forked from OSchip/llvm-project
c5348aecd7
Summary: Retaining per device primary context is preferred to creating a context owned by the plugin. From CUDA documentation 1. Note that the use of multiple CUcontext s per device within a single process will substantially degrade performance and is strongly discouraged. Instead, it is highly recommended that the implicit one-to-one device-to-context mapping for the process provided by the CUDA Runtime API be used." from https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__DRIVER.html 2. Right under cuCtxCreate. In most cases it is recommended to use cuDevicePrimaryCtxRetain. https://docs.nvidia.com/cuda/cuda-driver-api/group__CUDA__CTX.html#group__CUDA__CTX_1g65dc0012348bc84810e2103a40d8e2cf 3. The primary context is unique per device and shared with the CUDA runtime API. These functions allow integration with other libraries using CUDA. https://docs.nvidia.com/cuda/cuda-driver-api/group__CUDA__PRIMARY__CTX.html#group__CUDA__PRIMARY__CTX Two issues are addressed by this patch: 1. Not using the primary context caused interoperability issue with libraries like cublas, cusolver. CUBLAS_STATUS_EXECUTION_FAILED and cudaErrorInvalidResourceHandle 2. On OLCF summit, "Error returned from cuCtxCreate" and "CUDA error is: invalid device ordinal" Regarding the flags of the primary context. If it is inactive, we set CU_CTX_SCHED_BLOCKING_SYNC. If it is already active, we respect the current flags. Reviewers: grokos, ABataev, jdoerfert, protze.joachim, AndreyChurbanov, Hahnfeld Reviewed By: jdoerfert Subscribers: openmp-commits, yaxunl, guansong, sstefan1, tianshilei1992 Tags: #openmp Differential Revision: https://reviews.llvm.org/D82718 |
||
|---|---|---|
| .. | ||
| cmake/Modules | ||
| deviceRTLs | ||
| include | ||
| plugins | ||
| src | ||
| test | ||
| CMakeLists.txt | ||
| README.txt | ||
README.txt
README for the LLVM* OpenMP* Offloading Runtime Library (libomptarget)
======================================================================
How to Build the LLVM* OpenMP* Offloading Runtime Library (libomptarget)
========================================================================
In-tree build:
$ cd where-you-want-to-live
Check out openmp (libomptarget lives under ./libomptarget) into llvm/projects
$ cd where-you-want-to-build
$ mkdir build && cd build
$ cmake path/to/llvm -DCMAKE_C_COMPILER=<C compiler> -DCMAKE_CXX_COMPILER=<C++ compiler>
$ make omptarget
Out-of-tree build:
$ cd where-you-want-to-live
Check out openmp (libomptarget lives under ./libomptarget)
$ cd where-you-want-to-live/openmp/libomptarget
$ mkdir build && cd build
$ cmake path/to/openmp -DCMAKE_C_COMPILER=<C compiler> -DCMAKE_CXX_COMPILER=<C++ compiler>
$ make
For details about building, please look at README.rst in the parent directory.
Architectures Supported
=======================
The current library has been only tested in Linux operating system and the
following host architectures:
* Intel(R) 64 architecture
* IBM(R) Power architecture (big endian)
* IBM(R) Power architecture (little endian)
* ARM(R) AArch64 architecture (little endian)
The currently supported offloading device architectures are:
* Intel(R) 64 architecture (generic 64-bit plugin - mostly for testing purposes)
* IBM(R) Power architecture (big endian) (generic 64-bit plugin - mostly for testing purposes)
* IBM(R) Power architecture (little endian) (generic 64-bit plugin - mostly for testing purposes)
* ARM(R) AArch64 architecture (little endian) (generic 64-bit plugin - mostly for testing purposes)
* CUDA(R) enabled 64-bit NVIDIA(R) GPU architectures
Supported RTL Build Configurations
==================================
Supported Architectures: Intel(R) 64, IBM(R) Power 7 and Power 8
---------------------------
| gcc | clang |
--------------|------------|------------|
| Linux* OS | Yes(1) | Yes(2) |
-----------------------------------------
(1) gcc version 4.8.2 or later is supported.
(2) clang version 3.7 or later is supported.
Front-end Compilers that work with this RTL
===========================================
The following compilers are known to do compatible code generation for
this RTL:
- clang (from https://github.com/clang-ykt )
- clang (development branch at http://clang.llvm.org - several features still
under development)
-----------------------------------------------------------------------
Notices
=======
This library and related compiler support is still under development, so the
employed interface is likely to change in the future.
*Other names and brands may be claimed as the property of others.