FP16 python module error at make

[agoston-mc]

Since the update to FP16, the NNPACK install fails on latest, as in the build process "Generating src/x86_64-fma/blas/shdotxf.py.o" would use the python module fp16.avx, which is no longer generated in FP16.

build process:

git clone https://github.com/Maratyszcza/NNPACK NNPACK
git clone https://github.com/Maratyszcza/pthreadpool  NNPACK/pthreadpool

mkdir -p NNPACK/build
cd NNPACK/build
cmake -DCMAKE_INSTALL_PREFIX:PATH=. -DNNPACK_INFERENCE_ONLY=OFF -DNNPACK_CONVOLUTION_ONLY=OFF -DNNPACK_BUILD_TESTS=OFF -DCMAKE_POSITION_INDEPENDENT_CODE=ON -DPTHREADPOOL_SOURCE_DIR=pthreadpool ..
make -j4

The make error I get:

[  1%] Building C object deps/pthreadpool/CMakeFiles/pthreadpool.dir/src/legacy-api.c.o
[  2%] Building C object deps/pthreadpool/CMakeFiles/pthreadpool.dir/src/portable-api.c.o
[  3%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/api.c.o
[  4%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/api.c.o
[  6%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/cache.c.o
[  7%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/cache.c.o
[  8%] Building C object deps/pthreadpool/CMakeFiles/pthreadpool.dir/src/memory.c.o
[ 10%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/init.c.o
[ 10%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/init.c.o
[ 12%] Building C object deps/pthreadpool/CMakeFiles/pthreadpool.dir/src/pthreads.c.o
[ 13%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/log.c.o
[ 14%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/log.c.o
[ 15%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/x86/init.c.o
[ 16%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/x86/init.c.o
[ 18%] Building C object deps/pthreadpool/CMakeFiles/pthreadpool.dir/src/fastpath.c.o
[ 20%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/x86/info.c.o
[ 20%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/x86/info.c.o
[ 22%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/x86/vendor.c.o
[ 22%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/x86/vendor.c.o
[ 24%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/x86/uarch.c.o
[ 25%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/x86/uarch.c.o
[ 26%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/x86/name.c.o
[ 27%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/x86/topology.c.o
[ 28%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/x86/name.c.o
[ 30%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/x86/isa.c.o
[ 31%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/x86/cache/init.c.o
[ 32%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/x86/topology.c.o
[ 34%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/x86/cache/descriptor.c.o
[ 34%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/x86/isa.c.o
[ 36%] Linking C static library libpthreadpool.a
[ 37%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/x86/cache/init.c.o
[ 38%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/x86/cache/descriptor.c.o
[ 39%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/x86/cache/deterministic.c.o
[ 40%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/x86/cache/deterministic.c.o
[ 40%] Built target pthreadpool
[ 42%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/x86/linux/init.c.o
[ 43%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/x86/linux/cpuinfo.c.o
[ 44%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/x86/linux/init.c.o
[ 45%] Building C object CMakeFiles/nnpack_reference_layers.dir/src/ref/convolution-output.c.o
[ 46%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/linux/smallfile.c.o
[ 48%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/linux/multiline.c.o
[ 49%] Building C object CMakeFiles/nnpack_reference_layers.dir/src/ref/convolution-input-gradient.c.o
[ 50%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/x86/linux/cpuinfo.c.o
[ 51%] Building C object CMakeFiles/nnpack_reference_layers.dir/src/ref/convolution-kernel.c.o
[ 53%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/linux/cpulist.c.o
[ 54%] Building C object CMakeFiles/nnpack_reference_layers.dir/src/ref/fully-connected-output.c.o
[ 55%] Building C object CMakeFiles/nnpack_reference_layers.dir/src/ref/max-pooling-output.c.o
[ 56%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/linux/smallfile.c.o
[ 57%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo_internals.dir/src/linux/processors.c.o
[ 59%] Building C object CMakeFiles/nnpack_reference_layers.dir/src/ref/softmax-output.c.o
[ 60%] Building C object CMakeFiles/nnpack_reference_layers.dir/src/ref/relu-output.c.o
[ 61%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/linux/multiline.c.o
[ 62%] Building C object CMakeFiles/nnpack_reference_layers.dir/src/ref/relu-input-gradient.c.o
[ 63%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/linux/cpulist.c.o
[ 65%] Building C object deps/cpuinfo/CMakeFiles/cpuinfo.dir/src/linux/processors.c.o
[ 66%] Linking C static library libcpuinfo_internals.a
[ 67%] Linking C static library libnnpack_reference_layers.a
[ 68%] Linking C static library libcpuinfo.a
[ 68%] Built target cpuinfo_internals
[ 68%] Built target nnpack_reference_layers
[ 68%] Built target cpuinfo
[ 69%] Generating src/x86_64-fma/2d-fourier-8x8.py.o
[ 71%] Generating src/x86_64-fma/2d-fourier-16x16.py.o
[ 72%] Generating src/x86_64-fma/2d-winograd-8x8-3x3.py.o
[ 73%] Generating src/x86_64-fma/blas/s8gemm.py.o
[ 74%] Generating src/x86_64-fma/blas/c8gemm.py.o
[ 75%] Generating src/x86_64-fma/blas/s4c6gemm.py.o
[ 77%] Generating src/x86_64-fma/blas/conv1x1.py.o
[ 78%] Generating src/x86_64-fma/blas/sgemm.py.o
[ 79%] Generating src/x86_64-fma/max-pooling.py.o
[ 80%] Generating src/x86_64-fma/relu.py.o
[ 81%] Generating src/x86_64-fma/softmax.py.o
[ 83%] Generating src/x86_64-fma/blas/sdotxf.py.o
[ 84%] Generating src/x86_64-fma/blas/shdotxf.py.o
Traceback (most recent call last):
  File "/usr/lib/python3.10/runpy.py", line 196, in _run_module_as_main
    return _run_code(code, main_globals, None,
  File "/usr/lib/python3.10/runpy.py", line 86, in _run_code
    exec(code, run_globals)
  File "/home/ad.adasworks.com/agoston.czobor/NNPACK/deps/peachpy/peachpy/x86_64/__main__.py", line 282, in <module>
    main()
  File "/home/ad.adasworks.com/agoston.czobor/NNPACK/deps/peachpy/peachpy/x86_64/__main__.py", line 269, in main
    execute_script(writers, options.input[0])
  File "/home/ad.adasworks.com/agoston.czobor/NNPACK/deps/peachpy/peachpy/x86_64/__main__.py", line 199, in execute_script
    with writer:
  File "/home/ad.adasworks.com/agoston.czobor/NNPACK/deps/peachpy/peachpy/x86_64/__main__.py", line 200, in execute_script
    execute_script(writers, source_filename)
  File "/home/ad.adasworks.com/agoston.czobor/NNPACK/deps/peachpy/peachpy/x86_64/__main__.py", line 204, in execute_script
    exec(code, globals())
  File "/home/ad.adasworks.com/agoston.czobor/NNPACK/src/x86_64-fma/blas/shdotxf.py", line 4, in <module>
    from fp16.avx import fp16_alt_xmm_to_fp32_xmm
ModuleNotFoundError: No module named 'fp16.avx'
make[2]: *** [CMakeFiles/nnpack.dir/build.make:188: src/x86_64-fma/blas/shdotxf.py.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:175: CMakeFiles/nnpack.dir/all] Error 2
make: *** [Makefile:136: all] Error 2

[justin-Kor]

I also get the same error.

@Maratyszcza
Please give me some advice.

[justin-Kor]

If you comment out only the relevant code as shown below, it will build normally.
@Maratyszcza
Please check for build errors.

---