OpenBLAS 0.3.12 skylakex build fail

[minchaoyue]

In file included from ../kernel/x86_64/dasum.c:8:0:
../kernel/x86_64/dasum_microk_skylakex-2.c: In function ‘dasum_kernel’:
../kernel/x86_64/dasum_microk_skylakex-2.c:42:38: error: incompatible type for argument 1 of ‘mm512_abs_p’
accum_0 += mm512_abs_pd(mm512_load_pd(&x1[i + 0]));
^~~~~~~~~~~~~~
cc -c -O2 -DMAX_STACK_ALLOC=2048 -Wall -m64 -DF_INTERFACE_GFORT -fPIC -DSMP_SERVER -DNO_WARMUP -DMAX_CPU_NUMBER=64 -DMAX_PARALLEL_NUMBER=1 -DBUILD_SINGLE=1 -DBUILD_DOUBLE=1 -DBUILD_COMPLEX=1 -DBUILD_COMPLEX16=1 -DVERSION="0.3.12" -msse3 -mssse3 -msse4.1 -march=skylake-avx512 -mavx2 -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME -DASMNAME=camax_k -DASMFNAME=camax_k -DNAME=camax_k -DCNAME=camax_k -DCHAR_NAME="camax_k" -DCHAR_CNAME="camax_k" -DNO_AFFINITY -I.. -UDOUBLE -DCOMPLEX -DCOMPLEX -UDOUBLE -DUSE_ABS -UUSE_MIN ../kernel/x86_64/zamax_sse.S -o camax_k.o
In file included from /opt/rh/devtoolset-7/root/usr/lib/gcc/x86_64-redhat-linux/7/include/immintrin.h:45:0,
from ../kernel/x86_64/dasum_microk_skylakex-2.c:6,
from ../kernel/x86_64/dasum.c:8:
/opt/rh/devtoolset-7/root/usr/lib/gcc/x86_64-redhat-linux/7/include/avx512fintrin.h:7301:1: note: expected ‘__m512 {aka __vector(16) float}’ but argument is of type ‘__m512d {aka __vector(8) double}’
_mm512_abs_pd (_m512 A)
^~~~~~~~~~~~~
cc -c -O2 -DMAX_STACK_ALLOC=2048 -Wall -m64 -DF_INTERFACE_GFORT -fPIC -DSMP_SERVER -DNO_WARMUP -DMAX_CPU_NUMBER=64 -DMAX_PARALLEL_NUMBER=1 -DBUILD_SINGLE=1 -DBUILD_DOUBLE=1 -DBUILD_COMPLEX=1 -DBUILD_COMPLEX16=1 -DVERSION="0.3.12" -msse3 -mssse3 -msse4.1 -march=skylake-avx512 -mavx2 -UASMNAME -UASMFNAME -UNAME -UCNAME -UCHAR_NAME -UCHAR_CNAME -DASMNAME=camin_k -DASMFNAME=camin_k -DNAME=camin_k -DCNAME=camin_k -DCHAR_NAME="camin_k" -DCHAR_CNAME="camin_k" -DNO_AFFINITY -I.. -UDOUBLE -DCOMPLEX -DCOMPLEX -UDOUBLE -DUSE_ABS -DUSE_MIN ../kernel/x86_64/zamax_sse.S -o camin_k.o
In file included from ../kernel/x86_64/dasum.c:8:0:
../kernel/x86_64/dasum_microk_skylakex-2.c:43:38: error: incompatible type for argument 1 of ‘_mm512_abs_p’
accum_1 += _mm512_abs_pd(_mm512_load_pd(&x1[i + 8]));
^~~~~~~~~~~~~~
In file included from /opt/rh/devtoolset-7/root/usr/lib/gcc/x86_64-redhat-linux/7/include/immintrin.h:45:0,
from ../kernel/x86_64/dasum_microk_skylakex-2.c:6,
from ../kernel/x86_64/dasum.c:8:
/opt/rh/devtoolset-7/root/usr/lib/gcc/x86_64-redhat-linux/7/include/avx512fintrin.h:7301:1: note: expected ‘__m512 {aka __vector(16) float}’ but argument is of type ‘__m512d {aka __vector(8) double}’
_mm512_abs_pd (__m512 __A)
^~~~~~~~~~~~~
In file included from ../kernel/x86_64/dasum.c:8:0:
../kernel/x86_64/dasum_microk_skylakex-2.c:44:38: error: incompatible type for argument 1 of ‘_mm512_abs_p’
accum_2 += _mm512_abs_pd(_mm512_load_pd(&x1[i +16]));
^~~~~~~~~~~~~~
In file included from /opt/rh/devtoolset-7/root/usr/lib/gcc/x86_64-redhat-linux/7/include/immintrin.h:45:0,
from ../kernel/x86_64/dasum_microk_skylakex-2.c:6,
from ../kernel/x86_64/dasum.c:8:
/opt/rh/devtoolset-7/root/usr/lib/gcc/x86_64-redhat-linux/7/include/avx512fintrin.h:7301:1: note: expected ‘__m512 {aka __vector(16) float}’ but argument is of type ‘__m512d {aka __vector(8) double}’
_mm512_abs_pd (__m512 __A)
^~~~~~~~~~~~~
In file included from ../kernel/x86_64/dasum.c:8:0:
../kernel/x86_64/dasum_microk_skylakex-2.c:45:38: error: incompatible type for argument 1 of ‘_mm512_abs_p’
accum_3 += _mm512_abs_pd(_mm512_load_pd(&x1[i +24]));
^~~~~~~~~~~~~~
In file included from /opt/rh/devtoolset-7/root/usr/lib/gcc/x86_64-redhat-linux/7/include/immintrin.h:45:0,
from ../kernel/x86_64/dasum_microk_skylakex-2.c:6,
from ../kernel/x86_64/dasum.c:8:
/opt/rh/devtoolset-7/root/usr/lib/gcc/x86_64-redhat-linux/7/include/avx512fintrin.h:7301:1: note: expected ‘__m512 {aka __vector(16) float}’ but argument is of type ‘__m512d {aka __vector(8) double}’
_mm512_abs_pd (__m512 __A)
^~~~~~~~~~~~~
In file included from ../kernel/x86_64/dasum.c:8:0:
../kernel/x86_64/dasum_microk_skylakex-2.c:61:64: warning: passing argument 1 of ‘_mm_loadu_si128’ from incompatible pointer type [-Wincompatible-pointer-types]
accum_20 += (__m128d)_mm_and_si128(_mm_loadu_si128(&x1[i + 0]), abs_mask2);

[martin-frbg]

Which gcc version is this ? This may not be the cleanest way to code it, but it should not raise an error. (Definitely builds with versions 7.5.0 and 10.2)

[brada4]

Should be 7.1.1 or 7.3.1 as per softwarecollections.org

[minchaoyue]

Used gcc version 7.3.1 20180303 (Red Hat 7.3.1-5) (GCC)

[brada4]

Does it continue with devtoolset-8?

[martin-frbg]

This appears to be a genuine compiler (or header) bug - https://gcc.gnu.org/legacy-ml/gcc-patches/2018-01/msg01962.html

[minchaoyue]

@martin-frbg thanks for your help!!