Vectorize the CRC-32 calculation on the s390x.

Use vector extensions when compiling for s390x and binutils knows
about them. At runtime, check whether kernel supports vector
extensions (it has to be not just the CPU, but also the kernel)
and choose between the regular and the vectorized implementations.

Co-authored-by: Eduard Stefes <eddy@linux.ibm.com>

1 files changed, 70 insertions, 0 deletions

diff --git a/configure b/configure
index 3770b03d..05bee1f4 100755
--- a/configure
+++ b/configure
@@ -95,6 +95,7 @@ memory=0
 undefined=0
 insecure=0
 unknown=0
+enable_crcvx=1
 old_cc="$CC"
 old_cflags="$CFLAGS"
 OBJC='$(OBJZ) $(OBJG)'
@@ -122,6 +123,7 @@ case "$1" in
       echo '  configure [--const] [--zprefix] [--prefix=PREFIX]  [--eprefix=EXPREFIX]' | tee -a configure.log
       echo '    [--insecure] [--static] [--64] [--libdir=LIBDIR] [--sharedlibdir=LIBDIR]' | tee -a configure.log
       echo '    [--includedir=INCLUDEDIR] [--archs="-arch i386 -arch x86_64"]' | tee -a configure.log
+      echo '    [--disable-crcvx]' | tee -a configure.log
         exit 0 ;;
     -p*=* | --prefix=*) prefix=`echo $1 | sed 's/.*=//'`; shift ;;
     -e*=* | --eprefix=*) exec_prefix=`echo $1 | sed 's/.*=//'`; shift ;;
@@ -150,6 +152,7 @@ case "$1" in
     --memory) memory=1; shift ;;
     --undefined) undefined=1; shift ;;
     --insecure) insecure=1; shift ;;
+    --disable-crcvx) enable_crcvx=0; shift ;;
     *) unknown=1; echo "unknown option ignored: $1" | tee -a configure.log; shift;;
     esac
 done
@@ -888,6 +891,70 @@ EOF
   fi
 fi
 
+# check for ibm s390x build
+HAVE_S390X=0
+cat > $test.c << EOF
+#ifndef __s390x__
+    #error
+#endif
+EOF
+if try $CC -c $CFLAGS $test.c; then 
+    echo "Checking for s390x build ... Yes." | tee -a configure.log
+    HAVE_S390X=1
+else
+    echo "Checking for s390x build ... No." | tee -a configure.log
+fi
+
+# check for ibm s390x vx vector extensions
+HAVE_S390X_VX=0
+if test $HAVE_S390X -eq 1 && test $enable_crcvx -eq 1 ; then
+    # preset the compiler specific flags
+    if test $clang -eq 1; then
+        VGFMAFLAG=-fzvector
+    else
+        VGFMAFLAG=-mzarch
+    fi
+
+    cat > $test.c <<EOF
+#ifndef __s390x__
+#error
+#endif
+#include <vecintrin.h>
+int main(void) {
+    unsigned long long a __attribute__((vector_size(16))) = { 0 };
+    unsigned long long b __attribute__((vector_size(16))) = { 0 };
+    unsigned char c __attribute__((vector_size(16))) = { 0 };
+    c = vec_gfmsum_accum_128(a, b, c);
+    return c[0];
+}
+EOF
+
+    # cflags already contains a valid march
+    if try $CC -c $CFLAGS $VGFMAFLAG $test.c; then
+        echo "Checking for s390x vx vector extension ... Yes." | tee -a configure.log
+        HAVE_S390X_VX=1
+    # or set march for our compile units
+    elif try $CC -c $CFLAGS $VGFMAFLAG -march=z13 $test.c; then
+        echo "Checking for s390x vx vector extension (march=z13) ... Yes." | tee -a configure.log
+        HAVE_S390X_VX=1
+        VGFMAFLAG="$VGFMAFLAG -march=z13"
+    # else we are not on s390x
+    else 
+        echo "Checking for s390x vx vector extension ... No." | tee -a configure.log
+    fi
+
+    # prepare compiling for s390x
+    if test $HAVE_S390X_VX -eq 1; then
+        CFLAGS="$CFLAGS -DHAVE_S390X_VX"
+        SFLAGS="$SFLAGS -DHAVE_S390X_VX"
+        OBJC="$OBJC crc32_vx.o"
+        PIC_OBJC="$PIC_OBJC crc32_vx.lo"
+    else
+        # target has no vx extension
+        VGFMAFLAG=""
+    fi
+fi
+
 # show the results in the log
 echo >> configure.log
 echo ALL = $ALL >> configure.log
@@ -919,6 +986,9 @@ echo mandir = $mandir >> configure.log
 echo prefix = $prefix >> configure.log
 echo sharedlibdir = $sharedlibdir >> configure.log
 echo uname = $uname >> configure.log
+echo HAVE_S390X = $HAVE_S390X >> configure.log
+echo HAVE_S390X_VX = $HAVE_S390X_VX >> configure.log
+echo VGFMAFLAG = $VGFMAFLAG >> configure.log
 
 # update Makefile with the configure results
 sed < ${SRCDIR}Makefile.in "