More asm improvements

- Only use the asm code if we're on x86_64.
- More changes to decouple asm from simd.
- Check if the -Wa,--noexecstack option works.
- Support --disable-asm configure option.

diff --git a/Makefile.in b/Makefile.in
index 2d76984606bb88e2550bb56122a88a670e1e607a..f52976e90c9129781b51712debe64e6f2f66bbbb 100644 (file)
--- a/Makefile.in
+++ b/Makefile.in
@@ -33,6 +33,7 @@ VERSION=@RSYNC_VERSION@
 .SUFFIXES: .c .o
 
 SIMD_x86_64=simd-checksum-x86_64.o
+ASM_x86_64=lib/md5-asm-x86_64.o
 
 GENFILES=configure.sh aclocal.m4 config.h.in proto.h proto.h-tstamp rsync.1 rsync.1.html \
         rsync-ssl.1 rsync-ssl.1.html rsyncd.conf.5 rsyncd.conf.5.html
@@ -46,7 +47,7 @@ OBJS1=flist.o rsync.o generator.o receiver.o cleanup.o sender.o exclude.o \
        util.o util2.o main.o checksum.o match.o syscall.o log.o backup.o delete.o
 OBJS2=options.o io.o compat.o hlink.o token.o uidlist.o socket.o hashtable.o \
        fileio.o batch.o clientname.o chmod.o acls.o xattrs.o
-OBJS3=progress.o pipe.o lib/md5-asm-x86_64.o
+OBJS3=progress.o pipe.o @ASM@
 DAEMON_OBJ = params.o loadparm.o clientserver.o access.o connection.o authenticate.o
 popt_OBJS=popt/findme.o  popt/popt.o  popt/poptconfig.o \
        popt/popthelp.o popt/poptparse.o
@@ -136,7 +137,7 @@ simd-checksum-x86_64.o: simd-checksum-x86_64.cpp
        $(CXX) -I. $(CXXFLAGS) $(CPPFLAGS) -c -o $@ $(srcdir)/simd-checksum-x86_64.cpp
 
 lib/md5-asm-x86_64.o: lib/md5-asm-x86_64.S config.h lib/md-defines.h
-       $(CC) -I. -Wa,--noexecstack -c -o $@ $(srcdir)/lib/md5-asm-x86_64.S
+       $(CC) -I. @NOEXECSTACK@ -c -o $@ $(srcdir)/lib/md5-asm-x86_64.S
 
 tls$(EXEEXT): $(TLS_OBJ)
        $(CC) $(CFLAGS) $(LDFLAGS) -o $@ $(TLS_OBJ) $(LIBS)

diff --git a/configure.ac b/configure.ac
index 208d4de7977df881aaaabb54599e91ab490f98e6..c99fe0763797cc6e2514eb3d001ed6f68d7c3fac 100644 (file)
--- a/configure.ac
+++ b/configure.ac
@@ -253,6 +253,41 @@ fi
 
 AC_SUBST(SIMD)
 
+AC_MSG_CHECKING([if assembler accepts noexecstack])
+OLD_CFLAGS="$CFLAGS"
+CFLAGS="$CFLAGS -Wa,--noexecstack"
+AC_COMPILE_IFELSE([AC_LANG_PROGRAM([[ ]], [[return 0;]])],
+[ NOEXECSTACK='-Wa,--noexecstack' ; AC_MSG_RESULT(yes) ],
+[ NOEXECSTACK='' ; AC_MSG_RESULT(no) ])
+CFLAGS="$OLD_CFLAGS"
+AC_SUBST(NOEXECSTACK)
+
+ASM=
+
+AC_MSG_CHECKING([whether to enable ASM optimizations])
+AC_ARG_ENABLE(asm,
+    AS_HELP_STRING([--disable-asm],[disable ASM optimizations]))
+
+if test x"$enable_asm" != x"no"; then
+    if test x"$build_cpu" = x"x86_64"; then
+       ASM="$build_cpu"
+    elif test x"$enable_asm" = x"yes"; then
+        AC_MSG_RESULT(unavailable)
+        AC_MSG_ERROR(The ASM optimizations are currently x86_64 only.
+Omit --enable-asm to continue without it.)
+    fi
+fi
+
+if test x"$ASM" != x""; then
+    AC_MSG_RESULT([yes ($ASM)])
+    AC_DEFINE(HAVE_ASM, 1, [Define to 1 to enable ASM optimizations])
+    ASM='$(ASM_'"$ASM)"
+else
+    AC_MSG_RESULT(no)
+fi
+
+AC_SUBST(ASM)
+
 # arrgh. libc in some old debian version screwed up the largefile
 # stuff, getting byte range locking wrong
 AC_CACHE_CHECK([for broken largefile support],rsync_cv_HAVE_BROKEN_LARGEFILE,[

diff --git a/lib/md5.c b/lib/md5.c
index a2eebdb0e0db96096c3446bb039d2a415a83c816..41f158b82f7c90ee117d3e2689b4193e08aad6cd 100644 (file)
--- a/lib/md5.c
+++ b/lib/md5.c
@@ -148,7 +148,7 @@ static void md5_process(md_context *ctx, const uchar data[CSUM_CHUNK])
        ctx->D += D;
 }
 
-#if defined(HAVE_SIMD) && (CSUM_CHUNK == 64)
+#if defined HAVE_ASM && CSUM_CHUNK == 64
 extern void md5_process_asm(md_context *ctx, const void *data, size_t num);
 #endif
 
@@ -176,7 +176,7 @@ void md5_update(md_context *ctx, const uchar *input, uint32 length)
                left = 0;
        }
 
-#if defined(HAVE_SIMD) && (CSUM_CHUNK == 64)
+#if defined HAVE_ASM && CSUM_CHUNK == 64
        if (length >= CSUM_CHUNK) {
                uint32 chunks = length / CSUM_CHUNK;
                md5_process_asm(ctx, input, chunks);