Getting "floating point exception" with CUDA 9.0

NervanaSystems / neon

Intel® Nervana™ reference deep learning framework committed to best performance on all hardware

http://neon.nervanasys.com/docs/latest

Apache License 2.0

3.87k stars 811 forks source link

Getting "floating point exception" with CUDA 9.0 #416

Closed shaofushih closed 6 years ago

shaofushih commented 6 years ago

Hi all,

I just ran into an issue with python examples/mnist_mlp.py -b gpu while using neon 2.3.0

RuntimeError(136); ptxas -v -arch sm_61 -o /home/midearoot/.cache/neon/kernels/cubin/sgemm_nn_128x128_vec.cubin /home/midearoot/.cache/neon/kernels/ptx/sgemm_nn_128x128_vec.ptx Floating point exception (core dumped)

Did anyone have the same problem after upgrading to neon 2.3.0

shaofushih commented 6 years ago

We changed the make file to support sm_61 and that worked

wei-v-wang commented 6 years ago

@shaofushih Can you please provide your changes so that issue#413 could benefit and be resolved? Thanks!

wei-v-wang commented 6 years ago

https://github.com/xingjinglu/PerfAILibs/blob/master/README.md seems to already get the workaround by @xingjinglu.