Skip to content

Commit 3203b0c

Browse files
ggeorgakoudisggeorgakoudis
committed
[RTD] Update docs and configuration
- Bump up: sphinx 8.2.3, rtd theme 3.1.0 - Add copybutton extensions to easy copy from code blocks - Update documentation for latest information and readability
1 parent 8d2e653 commit 3203b0c

5 files changed

Lines changed: 229 additions & 101 deletions

File tree

‎docs/requirements.txt‎

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -1,2 +1,3 @@
1-
sphinx==7.1.2
2-
sphinx-rtd-theme==1.3.0rc1
1+
sphinx==8.2.3
2+
sphinx-rtd-theme==3.1.0
3+
sphinx-copybutton==0.5.2

‎docs/source/conf.py‎

Lines changed: 6 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -38,6 +38,7 @@
3838
"sphinx.ext.autodoc",
3939
"sphinx.ext.autosummary",
4040
"sphinx.ext.intersphinx",
41+
"sphinx_copybutton",
4142
]
4243

4344
intersphinx_mapping = {
@@ -46,6 +47,11 @@
4647
}
4748
intersphinx_disabled_domains = ["std"]
4849

50+
# -- Options for sphinx_copybutton
51+
52+
sphinx_copybutton_prompt_text = r">>> |\.\.\. "
53+
sphinx_copybutton_prompt_is_regexp = True
54+
4955
templates_path = ["_templates"]
5056

5157
# -- Options for HTML output

‎docs/source/index.rst‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -1,7 +1,7 @@
11
PyOMP documentation
22
===================
33

4-
**PyOMP** implements OpenMP parallel CPU/GPU programming abstraction in Python using Numba.
4+
**PyOMP** implements OpenMP parallel CPU/GPU programming abstractions in Python using Numba.
55

66
Check out the :doc:`installation` section for how to install, the :doc:`usage`
77
section on information regarding using PyOMP, and :doc:`openmp` for the list of

‎docs/source/openmp.rst‎

Lines changed: 144 additions & 65 deletions
Original file line numberDiff line numberDiff line change
@@ -1,7 +1,10 @@
11
OpenMP support
22
==============
33

4-
The following tables show supported OpenMP directives and the support status of
4+
OpenMP directives and clauses
5+
-----------------------------
6+
7+
The following section shows supported OpenMP directives and the support status of
58
their clauses.
69

710
.. note::
@@ -10,103 +13,179 @@ their clauses.
1013
🔶 = partial support
1114

1215

13-
.. csv-table::
14-
15-
,
16-
**barrier**,
16+
barrier
17+
~~~~~~~
18+
19+
No clauses.
20+
21+
critical
22+
~~~~~~~~
23+
24+
No clauses.
25+
26+
for
27+
~~~
28+
29+
* ✅ collapse, firstprivate, lastprivate, private
30+
* 🔶 reduction
31+
* ❌ allocate, linear, nowait, order, ordered, schedule
32+
33+
parallel
34+
~~~~~~~~
35+
36+
* ✅ default, firstprivate, if, num_threads, private, shared
37+
* 🔶 reduction
38+
* ❌ allocate, copyin, proc_bind
39+
40+
parallel for
41+
~~~~~~~~~~~~
1742

18-
.. csv-table::
43+
Combines ``parallel`` and ``for`` directives. See clauses for `for`_ and `parallel`_ above.
1944

20-
,
21-
**critical**,
45+
single
46+
~~~~~~
2247

23-
.. csv-table::
48+
* ❌ allocate, copyprivate, firstprivate, nowait, private
2449

25-
,allocate,collapse,firstprivate,lastprivate,linear,nowait,order,ordered,private,reduction,schedule
26-
**for**,❌,✅,✅,✅,❌,❌,❌,❌,✅,🔶,❌
50+
task
51+
~~~~
2752

28-
.. csv-table::
29-
30-
,allocate,copyin,default,firstprivate,if,num_threads,private,proc_bind,reduction,shared
31-
**parallel**,❌,❌,✅,✅,✅,✅,✅,❌,🔶,✅
53+
* ✅ default, firstprivate, private, shared
54+
* ❌ affinity, allocate, detach, if, in_reduction, final, mergeable, priority, untied
3255

33-
.. csv-table::
34-
35-
,*See clauses for* **parallel** *and* **for** directives
36-
**parallel for**,
56+
taskwait
57+
~~~~~~~~
3758

38-
.. csv-table::
59+
* ❌ depend, nowait
3960

40-
,allocate,copyprivate,firstprivate,nowait,private
41-
**single**,❌,❌,❌,❌,❌
61+
target
62+
~~~~~~
4263

43-
.. csv-table::
64+
* ✅ device, firstprivate, map, private, thread_limit
65+
* ❌ allocate, defaultmap, depend, has_device_addr, if, in_reduction, is_device_ptr, nowait, uses_allocators
4466

45-
,affinity,allocate,default,detach,if,in_reduction,final,firstprivate,mergeable,priority,private,shared,untied
46-
**task**,❌,❌,✅,❌,❌,❌,❌,✅,❌,❌,✅,✅,❌
67+
teams
68+
~~~~~
4769

48-
.. csv-table::
70+
* ✅ default, firstprivate, num_teams, private, shared, thread_limit
71+
* 🔶 reduction
4972

50-
,depend,nowait
51-
**taskwait**,❌,❌
73+
distribute
74+
~~~~~~~~~~
5275

53-
.. csv-table::
76+
* ✅ firstprivate, lastprivate, private
77+
* ❌ allocate, collapse, dist_schedule, order
5478

55-
,allocate,defaultmap,depend,device,firstprivate,has_device_addr,if,in_reduction,is_device_ptr,map,nowait,private,thread_limit,uses_allocators
56-
**target**,❌,❌,❌,✅,✅,❌,❌,❌,❌,✅,❌,✅,✅,❌
79+
teams distribute
80+
~~~~~~~~~~~~~~~~
5781

58-
.. csv-table::
82+
Combines ``teams`` and ``distribute`` directives. See clauses for `teams`_ and `distribute`_ above.
5983

60-
,private,firstprivate,shared,reduction,default,num_teams,thread_limit
61-
**teams**,✅,✅,✅,🔶,✅,✅,✅
84+
target teams
85+
~~~~~~~~~~~~
6286

63-
.. csv-table::
87+
Combines ``target`` and ``teams`` directives. See clauses for `target`_ and `teams`_ above.
6488

65-
,allocate,collapse,dist_schedule,firstprivate,lastprivate,order,private
66-
**distribute**,❌,❌,❌,✅,✅,❌,✅
89+
target data
90+
~~~~~~~~~~~
6791

68-
.. csv-table::
92+
* ✅ device, map
93+
* ❌ if, use_device_ptr, use_device_addr
6994

70-
,*See clauses for* **teams** *and* **distribute** directives
71-
**teams distribute**,
95+
target enter data
96+
~~~~~~~~~~~~~~~~~
7297

73-
.. csv-table::
98+
* ✅ device, map
99+
* ❌ depend, if, nowait
74100

75-
,*See clauses for* **teams** *and* **teams** directives
76-
**target teams**,
101+
target exit data
102+
~~~~~~~~~~~~~~~~
77103

104+
Same clauses as `target enter data`_. See above.
78105

79-
.. csv-table::
106+
target update
107+
~~~~~~~~~~~~~
80108

81-
,device,if,map,use_device_ptr,use_device_addr
82-
**target data**,✅,❌,✅,❌,❌
83-
84-
.. csv-table::
109+
* ✅ device, from, to
110+
* ❌ depend, if, nowait
85111

86-
,depend,device,if,map,nowait
87-
**target enter data**,❌,✅,❌,✅,❌
112+
target teams distribute
113+
~~~~~~~~~~~~~~~~~~~~~~~
88114

89-
.. csv-table::
115+
Combines ``target``, ``teams``, and ``distribute`` directives. See clauses for `target`_, `teams`_, and `distribute`_ above.
90116

91-
,*See clauses for the* **target enter data** directive
92-
**target exit data**,
117+
distribute parallel for
118+
~~~~~~~~~~~~~~~~~~~~~~~
93119

94-
.. csv-table::
120+
Combines ``distribute`` and ``parallel for`` directives. See clauses for `distribute`_, `parallel`_, and `for`_ above.
95121

96-
,nowait,depend,device,from,if,to
97-
**target update**,❌,❌,✅,✅,❌,✅
122+
target teams distribute parallel for
123+
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
98124

99-
.. csv-table::
125+
Combines ``target``, ``teams``, ``distribute``, and ``parallel for`` directives. See clauses for `target`_, `teams`_, `parallel`_, and `for`_ above.
100126

101-
,*See clauses for* **target** *and* **teams distribute** directives
102-
**target teams distribute**,
103127

104-
.. csv-table::
128+
OpenMP runtime functions
129+
-------------------------
130+
**Thread and team information:**
105131

106-
,*See clauses for* **distribute** *and* **parallel for** directives
107-
**distribute parallel for**,
132+
* ``omp_get_thread_num()`` - Returns the unique identifier of the calling thread
133+
* ``omp_get_num_threads()`` - Returns the total number of threads in the current parallel region
134+
* ``omp_set_num_threads(n)`` - Sets the number of threads for subsequent parallel regions
135+
* ``omp_get_max_threads()`` - Returns the maximum number of threads available
136+
* ``omp_get_num_procs()`` - Returns the number of processors in the system
137+
* ``omp_get_thread_limit()`` - Returns the thread limit for the parallel region
138+
* ``omp_in_parallel()`` - Returns 1 if called within a parallel region, 0 otherwise
139+
* ``omp_get_team_num()`` - Returns the team number in a target region
140+
* ``omp_get_num_teams()`` - Returns the number of teams in a target region
108141

109-
.. csv-table::
142+
**Timing:**
110143

111-
,*See clauses for* **target** *and* **teams** *and* **distribute parallel for** directives
112-
**target teams distribute parallel for**,
144+
* ``omp_get_wtime()`` - Returns elapsed wall-clock time (useful for performance profiling)
145+
146+
**Nested and hierarchical parallelism:**
147+
148+
* ``omp_set_nested(flag)`` - Enables or disables nested parallelism
149+
* ``omp_set_dynamic(flag)`` - Enables or disables dynamic thread adjustment
150+
* ``omp_set_max_active_levels(n)`` - Sets the maximum number of nested parallel levels
151+
* ``omp_get_max_active_levels()`` - Returns the maximum number of nested parallel levels
152+
* ``omp_get_level()`` - Returns the current nesting level
153+
* ``omp_get_active_level()`` - Returns the current active nesting level
154+
* ``omp_get_ancestor_thread_num(level)`` - Returns the thread number at a given nesting level
155+
* ``omp_get_team_size(level)`` - Returns the team size at a given nesting level
156+
* ``omp_get_supported_active_levels()`` - Returns the supported number of nested active levels
157+
158+
**Advanced features:**
159+
160+
* ``omp_get_proc_bind()`` - Returns the processor binding policy
161+
* ``omp_get_num_places()`` - Returns the number of available places
162+
* ``omp_get_place_num_procs(place)`` - Returns the number of processors in a place
163+
* ``omp_get_place_num()`` - Returns the current place number
164+
165+
Supported features and platforms
166+
---------------------------------
167+
168+
PyOMP builds on `Numba <https://numba.pydata.org/>`_ Just-In-Time (JIT)
169+
compilation extensions to implement portable parallel execution using LLVM's
170+
OpenMP implementation.
171+
172+
OpenMP support
173+
~~~~~~~~~~~~~~
174+
175+
PyOMP supports OpenMP features that are officially supported by the LLVM OpenMP
176+
runtime. The specific set of supported features depends on the version of the
177+
LLVM OpenMP runtime and the version of Numba you are using. As both LLVM and
178+
Numba evolve, support for additional OpenMP features may be added in future
179+
PyOMP releases. For detailed information about Numba and LLVM version
180+
compatibility, see the `Numba support info
181+
<https://numba.readthedocs.io/en/stable/user/installing.html#numba-support-info>`_
182+
in the Numba documentation.
183+
184+
GPU offloading support
185+
~~~~~~~~~~~~~~~~~~~~~~
186+
187+
PyOMP supports GPU offloading exclusively for NVIDIA GPU architectures. The
188+
specific GPU architecture versions supported (e.g., compute capability levels)
189+
depend on the LLVM OpenMP runtime support in your environment. Consult the
190+
LLVM OpenMP documentation for your specific version to determine which NVIDIA
191+
GPU architectures are supported.

0 commit comments

Comments
 (0)