[BUG] cuDF build error #1243

huangsongjue · 2019-03-20T02:55:55Z

cuDF (branch-0.7) build failed while make -j

Error Message:

/root/cudf/cpp/src/string/nvcategory_util.cpp: In function 'NVCategory* {anonymous}::combine_column_categories(gdf_column**, int)':
/root/cudf/cpp/src/string/nvcategory_util.cpp:19:50: error: 'class NVCategory' has no member named 'merge_and_remap'
combined_category = combined_category->merge_and_remap(
^
/root/cudf/cpp/src/string/nvcategory_util.cpp:28:35: error: 'class NVCategory' has no member named 'copy'
return combined_category->copy();
^
/root/cudf/cpp/src/string/nvcategory_util.cpp: In function 'gdf_error nvcategory_gather(gdf_column*, NVCategory*)':
/root/cudf/cpp/src/string/nvcategory_util.cpp:76:34: error: 'class NVCategory' has no member named 'add_keys_and_remap'
nv_category = nv_category->add_keys_and_remap(strs);
^
/root/cudf/cpp/src/string/nvcategory_util.cpp:109:44: error: 'class NVCategory' has no member named 'gather'
NVCategory * new_category = nv_category->gather(static_cast<nv_category_index_type >(column->data),
^
/root/cudf/cpp/src/string/nvcategory_util.cpp:117:44: error: invalid conversion from 'nv_category_index_type {aka int}' to 'unsigned int*' [-fpermissive]
DEVICE_ALLOCATED);
^
In file included from /root/cudf/cpp/src/string/nvcategory_util.cpp:5:0:
/root/cudf/cpp/build/include/nvstrings/NVCategory.h:94:9: note: initializing argument 1 of 'int NVCategory::get_values(unsigned int*, bool)'
int get_values( unsigned int* results, bool devmem=true );
^
/root/cudf/cpp/src/string/nvcategory_util.cpp: In function 'gdf_error concat_categories(gdf_column**, gdf_column*, int)':
/root/cudf/cpp/src/string/nvcategory_util.cpp:154:13: error: invalid conversion from 'nv_category_index_type* {aka int*}' to 'unsigned int*' [-fpermissive]
true);
^
In file included from /root/cudf/cpp/src/string/nvcategory_util.cpp:5:0:
/root/cudf/cpp/build/include/nvstrings/NVCategory.h:94:9: note: initializing argument 1 of 'int NVCategory::get_values(unsigned int*, bool)'
int get_values( unsigned int* results, bool devmem=true );
^
In file included from /root/cudf/cpp/include/types.hpp:19:0,
from /root/cudf/cpp/src/string/nvcategory_util.hpp:6,
from /root/cudf/cpp/src/string/nvcategory_util.cpp:2:
/root/cudf/cpp/src/string/nvcategory_util.cpp: In function 'gdf_error sync_column_categories(gdf_column**, gdf_column**, int)':
/root/cudf/cpp/src/string/nvcategory_util.cpp:185:46: error: 'class NVCategory' has no member named 'values_cptr'
combined_category->values_cptr() + current_column_start_position,
^
/root/cudf/cpp/src/utilities/error_utils.hpp:120:33: note: in definition of macro 'CUDA_TRY'
cudaError_t const status = (call);
^
/root/cudf/cpp/src/string/nvcategory_util.cpp:191:76: error: 'class NVCategory' has no member named 'copy'
output_columns[column_index]->dtype_info.category = combined_category->copy();
^
CMakeFiles/cudf.dir/build.make:101: recipe for target 'CMakeFiles/cudf.dir/src/string/nvcategory_util.cpp.o' failed

Steps/Code to reproduce bug
Follow this guide https://github.com/rapidsai/cudf/blob/branch-0.7/CONTRIBUTING.md#setting-up-your-build-environment, failed in the make -j stage.

Environment details (please complete the following information):
git*
commit abea679 (HEAD -> branch-0.7, origin/branch-0.7, origin/HEAD)
Merge: d5ca4b1 11dda6e
Author: gpuCI 38199262+GPUtester@users.noreply.github.com
Date: Mon Mar 18 16:39:33 2019 -0400

Merge pull request #1225 from rapidsai/branch-0.6

[gpuCI] Auto-merge branch-0.6 to branch-0.7 [skip ci]

OS Information
DISTRIB_ID=Ubuntu
DISTRIB_RELEASE=16.04
DISTRIB_CODENAME=xenial
DISTRIB_DESCRIPTION="Ubuntu 16.04.5 LTS"
NAME="Ubuntu"
VERSION="16.04.5 LTS (Xenial Xerus)"
ID=ubuntu
ID_LIKE=debian
PRETTY_NAME="Ubuntu 16.04.5 LTS"
VERSION_ID="16.04"
HOME_URL="http://www.ubuntu.com/"
SUPPORT_URL="http://help.ubuntu.com/"
BUG_REPORT_URL="http://bugs.launchpad.net/ubuntu/"
VERSION_CODENAME=xenial
UBUNTU_CODENAME=xenial
Linux 1edfd97799d5 3.10.0-862.14.4.el7.x86_64 #1 SMP Wed Sep 26 15:12:11 UTC 2018 x86_64 x86_64 x86_64 GNU/Linux

+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| No running processes found |
+-----------------------------------------------------------------------------+

CPU
Architecture: x86_64
CPU op-mode(s): 32-bit, 64-bit
Byte Order: Little Endian
CPU(s): 56
On-line CPU(s) list: 0-55
Thread(s) per core: 2
Core(s) per socket: 14
Socket(s): 2
NUMA node(s): 2
Vendor ID: GenuineIntel
CPU family: 6
Model: 85
Model name: Intel(R) Xeon(R) Gold 5120 CPU @ 2.20GHz
Stepping: 4
CPU MHz: 1000.000
CPU max MHz: 2201.0000
CPU min MHz: 1000.0000
BogoMIPS: 4405.23
Virtualization: VT-x
Hypervisor vendor: vertical
Virtualization type: full
L1d cache: 32K
L1i cache: 32K
L2 cache: 1024K
L3 cache: 19712K
NUMA node0 CPU(s): 0-13,28-41
NUMA node1 CPU(s): 14-27,42-55
Flags: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 ss ht tm pbe syscall nx pdpe1gb rdtscp lm constant_tsc art arch_perfmon pebs bts rep_good nopl xtopology nonstop_tsc aperfmperf eagerfpu pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 sdbg fma cx16 xtpr pdcm pcid dca sse4_1 sse4_2 x2apic movbe popcnt tsc_deadline_timer aes xsave avx f16c rdrand lahf_lm abm 3dnowprefetch epb cat_l3 cdp_l3 intel_ppin intel_pt ssbd mba ibrs ibpb stibp tpr_shadow vnmi flexpriority ept vpid fsgsbase tsc_adjust bmi1 hle avx2 smep bmi2 erms invpcid rtm cqm mpx rdt_a avx512f avx512dq rdseed adx smap clflushopt clwb avx512cd avx512bw avx512vl xsaveopt xsavec xgetbv1 cqm_llc cqm_occup_llc cqm_mbm_total cqm_mbm_local dtherm ida arat pln pts pku ospke spec_ctrl intel_stibp flush_l1d

CMake
/conda/envs/cudf_dev/bin/cmake
cmake version 3.14.0

CMake suite maintained and supported by Kitware (kitware.com/cmake).

g++
/usr/bin/g++
g++ (Ubuntu 5.4.0-6ubuntu1~16.04.10) 5.4.0 20160609
Copyright (C) 2015 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

nvcc
/usr/local/cuda/bin/nvcc
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2018 NVIDIA Corporation
Built on Tue_Jun_12_23:07:04_CDT_2018
Cuda compilation tools, release 9.2, V9.2.148

Python
/conda/envs/cudf_dev/bin/python
Python 3.7.2

Environment Variables
PATH : /conda/envs/cudf_dev/bin:/usr/local/nvidia/bin:/usr/local/cuda/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin:/conda/bin
LD_LIBRARY_PATH : /usr/local/nvidia/lib:/usr/local/nvidia/lib64:/usr/local/cuda/lib64
NUMBAPRO_NVVM : /usr/local/cuda/nvvm/lib64/libnvvm.so
NUMBAPRO_LIBDEVICE : /usr/local/cuda/nvvm/libdevice
CONDA_PREFIX : /conda/envs/cudf_dev
PYTHON_PATH :

conda packages
/conda/bin/conda

packages in environment at /conda/envs/cudf_dev:

Additional context
Add any other context about the problem here.

The text was updated successfully, but these errors were encountered:

harrism · 2019-03-20T02:59:40Z

The problem is probably that you don't have the latest nvstrings installed. Can you please activate your conda environment and run conda install -c nvidia/label/cuda10.0 -c rapidsai/label/cuda10.0 -c numba -c conda-forge -c defaults nvstrings=0.3 python=3.7?

After that, remove your build directory and rebuild.

We are working on updating the documentation and conda environment configuration to make this smoother.

eyalroz · 2019-03-20T08:50:43Z

Also encountered this bug just now.

@harrism : I'm manually downloaded and installed the latest nvstrings available on anaconda.org - but it's not 0.3; there is no version 0.3 there. 0.2 is the latest.

j-ieong · 2019-03-20T17:20:28Z

@eyalroz It seems that nvstrings-0.3 is under the rapidsai channel, rather than nvidia channel. I don't know why.
Additionally, the default uses CUDA9.2 so I had to manually install the CUDA10 version. https://anaconda.org/rapidsai/nvstrings

eyalroz · 2019-03-20T18:48:55Z

@j-ieong : Thanks, will check. However, CMakeLists.txt must check for the appropriate minimum version, and the conda environment should take care of actually getting the appropriate version.
...
So, apparently, at that link, there are only binaries for Python versions 3.6 and later, while before there were binaries for 3.5 and later. And since my distribution's Python version is 3.5.3, all of my conda environment has defaulted to Python-3.5-compatible packages. Trying to conda install python=3.5.6 to see if that works.

eyalroz · 2019-03-20T19:17:37Z

@j-ieong : The nvstrings package binaries are broken - at least for CUDA 9.2 and Python 3.6:

$ conda install -c rapidsai/label/cuda9.2 nvstrings
... etc. etc.. 

Downloading and Extracting Packages
nvstrings-0.3.0      | 14.8 MB   | ####################################################################################################################################### | 100% 
CondaFileIOError: '/home/eyalroz/miniconda2/pkgs/nvstrings-0.3.0-cuda9.2_py36_18.tar.bz2'. contains unsafe path: lib/libNVCategory.so

eyalroz · 2019-03-20T19:18:31Z

@huangsongjue : Would you mind editing your opening comment so as to properly format the quoted text? Put it within ``` ``` marks please.

harrism · 2019-03-20T20:16:26Z

@eyalroz cuDF requires Python 3.6 or later: https://github.com/rapidsai/cudf/tree/branch-0.6#customizing-the-build

eyalroz · 2019-03-20T20:23:25Z

@eyalroz cuDF requires Python 3.6 or later: https://github.com/rapidsai/cudf/tree/branch-0.6#customizing-the-build

Then CMake should fail the build configuration when one only has Python 3.5.x . Also - the error message I quoted (about the CondaFileIOError is for when I've installed Python 3.6 (in my conda environment).

j-ieong · 2019-03-20T20:43:09Z

@eyalroz Did you update conda recently? I had that message when I updated conda prior to installing. I rebooted my system, tried again and that went away.

If that doesn't work, @vuule had to modify conda itself to get past it.

eyalroz · 2019-03-20T20:44:47Z

@j-ieong : I don't reboot my system unless it hangs or I need to update the kernel. But I'm trying a further update which @harrism suggested to me in bug #1248 , and that might work.

j-ieong · 2019-03-20T20:47:50Z

@eyalroz Yeah, understandable. For me, that was the fastest way to get past it after trying several things.

eyalroz · 2019-03-20T20:55:23Z

Ok, this solved things for me:

conda install -c nvidia/label/cuda10.0 -c rapidsai/label/cuda10.0 -c numba -c conda-forge -c defaults nvstrings=0.3 python=3.7

or replace 10.0 with 9.2 depending on your version of CUDA.

I still feel CMake should do more version checking.

harrism · 2019-03-20T21:38:30Z

Possibly. Maybe file an issue for that.

huangsongjue · 2019-03-21T06:09:59Z

The problem is probably that you don't have the latest nvstrings installed. Can you please activate your conda environment and run conda install -c nvidia/label/cuda10.0 -c rapidsai/label/cuda10.0 -c numba -c conda-forge -c defaults nvstrings=0.3 python=3.7?

After that, remove your build directory and rebuild.

We are working on updating the documentation and conda environment configuration to make this smoother.

@harrism thanks, it works, and I agreed with @eyalroz cmake should do the vesion check.

harrism · 2019-03-22T00:12:22Z

Note that PR #1252 should fix this, if you use the conda/envs/cudf_dev*.yml files to configure your environment.

atlasgurus · 2019-03-22T21:53:14Z

Thank you @harrism . I see that #1252 is already fixed in branch-0.6, so I tried it as follows and still got similar build errors. Is there a stable branch that I can use to build it?
git checkout branch-0.6
conda env create --name cudf_dev3 --file conda/environments/cudf_dev_cuda9.2.yml
source activate cudf_dev3
cd cpp/build
cmake .. -DCMAKE_INSTALL_PREFIX=/install/path -DCMAKE_BUILD_TYPE=Debug
make -j
......
/home/ubuntu/projects/cudf/cpp/src/string/nvcategory_util.cpp: In function ‘gdf_error concat_categories(gdf_column**, gdf_column*, int)’:
/home/ubuntu/projects/cudf/cpp/src/string/nvcategory_util.cpp:153:9: error: invalid conversion from ‘nv_category_index_type* {aka int*}’ to ‘unsigned int*’ [-fpermissive]
static_cast<nv_category_index_type >(output_column->data),
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In file included from /home/ubuntu/projects/cudf/cpp/src/string/nvcategory_util.cpp:5:0:
/home/ubuntu/projects/cudf/cpp/build/include/nvstrings/NVCategory.h:94:9: note: initializing argument 1 of ‘int NVCategory::get_values(unsigned int, bool)’
int get_values( unsigned int* results, bool devmem=true );
.....

atlasgurus · 2019-03-22T21:57:45Z

I have run the command below too for goo measure

conda install -c nvidia/label/cuda9.2 -c rapidsai/label/cuda9.2 -c numba -c conda-forge -c defaults nvstrings=0.3 python=3.7

but still got the same errors as listed in this issue:
/home/ubuntu/projects/cudf/cpp/src/string/nvcategory_util.cpp: In function ‘NVCategory* {anonymous}::combine_column_categories(gdf_column**, int)’: /home/ubuntu/projects/cudf/cpp/src/string/nvcategory_util.cpp:19:50: error: ‘class NVCategory’ has no member named ‘merge_and_remap’ combined_category = combined_category->merge_and_remap( ^~~~~~~~~~~~~~~ /home/ubuntu/projects/cudf/cpp/src/string/nvcategory_util.cpp:28:35: error: ‘class NVCategory’ has no member named ‘copy’ return combined_category->copy();

eyalroz · 2019-03-22T21:59:40Z

@atlasgurus : Did you delete your build directory, recreate it and rebuilt?

atlasgurus · 2019-03-22T23:23:52Z

@eyalroz , thank you very much for the suggestion. It did help. I think I am almost there. It now fails during link step:

/usr/bin/x86_64-linux-gnu-ld: warning: libcudart.so.9.2, needed by /home/ubuntu/projects/cudf/cpp/build/lib/libNVCategory.so, not found (try using -rpath or -rpath-link) /home/ubuntu/projects/cudf/cpp/build/lib/libNVCategory.so: undefined reference to cudaLaunchKernel@libcudart.so.9.2'
/home/ubuntu/projects/cudf/cpp/build/lib/libNVCategory.so: undefined reference to cudaMemcpy@libcudart.so.9.2' /home/ubuntu/projects/cudf/cpp/build/lib/libNVCategory.so: undefined reference to cudaFree@libcudart.so.9.2'
/home/ubuntu/projects/cudf/cpp/build/lib/libNVCategory.so: undefined reference to __cudaRegisterFatBinary@libcudart.so.9.2' /home/ubuntu/projects/cudf/cpp/build/lib/libNVCategory.so: undefined reference to __cudaRegisterVar@libcudart.so.9.2'
`

eyalroz · 2019-03-22T23:37:09Z

@atlasgurus : Something is messed up with the directories in which the linker is trying to find libraries. I suggest you open a separate issue, and that you provide a full command-line of a failing link command with the -L flags.

harrism · 2019-03-24T22:08:00Z

@atlasgurus Are you sure you have CUDA 9.2 installed and not CUDA 10? It's possible that you have a version mismatch here.

huangsongjue added Needs Triage Need team to review and classify bug Something isn't working labels Mar 20, 2019

eyalroz mentioned this issue Mar 20, 2019

[BUG] branch-0.6 fails its HASH_TEST in Debug build type #1189

Closed

harrism closed this as completed Mar 22, 2019

eyalroz mentioned this issue Mar 22, 2019

[BUG] CMake build config should check for versions of dependency libs (nvstrings etc.) #1262

Closed

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[BUG] cuDF build error #1243

[BUG] cuDF build error #1243

huangsongjue commented Mar 20, 2019

harrism commented Mar 20, 2019

eyalroz commented Mar 20, 2019

j-ieong commented Mar 20, 2019

eyalroz commented Mar 20, 2019 •

edited

Loading

eyalroz commented Mar 20, 2019

eyalroz commented Mar 20, 2019

harrism commented Mar 20, 2019

eyalroz commented Mar 20, 2019 •

edited

Loading

j-ieong commented Mar 20, 2019

eyalroz commented Mar 20, 2019

j-ieong commented Mar 20, 2019

eyalroz commented Mar 20, 2019

harrism commented Mar 20, 2019

huangsongjue commented Mar 21, 2019

harrism commented Mar 22, 2019

atlasgurus commented Mar 22, 2019

atlasgurus commented Mar 22, 2019

eyalroz commented Mar 22, 2019

atlasgurus commented Mar 22, 2019

eyalroz commented Mar 22, 2019

harrism commented Mar 24, 2019 •

edited

Loading

[BUG] cuDF build error #1243

[BUG] cuDF build error #1243

Comments

huangsongjue commented Mar 20, 2019

packages in environment at /conda/envs/cudf_dev:

Name Version Build Channel

harrism commented Mar 20, 2019

eyalroz commented Mar 20, 2019

j-ieong commented Mar 20, 2019

eyalroz commented Mar 20, 2019 • edited Loading

eyalroz commented Mar 20, 2019

eyalroz commented Mar 20, 2019

harrism commented Mar 20, 2019

eyalroz commented Mar 20, 2019 • edited Loading

j-ieong commented Mar 20, 2019

eyalroz commented Mar 20, 2019

j-ieong commented Mar 20, 2019

eyalroz commented Mar 20, 2019

harrism commented Mar 20, 2019

huangsongjue commented Mar 21, 2019

harrism commented Mar 22, 2019

atlasgurus commented Mar 22, 2019

atlasgurus commented Mar 22, 2019

eyalroz commented Mar 22, 2019

atlasgurus commented Mar 22, 2019

eyalroz commented Mar 22, 2019

harrism commented Mar 24, 2019 • edited Loading

eyalroz commented Mar 20, 2019 •

edited

Loading

eyalroz commented Mar 20, 2019 •

edited

Loading

harrism commented Mar 24, 2019 •

edited

Loading