Build fails with strict-aliasing violations

[eli-schwartz]

I tried to build with the following *FLAGS to optimize the build: -flto=4 -Werror=odr -Werror=lto-type-mismatch -Werror=strict-aliasing

The -Werror=* flags are important to detect cases where the compiler can try to optimize based on assuming Undefined Behavior (UB) cannot happen, and miscompile code that has UB in it. strict-aliasing issues are always bad but LTO can make them even worse.

I got this error:

[322/356] /usr/bin/ccache /usr/bin/x86_64-pc-linux-gnu-gcc -DXNN_ENABLE_ARM_BF16=0 -DXNN_ENABLE_ARM_DOTPROD=0 -DXNN_ENABLE_ARM_FP16_SCALAR=0 -DXNN_ENABLE_ARM_FP16_VECTOR=0 -DXNN_ENABLE_ARM_I8MM=0 -DXNN_ENABLE_ARM_SME2=0 -DXNN_ENABLE_ARM_SME=0 -DXNN_ENABLE_ASSEMBLY=1 -DXNN_ENABLE_AVX256SKX=1 -DXNN_ENABLE_AVX256VNNI=1 -DXNN_ENABLE_AVX256VNNIGFNI=1 -DXNN_ENABLE_AVX512AMX=1 -DXNN_ENABLE_AVX512F=1 -DXNN_ENABLE_AVX512FP16=1 -DXNN_ENABLE_AVX512SKX=1 -DXNN_ENABLE_AVX512VBMI=1 -DXNN_ENABLE_AVX512VNNI=1 -DXNN_ENABLE_AVX512VNNIGFNI=1 -DXNN_ENABLE_AVXVNNI=1 -DXNN_ENABLE_AVXVNNIINT8=0 -DXNN_ENABLE_CPUINFO=1 -DXNN_ENABLE_DWCONV_MULTIPASS=0 -DXNN_ENABLE_GEMM_M_SPECIALIZATION=1 -DXNN_ENABLE_HVX=1 -DXNN_ENABLE_KLEIDIAI=0 -DXNN_ENABLE_MEMOPT=1 -DXNN_ENABLE_RISCV_VECTOR=1 -DXNN_ENABLE_SPARSE=1 -DXNN_ENABLE_VSX=1 -I/var/tmp/portage/sci-ml/XNNPACK-2024.12.03/work/XNNPACK-51a0103656eff6fc9bfd39a4597923c4b542c883/include -I/var/tmp/portage/sci-ml/XNNPACK-2024.12.03/work/XNNPACK-51a0103656eff6fc9bfd39a4597923c4b542c883/src -I/include  -pipe -ggdb -march=native -fstack-protector-all -O2 -fdiagnostics-color=always -frecord-gcc-switches -Werror=odr -Werror=lto-type-mismatch -Werror=strict-aliasing -Wformat -Werror=format-security -std=c99 -fPIC -Wno-psabi -O2  -mf16c -mfma -mavx512f -mavx512cd -mavx512bw -mavx512dq -mavx512vl -mavx512vnni -mgfni -MD -MT CMakeFiles/microkernels-all.dir/src/qd8-f16-qc4w-gemm/gen/qd8-f16-qc4w-gemm-14x8c8-minmax-avx256vnnigfni.c.o -MF CMakeFiles/microkernels-all.dir/src/qd8-f16-qc4w-gemm/gen/qd8-f16-qc4w-gemm-14x8c8-minmax-avx256vnnigfni.c.o.d -o CMakeFiles/microkernels-all.dir/src/qd8-f16-qc4w-gemm/gen/qd8-f16-qc4w-gemm-14x8c8-minmax-avx256vnnigfni.c.o -c /var/tmp/portage/sci-ml/XNNPACK-2024.12.03/work/XNNPACK-51a0103656eff6fc9bfd39a4597923c4b542c883/src/qd8-f16-qc4w-gemm/gen/qd8-f16-qc4w-gemm-14x8c8-minmax-avx256vnnigfni.c
FAILED: CMakeFiles/microkernels-all.dir/src/qd8-f16-qc4w-gemm/gen/qd8-f16-qc4w-gemm-14x8c8-minmax-avx256vnnigfni.c.o 
/usr/bin/ccache /usr/bin/x86_64-pc-linux-gnu-gcc -DXNN_ENABLE_ARM_BF16=0 -DXNN_ENABLE_ARM_DOTPROD=0 -DXNN_ENABLE_ARM_FP16_SCALAR=0 -DXNN_ENABLE_ARM_FP16_VECTOR=0 -DXNN_ENABLE_ARM_I8MM=0 -DXNN_ENABLE_ARM_SME2=0 -DXNN_ENABLE_ARM_SME=0 -DXNN_ENABLE_ASSEMBLY=1 -DXNN_ENABLE_AVX256SKX=1 -DXNN_ENABLE_AVX256VNNI=1 -DXNN_ENABLE_AVX256VNNIGFNI=1 -DXNN_ENABLE_AVX512AMX=1 -DXNN_ENABLE_AVX512F=1 -DXNN_ENABLE_AVX512FP16=1 -DXNN_ENABLE_AVX512SKX=1 -DXNN_ENABLE_AVX512VBMI=1 -DXNN_ENABLE_AVX512VNNI=1 -DXNN_ENABLE_AVX512VNNIGFNI=1 -DXNN_ENABLE_AVXVNNI=1 -DXNN_ENABLE_AVXVNNIINT8=0 -DXNN_ENABLE_CPUINFO=1 -DXNN_ENABLE_DWCONV_MULTIPASS=0 -DXNN_ENABLE_GEMM_M_SPECIALIZATION=1 -DXNN_ENABLE_HVX=1 -DXNN_ENABLE_KLEIDIAI=0 -DXNN_ENABLE_MEMOPT=1 -DXNN_ENABLE_RISCV_VECTOR=1 -DXNN_ENABLE_SPARSE=1 -DXNN_ENABLE_VSX=1 -I/var/tmp/portage/sci-ml/XNNPACK-2024.12.03/work/XNNPACK-51a0103656eff6fc9bfd39a4597923c4b542c883/include -I/var/tmp/portage/sci-ml/XNNPACK-2024.12.03/work/XNNPACK-51a0103656eff6fc9bfd39a4597923c4b542c883/src -I/include  -pipe -ggdb -march=native -fstack-protector-all -O2 -fdiagnostics-color=always -frecord-gcc-switches -Werror=odr -Werror=lto-type-mismatch -Werror=strict-aliasing -Wformat -Werror=format-security -std=c99 -fPIC -Wno-psabi -O2  -mf16c -mfma -mavx512f -mavx512cd -mavx512bw -mavx512dq -mavx512vl -mavx512vnni -mgfni -MD -MT CMakeFiles/microkernels-all.dir/src/qd8-f16-qc4w-gemm/gen/qd8-f16-qc4w-gemm-14x8c8-minmax-avx256vnnigfni.c.o -MF CMakeFiles/microkernels-all.dir/src/qd8-f16-qc4w-gemm/gen/qd8-f16-qc4w-gemm-14x8c8-minmax-avx256vnnigfni.c.o.d -o CMakeFiles/microkernels-all.dir/src/qd8-f16-qc4w-gemm/gen/qd8-f16-qc4w-gemm-14x8c8-minmax-avx256vnnigfni.c.o -c /var/tmp/portage/sci-ml/XNNPACK-2024.12.03/work/XNNPACK-51a0103656eff6fc9bfd39a4597923c4b542c883/src/qd8-f16-qc4w-gemm/gen/qd8-f16-qc4w-gemm-14x8c8-minmax-avx256vnnigfni.c
/var/tmp/portage/sci-ml/XNNPACK-2024.12.03/work/XNNPACK-51a0103656eff6fc9bfd39a4597923c4b542c883/src/qd8-f16-qc4w-gemm/gen/qd8-f16-qc4w-gemm-14x8c8-minmax-avx256vnnigfni.c: In function ‘xnn_qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256vnnigfni’:
/var/tmp/portage/sci-ml/XNNPACK-2024.12.03/work/XNNPACK-51a0103656eff6fc9bfd39a4597923c4b542c883/src/qd8-f16-qc4w-gemm/gen/qd8-f16-qc4w-gemm-14x8c8-minmax-avx256vnnigfni.c:138:62: error: dereferencing type-punned pointer will break strict-aliasing rules [-Werror=strict-aliasing]
  138 |   const __m256 voutput_min = _mm256_cvtph_ps(_mm_set1_epi16(*(const uint16_t*) &params->scalar.min));
      |                                                              ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
/var/tmp/portage/sci-ml/XNNPACK-2024.12.03/work/XNNPACK-51a0103656eff6fc9bfd39a4597923c4b542c883/src/qd8-f16-qc4w-gemm/gen/qd8-f16-qc4w-gemm-14x8c8-minmax-avx256vnnigfni.c:139:62: error: dereferencing type-punned pointer will break strict-aliasing rules [-Werror=strict-aliasing]
  139 |   const __m256 voutput_max = _mm256_cvtph_ps(_mm_set1_epi16(*(const uint16_t*) &params->scalar.max));
      |                                                              ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
cc1: some warnings being treated as errors
ninja: build stopped: cannot make progress due to previous errors.
 * ERROR: sci-ml/XNNPACK-2024.12.03::gentoo failed (compile phase):
 *   ninja -v -k0 failed

... repeated for many files.

Originally reported upstream: https://bugs.gentoo.org/953467
Exhaustive build log: build.log