Caffe: `/usr/bin/ld: cannot find -lproto` when opencl branch building

hi,I think u need install googleprotobuffer

Use caffe on ubuntu u would better read caffe wiki on ubuntu or mac etc. if u run on windows the caffe source code need fix an error that cannot find some layer .the code I have written on my question at this issues,but still another error I cann't fix it.

It's a bug, will fix asap.

Should be fixed.

This looks like commit 30a2ab7e50 if broken.

Looks like we also need rename proto to caffeproto in cmake/Targets.cmake

Something like this:
--- a/cmake/Targets.cmake
+++ b/cmake/Targets.cmake
@@ -12,14 +12,14 @@ macro(caffe_set_caffe_link)
endforeach()
endif()
if(BUILD_SHARED_LIBS)
- set(Caffe_LINK caffe proto)
+ set(Caffe_LINK caffe caffeproto)
else()
if("${CMAKE_CXX_COMPILER_ID}" STREQUAL "Clang")
- set(Caffe_LINK -Wl,-force_load caffe -Wl,-force_load proto)
+ set(Caffe_LINK -Wl,-force_load caffe -Wl,-force_load caffeproto)
elseif("${CMAKE_CXX_COMPILER_ID}" STREQUAL "GNU")
- set(Caffe_LINK -Wl,--whole-archive caffe proto -Wl,--no-whole-archive)
+ set(Caffe_LINK -Wl,--whole-archive caffe caffeproto -Wl,--no-whole-archive)
elseif(MSVC)
- set(Caffe_LINK caffe proto)
+ set(Caffe_LINK caffe caffeproto)
endif()
endif()
endmacro()

Thanks,
Val.

1. make -j4 succeeded with @naibaf7 commit https://github.com/BVLC/caffe/commit/e0f77c3b5f4837615f05b097ba3f2a05d7413e58 Thank you.
2. make runtest -j4 failed because EXAMPLES_SOURCE_DIR macro is not defined.
   EXAMPLES_SOURCE_DIR is defined here but when cmake building it isn't defined.
   So I need sed -e 's/\#define CUDA_TEST_DEVICE/\#define CUDA_TEST_DEVICE \n\#define EXAMPLES_SOURCE_DIR "\/home\/delta\/git\/caffe\/examples\/"/g' -i ./caffe_config.h and test works.
3. some test failed. The most serious is Segmentation fault on clFinish.

hardware: Surface Pro3
os: ubuntu14.04
SDK: OpenCL 1.2, beignet 1.4 (git-beaf26f)
CPU: Core i5 4300U HD4400 (Haswell)

$ ./tools/caffe device_query -gpu all
I0518 16:19:30.288357 13602 common.cpp:433] Total devices: 1
I0518 16:19:30.288759 13602 common.cpp:434] CUDA devices: 0
I0518 16:19:30.288827 13602 common.cpp:435] OpenCL devices: 1
I0518 16:19:30.288887 13602 common.cpp:459] Device id:                     0
I0518 16:19:30.288947 13602 common.cpp:461] Device backend:                OpenCL
I0518 16:19:30.289026 13602 common.cpp:463] Backend details:               Intel: OpenCL 1.2 beignet 1.4 (git-beaf26f)
I0518 16:19:30.289121 13602 common.cpp:465] Device vendor:                 Intel
I0518 16:19:30.289168 13602 common.cpp:467] Name:                          Intel(R) HD Graphics Haswell Ultrabook GT2 Mobile
I0518 16:19:30.289214 13602 common.cpp:469] Total global memory:           2021654528

SyncedMemoryTest

[ RUN      ] SyncedMemoryTest.TestGPURead
/home/delta/git/caffe/src/caffe/test/test_syncedmem.cpp:129: Failure
Value of: 2
Expected: (static_cast<char*>(recovered_value))[i]
Which is: '\x1' (1)
/home/delta/git/caffe/src/caffe/test/test_syncedmem.cpp:129: Failure
Value of: 2
Expected: (static_cast<char*>(recovered_value))[i]
Which is: '\x1' (1)
/home/delta/git/caffe/src/caffe/test/test_syncedmem.cpp:129: Failure
Value of: 2
Expected: (static_cast<char*>(recovered_value))[i]
Which is: '\x1' (1)
/home/delta/git/caffe/src/caffe/test/test_syncedmem.cpp:129: Failure
Value of: 2
Expected: (static_cast<char*>(recovered_value))[i]
Which is: '\x1' (1)
/home/delta/git/caffe/src/caffe/test/test_syncedmem.cpp:129: Failure
Value of: 2
Expected: (static_cast<char*>(recovered_value))[i]
Which is: '\x1' (1)
/home/delta/git/caffe/src/caffe/test/test_syncedmem.cpp:129: Failure
Value of: 2
Expected: (static_cast<char*>(recovered_value))[i]
Which is: '\x1' (1)
/home/delta/git/caffe/src/caffe/test/test_syncedmem.cpp:129: Failure
Value of: 2
Expected: (static_cast<char*>(recovered_value))[i]
Which is: '\x1' (1)
/home/delta/git/caffe/src/caffe/test/test_syncedmem.cpp:129: Failure
Value of: 2
Expected: (static_cast<char*>(recovered_value))[i]
Which is: '\x1' (1)
/home/delta/git/caffe/src/caffe/test/test_syncedmem.cpp:129: Failure
Value of: 2
Expected: (static_cast<char*>(recovered_value))[i]
Which is: '\x1' (1)
/home/delta/git/caffe/src/caffe/test/test_syncedmem.cpp:129: Failure
Value of: 2
Expected: (static_cast<char*>(recovered_value))[i]
Which is: '\x1' (1)
[  FAILED  ] SyncedMemoryTest.TestGPURead (1 ms)
[----------] 7 tests from SyncedMemoryTest (2 ms total)

CPUDevice

https://pastebin.com/1Xk3hi9a

[----------] 26 tests from NetTest/1, where TypeParam = caffe::CPUDevice<double>
[ RUN      ] NetTest/1.TestGetBlob
[       OK ] NetTest/1.TestGetBlob (0 ms)
[ RUN      ] NetTest/1.TestBottomNeedBackwardForce
[       OK ] NetTest/1.TestBottomNeedBackwardForce (1 ms)
[ RUN      ] NetTest/1.TestReshape
Build Status = -2 ( Err = -11 )
Log: stringInput.cl:111:49: error: use of type 'double' requires cl_khr_fp64 extension to be enabled
stringInput.cl:20:15: note: expanded from macro 'Dtype'
...
unknown file: Failure
C++ exception with description "ViennaCL: FATAL ERROR: CL_INVALID_PROGRAM_EXECUTABLE.
If you think that this is a bug in ViennaCL, please report it at [email protected] and supply at least the following information:
 * Operating System
 * Which OpenCL implementation (AMD, NVIDIA, etc.)
 * ViennaCL version
Many thanks in advance!" thrown in the test body.
[  FAILED  ] NetTest/1.TestReshape, where TypeParam = caffe::CPUDevice<double> (159 ms)

Segmentation fault on clFinish

[----------] Global test environment tear-down
[==========] 2100 tests from 276 test cases ran. (1667193 ms total)
[  PASSED  ] 2098 tests.
[  FAILED  ] 2 tests, listed below:
[  FAILED  ] NetTest/1.TestReshape, where TypeParam = caffe::CPUDevice<double>
[  FAILED  ] SyncedMemoryTest.TestGPURead

 2 FAILED TESTS
*** Aborted at 1495090803 (unix time) try "date -d @1495090803" if you are using GNU date ***
PC: @     0x7f70632e9688 clFinish
*** SIGSEGV (@0x400000218) received by PID 12814 (TID 0x7f7065513ac0) from PID 536; stack trace: ***
    @     0x7f70642ff390 (unknown)
    @     0x7f70632e9688 clFinish
    @     0x7f7064f9d62a caffe::SyncedMemory::~SyncedMemory()
    @           0x9109c2 boost::detail::sp_counted_impl_p<>::dispose()
    @           0x8e3f3a boost::detail::sp_counted_base::release()
    @           0x912222 boost::detail::sp_counted_impl_p<>::dispose()
    @     0x7f7064ce6b59 boost::detail::sp_counted_impl_p<>::dispose()
    @     0x7f7064ce3fe8 std::vector<>::~vector()
    @     0x7f7061ec136a __cxa_finalize
    @     0x7f7064cddef3 (unknown)
Segmentation fault (core dumped)

also benchmark

$ ./tools/caffe time -model ../models/bvlc_alexnet/deploy.prototxt -gpu 0
I0518 14:28:17.989312  7905 caffe.cpp:391] Use GPU with device ID 0
I0518 14:28:18.069370  7905 device.cpp:62] CL_DEVICE_HOST_UNIFIED_MEMORY: 1
I0518 14:28:19.164757  7905 net.cpp:57] Initializing net from parameters: 
name: "AlexNet"
...
I0518 14:31:27.541334  7905 caffe.cpp:469] Average Forward pass: 3.68935e+11 ms.
I0518 14:31:27.541369  7905 caffe.cpp:471] Average Backward pass: 1112.85 ms.
I0518 14:31:27.541389  7905 caffe.cpp:473] Average Forward-Backward: 3.68935e+11 ms.
I0518 14:31:27.541406  7905 caffe.cpp:475] Total Time: 1.84467e+13 ms.
I0518 14:31:27.541422  7905 caffe.cpp:476] *** Benchmark ends ***
*** Aborted at 1495085487 (unix time) try "date -d @1495085487" if you are using GNU date ***
PC: @     0x7f1fec302685 clFinish
*** SIGSEGV (@0xffffffffffffffff) received by PID 7905 (TID 0x7f1fed94bac0) from PID 18446744073709551615; stack trace: ***
    @     0x7f1feb9d34b0 (unknown)
    @     0x7f1fec302685 clFinish
    @     0x7f1fed3d662a caffe::SyncedMemory::~SyncedMemory()
    @     0x7f1fed18a262 boost::detail::sp_counted_impl_p<>::dispose()
    @     0x7f1fed12550a boost::detail::sp_counted_base::release()
    @     0x7f1fed125a2d boost::detail::sp_counted_impl_p<>::dispose()
    @     0x7f1fed11fb59 boost::detail::sp_counted_impl_p<>::dispose()
    @     0x7f1fed11cfe8 std::vector<>::~vector()
    @     0x7f1feb9d836a __cxa_finalize
    @     0x7f1fed116ef3 (unknown)
zsh: segmentation fault (core dumped)  ./tools/caffe time -model ../models/bvlc_alexnet/deploy.prototxt -gpu 0

Should We report a new issue?
Or it may be a beignet bug...?

Thanks.

@legokichi Ok I'll look into it. Yes you could try the proprietary Intel drivers as an alternative to check if they run any better meanwhile.

@naibaf7 I change Intel driver (here is the clinfo result)

$ ./tools/caffe device_query
I0519 20:50:38.733840 11750 common.cpp:433] Total devices: 1
I0519 20:50:38.734081 11750 common.cpp:434] CUDA devices: 0
I0519 20:50:38.734099 11750 common.cpp:435] OpenCL devices: 1
I0519 20:50:38.734107 11750 common.cpp:459] Device id:                     0
I0519 20:50:38.734114 11750 common.cpp:461] Device backend:                OpenCL
I0519 20:50:38.734138 11750 common.cpp:463] Backend details:               Intel(R) Corporation: OpenCL 1.2 LINUX
I0519 20:50:38.734163 11750 common.cpp:465] Device vendor:                 Intel(R) Corporation
I0519 20:50:38.734176 11750 common.cpp:467] Name:                          Intel(R) Core(TM) i5-4300U CPU @ 1.90GHz
I0519 20:50:38.734329 11750 common.cpp:469] Total global memory:           4044312576

benchmark works without segfault.

$ tools/caffe time -model ../models/bvlc_alexnet/deploy.prototxt -gpu 0
I0519 19:48:51.957670  9651 caffe.cpp:391] Use GPU with device ID 0
I0519 19:48:52.070819  9651 device.cpp:62] CL_DEVICE_HOST_UNIFIED_MEMORY: 1
I0519 19:49:02.506319  9651 net.cpp:57] Initializing net from parameters: 
name: "AlexNet"
...
I0519 19:58:50.314121  9651 caffe.cpp:469] Average Forward pass: 5247.06 ms.
I0519 19:58:50.314211  9651 caffe.cpp:471] Average Backward pass: 6389.42 ms.
I0519 19:58:50.314231  9651 caffe.cpp:473] Average Forward-Backward: 11636.9 ms.
I0519 19:58:50.314245  9651 caffe.cpp:475] Total Time: 581843 ms.
I0519 19:58:50.314256  9651 caffe.cpp:476] *** Benchmark ends ***

However, TestReshape stopped with 100% CPU usage.

$ make runtest
[  1%] Built target caffeproto
[ 69%] Built target caffe
[ 70%] Built target gtest
[100%] Built target test.testbin
Setting to use device 0
Note: Randomizing tests' orders with a seed of 96580 .
[==========] Running 2100 tests from 276 test cases.
[----------] Global test environment set-up.
[----------] 26 tests from NetTest/3, where TypeParam = caffe::GPUDevice<double>
...
[ RUN      ] NetTest/3.TestReshape

Hi, @legokichi , it seems I met the same problem as you.
The simple phenomenon occurs when I run cmake:

`CMake Warning (dev) in src/caffe/CMakeLists.txt:
Policy CMP0022 is not set: INTERFACE_LINK_LIBRARIES defines the link
interface. Run "cmake --help-policy CMP0022" for policy details. Use the
cmake_policy command to set the policy and suppress this warning.

Target "caffe" has an INTERFACE_LINK_LIBRARIES property which differs from
its LINK_INTERFACE_LIBRARIES properties.

INTERFACE_LINK_LIBRARIES:

caffeproto;/home/liuxiang/anaconda2/lib/libboost_system.so;/home/liuxiang/anaconda2/lib/libboost_thread.so;/home/liuxiang/anaconda2/lib/libboost_filesystem.so;/home/liuxiang/anaconda2/lib/libboost_chrono.so;/home/liuxiang/anaconda2/lib/libboost_date_time.so;/home/liuxiang/anaconda2/lib/libboost_atomic.so;/usr/lib/x86_64-linux-gnu/libpthread.so;/home/liuxiang/anaconda2/lib/libglog.so;/home/liuxiang/anaconda2/lib/libgflags.so;$<$<NOT:$<CONFIG:DEBUG>>:/usr/lib/x86_64-linux-gnu/libprotobuf.so>;$<$<CONFIG:DEBUG>:/usr/lib/x86_64-linux-gnu/libprotobuf.so>;-lpthread;/home/liuxiang/anaconda2/lib/libhdf5_hl.so;/home/liuxiang/anaconda2/lib/libhdf5.so;/home/liuxiang/anaconda2/lib/libhdf5_hl.so;/home/liuxiang/anaconda2/lib/libhdf5.so;/usr/lib/x86_64-linux-gnu/librt.so;/usr/lib/x86_64-linux-gnu/libpthread.so;/home/liuxiang/anaconda2/lib/libz.so;/usr/lib/x86_64-linux-gnu/libdl.so;/usr/lib/x86_64-linux-gnu/libm.so;/home/liuxiang/anaconda2/lib/libhdf5_hl.so;/home/liuxiang/anaconda2/lib/libhdf5.so;/home/liuxiang/anaconda2/lib/libhdf5_hl.so;/home/liuxiang/anaconda2/lib/libhdf5.so;/usr/lib/x86_64-linux-gnu/librt.so;/usr/lib/x86_64-linux-gnu/libpthread.so;/home/liuxiang/anaconda2/lib/libz.so;/usr/lib/x86_64-linux-gnu/libdl.so;/usr/lib/x86_64-linux-gnu/libm.so;/home/liuxiang/anaconda2/lib/liblmdb.so;/home/liuxiang/anaconda2/lib/libleveldb.so;/usr/local/cuda-8.0/lib64/libcudart.so;/usr/local/cuda-8.0/lib64/libcurand.so;/usr/local/cuda-8.0/lib64/libcublas.so;/usr/local/cuda-8.0/lib64/libcudnn.so;opencv_core;opencv_highgui;opencv_imgproc;opencv_imgcodecs;/usr/lib/liblapack.so;/usr/lib/libcblas.so;/usr/lib/libatlas.so;/home/liuxiang/anaconda2/lib/libboost_python.so

LINK_INTERFACE_LIBRARIES:

caffeproto;/home/liuxiang/anaconda2/lib/libboost_system.so;/home/liuxiang/anaconda2/lib/libboost_thread.so;/home/liuxiang/anaconda2/lib/libboost_filesystem.so;/home/liuxiang/anaconda2/lib/libboost_chrono.so;/home/liuxiang/anaconda2/lib/libboost_date_time.so;/home/liuxiang/anaconda2/lib/libboost_atomic.so;/usr/lib/x86_64-linux-gnu/libpthread.so;/home/liuxiang/anaconda2/lib/libglog.so;/home/liuxiang/anaconda2/lib/libgflags.so;/usr/lib/x86_64-linux-gnu/libprotobuf.so;-lpthread;/home/liuxiang/anaconda2/lib/libhdf5_hl.so;/home/liuxiang/anaconda2/lib/libhdf5.so;/home/liuxiang/anaconda2/lib/libhdf5_hl.so;/home/liuxiang/anaconda2/lib/libhdf5.so;/usr/lib/x86_64-linux-gnu/librt.so;/usr/lib/x86_64-linux-gnu/libpthread.so;/home/liuxiang/anaconda2/lib/libz.so;/usr/lib/x86_64-linux-gnu/libdl.so;/usr/lib/x86_64-linux-gnu/libm.so;/home/liuxiang/anaconda2/lib/libhdf5_hl.so;/home/liuxiang/anaconda2/lib/libhdf5.so;/home/liuxiang/anaconda2/lib/libhdf5_hl.so;/home/liuxiang/anaconda2/lib/libhdf5.so;/usr/lib/x86_64-linux-gnu/librt.so;/usr/lib/x86_64-linux-gnu/libpthread.so;/home/liuxiang/anaconda2/lib/libz.so;/usr/lib/x86_64-linux-gnu/libdl.so;/usr/lib/x86_64-linux-gnu/libm.so;/home/liuxiang/anaconda2/lib/liblmdb.so;/home/liuxiang/anaconda2/lib/libleveldb.so;/usr/local/cuda-8.0/lib64/libcudart.so;/usr/local/cuda-8.0/lib64/libcurand.so;/usr/local/cuda-8.0/lib64/libcublas.so;/usr/local/cuda-8.0/lib64/libcudnn.so;opencv_core;opencv_highgui;opencv_imgproc;opencv_imgcodecs;/usr/lib/liblapack.so;/usr/lib/libcblas.so;/usr/lib/libatlas.so;/home/liuxiang/anaconda2/lib/libboost_python.so

This warning is for project developers. Use -Wno-dev to suppress it.
`
I have done all I could following answers this page, yet them do not work.
So can you give me some advices?