No GPU devices error on AMD ATI 7600M series

[kkarnatak]

Hi,

I installed OpenCL on AMD ATI Radeon HD 7600M series. As far as I know my GPU is configured and installed correctly. But, I am getting "No GPU devices error" while running make runtest.

~/OpenCL-caffe$ clinfo | grep CPU
  Device Type:					 CL_DEVICE_TYPE_CPU
~/OpenCL-caffe$ clinfo | grep GPU // No output

However, glxinfo displays the AMD GPU.

~/OpenCL-caffe$ glxinfo  | grep OpenGL
OpenGL vendor string: Advanced Micro Devices, Inc.
OpenGL renderer string: AMD Radeon HD 7600M Series
OpenGL core profile version string: 4.3.13416 Core Profile Context 15.201.1151

ERROR:

~/OpenCL-caffe$ make runtest
[100%] Built target proto
[100%] Built target caffe
[100%] Built target gtest
[100%] Built target test.testbin
Current device id: 0
F0304 18:49:09.351958  1995 device.cpp:75] Err: No GPU devices
*** Check failure stack trace: ***
    @     0x2b783ee9cdaa  (unknown)
    @     0x2b783ee9cce4  (unknown)
    @     0x2b783ee9c6e6  (unknown)
    @     0x2b783ee9f687  (unknown)
    @     0x2b783e8322a4  caffe::Device::Init()
    @           0x6f4468  main
    @     0x2b7840bb5f45  (unknown)
    @           0x6f7d02  (unknown)
    @              (nil)  (unknown)
Aborted (core dumped)

[naibaf7]

Try this: https://github.com/BVLC/caffe/tree/opencl

[kkarnatak]

Thanks. however, I have a question. I cannot see any flag in the makefile.config to enable/disable opencl. How can I configure it?

[naibaf7]

It is enabled by default (USE_GREENTEA := 1) and CUDA should be disabled by default. This is the excert from Makefile.config.example that you can copy to your own Makefile.config, just like normal Caffe:

# Enable the CUDA backend
# USE_CUDA := 1
# Enable the OpenCL/Greentea backend
USE_GREENTEA := 1

It's a bit of an inside joke, since Caffe is the "CUDA Caffe" and Greentea (the refreshing alternative beverage to it) is the "OpenCL Caffe" ;)

[kkarnatak]

I compiled it without changing anything in the example Makefile.config. Just commented out the CUDA lines. My compilation and runtest were successful ( although I am not sure if its considering opencl and my gpu). I didnt even enable the CPU mode. The CPU line is commented.

Now, when I run the example training script, I get following error:

$ examples/cifar10/train_quick.sh 
I0305 03:02:22.887403 28627 caffe.cpp:218] Using GPUs 0
F0305 03:02:22.887708 28627 common.cpp:66] Cannot use GPU in CPU-only Caffe: check mode.
*** Check failure stack trace: ***
    @     0x7f765e621daa  (unknown)
    @     0x7f765e621ce4  (unknown)
    @     0x7f765e6216e6  (unknown)
    @     0x7f765e624687  (unknown)
    @     0x7f765ec4f800  caffe::Caffe::SetDevice()
    @           0x40a7fe  train()
    @           0x40818c  main
    @     0x7f765d610f45  (unknown)
    @           0x4089fb  (unknown)
    @              (nil)  (unknown)

[naibaf7]

Hm that should not happen, and you shouldn't have to change anything in the Makefile.config...
the NO_GPU error is usually occuring if you try to use a GPU but you compiled Caffe with CPU_ONLY uncommented in the Makefile.config. So check that and rebuild using make clean; make all -j4

[kkarnatak]

Sorry, my bad. I didnt clone from the right branch and didnt used the correct code/Makefile.config. I wasnt expecting such a late night reply and thus didnt even read your comment before posting my previous post :D
I will try now with new file and repost thanks!

[kkarnatak]

I tried again and now I am not getting any of the errors I mentioned before. However, I am getting another error. The code is trying to look for a file "solver_data_list.txt.gen.cmake" which isnt generated. I believe somehow the file isnt created or so.

 22  protected:
 23   HDF5DataLayerTest()
 24       : filename(NULL),
 33     // Check out generate_sample_data.py in the same directory.
 34     filename = new string(
 35     CMAKE_SOURCE_DIR "caffe/test/test_data/sample_data_list.txt" CMAKE_EXT);
 36     LOG(INFO)<< "Using sample HDF5 data file " << filename;
 37   }

 AdaDeltaSolverTest/3.TestAdaDeltaLeastSquaresUpdate
F0305 04:16:52.231627  8628 hdf5_data_layer.cpp:89] Failed to open source file: /home/kartikeya/Caffe/caffe/src/caffe/test/test_data/solver_data_list.txt.gen.cmake
*** Check failure stack trace: ***
    @     0x2af6faff3daa  (unknown)
    @     0x2af6faff3ce4  (unknown)
    @     0x2af6faff36e6  (unknown)
    @     0x2af6faff6687  (unknown)
    @     0x2af6fa209b66  caffe::HDF5DataLayer<>::LayerSetUp()
    @     0x2af6fa303ac3  caffe::Net<>::Init()
    @     0x2af6fa305699  caffe::Net<>::Net()
    @     0x2af6fa2d84c7  caffe::Solver<>::InitTrainNet()
    @     0x2af6fa2d89f7  caffe::Solver<>::Init()
    @     0x2af6fa2d8cbb  caffe::Solver<>::Solver()
    @           0x948572  caffe::AdaDeltaSolverTest<>::InitSolver()
    @           0x946e43  caffe::GradientBasedSolverTest<>::InitSolverFromProtoString()
    @           0x90270d  caffe::GradientBasedSolverTest<>::RunLeastSquaresSolver()
    @           0x905d19  caffe::GradientBasedSolverTest<>::TestLeastSquaresUpdate()
    @           0xda2d83  testing::internal::HandleExceptionsInMethodIfSupported<>()
    @           0xd99437  testing::Test::Run()
    @           0xd994de  testing::TestInfo::Run()
    @           0xd995e5  testing::TestCase::Run()
    @           0xd9a458  testing::internal::UnitTestImpl::RunAllTests()
    @           0xd9a707  testing::UnitTest::Run()
    @           0x8caf35  main
    @     0x2af6fd136f45  (unknown)
    @           0x8ce792  (unknown)
    @              (nil)  (unknown)

[kkarnatak]

I manually renamed the file sample_data_list.txt to the expected name ( with cmake ext) and it passed through. The overall test perf:

----------] Global test environment tear-down
[==========] 2062 tests from 274 test cases ran. (1441081 ms total)
[ PASSED ] 2060 tests.
[ FAILED ] 2 tests, listed below:
[ FAILED ] NetTest/0.TestSharedWeightsUpdate, where TypeParam = caffe::CPUDevice
[ FAILED ] NetTest/2.TestSharedWeightsUpdate, where TypeParam = caffe::GPUDevice