Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

make test stalled #28

Closed
uu-praveeng opened this issue Feb 24, 2017 · 5 comments
Closed

make test stalled #28

uu-praveeng opened this issue Feb 24, 2017 · 5 comments
Labels
bug A confirmed library bug

Comments

@uu-praveeng
Copy link

Hi,
I built the mkl-dnn but make test wont run.
Running tests...
Test project /home/gaurav/GPraveen/mkl-dnn/build
Start 1: simple-net-c
It is halting with the above error.

Intel processor that my PC has in Intel(R) Core(TM) i3-2120 CPU @ 3.30GHz with out avx2.
Is this the issue will it work only for intel Xeon processors that they have mentioned?

Thanks in advance

@vpirogov
Copy link
Member

Thanks for your report. This library is not yet optimized for processors without Intel(R) AVX2 support. It will run reference implementation that is rather slow and tests might take long time to complete. This might be what you are experiencing.

I also do not see specific error message in the log you provide. Is this all the output you get?

@vpirogov vpirogov added the bug A confirmed library bug label Feb 24, 2017
@uu-praveeng
Copy link
Author

There are no error messages as such. But it is taking long enough time. Dint wait that much long to see the output to come up.
I thought there could be some thing that I was missing hence I posted.
How much time will it take to run the test if the processor in non Intel Xeon?

@rsdubtso
Copy link

mkl-dnn is only optimized for AVX2 and (partially) AVX512. This means that it should work on any mobile/desktop/server CPU that supports those instruction sets, so it does not require a Xeon CPU.

For older CPUs that do not support at least AVX2, mkl-dnn falls back to reference implementations that are very slow. We also seem to have a bug where we do not correctly check for supported ISA and try executing AVX2 code on CPUs that do not support it.

@liyancas
Copy link

liyancas commented Apr 1, 2017

@rsdubtso Does mkl-dnn will be optimized for CPU with AVX in the future? I tested the inference results with mkl-dnn on AVX CPU, but it's much too slower than im2col + MKL GEMM method.

@rsdubtso
Copy link

rsdubtso commented Apr 2, 2017

There are no such plans at the moment. Of course we will consider pull requests that add such optimizations.

@Darwinian2 Darwinian2 mentioned this issue Aug 9, 2017
m-ymzk pushed a commit to m-ymzk/oneDNN that referenced this issue Nov 29, 2020
Fix: some errors (benchdnn + test_conv_all)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
bug A confirmed library bug
Projects
None yet
Development

No branches or pull requests

4 participants