HP aC++/HP C A.06.25 Programmer's Guide

for each thread is allocated from the heap. The heap defaults 1 gigabyte, and the
default stack size is 8 megabytes. See the linker option-N for increasing data area
size to 2 gigabytes.
If the dynamic threads mechanism is disabled, requests for additional threads will
result in no additional threads being created. Programs should not assume that a
request will result in additional threads for all requests. If the dynamic threads
mechanism is enabled, requests for more threads than an implementation can
support are satisfied by creating additional pthreads which are then scheduled by
the HP-UX scheduler using a smaller number of threads.
• Number of processors: The number of physical processors actually hosting the
threads at any time is the lesser of the number of threads or the number of physical
processors on the system.
• Creating teams of threads: The number of threads in a team that executes a nested
parallel region is dependent on the number of threads created when the application
is started and the number of threads already in use. The number of threads created
at application startup defaults to the number of processors on the system, but may
be increased or decreased using the OMP_NUM_THREADS environment variable. If
all threads are already in use when a nested parallel region is encountered, the
number of threads in the team that executes the parallel region is one. If all threads
are not already in use when a nested parallel region is encountered, the number
of threads in the team used to execute the parallel region will be the lesser of the
number of threads requested by a num_threads clause or the most
omp_set_num_threads call, if any, or the number of threads not already in use.
(Section 2.3, page 10 of OpenMP C/C++ specs).
• When schedule(runtime) clause is specified, the decision regarding scheduling
is deferred until run time. The schedule type and chunk size can be chosen at run
time by setting the OMP_SCHEDULE environment variable. If this environment
variable is not set, the resulting schedule defaults to static schedule with a chunk
size of 1.
• In the absence of the schedule clause, the default schedule is static with chunksize
computed as iterations.
• An implementation can replace all ATOMIC directives by enclosing the statement
in a critical section. HP implements the ATOMIC clause using a slightly more
efficient form of critical section roughly 60-70% faster than critical, although there
is still a runtime call.
• omp_get_num_threads: If the number of threads has not been explicitly set by
the user, the default is the number of physical processors on the system.
• omp_set_dynamic: The default for dynamic thread adjustment is 0 (disabled).
• omp_set_nested: When nested parallelism is enabled, the number of threads
used to execute nested parallel regions is determined at runtime by the underlying
OpenMP parallel library.
220 Exception Handling