This is an automated email from the ASF dual-hosted git repository.
janardhan pushed a commit to branch master
in repository https://gitbox.apache.org/repos/asf/systemds.git
The following commit(s) were added to refs/heads/master by this push:
new ee4b19c [MINOR] Add PTX JIT option instructions for sm_*
ee4b19c is described below
commit ee4b19c5497ae9c3cb58057ed4d948ecc12a5dc3
Author: Janardhan Pulivarthi <[email protected]>
AuthorDate: Mon May 10 05:21:40 2021 +0530
[MINOR] Add PTX JIT option instructions for sm_*
---
docs/site/gpu.md | 11 ++++++++++-
1 file changed, 10 insertions(+), 1 deletion(-)
diff --git a/docs/site/gpu.md b/docs/site/gpu.md
index 20c876e..80a3e5e 100644
--- a/docs/site/gpu.md
+++ b/docs/site/gpu.md
@@ -43,7 +43,16 @@ The following GPUs are supported:
For CUDA enabled gpu cards at [CUDA
GPUs](https://developer.nvidia.com/cuda-gpus)
* For GPUs with unsupported CUDA architectures, or to avoid JIT compilation
from PTX, or to
use difference versions of the NVIDIA libraries, build on Linux from source
code.
-* Release artifacts contain PTX code for the latest supported CUDA
architecture.
+* Release artifacts contain PTX code for the latest supported CUDA
architecture. In case your
+architecture specific PTX is not available enable JIT PTX with instructions
compiler driver `nvcc`
+[GPU
Compilation](https://docs.nvidia.com/cuda/cuda-compiler-driver-nvcc/index.html#gpu-compilation).
+
+ > For example, with `--gpu-code` use actual gpu names, `--gpu-architecture`
is the name of virtual
+ > compute architecture
+ >
+ > ```sh
+ > nvcc SystemDS.cu --gpu-architecture=compute_50 --gpu-code=sm_50,sm_52
+ > ```
### Software
