Install pytorch commit fd25a2a

[junfu1115]

Install pytorch commit fd25a2a

[huanghoujing]

There are some environment variables in setup.py to set for customizing your CUDA, cudnn, NCCL paths. Thanks to God, it is in Python logic. The example in my case:

1. Install CUDA 8.0 and cudnn 7
2. Install Anaconda 3
3. Install Pytorch from source

# Install certain version of pytorch from source

INSTALL_PYTORCH_SCRIPT_DIR=${HOME}/Software/install_pytorch

pip uninstall -y torch

# Install basic dependencies
conda install --yes numpy pyyaml mkl mkl-include setuptools cmake cffi typing
conda install --yes -c mingfeima mkldnn
# Add LAPACK support for the GPU
conda install --yes -c pytorch magma-cuda80 # or magma-cuda90 if CUDA 9

cd ${HOME}/Software
git clone --recursive https://github.com/pytorch/pytorch
cd pytorch
git checkout fd25a2a
git submodule init # There may be submodules no longer existing in new commits
git submodule update --recursive

rm -rf build
rm -rf torch.egg-info
export CUDA_HOME=/mnt/data-1/data/houjing.huang/Software/cuda-8.0
export USE_SYSTEM_NCCL=1
export NCCL_LIB_DIR=${CUDA_HOME}/lib64
export NCCL_INCLUDE_DIR=${CUDA_HOME}/include
export CMAKE_PREFIX_PATH="$(dirname $(which conda))/../" # [anaconda root directory]

python setup.py install 2>&1 | tee ${INSTALL_PYTORCH_SCRIPT_DIR}/install_pytorch.log
cd ${INSTALL_PYTORCH_SCRIPT_DIR}
python test_data_parallel.py 2>&1 | tee test_pytorch_data_parallel.log

# test_data_parallel.py
import torch
import torch.nn as nn
from torch.nn.parallel import DataParallel

model = nn.Linear(10, 20).cuda()
x = torch.ones(100, 10).float().cuda()
model_w = DataParallel(model, device_ids=[0,1,2,3])
x = model_w(x)
# x = model(x)
print(x.size())

[junfu1115]

Thanks, we have corrected it. c20e4b8

[sbharadwajj]

This is the result I got for cityscapes dataset without multiscale:

Mean IoU over 19 classes: 0.6273
Pixel-wise Accuracy: 91.79%

[huanghoujing]

@chichilicious Test the provided model on val set, I got pixAcc: 0.9596, mIoU: 0.7991: ...125/125 without multiscale.

[sbharadwajj]

That is exactly what I have done. I have tested it on the val folder only. I have gt also since its cityscapes dataset.

[sbharadwajj]

But i had to modify this part in the eval_batch as it did not work properly and i got error.

def eval_batch(image, dst, evaluator, eval_mode):
        if eval_mode:
            # evaluation mode on validation set
            targets = dst
            outputs = evaluator.parallel_forward(image)
            
            batch_inter, batch_union, batch_correct, batch_label = 0, 0, 0, 0
            for output in outputs:
                for out, target in zip(output, targets):                
                    correct, labeled = utils.batch_pix_accuracy(out.data.cpu(), target)
                    inter, union = utils.batch_intersection_union(
                        out.data.cpu(), target, testset.num_class)

[huanghoujing]

I use DANet commit f34f8fd and pytorch commit fd25a2a and I did not modify codes of this project. The only problem is args.test_scale, but it does not affect printing the scores during testing.

What error do you have if you do not modify this part:

            for output in outputs:
                for out, target in zip(output, targets): 

[sbharadwajj]

correct, labeled = utils.batch_pix_accuracy(output.data.cpu(), target)
AttributeError: 'list' object has no attribute 'data'

This is the error.
But I just realized for some reason my performance of test.py is not up to the mark. I just visualized the results and they are not great. And this visualization is for the unmodified code. Will the torch version have an effect on this? When my code runs it always executes this part

with torch.no_grad():
                correct, labeled, inter, union = eval_batch(image, dst, evaluator, args.eval)

[huanghoujing]

I think torch version is a matter. Try to follow the same version to avoid invisible inconsistency.

[sbharadwajj]

@junfu1115 will the performance vary with the pytorch version and by installing it with source/pip?
I have done pip install for pytorch.

[sbharadwajj]

@huanghoujing while building torch from source, did you follow https://github.com/pytorch/pytorch/tree/fd25a2a86c6afa93c7062781d013ad5f41e0504b#from-source?I'm getting a few errors while building it, so just wanted to confirm.

[huanghoujing]

Yes, I follow the README of that certain commit of Pytorch.

[junfu1115]

@chichilicious We don't try other pytorch version, you can install the pytorch commit fd25a2a.

[sbharadwajj]

Hi @huanghoujing I don't use conda as I'm running this on cloud. But i will try normal install and if that doesn't go well, I'll try the conda one.
@junfu1115 , I will update the results once I run it on the other pytorch version.

[huanghoujing]

@chichilicious I just found another problem in your procedure. You said you use this repository https://github.com/freedrone/cityscapes-scripts to transform cityscapes annotations, which comes with a bug. You can look at what's been changed in line 97 of cityscapesscripts/helpers/labels.py. This modified version changes license plate id from -1 to 34. Then look at cityscapesscripts/preparation/json2labelImg.py line 105 and line 119. The region of license plate would be ignore in the official labels.py, but in the modified version, it would be filled with value -1 which in fact would eventually be 0! The following PIL example would explain why -1 would become 0 in the drawn image. trainID=0 is for class road. As a result, your generated label ground truth image would not be correct for road.

Example of: Drawing with value -1 eventually draws 0:

import PIL.Image as Image
import PIL.ImageDraw as ImageDraw
import numpy as np

# im_size is (w, h)
im = Image.new("L", (16, 16), 255)
drawer = ImageDraw.Draw(im)
# a list of (x, y)
polygon = [(2, 2), (4, 4), (3, 7)]
drawer.polygon(polygon, fill=-1)
im = np.array(im)
# The region filled with -1 eventually has value 0. In the official code,
# this region should have value 255, i.e. `background`.
# Class `road` has trainID 0, so the evaluation result for `road` is not correct.
print(im)

You can now regenerate your cityscape annotation images with the correct labels.py and re-evaluate your results. If it solves the problem, it would be great. BTW, I found the official labels.py and the one used by DANet have the same contents.

[sbharadwajj]

@huanghoujing I will check this out. Maybe this is the error.
But this is an example of my output image, are you getting a similar result or a better one? Because if you are getting a better result, then I will have to try the pytorch installation fix aswell.

[huanghoujing]

@chichilicious I have much better visual results. For this frankfurt_000000_000576_leftImg8bit.png

Other examples:
frankfurt_000000_000294_leftImg8bit.png:

frankfurt_000000_007365_leftImg8bit.png:

frankfurt_000000_009291_leftImg8bit.png:

[sbharadwajj]

@huanghoujing I successfully installed that pytorch commit and ran setup.py again in my directory after removing the old build& torch* file.
But my results are still the same. Any other way to debug this?
I git cloned this repository again and did not change any part of the code.

This is my git status for pytorch:
HEAD detached at fd25a2a

But for some reason when I ran setup.py in the DAnet dir, this is what I got.:

running install
-- Building version 0.4.5+c20e4b8
running build
running build_py

This is another different thing that happens and my model runs on the next loop.

MultiEvalModule: base_size 608, crop_size 768
  0%|                                                                              | 0/500 [00:00<?, ?it/s]/datadrive/virtualenvs/torchDA/lib/python3.6/site-packages/torch/nn/functional.py:1909: UserWarning: nn.functional.upsample is deprecated. Use nn.functional.interpolate instead.
  warnings.warn("nn.functional.upsample is deprecated. Use nn.functional.interpolate instead.")

Result for:

python3.6 -c "import torch; print(torch.__version__)"
>>0.5.0a0+fd25a2a

[sbharadwajj]

@huanghoujing Thanks for all the help. It works now. I'm getting around 79+ mIoU.
@junfu1115 what is the time taken for running inference for each image and on what GPU? Mine is taking 14secs to run each image on k80.

[zebinc0409]

when compile the project as you discuss above,error stills occur.it just cannot find libmkldnn.so.0 and libnccl.so.1,and i found in other places and link to /usr/lib,error shows that no rule to make target libmkldnn.so.0,can you tell me what happen. environment:cudnn7 cuda8.0 conda3.6

[huanghoujing]

@mysweetie I can find libmkldnn.so in following places

${Anaconda_HOME}/pkgs/mkldnn-0.13.0-0/lib/libmkldnn.so.0.13.0
${Anaconda_HOME}/pkgs/mkldnn-0.13.0-0/lib/libmkldnn.so.0
${Anaconda_HOME}/pkgs/mkldnn-0.13.0-0/lib/libmkldnn.so
${Anaconda_HOME}/lib/libmkldnn.so
${Anaconda_HOME}/lib/libmkldnn.so.0
${Anaconda_HOME}/lib/libmkldnn.so.0.13.0

and libnccl.so in following places

${CUDA_HOME}/lib64/libnccl.so
${CUDA_HOME}/lib64/libnccl.so.2
${CUDA_HOME}/lib64/libnccl.so.2.1.4

[zebinc0409]

I can find libmkdlnn.so series but not libnccl.so.

[zebinc0409]

Does libnccl.so produce by during compiling cuda？after I linked libmkldnn.so to /usr/lib，it said no rule to make it,maybe it is the problem of gcc version or cmake？can you tell me some more details about it
@huanghoujing

[huanghoujing]

@mysweetie

1. My NCCL lib files are installed by cuda-8.0 automatically. CMake is specified by export CMAKE_PREFIX_PATH="$(dirname $(which conda))/../". The following is needed, do you export them?

export CUDA_HOME=/mnt/data-1/data/houjing.huang/Software/cuda-8.0
export USE_SYSTEM_NCCL=1
export NCCL_LIB_DIR=${CUDA_HOME}/lib64
export NCCL_INCLUDE_DIR=${CUDA_HOME}/include
export CMAKE_PREFIX_PATH="$(dirname $(which conda))/../" # [anaconda root directory]

2. You can't find ${CUDA_HOME}/lib64/libnccl.so.* after installing CUDA?

[zebinc0409]

@huanghoujing
1.there is no nccl lib, and i compile it just now as follow:
git clone https://github.com/NVIDIA/nccl.git
cd nccl
export CUDA_HOME=/usr/local/cuda-8.0
make
2.the problem changed:
[ 96%] Building CXX object caffe2/CMakeFiles/scalar_test.dir/__/aten/src/ATen/test/scalar_test.cpp.o Scanning dependencies of target cuda_rng_test [ 96%] Building CXX object caffe2/CMakeFiles/cuda_rng_test.dir/__/aten/src/ATen/test/cuda_rng_test.cpp.o /usr/bin/ld: /usr/local/cuda-8.0/lib64/libcudart_static.a(libcudart_static.a.o): undefined reference to symbol 'shm_unlink@@GLIBC_2.2.5' //lib/x86_64-linux-gnu/librt.so.1: error adding symbols: DSO missing from command line collect2: error: ld returned 1 exit status caffe2/CMakeFiles/integer_divider_test.dir/build.make:397: recipe for target 'bin/integer_divider_test' failed make[2]: *** [bin/integer_divider_test] Error 1 CMakeFiles/Makefile2:933: recipe for target 'caffe2/CMakeFiles/integer_divider_test.dir/all' failed make[1]: *** [caffe2/CMakeFiles/integer_divider_test.dir/all] Error 2 make[1]: *** Waiting for unfinished jobs.... [ 96%] Linking CXX executable ../bin/scalar_test [ 96%] Built target scalar_test [ 96%] Linking CXX executable ../bin/cuda_rng_test [ 96%] Linking CXX executable ../bin/test_parallel /usr/bin/ld: /usr/local/cuda-8.0/lib64/libcudart_static.a(libcudart_static.a.o): undefined reference to symbol 'shm_unlink@@GLIBC_2.2.5' //lib/x86_64-linux-gnu/librt.so.1: error adding symbols: DSO missing from command line collect2: error: ld returned 1 exit status caffe2/CMakeFiles/cuda_rng_test.dir/build.make:92: recipe for target 'bin/cuda_rng_test' failed make[2]: *** [bin/cuda_rng_test] Error 1 CMakeFiles/Makefile2:1393: recipe for target 'caffe2/CMakeFiles/cuda_rng_test.dir/all' failed make[1]: *** [caffe2/CMakeFiles/cuda_rng_test.dir/all] Error 2 [ 96%] Built target test_parallel [ 97%] Linking CXX executable ../bin/dlconvertor_test [ 98%] Linking CXX executable ../bin/wrapdim_test [ 98%] Built target dlconvertor_test [ 98%] Built target wrapdim_test [ 99%] Linking CXX executable ../bin/native_test [ 99%] Built target native_test [ 99%] Linking CXX executable ../bin/apply_utils_test [ 99%] Built target apply_utils_test Makefile:140: recipe for target 'all' failed make: *** [all] Error 2 Failed to run 'bash tools/build_pytorch_libs.sh --use-cuda --use-nnpack --use-mkldnn nccl caffe2 nanopb libshm gloo THD c10d'

[zebinc0409]

i also have export those environment vars

[huanghoujing]

You can simply disable distributed and caffe2 to save time and avoid unnecessary errors. So also export these:

export USE_DISTRIBUTED=0
export BUILD_CAFFE2_OPS=0

[huanghoujing]

Also look at the Summary in the log during installing pytorch, where you can find what paths it uses for finding packages.

[emma-sjwang]

Bugs.

The error logs are below:
error: command 'gcc' failed with exit status 1

`Performing C++ SOURCE FILE Test CAFFE2_NEED_TO_TURN_OFF_DEPRECATION_WARNING failed with the following output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_b723d/fast
/usr/bin/gmake -f CMakeFiles/cmTC_b723d.dir/build.make CMakeFiles/cmTC_b723d.dir/build
gmake[1]: Entering directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp' Building CXX object CMakeFiles/cmTC_b723d.dir/src.cxx.o /usr/bin/c++ -DCAFFE2_NEED_TO_TURN_OFF_DEPRECATION_WARNING -std=c++11 -o CMakeFiles/cmTC_b723d.dir/src.cxx.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.cxx In file included from /usr/include/c++/4.8.2/ext/hash_set:60:0, from /usr/include/glog/stl_logging.h:54, from /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.cxx:1: /usr/include/c++/4.8.2/backward/backward_warning.h:32:2: warning: #warning This file includes at least one deprecated or antiquated header which may be removed without further notice at a future date. Please use a non-deprecated interface with equivalent functionality instead. For a listing of replacement headers and interfaces, consult the file backward_warning.h. To disable this warning use -Wno-deprecated. [-Wcpp] #warning \ ^ Linking CXX executable cmTC_b723d /research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/bin/cmake -E cmake_link_script CMakeFiles/cmTC_b723d.dir/link.txt --verbose=1 /usr/bin/c++ -DCAFFE2_NEED_TO_TURN_OFF_DEPRECATION_WARNING -std=c++11 -L"/research/pheng4/sjwang/program/accumulation/pytorch/torch/lib/tmp_install/lib" -Wl,-rpath,$ORIGIN CMakeFiles/cmTC_b723d.dir/src.cxx.o -o cmTC_b723d gmake[1]: Leaving directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'

Source file was:
#include <glog/stl_logging.h>
int main(int argc, char** argv) {
return 0;
}
Determining if the pthread_create exist failed with the following output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_b2372/fast
/usr/bin/gmake -f CMakeFiles/cmTC_b2372.dir/build.make CMakeFiles/cmTC_b2372.dir/build
gmake[1]: Entering directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp' Building C object CMakeFiles/cmTC_b2372.dir/CheckSymbolExists.c.o /usr/bin/cc -fPIE -o CMakeFiles/cmTC_b2372.dir/CheckSymbolExists.c.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/CheckSymbolExists.c Linking C executable cmTC_b2372 /research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/bin/cmake -E cmake_link_script CMakeFiles/cmTC_b2372.dir/link.txt --verbose=1 /usr/bin/cc -rdynamic CMakeFiles/cmTC_b2372.dir/CheckSymbolExists.c.o -o cmTC_b2372 CMakeFiles/cmTC_b2372.dir/CheckSymbolExists.c.o: In function main':
CheckSymbolExists.c:(.text+0x1b): undefined reference to pthread_create' collect2: error: ld returned 1 exit status gmake[1]: *** [cmTC_b2372] Error 1 gmake[1]: Leaving directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
gmake: *** [cmTC_b2372/fast] Error 2

File /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/CheckSymbolExists.c:
/* */
#include <pthread.h>

int main(int argc, char** argv)
{
(void)argv;
#ifndef pthread_create
return ((int*)(&pthread_create))[argc];
#else
(void)argc;
return 0;
#endif
}

Determining if the function pthread_create exists in the pthreads failed with the following output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_75dd3/fast
/usr/bin/gmake -f CMakeFiles/cmTC_75dd3.dir/build.make CMakeFiles/cmTC_75dd3.dir/build
gmake[1]: Entering directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp' Building C object CMakeFiles/cmTC_75dd3.dir/CheckFunctionExists.c.o /usr/bin/cc -DCHECK_FUNCTION_EXISTS=pthread_create -fPIE -o CMakeFiles/cmTC_75dd3.dir/CheckFunctionExists.c.o -c /research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/share/cmake-3.14/Modules/CheckFunctionExists.c Linking C executable cmTC_75dd3 /research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/bin/cmake -E cmake_link_script CMakeFiles/cmTC_75dd3.dir/link.txt --verbose=1 /usr/bin/cc -DCHECK_FUNCTION_EXISTS=pthread_create -rdynamic CMakeFiles/cmTC_75dd3.dir/CheckFunctionExists.c.o -o cmTC_75dd3 -lpthreads /usr/bin/ld: cannot find -lpthreads collect2: error: ld returned 1 exit status gmake[1]: *** [cmTC_75dd3] Error 1 gmake[1]: Leaving directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
gmake: *** [cmTC_75dd3/fast] Error 2

Performing C SOURCE FILE Test C_HAS_SSE3_1 failed with the following compile output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_93f23/fast
/usr/bin/gmake -f CMakeFiles/cmTC_93f23.dir/build.make CMakeFiles/cmTC_93f23.dir/build
gmake[1]: Entering directory `/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
Building C object CMakeFiles/cmTC_93f23.dir/src.c.o
/usr/bin/cc -I/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/include -I/usr/local/cuda-9.0/include -fopenmp -DC_HAS_SSE3_1 -fPIE -o CMakeFiles/cmTC_93f23.dir/src.c.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c
In file included from /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:2:0:
/usr/lib/gcc/x86_64-redhat-linux/4.8.5/include/pmmintrin.h:31:3: error: #error "SSE3 instruction set not enabled"

error "SSE3 instruction set not enabled"

^
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c: In function ‘main’:
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:7:5: error: unknown type name ‘__m128i’
__m128i a;
^
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:8:5: error: unknown type name ‘__m128i’
a = _mm_lddqu_si128( (const __m128i*)vals );
^
gmake[1]: *** [CMakeFiles/cmTC_93f23.dir/src.c.o] Error 1
gmake[1]: Leaving directory `/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
gmake: *** [cmTC_93f23/fast] Error 2

...and run output:

Return value: 1
Source file was:

#include <pmmintrin.h>

int main( )
{
const int vals[4] = {0,0,0,0};
__m128i a;
a = _mm_lddqu_si128( (const __m128i*)vals );
return 0;
}
Performing C SOURCE FILE Test C_HAS_SSE4_1_1 failed with the following compile output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_c81ba/fast
/usr/bin/gmake -f CMakeFiles/cmTC_c81ba.dir/build.make CMakeFiles/cmTC_c81ba.dir/build
gmake[1]: Entering directory `/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
Building C object CMakeFiles/cmTC_c81ba.dir/src.c.o
/usr/bin/cc -I/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/include -I/usr/local/cuda-9.0/include -fopenmp -DC_HAS_SSE4_1_1 -fPIE -o CMakeFiles/cmTC_c81ba.dir/src.c.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c
In file included from /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:2:0:
/usr/lib/gcc/x86_64-redhat-linux/4.8.5/include/smmintrin.h:31:3: error: #error "SSE4.1 instruction set not enabled"

error "SSE4.1 instruction set not enabled"

^
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c: In function ‘main’:
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: error: unknown type name ‘__m128i’
__m128i a = {0,0,0,0}, b = {0,0,0,0};
^
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘a’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘a’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘a’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘b’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘b’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘b’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:7:5: error: unknown type name ‘__m128i’
__m128i res = _mm_max_epi8(a, b);
^
gmake[1]: *** [CMakeFiles/cmTC_c81ba.dir/src.c.o] Error 1
gmake[1]: Leaving directory `/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
gmake: *** [cmTC_c81ba/fast] Error 2

...and run output:

Return value: 1
Source file was:

#include <smmintrin.h>

int main ()
{
__m128i a = {0,0,0,0}, b = {0,0,0,0};
__m128i res = _mm_max_epi8(a, b);

return 0;

}

Performing C SOURCE FILE Test C_HAS_SSE4_2_1 failed with the following compile output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_c7f76/fast
/usr/bin/gmake -f CMakeFiles/cmTC_c7f76.dir/build.make CMakeFiles/cmTC_c7f76.dir/build
gmake[1]: Entering directory `/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
Building C object CMakeFiles/cmTC_c7f76.dir/src.c.o
/usr/bin/cc -I/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/include -I/usr/local/cuda-9.0/include -fopenmp -DC_HAS_SSE4_2_1 -fPIE -o CMakeFiles/cmTC_c7f76.dir/src.c.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c
In file included from /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:2:0:
/usr/lib/gcc/x86_64-redhat-linux/4.8.5/include/nmmintrin.h:31:3: error: #error "SSE4.2 instruction set not enabled"

error "SSE4.2 instruction set not enabled"

^
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c: In function ‘main’:
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: error: unknown type name ‘__m128i’
__m128i a = {0,0,0,0}, b = {0,0,0,0}, c = {0,0,0,0};
^
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘a’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘a’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘a’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘b’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘b’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘b’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘c’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘c’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘c’) [enabled by default]
gmake[1]: *** [CMakeFiles/cmTC_c7f76.dir/src.c.o] Error 1
gmake[1]: Leaving directory `/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
gmake: *** [cmTC_c7f76/fast] Error 2

...and run output:

Return value: 1
Source file was:

#include <nmmintrin.h>

int main()
{
__m128i a = {0,0,0,0}, b = {0,0,0,0}, c = {0,0,0,0};
c = _mm_cmpgt_epi64(a, b);
return 0;
}

Performing C SOURCE FILE Test C_HAS_AVX_1 failed with the following compile output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_f6eae/fast
/usr/bin/gmake -f CMakeFiles/cmTC_f6eae.dir/build.make CMakeFiles/cmTC_f6eae.dir/build
gmake[1]: Entering directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp' Building C object CMakeFiles/cmTC_f6eae.dir/src.c.o /usr/bin/cc -I/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/include -I/usr/local/cuda-9.0/include -fopenmp -DC_HAS_AVX_1 -fPIE -o CMakeFiles/cmTC_f6eae.dir/src.c.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c: In function ‘main’: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: error: unknown type name ‘__m256’ __m256 a; ^ gmake[1]: *** [CMakeFiles/cmTC_f6eae.dir/src.c.o] Error 1 gmake[1]: Leaving directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
gmake: *** [cmTC_f6eae/fast] Error 2

...and run output:

Return value: 1
Source file was:

#include <immintrin.h>

int main()
{
__m256 a;
a = _mm256_set1_ps(0);
return 0;
}

Performing C SOURCE FILE Test C_HAS_AVX2_1 failed with the following compile output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_4b020/fast
/usr/bin/gmake -f CMakeFiles/cmTC_4b020.dir/build.make CMakeFiles/cmTC_4b020.dir/build
gmake[1]: Entering directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp' Building C object CMakeFiles/cmTC_4b020.dir/src.c.o /usr/bin/cc -I/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/include -I/usr/local/cuda-9.0/include -fopenmp -DC_HAS_AVX2_1 -fPIE -o CMakeFiles/cmTC_4b020.dir/src.c.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c: In function ‘main’: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: error: unknown type name ‘__m256i’ __m256i a = {0}; ^ gmake[1]: *** [CMakeFiles/cmTC_4b020.dir/src.c.o] Error 1 gmake[1]: Leaving directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
gmake: *** [cmTC_4b020/fast] Error 2

...and run output:

Return value: 1
Source file was:

#include <immintrin.h>

int main()
{
__m256i a = {0};
a = _mm256_abs_epi16(a);
return 0;
}

Performing C SOURCE FILE Test CXX_HAS_SSE3_1 failed with the following compile output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_797fe/fast
/usr/bin/gmake -f CMakeFiles/cmTC_797fe.dir/build.make CMakeFiles/cmTC_797fe.dir/build
gmake[1]: Entering directory `/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
Building C object CMakeFiles/cmTC_797fe.dir/src.c.o
/usr/bin/cc -I/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/include -I/usr/local/cuda-9.0/include -fopenmp -DCXX_HAS_SSE3_1 -fPIE -o CMakeFiles/cmTC_797fe.dir/src.c.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c
In file included from /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:2:0:
/usr/lib/gcc/x86_64-redhat-linux/4.8.5/include/pmmintrin.h:31:3: error: #error "SSE3 instruction set not enabled"

error "SSE3 instruction set not enabled"

^
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c: In function ‘main’:
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:7:5: error: unknown type name ‘__m128i’
__m128i a;
^
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:8:5: error: unknown type name ‘__m128i’
a = _mm_lddqu_si128( (const __m128i*)vals );
^
gmake[1]: *** [CMakeFiles/cmTC_797fe.dir/src.c.o] Error 1
gmake[1]: Leaving directory `/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
gmake: *** [cmTC_797fe/fast] Error 2

...and run output:

Return value: 1
Source file was:

#include <pmmintrin.h>

int main( )
{
const int vals[4] = {0,0,0,0};
__m128i a;
a = _mm_lddqu_si128( (const __m128i*)vals );
return 0;
}
Performing C SOURCE FILE Test CXX_HAS_SSE4_1_1 failed with the following compile output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_fa7a9/fast
/usr/bin/gmake -f CMakeFiles/cmTC_fa7a9.dir/build.make CMakeFiles/cmTC_fa7a9.dir/build
gmake[1]: Entering directory `/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
Building C object CMakeFiles/cmTC_fa7a9.dir/src.c.o
/usr/bin/cc -I/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/include -I/usr/local/cuda-9.0/include -fopenmp -DCXX_HAS_SSE4_1_1 -fPIE -o CMakeFiles/cmTC_fa7a9.dir/src.c.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c
In file included from /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:2:0:
/usr/lib/gcc/x86_64-redhat-linux/4.8.5/include/smmintrin.h:31:3: error: #error "SSE4.1 instruction set not enabled"

error "SSE4.1 instruction set not enabled"

^
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c: In function ‘main’:
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: error: unknown type name ‘__m128i’
__m128i a = {0,0,0,0}, b = {0,0,0,0};
^
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘a’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘a’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘a’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘b’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘b’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘b’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:7:5: error: unknown type name ‘__m128i’
__m128i res = _mm_max_epi8(a, b);
^
gmake[1]: *** [CMakeFiles/cmTC_fa7a9.dir/src.c.o] Error 1
gmake[1]: Leaving directory `/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
gmake: *** [cmTC_fa7a9/fast] Error 2

...and run output:

Return value: 1
Source file was:

#include <smmintrin.h>

int main ()
{
__m128i a = {0,0,0,0}, b = {0,0,0,0};
__m128i res = _mm_max_epi8(a, b);

return 0;

}

Performing C SOURCE FILE Test CXX_HAS_SSE4_2_1 failed with the following compile output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_e8755/fast
/usr/bin/gmake -f CMakeFiles/cmTC_e8755.dir/build.make CMakeFiles/cmTC_e8755.dir/build
gmake[1]: Entering directory `/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
Building C object CMakeFiles/cmTC_e8755.dir/src.c.o
/usr/bin/cc -I/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/include -I/usr/local/cuda-9.0/include -fopenmp -DCXX_HAS_SSE4_2_1 -fPIE -o CMakeFiles/cmTC_e8755.dir/src.c.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c
In file included from /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:2:0:
/usr/lib/gcc/x86_64-redhat-linux/4.8.5/include/nmmintrin.h:31:3: error: #error "SSE4.2 instruction set not enabled"

error "SSE4.2 instruction set not enabled"

^
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c: In function ‘main’:
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: error: unknown type name ‘__m128i’
__m128i a = {0,0,0,0}, b = {0,0,0,0}, c = {0,0,0,0};
^
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘a’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘a’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘a’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘b’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘b’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘b’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘c’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘c’) [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: excess elements in scalar initializer [enabled by default]
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: warning: (near initialization for ‘c’) [enabled by default]
gmake[1]: *** [CMakeFiles/cmTC_e8755.dir/src.c.o] Error 1
gmake[1]: Leaving directory `/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
gmake: *** [cmTC_e8755/fast] Error 2

...and run output:

Return value: 1
Source file was:

#include <nmmintrin.h>

int main()
{
__m128i a = {0,0,0,0}, b = {0,0,0,0}, c = {0,0,0,0};
c = _mm_cmpgt_epi64(a, b);
return 0;
}

Performing C SOURCE FILE Test CXX_HAS_AVX_1 failed with the following compile output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_4e8b0/fast
/usr/bin/gmake -f CMakeFiles/cmTC_4e8b0.dir/build.make CMakeFiles/cmTC_4e8b0.dir/build
gmake[1]: Entering directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp' Building C object CMakeFiles/cmTC_4e8b0.dir/src.c.o /usr/bin/cc -I/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/include -I/usr/local/cuda-9.0/include -fopenmp -DCXX_HAS_AVX_1 -fPIE -o CMakeFiles/cmTC_4e8b0.dir/src.c.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c: In function ‘main’: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: error: unknown type name ‘__m256’ __m256 a; ^ gmake[1]: *** [CMakeFiles/cmTC_4e8b0.dir/src.c.o] Error 1 gmake[1]: Leaving directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
gmake: *** [cmTC_4e8b0/fast] Error 2

...and run output:

Return value: 1
Source file was:

#include <immintrin.h>

int main()
{
__m256 a;
a = _mm256_set1_ps(0);
return 0;
}

Performing C SOURCE FILE Test CXX_HAS_AVX2_1 failed with the following compile output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_8ba65/fast
/usr/bin/gmake -f CMakeFiles/cmTC_8ba65.dir/build.make CMakeFiles/cmTC_8ba65.dir/build
gmake[1]: Entering directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp' Building C object CMakeFiles/cmTC_8ba65.dir/src.c.o /usr/bin/cc -I/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/include -I/usr/local/cuda-9.0/include -fopenmp -DCXX_HAS_AVX2_1 -fPIE -o CMakeFiles/cmTC_8ba65.dir/src.c.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c: In function ‘main’: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:6:5: error: unknown type name ‘__m256i’ __m256i a = {0}; ^ gmake[1]: *** [CMakeFiles/cmTC_8ba65.dir/src.c.o] Error 1 gmake[1]: Leaving directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
gmake: *** [cmTC_8ba65/fast] Error 2

...and run output:

Return value: 1
Source file was:

#include <immintrin.h>

int main()
{
__m256i a = {0};
a = _mm256_abs_epi16(a);
return 0;
}

Performing C SOURCE FILE Test HAS_C11_ATOMICS failed with the following compile output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_63188/fast
/usr/bin/gmake -f CMakeFiles/cmTC_63188.dir/build.make CMakeFiles/cmTC_63188.dir/build
gmake[1]: Entering directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp' Building C object CMakeFiles/cmTC_63188.dir/src.c.o /usr/bin/cc -I/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/include -I/usr/local/cuda-9.0/include -msse3 -msse4.1 -msse4.2 -fopenmp -DHAS_C11_ATOMICS -fPIE -o CMakeFiles/cmTC_63188.dir/src.c.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:2:25: fatal error: stdatomic.h: No such file or directory #include <stdatomic.h> ^ compilation terminated. gmake[1]: *** [CMakeFiles/cmTC_63188.dir/src.c.o] Error 1 gmake[1]: Leaving directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
gmake: *** [cmTC_63188/fast] Error 2

...and run output:

Return value: 1
Source file was:

#include <stdatomic.h>
// ATOMIC_INT_LOCK_FREE is flaky on some older gcc versions
// so if this define is not usable a preprocessor definition
// we fail this check and fall back to GCC atomics
#if ATOMIC_INT_LOCK_FREE == 2
#define TH_ATOMIC_IPC_REFCOUNT 1
#endif
int main()
{
int a;
int oa;
atomic_store(&a, 1);
atomic_fetch_add(&a, 1);
oa = atomic_load(&a);
if(!atomic_compare_exchange_strong(&a, &oa, 3))
return -1;
return 0;
}

Performing C SOURCE FILE Test HAS_MSC_ATOMICS failed with the following compile output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_e4b92/fast
/usr/bin/gmake -f CMakeFiles/cmTC_e4b92.dir/build.make CMakeFiles/cmTC_e4b92.dir/build
gmake[1]: Entering directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp' Building C object CMakeFiles/cmTC_e4b92.dir/src.c.o /usr/bin/cc -I/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/include -I/usr/local/cuda-9.0/include -msse3 -msse4.1 -msse4.2 -fopenmp -DHAS_MSC_ATOMICS -fPIE -o CMakeFiles/cmTC_e4b92.dir/src.c.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:2:22: fatal error: intrin.h: No such file or directory #include <intrin.h> ^ compilation terminated. gmake[1]: *** [CMakeFiles/cmTC_e4b92.dir/src.c.o] Error 1 gmake[1]: Leaving directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
gmake: *** [cmTC_e4b92/fast] Error 2

...and run output:

Return value: 1
Source file was:

#include <intrin.h>
int main()
{
long a;
_InterlockedExchange(&a, 1);
_InterlockedExchangeAdd(&a, 1);
if(_InterlockedCompareExchange(&a, 3, 2) != 2)
return -1;
return 0;
}

Performing C SOURCE FILE Test BLAS_F2C_DOUBLE_WORKS failed with the following compile output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_6783d/fast
/usr/bin/gmake -f CMakeFiles/cmTC_6783d.dir/build.make CMakeFiles/cmTC_6783d.dir/build
gmake[1]: Entering directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp' Building C object CMakeFiles/cmTC_6783d.dir/src.c.o /usr/bin/cc -I/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/include -I/usr/local/cuda-9.0/include -msse3 -msse4.1 -msse4.2 -fopenmp -DBLAS_F2C_DOUBLE_WORKS -fPIE -o CMakeFiles/cmTC_6783d.dir/src.c.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c Linking C executable cmTC_6783d /research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/bin/cmake -E cmake_link_script CMakeFiles/cmTC_6783d.dir/link.txt --verbose=1 /usr/bin/cc -msse3 -msse4.1 -msse4.2 -fopenmp -DBLAS_F2C_DOUBLE_WORKS -L"/research/pheng4/sjwang/program/accumulation/pytorch/torch/lib/tmp_install/lib" -Wl,-rpath,$ORIGIN CMakeFiles/cmTC_6783d.dir/src.c.o -o cmTC_6783d -L/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/lib -Wl,-rpath,/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/lib -lmkl_intel_lp64 -lmkl_gnu_thread -lmkl_core -fopenmp -lpthread /usr/lib64/libm.so /usr/lib64/libdl.so gmake[1]: Leaving directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'

...and run output:

Return value: 1
Source file was:

#include <stdlib.h>
#include <stdio.h>
float x[4] = { 1, 2, 3, 4 };
float y[4] = { .1, .01, .001, .0001 };
int four = 4;
int one = 1;
extern double sdot_();
int main() {
int i;
double r = sdot_(&four, x, &one, y, &one);
exit((float)r != (float).1234);
}
Performing C SOURCE FILE Test COMPILER_SUPPORTS_LONG_DOUBLE failed with the following output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_59ac9/fast
/usr/bin/gmake -f CMakeFiles/cmTC_59ac9.dir/build.make CMakeFiles/cmTC_59ac9.dir/build
gmake[1]: Entering directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp' Building C object CMakeFiles/cmTC_59ac9.dir/src.c.o /usr/bin/cc -I/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/include -I/usr/local/cuda-9.0/include -msse3 -msse4.1 -msse4.2 -fopenmp -Wno-ignored-qualifiers -Wno-absolute-value -DCOMPILER_SUPPORTS_LONG_DOUBLE -fPIE -o CMakeFiles/cmTC_59ac9.dir/src.c.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c: In function ‘vcast_vl_l’: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:3:3: internal compiler error: in type_natural_mode, at config/i386/i386.c:6065 vlongdouble vcast_vl_l(long double d) { return (vlongdouble) { d, d }; } ^ Please submit a full bug report, with preprocessed source if appropriate. See <http://bugzilla.redhat.com/bugzilla> for instructions. Preprocessed source stored into /tmp/ccMtf7yB.out file, please attach this to your bugreport. gmake[1]: *** [CMakeFiles/cmTC_59ac9.dir/src.c.o] Error 1 gmake[1]: Leaving directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
gmake: *** [cmTC_59ac9/fast] Error 2

Source file was:

typedef long double vlongdouble attribute((vector_size(sizeof(long double)*2)));
vlongdouble vcast_vl_l(long double d) { return (vlongdouble) { d, d }; }
int main() { vlongdouble vld = vcast_vl_l(0);
}
Performing C SOURCE FILE Test COMPILER_SUPPORTS_SVE failed with the following output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_8b0ee/fast
/usr/bin/gmake -f CMakeFiles/cmTC_8b0ee.dir/build.make CMakeFiles/cmTC_8b0ee.dir/build
gmake[1]: Entering directory `/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
Building C object CMakeFiles/cmTC_8b0ee.dir/src.c.o
/usr/bin/cc -I/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/include -I/usr/local/cuda-9.0/include -msse3 -msse4.1 -msse4.2 -fopenmp -Wno-ignored-qualifiers -Wno-absolute-value -DCOMPILER_SUPPORTS_SVE -march=armv8-a+sve -fPIE -o CMakeFiles/cmTC_8b0ee.dir/src.c.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c
/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c:1:0: error: bad value (armv8-a+sve) for -march= switch

^
cc1: warning: unrecognized command line option "-Wno-absolute-value" [enabled by default]
gmake[1]: *** [CMakeFiles/cmTC_8b0ee.dir/src.c.o] Error 1
gmake[1]: Leaving directory `/research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
gmake: *** [cmTC_8b0ee/fast] Error 2

Source file was:

#include <arm_sve.h>
int main() {
svint32_t r = svdup_n_s32(1); }
Performing C SOURCE FILE Test COMPILER_SUPPORTS_AVX512F failed with the following output:
Change Dir: /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp

Run Build Command(s):/usr/bin/gmake cmTC_20ccb/fast
/usr/bin/gmake -f CMakeFiles/cmTC_20ccb.dir/build.make CMakeFiles/cmTC_20ccb.dir/build
gmake[1]: Entering directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp' Building C object CMakeFiles/cmTC_20ccb.dir/src.c.o /usr/bin/cc -I/research/pheng4/sjwang/software/anaconda3/envs/pytorch36-source2/include -I/usr/local/cuda-9.0/include -msse3 -msse4.1 -msse4.2 -fopenmp -Wno-ignored-qualifiers -Wno-absolute-value -DCOMPILER_SUPPORTS_AVX512F -mavx512f -fPIE -o CMakeFiles/cmTC_20ccb.dir/src.c.o -c /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp/src.c cc: error: unrecognized command line option ‘-mavx512f’ gmake[1]: *** [CMakeFiles/cmTC_20ccb.dir/src.c.o] Error 1 gmake[1]: Leaving directory /research/pheng4/sjwang/program/accumulation/pytorch/build/CMakeFiles/CMakeTmp'
gmake: *** [cmTC_20ccb/fast] Error 2

Source file was:

#if defined(_MSC_VER)
#include <intrin.h>
#else
#include <x86intrin.h>
#endif
__m512 addConstant(__m512 arg) {
return _mm512_add_ps(arg, _mm512_set1_ps(1.f));
}
int main() {
__m512i a = _mm512_set1_epi32(1);
__m256i ymm = _mm512_extracti64x4_epi64(a, 0);
__mmask16 m = _mm512_cmp_epi32_mask(a, a, _MM_CMPINT_EQ);
__m512i r = _mm512_andnot_si512(a, a); }`

Anyone could help me check this out?

[MSC19950601]

When installing pytorch@fd25a2a, there is no third_party nervanagpu[https://github.com/NervanaSystems/nervanagpu]!
How to solve it~=!

[emma-sjwang]

Every thing was solved by re-download the pytorch commit and 3th-party source files.....

[wmj1238]

When installing pytorch@fd25a2a, there is no third_party nervanagpu[https://github.com/NervanaSystems/nervanagpu]!
How to solve it~=!

have you solve this problem ?

[emma-sjwang]

@wmj1238
Actually, after I successfully installed pytorch@fd25a2a, I did not find that I need this 'nervanagpu' third_party. Meybe you forget to change to the correct pytorch version by:

git checkout fd25a2a