marcoabreu commented on a change in pull request #10760: [MXNET-372] Add build
flag for USE_F16C in CMake and clarify flag in make
URL: https://github.com/apache/incubator-mxnet/pull/10760#discussion_r185213730
##########
File path: make/config.mk
##########
@@ -132,10 +132,19 @@ endif
ARCH := $(shell uname -a)
ifneq (,$(filter $(ARCH), armv6l armv7l powerpc64le ppc64le aarch64))
USE_SSE=0
+ USE_F16C=0
else
USE_SSE=1
endif
+#----------------------------
+# F16C instruction support for faster arithmetic of fp16 on CPU
+#----------------------------
+# For distributed training with fp16, this helps even if training on GPUs
+# If left empty, checks CPU support and turns it on.
+# For cross compilation, please check support for F16C on target device and
turn off if necessary.
+USE_F16C =
Review comment:
I think this is hard to pull off, considering it's a compile flag, and
overcomplicates our build-process even further. As @hen pointed out on dev@, we
should rather find a more permanent solution rather than trying to add even
more hacks
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
