ptrendx commented on issue #14067: Increase perfomance of BulkAppend and
BulkFlush
URL: https://github.com/apache/incubator-mxnet/pull/14067#issuecomment-460842410
@yuxihu It is actually the other way around - single process per GPU
alleviates some of those issues, because each GPU is hand
ptrendx commented on issue #14067: Increase perfomance of BulkAppend and
BulkFlush
URL: https://github.com/apache/incubator-mxnet/pull/14067#issuecomment-461145342
@junrushao1994 2 reasons:
- std::move to lambda is not supported before C++14 without some arcane
template tricks, which
ptrendx commented on issue #14067: Increase perfomance of BulkAppend and
BulkFlush
URL: https://github.com/apache/incubator-mxnet/pull/14067#issuecomment-461145731
But I agree unique_ptr would be the ultimate solution there.