CMake build error on armv7 with latest gcc

[yukyon]

Hi, thank you for your great job!

I'm trying to build tensorflow lite 2.6.0 on armv7 environment with xnnpack.

But, cmake build error occurred with gcc 9.3 or higher.
(gcc 8.3, 9.2 is ok)

I also tried to build the same test in yocto(dunfell and hardknott)
(dunfell uses gcc 9.3.1, hardknott uses gcc 10.2)

All the same, a build error occurred in neondot.

my build option

ARMCC_FLAGS="-march=armv7-a -mfpu=neon-vfpv4 -funsafe-math-optimizations"
ARMCC_PREFIX=/home/worker/work/toolchains/gcc-arm-10.2-2020.11-x86_64-arm-none-linux-gnueabihf/bin/arm-none-linux-gnueabihf-
cmake -DCMAKE_C_COMPILER=${ARMCC_PREFIX}gcc   -DCMAKE_CXX_COMPILER=${ARMCC_PREFIX}g++   -DCMAKE_C_FLAGS="${ARMCC_FLAGS}"   -DCMAKE_CXX_FLAGS="${ARMCC_FLAGS}"   -DCMAKE_VERBOSE_MAKEFILE:BOOL=ON   -DCMAKE_SYSTEM_NAME=Linux   -DCMAKE_SYSTEM_PROCESSOR=armv7   -DBUILD_SHARED_LIBS=ON   ../tensorflow/lite/

build error log

[ 86%] Building C object _deps/xnnpack-build/CMakeFiles/XNNPACK.dir/src/qc8-gemm/gen/1x8c4-minmax-fp32-neondot.c.o
cd /home/worker/git/tensorflow/build_armv7/_deps/xnnpack-build && /home/worker/work/toolchains/gcc-arm-10.2-2020.11-x86_64-arm-none-linux-gnueabihf/bin/arm-none-linux-gnueabihf-gcc -DCPUINFO_SUPPORTED_PLATFORM=1 -DEIGEN_MPL2_ONLY -DFXDIV_USE_INLINE_ASSEMBLY=0 -DNOMINMAX=1 -DPTHREADPOOL_NO_DEPRECATED_API=1 -DXNNPACK_EXPORTS -DXNN_ENABLE_ASSEMBLY=1 -DXNN_ENABLE_MEMOPT=1 -DXNN_ENABLE_SPARSE=1 -DXNN_LOG_LEVEL=0 -I/home/worker/git/tensorflow/build_armv7/xnnpack/include -I/home/worker/git/tensorflow/build_armv7/xnnpack/src -I/home/worker/git/tensorflow/build_armv7/clog-source/deps/clog/include -I/home/worker/git/tensorflow/build_armv7/cpuinfo-source/include -I/home/worker/git/tensorflow/build_armv7/pthreadpool-source/include -I/home/worker/git/tensorflow/build_armv7/FXdiv-source/include -I/home/worker/git/tensorflow/build_armv7/FP16-source/include -march=armv7-a -mfpu=neon-vfpv4 -funsafe-math-optimizations -O3 -DNDEBUG -fPIC -Wno-psabi -pthread -std=gnu99  -marm  -march=armv8.2-a+dotprod -mfpu=neon-fp-armv8  -O2  -MD -MT _deps/xnnpack-build/CMakeFiles/XNNPACK.dir/src/qc8-gemm/gen/1x8c4-minmax-fp32-neondot.c.o -MF CMakeFiles/XNNPACK.dir/src/qc8-gemm/gen/1x8c4-minmax-fp32-neondot.c.o.d -o CMakeFiles/XNNPACK.dir/src/qc8-gemm/gen/1x8c4-minmax-fp32-neondot.c.o -c 
/home/worker/git/tensorflow/build_armv7/xnnpack/src/qc8-gemm/gen/1x8c4-minmax-fp32-neondot.c
/tmp/cccuMfcB.s: Assembler messages:
/tmp/cccuMfcB.s:60: Error: selected processor does not support `vsdot.s8 q8,q11,d7[0]' in ARM mode
/tmp/cccuMfcB.s:62: Error: selected processor does not support `vsdot.s8 q10,q9,d7[0]' in ARM mode
/tmp/cccuMfcB.s:65: Error: selected processor does not support `vsdot.s8 q8,q9,d7[1]' in ARM mode
/tmp/cccuMfcB.s:69: Error: selected processor does not support `vsdot.s8 q10,q9,d7[1]' in ARM mode
/tmp/cccuMfcB.s:140: Error: selected processor does not support `vsdot.s8 q8,q11,d7[0]' in ARM mode
/tmp/cccuMfcB.s:142: Error: selected processor does not support `vsdot.s8 q10,q9,d7[0]' in ARM mode
_deps/xnnpack-build/CMakeFiles/XNNPACK.dir/build.make:20980: recipe for target '_deps/xnnpack-build/CMakeFiles/XNNPACK.dir/src/qc8-gemm/gen/1x8c4-minmax-fp32-neondot.c.o' failed

I would like to know if there are any plans to support the latest gcc version.
Or Please help if there is something I did wrong.

Thanks,

[Maratyszcza]

This is bug in gcc, and it is fixed there. See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=101723 for details.