Shared Memory Parallelism -
OpenMP
Sathish Vadhiyar
Credits/Sources:
OpenMP C/C++ standard (openmp.org)
OpenMP tutorial (http://www.llnl.gov/computing/tutorials/openMP/#Introduction)
OpenMP sc99 tutorial presentation (openmp.org)
Dr. Eric Strohmaier (University of Tennessee, CS594 class, Feb 9, 2000)
Introduction
A portable programming model and standard for
shared memory programming using compiler
directives
Directives?: constructs or statements in the program
applying some action on a block of code
A specification for a set of compiler directives, library
routines, and environment variables – standardizing
pragmas
Easy to program; easy for code developer to convert
his sequential to parallel program by throwing
directives
First version in 1997, development over the years till
the latest 4.5 in 2015
Fork-Join Model
Begins as a single thread
called master thread
Fork: When parallel construct
is encountered, team of
threads are created
Statements in the parallel
region are executed in parallel
Join: At the end of the parallel
region, the team threads
synchronize and terminate
OpenMP consists of…
Work-sharing constructs
Synchronization constructs
Data environment constructs
Library calls, environment variables
Introduction
Mainly supports loop-level parallelism
Specifies parallelism for a region of code: fine-
level parallelism
The number of threads can be varied from one
region to another – dynamic parallelism
Follows Amdahl’s law – sequential portions in the
code
Applications have varying phases of parallelism
Also supports
Coarse-level parallelism – sections and tasks
Executions on accelerators
SIMD vectorizations
task-core affinity
parallel construct
#pragma omp parallel [clause [, clause] …] new-line
structured-block
Clause: Can support nested
parallelism
Parallel construct - Example
#include <omp.h>
main () {
int nthreads, tid;
#pragma omp parallel private(nthreads, tid) {
printf("Hello World \n);
}
}
Work sharing construct
For distributing the execution among the threads
that encounter it
3 types of work sharing constructs – loops,
sections, single
for construct
For distributing the iterations among the threads
#pragma omp for [clause [, clause] …] new-
line
for-loop
Clause:
for construct
Restriction in the structure of the for
loop so that the compiler can
determine the number of iterations –
e.g. no branching out of loop
The assignment of iterations to
threads depends on the schedule
clause
Implicit barrier at the end of for if not
nowait
schedule clause
1. schedule(static, chunk_size) –
iterations/chunk_size chunks distributed
in round-robin
2. Schedule(dynamic, chunk_size) – same
as above, but chunks distributed
dynamically.
3. schedule(runtime) – decision at runtime.
Implementation dependent
for - Example
include <omp.h>
#define CHUNKSIZE 100
#define N 1000
main () {
int i, chunk; float a[N], b[N], c[N];
/* Some initializations */
for (i=0; i < N; i++)
a[i] = b[i] = i * 1.0;
chunk = CHUNKSIZE;
#pragma omp parallel shared(a,b,c,chunk) private(i) {
#pragma omp for schedule(dynamic,chunk) nowait
for (i=0; i < N; i++)
c[i] = a[i] + b[i];
} /* end of parallel section */
}
Coarse level parallelism – sections and
tasks
sections
tasks – dynamic mechanism
depend clause for task
Synchronization directives
flush directive
Point where consistent view of memory is
provided among the threads
Thread-visible variables (global variables,
shared variables etc.) are written to memory
If var-list is used, only variables in the list are
flushed
flush - Example
flush – Example (Contd…)
Data Scope Attribute Clauses
Most variables are shared by default
Data scopes explicitly specified by data scope attribute clauses
Clauses:
1. private
2. firstprivate
3. lastprivate
4. shared
5. default
6. reduction
7. copyin
8. copyprivate
threadprivate
• Global variable-list declared are made private to a thread
• Each thread gets its own copy
• Persist between different parallel regions
#include <omp.h>
int alpha[10], beta[10], i;
#pragma omp threadprivate(alpha)
main () {
/* Explicitly turn off dynamic threads */
omp_set_dynamic(0);
/* First parallel region */
#pragma omp parallel private(i,beta)
for (i=0; i < 10; i++) alpha[i] = beta[i] = i;
/* Second parallel region */
#pragma omp parallel
printf("alpha[3]= %d and beta[3]= %d\n",alpha[3],beta[3]);}
private, firstprivate & lastprivate
private (variable-list)
variable-list private to each thread
A new object with automatic storage duration allocated for the
construct
firstprivate (variable-list)
The new object is initialized with the value of the old object that
existed prior to the construct
lastprivate (variable-list)
The value of the private object corresponding to the last iteration
or the last section is assigned to the original object
shared, default, reduction
shared(variable-list)
default(shared | none)
Specifies the sharing behavior of all of the variables visible in the
construct
Reduction(op: variable-list)
Private copies of the variables are made for each thread
The final object value at the end of the reduction will be
combination of all the private object values
default - Example
Library Routines (API)
Querying function (number of threads etc.)
General purpose locking routines
Setting execution environment (dynamic
threads, nested parallelism etc.)
API
OMP_SET_NUM_THREADS(num_threads)
OMP_GET_NUM_THREADS()
OMP_GET_MAX_THREADS()
OMP_GET_THREAD_NUM()
OMP_GET_NUM_PROCS()
OMP_IN_PARALLEL()
OMP_SET_DYNAMIC(dynamic_threads)
OMP_GET_DYNAMIC()
OMP_SET_NESTED(nested)
OMP_GET_NESTED()
API(Contd..)
omp_init_lock(omp_lock_t *lock)
omp_init_nest_lock(omp_nest_lock_t *lock)
omp_destroy_lock(omp_lock_t *lock)
omp_destroy_nest_lock(omp_nest_lock_t *lock)
omp_set_lock(omp_lock_t *lock)
omp_set_nest_lock(omp_nest_lock_t *lock)
omp_unset_lock(omp_lock_t *lock)
omp_unset_nest__lock(omp_nest_lock_t *lock)
omp_test_lock(omp_lock_t *lock)
omp_test_nest_lock(omp_nest_lock_t *lock)
omp_get_wtime()
omp_get_wtick()
omp_get_thread_num()
omp_get_num_proc()
omp_get_num_devices()
Lock details
Simple locks and nestable locks
Simple locks are not locked if they are already in a
locked state
Nestable locks can be locked multiple times by the
same thread
Simple locks are available if they are unlocked
Nestable locks are available if they are unlocked or
owned by a calling thread
Example – Nested lock
Example – Nested lock (Contd..)
Example 1: Jacobi Solver
Example 2: BFS Version 1
(Nested Parallelism)
Example 3: BFS Version 3
(Using Task Construct)
Hybrid Programming – Combining MPI and
OpenMP benefits
MPI
- explicit parallelism, no synchronization problems
- suitable for coarse grain
OpenMP
- easy to program, dynamic scheduling allowed
- only for shared memory, data synchronization problems
MPI/OpenMP Hybrid
- Can combine MPI data placement with OpenMP fine-grain
parallelism
- Suitable for cluster of SMPs (Clumps)
- Can implement hierarchical model
END
Definitions
Construct – statement containing directive and
structured block
Directive – Based on C #pragma directives
#pragma <omp id> <other text>
#pragma omp directive-name [clause [, clause] …]
new-line
Example:
#pragma omp parallel default(shared) private(beta,pi)
Parallel construct
Parallel region executed by multiple threads
If num_threads, omp_set_num_threads(),
OMP_SET_NUM_THREADS not used, then
number of created threads is implementation
dependent
Number of physical processors hosting the
thread also implementation dependent
Threads numbered from 0 to N-1
Nested parallelism by embedding one parallel
construct inside another
