP(timems), SLEEP( timemc),
where time is an integer that specifies an amount of time, and s, ms, and mc specify the time unit (seconds, milli-seconds,
and micro-seconds, respectively).
ulimit -s <n>
Sets the stack size to n kbytes, or "unlimited" to allow the stack size to grow without limit.
Operating systems: Solaris 11
Last updated: 29-Mar-2012 gr
The text for many of the descriptions below was taken from the Oracle Studio Compiler Documentation, which is copyright © 2007-2012 Oracle Corporation, Inc. The original documentation can be found at docs.sun.com.
This document has both optimization flags (in the immediately following section) and a description of Platform Settings
]]>Unlike regular data dependence analysis, this analyzes detailed array sections accessed in a loop, so the analysis can be expensive in terms of compilation time.
]]>Note that this is a flag to the "iropt" component of the compilation system. In general, flags may be sent to iropt using "Qoption iropt" from the "f90" and "CC" commands; or using "-W2," from the "cc" command.
]]>The -R flag specifies library search directories to the runtime linker. The information is recorded in the object file and passed to the runtime linker.
The version of SmartHeap that is used can be one of:
The default is -verbose=%none.
]]>At -xipo=2, the compiler performs inter-procedural aliasing analysis as well as optimization of memory allocation and layout to improve cache performance.
]]>If only -xprefetch is specified, -xprefetch=auto,explicit is assumed.
The default is -xprefetch=auto,explicit.
]]>When FDO is used, the training run gathers information regarding execution paths. As of the Sun Studio 12 version of the compiler suite, the training run gathers information about data values on SPARC systems, but not on x86 systems. Hardware performance counters are not used. FDO improves existing optimizations but does not introduce new classes of optimization.
]]>Synonym for -unroll=n
]]>