USE_NNPACK build flag not honored.

[unrelatedlabs]

Description

USE_NNPACK build flag not honored.

git clone --recursive https://github.com/apache/incubator-mxnet.git --branch 1.5.0 mxnet
mkdir -p mxnet/build
cd mxnet/build

cmake -GNinja -DVERBOSE=1  -DUSE_NNPACK=1 \
 -DADD_LDFLAGS="-L${NNPACK}/build/ -lnnpack -lpthreadpool"  \
 -DADD_CFLAGS="-I${NNPACK}/include/ -I${NNPACK}/deps/pthreadpool/include/" \
 -DUSE_BLAS=openblas \
 -DUSE_OPENCV=1 \
 -DUSE_CUDA=0 \
 -USE_MKLDNN=0 \
 ..

I get

-- Generating done
CMake Warning:
  Manually-specified variables were not used by the project:

    ADD_CFLAGS
    ADD_LDFLAGS
    USE_NNPACK
    VERBOSE

Environment info (Required)

Architecture:        armv7l
CPU op-mode(s):      32-bit, 64-bit
Byte Order:          Little Endian
Address sizes:       39 bits physical, 48 bits virtual
CPU(s):              8
On-line CPU(s) list: 0-7
Thread(s) per core:  1
Core(s) per socket:  1
Socket(s):           8
Vendor ID:           GenuineIntel
CPU family:          6
Model:               158
Model name:          Intel(R) Core(TM) i9-9980HK CPU @ 2.40GHz
Stepping:            13
CPU MHz:             2400.000
BogoMIPS:            4800.00
L1d cache:           32K
L1i cache:           32K
L2 cache:            256K
L3 cache:            16384K
Flags:               fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov pat pse36 clflush mmx fxsr sse sse2 ss ht pbe syscall nx pdpe1gb lm constant_tsc rep_good nopl xtopology nonstop_tsc pni pclmulqdq dtes64 ds_cpl ssse3 sdbg fma cx16 xtpr pcid sse4_1 sse4_2 movbe popcnt aes xsave avx f16c rdrand hypervisor lahf_lm abm 3dnowprefetch kaiser fsgsbase bmi1 hle avx2 bmi2 erms rtm xsaveopt arat
----------Python Info----------
Version      : 3.7.3
Compiler     : GCC 8.3.0
Build        : ('default', 'Apr  3 2019 05:39:12')
Arch         : ('32bit', 'ELF')
------------Pip Info-----------
Version      : 19.2.2
Directory    : /usr/local/lib/python3.7/dist-packages/pip
----------MXNet Info-----------
No MXNet installed.
----------System Info----------
Platform     : Linux-4.9.184-linuxkit-armv7l-with-debian-10.0
system       : Linux
node         : 2007a277dc67
release      : 4.9.184-linuxkit
version      : #1 SMP Tue Jul 2 22:58:16 UTC 2019
----------Hardware Info----------
machine      : armv7l
processor    : 
----------Network Test----------
Setting timeout: 10
Timing for MXNet: https://github.com/apache/incubator-mxnet, DNS: 0.0494 sec, LOAD: 0.8141 sec.
Timing for Gluon Tutorial(en): http://gluon.mxnet.io, DNS: 0.0411 sec, LOAD: 0.2910 sec.
Timing for Gluon Tutorial(cn): https://zh.gluon.ai, DNS: 0.0434 sec, LOAD: 0.2042 sec.
Timing for FashionMNIST: https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/dataset/fashion-mnist/train-labels-idx1-ubyte.gz, DNS: 0.0334 sec, LOAD: 0.2438 sec.
Timing for PYPI: https://pypi.python.org/pypi/pip, DNS: 0.0324 sec, LOAD: 0.5857 sec.
Timing for Conda: https://repo.continuum.io/pkgs/free/, DNS: 0.0043 sec, LOAD: 0.3265 sec.

Package used (Python/R/Scala/Julia):
Python 3.7

Build info (Required if built from source)

Compiler (gcc/clang/mingw/visual studio):
gcc version 8.3.0 (Debian 8.3.0-6)

MXNet commit hash:
75a9e18

Build config:

cmake -GNinja -DVERBOSE=1  -DUSE_NNPACK=1 \
 -DADD_LDFLAGS="-L${NNPACK}/build/ -lnnpack -lpthreadpool"  \
 -DADD_CFLAGS="-I${NNPACK}/include/ -I${NNPACK}/deps/pthreadpool/include/" \
 -DUSE_BLAS=openblas \
 -DUSE_OPENCV=1 \
 -DUSE_CUDA=0 \
 -USE_MKLDNN=0 \
 ..

Error Message:

-- The C compiler identification is GNU 8.3.0
-- The CXX compiler identification is GNU 8.3.0
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- CMAKE_CROSSCOMPILING FALSE
-- CMAKE_HOST_SYSTEM_PROCESSOR armv7l
-- CMAKE_SYSTEM_PROCESSOR armv7l
-- CMAKE_SYSTEM_NAME Linux
-- Performing Test SUPPORT_CXX11
-- Performing Test SUPPORT_CXX11 - Success
-- Performing Test SUPPORT_CXX0X
-- Performing Test SUPPORT_CXX0X - Success
-- Performing Test SUPPORT_MSSE3
-- Performing Test SUPPORT_MSSE3 - Failed
-- Performing Test SUPPORT_MSSE2
-- Performing Test SUPPORT_MSSE2 - Failed
-- Determining F16C support
-- Performing Test COMPILER_SUPPORT_MF16C
-- Performing Test COMPILER_SUPPORT_MF16C - Failed
-- Found PkgConfig: /usr/bin/pkg-config (found version "0.29")
-- Could NOT find Jemalloc (missing: JEMALLOC_LIBRARY JEMALLOC_INCLUDE_DIR)
-- Found OpenCV: /usr/local (found version "4.1.1") found components: core highgui imgproc imgcodecs
-- OpenCV 4.1.1 found (/usr/local/lib/cmake/opencv4)
-- OpenCV_LIBS=opencv_core;opencv_highgui;opencv_imgproc;opencv_imgcodecs
-- Found OpenMP_C: -fopenmp (found version "4.5")
-- Found OpenMP_CXX: -fopenmp (found version "4.5")
-- Found OpenMP: TRUE (found version "4.5")
USE_LAPACK is ON
-- Could NOT find Jemalloc (missing: JEMALLOC_LIBRARY JEMALLOC_INCLUDE_DIR)
CMake Warning at 3rdparty/googletest/googletest/CMakeLists.txt:47 (project):
VERSION keyword not followed by a value or was followed by a value that
expanded to nothing.

-- Found PythonInterp: /usr/bin/python (found version "3.7.3")
-- Looking for pthread.h
-- Looking for pthread.h - found
-- Looking for pthread_create
-- Looking for pthread_create - found
-- Found Threads: TRUE
-- Found GTest: gtest
-- Looking for clock_gettime in rt
-- Looking for clock_gettime in rt - found
-- Looking for fopen64
-- Looking for fopen64 - not found
-- Looking for C++ include cxxabi.h
-- Looking for C++ include cxxabi.h - found
-- Looking for nanosleep
-- Looking for nanosleep - found
-- Looking for backtrace
-- Looking for backtrace - found
-- backtrace facility detected in default set of libraries
-- Found Backtrace: /usr/include
-- Check if the system is big endian
-- Searching 16 bit integer
-- Looking for sys/types.h
-- Looking for sys/types.h - found
-- Looking for stdint.h
-- Looking for stdint.h - found
-- Looking for stddef.h
-- Looking for stddef.h - found
-- Check size of unsigned short
-- Check size of unsigned short - done
-- Using unsigned short
-- Check if the system is big endian - little endian
-- /mxnet/3rdparty/dmlc-core/cmake/build_config.h.in -> /mxnet/3rdparty/dmlc-core/include/dmlc/build_config.h
-- Found OpenMP_C: -fopenmp (found version "4.5")
-- Found OpenMP_CXX: -fopenmp (found version "4.5")
-- Configuring done
-- Generating done
CMake Warning:
Manually-specified variables were not used by the project:

ADD_CFLAGS
ADD_LDFLAGS
USE_NNPACK
VERBOSE

-- Build files have been written to: /mxnet/build

Minimum reproducible example

Just building from source

[mxnet-label-bot]

Hey, this is the MXNet Label Bot.
Thank you for submitting the issue! I will try and suggest some labels so that the appropriate MXNet community members can help resolve it.
Here are my recommended labels: Build

[leleamol]

@mxnet-label-bot add [Build]

[cyrusbehr]

Any updates on this issue @marcoabreu (it has been nearly 6 months). I am experiencing the same problem - although I am first exporting the library paths:

export YOUR_NNPACK_INSTALL_PATH=/home/nchafni/Cyrus/libs/NNPACK
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:$YOUR_NNPACK_INSTALL_PATH/lib
export CFLAGS="-I/home/nchafni/Cyrus/libs/NNPACK/include/ -I/home/nchafni/Cyrus/libs/NNPACK/third-party/pthreadpool/include/"

cmake -DUSE_CPP_PACKAGE=1 -DUSE_NNPACK=1 -DUSE_CUDA=0 -DUSE_MKLDNN=1 -DUSE_LAPACK=0 -DUSE_OPENCV=0 -DUSE_OPENMP=0 ..

cmake output:

...
-- Configuring done
-- Generating done
CMake Warning:
  Manually-specified variables were not used by the project:

    USE_NNPACK

How are people currently building with NNPACK??

[apkuhar]

I managed to build an older version of mxnet with nnpack, but it was a couple of times slower than the binary build ob 1.5.1 on raspberry pi 4. Running a resnet50 based model. So might make no sense to try to build with nnpack anymore.

mxnet was also 2-3 times faster than tensorflow on the same model architecture on rpi4.