search

apache / mxnet

Lightweight, Portable, Flexible Distributed/Mobile Deep Learning with Dynamic, Mutation-aware Dataflow Dep Scheduler; for Python, R, Julia, Scala, Go, Javascript and more
https://mxnet.apache.org
Apache License 2.0
20.77k stars 6.79k forks source link

Enable MKL_USE_ILP64 cause a lot of compile time errors on windws #14364

Open stereomatchingkiss opened 5 years ago

stereomatchingkiss commented 5 years ago

As the title mentioned, there are tons of compile time error when you switch this option on. I suspect without it, mxnet would build with mkl 32bits, which cause another issues happens, you can check the details from #14343, I write down the details at there(main issue is " Intel MKL FATAL ERROR: Cannot load mkl_intel_thread.dll.")

part of the error messages

Will open a pull request if I can fix those compile time error. Any plan to support travis build for mxnet?This kind of error should be easy to spot if we could build it on each platform when commit.

mxnet-label-bot commented 5 years ago

Hey, this is the MXNet Label Bot. Thank you for submitting the issue! I will try and suggest some labels so that the appropriate MXNet community members can help resolve it. Here are my recommended labels: Build

stereomatchingkiss commented 5 years ago

I "fix" those compile time errors, but this do not solve the issue, because this time it pop out new errors when compile the app, the error message is

"....\3rdLibs\mxnet\build_cpu_1_3_1\install\include\mxnet-cpp/monitor.hpp(47): error C3861: 'MXExecutorSetMonitorCallbackEX': identifier not found"

I give up on using mxnet cpp-package on windows + cpu support, it is too painful to build it.

piyushghai commented 5 years ago

@mxnet-label-bot Add [CPP, Windows, Build]

pengzhao-intel commented 5 years ago

@TaoLv @juliusshufan

stereomatchingkiss commented 5 years ago

Found a way to build cpu version of 1.3.1 without using anything related to mkl, you can check this post and #14380