[clang] 98d5ab7 - [OpenMP][Docs][NFC] Update documentation
via cfe-commits
cfe-commits at lists.llvm.org
Tue Sep 19 14:08:46 PDT 2023
Author: Johannes Doerfert
Date: 2023-09-19T22:08:43+01:00
New Revision: 98d5ab706e93b7a10eae4e0e3592cb11619715d6
URL: https://github.com/llvm/llvm-project/commit/98d5ab706e93b7a10eae4e0e3592cb11619715d6
DIFF: https://github.com/llvm/llvm-project/commit/98d5ab706e93b7a10eae4e0e3592cb11619715d6.diff
LOG: [OpenMP][Docs][NFC] Update documentation
Removed lots of outdated statements that were misleading.
Added:
Modified:
clang/docs/OpenMPSupport.rst
Removed:
################################################################################
diff --git a/clang/docs/OpenMPSupport.rst b/clang/docs/OpenMPSupport.rst
index 641ec78a56da8e4..293a56d68f2d6b9 100644
--- a/clang/docs/OpenMPSupport.rst
+++ b/clang/docs/OpenMPSupport.rst
@@ -16,15 +16,12 @@
OpenMP Support
==============
-Clang fully supports OpenMP 4.5. Clang supports offloading to X86_64, AArch64,
-PPC64[LE] and has `basic support for Cuda devices`_.
-
-* #pragma omp declare simd: :part:`Partial`. We support parsing/semantic
- analysis + generation of special attributes for X86 target, but still
- missing the LLVM pass for vectorization.
+Clang fully supports OpenMP 4.5, almost all of 5.0 and most of 5.1/2.
+Clang supports offloading to X86_64, AArch64, PPC64[LE], NVIDIA GPUs (all models) and AMD GPUs (all models).
In addition, the LLVM OpenMP runtime `libomp` supports the OpenMP Tools
Interface (OMPT) on x86, x86_64, AArch64, and PPC64 on Linux, Windows, and macOS.
+OMPT is also supported for NVIDIA and AMD GPUs.
For the list of supported features from OpenMP 5.0 and 5.1
see `OpenMP implementation details`_ and `OpenMP 51 implementation details`_.
@@ -36,17 +33,6 @@ General improvements
collapse clause by replacing the expensive remainder operation with
multiplications and additions.
-- The default schedules for the `distribute` and `for` constructs in a
- parallel region and in SPMD mode have changed to ensure coalesced
- accesses. For the `distribute` construct, a static schedule is used
- with a chunk size equal to the number of threads per team (default
- value of threads or as specified by the `thread_limit` clause if
- present). For the `for` construct, the schedule is static with chunk
- size of one.
-
-- Simplified SPMD code generation for `distribute parallel for` when
- the new default schedules are applicable.
-
- When using the collapse clause on a loop nest the default behavior
is to automatically extend the representation of the loop counter to
64 bits for the cases where the sizes of the collapsed loops are not
@@ -54,25 +40,10 @@ General improvements
at most 32 bits, compile your program with the
`-fopenmp-optimistic-collapse`.
-.. _basic support for Cuda devices:
-Cuda devices support
+GPU devices support
====================
-Directives execution modes
---------------------------
-
-Clang code generation for target regions supports two modes: the SPMD and
-non-SPMD modes. Clang chooses one of these two modes automatically based on the
-way directives and clauses on those directives are used. The SPMD mode uses a
-simplified set of runtime functions thus increasing performance at the cost of
-supporting some OpenMP features. The non-SPMD mode is the most generic mode and
-supports all currently available OpenMP features. The compiler will always
-attempt to use the SPMD mode wherever possible. SPMD mode will not be used if:
-
- - The target region contains user code (other than OpenMP-specific
- directives) in between the `target` and the `parallel` directives.
-
Data-sharing modes
------------------
@@ -82,8 +53,9 @@ performance and can be activated using the `-fopenmp-cuda-mode` flag. In
`Generic` mode all local variables that can be shared in the parallel regions
are stored in the global memory. In `Cuda` mode local variables are not shared
between the threads and it is user responsibility to share the required data
-between the threads in the parallel regions.
-
+between the threads in the parallel regions. Often, the optimizer is able to
+reduce the cost of `Generic` mode to the level of `Cuda` mode, but the flag,
+as well as other assumption flags, can be used for tuning.
Features not supported or with limited support for Cuda devices
---------------------------------------------------------------
@@ -96,9 +68,6 @@ Features not supported or with limited support for Cuda devices
- Nested parallelism: inner parallel regions are executed sequentially.
-- Automatic translation of math functions in target regions to device-specific
- math functions is not implemented yet.
-
- Debug information for OpenMP target regions is supported, but sometimes it may
be required to manually specify the address class of the inspected variables.
In some cases the local variables are actually allocated in the global memory,
@@ -139,7 +108,7 @@ implementation.
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
| memory management | allocate directive and allocate clause | :good:`done` | r355614,r335952 |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| OMPD | OMPD interfaces | :part:`done` | https://reviews.llvm.org/D99914 (Supports only HOST(CPU) and Linux |
+| OMPD | OMPD interfaces | :good:`done` | https://reviews.llvm.org/D99914 (Supports only HOST(CPU) and Linux |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
| OMPT | OMPT interfaces (callback support) | :good:`done` | |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
@@ -171,7 +140,7 @@ implementation.
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
| device | infer target functions from initializers | :part:`worked on` | |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device | infer target variables from initializers | :part:`done` | D146418 |
+| device | infer target variables from initializers | :good:`done` | D146418 |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
| device | OMP_TARGET_OFFLOAD environment variable | :good:`done` | D50522 |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
@@ -217,7 +186,7 @@ implementation.
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
| device | support close modifier on map clause | :good:`done` | D55719,D55892 |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device | teams construct on the host device | :part:`done` | r371553 |
+| device | teams construct on the host device | :good:`done` | r371553 |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
| device | support non-contiguous array sections for target update | :good:`done` | |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
@@ -235,7 +204,7 @@ implementation.
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
| misc | library shutdown (omp_pause_resource[_all]) | :good:`done` | D55078 |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc | metadirectives | :part:`worked on` | D91944 |
+| misc | metadirectives | :part:`mostly done` | D91944 |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
| misc | conditional modifier for lastprivate clause | :good:`done` | |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
@@ -243,7 +212,7 @@ implementation.
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
| misc | depobj directive and depobj dependency kind | :good:`done` | |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc | user-defined function variants | :part:`worked on` | D67294, D64095, D71847, D71830, D109635 |
+| misc | user-defined function variants | :good:`done`. | D67294, D64095, D71847, D71830, D109635 |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
| misc | pointer/reference to pointer based array reductions | :good:`done` | |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
@@ -298,7 +267,7 @@ implementation.
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
| device | indirect clause on declare target directive | :none:`unclaimed` | |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device | allow virtual functions calls for mapped object on device | :none:`unclaimed` | |
+| device | allow virtual functions calls for mapped object on device | :part:`partial` | |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
| device | interop construct | :part:`partial` | parsing/sema done: D98558, D98834, D98815 |
+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
More information about the cfe-commits
mailing list