samskalicky commented on a change in pull request #18904:
URL: https://github.com/apache/incubator-mxnet/pull/18904#discussion_r476063390
##########
File path: config/linux_gpu.cmake
##########
@@ -33,6 +33,14 @@
# Specify `cmake --build . --parallel N` to set the number of parallel
compilation jobs.
# Default is derived from CPUs available.
#
+# By default, cmake will try and discover which GPU architecture to use by
looking at
+# the available GPUs on the machine that you're building on. If you want to
build for
+# a specific GPU architecture or are building on a machine without a GPU, then
+# specify the MXNET_CUDA_ARCH option like:
+#
+# $ cmake .. -DMXNET_CUDA_ARCH=7.0
+#
+# In the example above we're building for sm_70 which is the Volta
architecture.
Review comment:
lol, ya i totally didnt see that below. Plus I dont think I would have
understood what it was saying. I needed a dumbed down version like "heres how
to compile for a specific GPU...".
Do we expect users will just build on the machine with the GPU that they
intend to use? So the simple `cmake ..` is all they need? Who are these
instructions for? If they were for me, they would say heres how to build for a
specific GPU :D
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
