This bugzilla service is closed. All entries have been migrated to https://gitlab.com/libeigen/eigen
Bug 1292 - cxx11_tensor_of_float16_cuda_3 failure
Summary: cxx11_tensor_of_float16_cuda_3 failure
Status: UNCONFIRMED
Alias: None
Product: Eigen
Classification: Unclassified
Component: Tensor (show other bugs)
Version: 3.3 (current stable)
Hardware: GPU (CUDA) Linux
: Normal Failed Unit Test
Assignee: Nobody
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2016-09-09 19:59 UTC by william.tambellini
Modified: 2019-12-04 16:12 UTC (History)
4 users (show)



Attachments

Description william.tambellini 2016-09-09 19:59:37 UTC
The cxx11_tensor_of_float16_cuda_3 fails in Eigen::GpuDevice::memcpyDeviceToHost() :

806: Test command: /home/ubuntu/eigen/build/unsupported/test/cxx11_tensor_of_float16_cuda_3
806: Test timeout computed to be: 1500
806: Initializing random number generator with seed 1473450378
806: Repeating each test 10 times
806: Reducing 13 by 13 tensor along dim 0 result size:13 sizeof(Eigen::half)=2 sizeof(float)=4
806: Reducing 13 by 13 tensor along dim 1 result size:13 sizeof(Eigen::half)=2 sizeof(float)=4
806: Reducing 35 by 36 tensor along dim 0 result size:36 sizeof(Eigen::half)=2 sizeof(float)=4
806: Reducing 35 by 36 tensor along dim 1 result size:35 sizeof(Eigen::half)=2 sizeof(float)=4
806: Reducing 36 by 35 tensor along dim 0 result size:35 sizeof(Eigen::half)=2 sizeof(float)=4
806: Reducing 36 by 35 tensor along dim 1 result size:36 sizeof(Eigen::half)=2 sizeof(float)=4
806: cuda error 77: an illegal memory access was encountered dst:0x1a96000 src:0x7024a0000 size:72
806: cxx11_tensor_of_float16_cuda_3: /home/ubuntu/eigen/unsupported/Eigen/CXX11/src/Tensor/TensorDeviceCuda.h:234: void Eigen::GpuDevice::memcpyDeviceToHost(void*, const void*, size_t) const: Assertion `err == cudaSuccess' failed.
1/1 Test #806: cxx11_tensor_of_float16_cuda_3 ...***Exception: Other  0.75 sec

steps to reproduce:
- using Ubuntu 14.04.4 LTS
- nvcc Cuda compilation tools, release 7.5, V7.5.17
- gcc 4.8
- hg clone https://bitbucket.org/eigen/eigen/
- mkdir build & cmake .. & make
- ./buildtests.sh _float16_cuda
- EIGEN_CTEST_ARGS='-VV' ./check.sh float16_cuda_3

656: CUDA device info:
656:   name:                        GRID K520
656:   capability:                  3.0
656:   multiProcessorCount:         8
656:   maxThreadsPerMultiProcessor: 2048
656:   warpSize:                    32
656:   regsPerBlock:                65536
656:   concurrentKernels:           1
656:   clockRate:                   797000
656:   canMapHostMemory:            1
656:   computeMode:                 0
Comment 1 Benoit Steiner 2016-09-19 18:46:57 UTC
I can't reproduce the problem on my side. I often get a failure that's due to the random number generator that generates infinite numbers, but that's a different issue.
Comment 2 Nobody 2019-12-04 16:12:47 UTC
-- GitLab Migration Automatic Message --

This bug has been migrated to gitlab.com's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.com/libeigen/eigen/issues/1292.

Note You need to log in before you can comment on or make changes to this bug.