Standard template implementation (#703)

[ROCm/rccl commit: 6e48e518d9]
这个提交包含在:
Saad Rahim
2023-03-13 11:00:57 -06:00
提交者 GitHub
父节点 216c83c39d
当前提交 8fdc4795fd
修改 21 个文件,包含 340 行新增478 行删除
+13 -2
查看文件
@@ -32,7 +32,7 @@ The root of this repository has a helper script 'install.sh' to build and instal
* `./install.sh --prefix` -- specify custom path to install RCCL to (default:/opt/rocm)
## Manual build
#### To build the library :
### To build the library :
```shell
$ git clone https://github.com/ROCmSoftwarePlatform/rccl.git
@@ -48,7 +48,7 @@ $ CXX=/opt/rocm/bin/hipcc cmake -DCMAKE_PREFIX_PATH=/opt/rocm/ -DCMAKE_INSTALL_P
```
Note: ensure rocm-cmake is installed, `apt install rocm-cmake`.
#### To build the RCCL package and install package :
### To build the RCCL package and install package :
Assuming you have already cloned this repository and built the library as shown in the previous section:
@@ -60,6 +60,17 @@ $ sudo dpkg -i *.deb
RCCL package install requires sudo/root access because it creates a directory called "rccl" under /opt/rocm/. This is an optional step and RCCL can be used directly by including the path containing librccl.so.
### How to build documentation
Please follow the instructions below to build the documentation.
```bash
cd docs
pip3 install -r .sphinx/requirements.txt
python3 -m sphinx -T -E -b html -d _build/doctrees -D language=en . _build/html
```
## Enabling peer-to-peer transport
In order to enable peer-to-peer access on machines with PCIe-connected GPUs, the HSA environment variable HSA_FORCE_FINE_GRAIN_PCIE=1 is required to be set, on top of requiring GPUs that support peer-to-peer access and proper large BAR addressing support.