Commit e8e65119d10f6c2a54e7404ddc6269d8be62ddbd
2019-11-05T19:18:32
add COMPILE_LTO to test link time optimization the library performs best with lto since many small functions can be inlined across objects.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102
diff --git a/.travis.yml b/.travis.yml
index a269c4e..64ab75c 100644
--- a/.travis.yml
+++ b/.travis.yml
@@ -163,6 +163,10 @@ matrix:
packages:
- clang-4.0
+ # Link time optimization
+ - env: SANITIZER=1 COMPILE_LTO=1 BUILDOPTIONS='--with-cc=gcc-5 --with-m64 --with-travis-valgrind'
+ - env: SANITIZER=1 COMPILE_LTO=1 BUILDOPTIONS='--with-cc=clang-7 --with-m64 --with-travis-valgrind'
+
# GCC for the x86-64 architecture with restricted limb sizes
# formerly started with the option "--with-low-mp" to testme.sh
# but testing all three in one run took to long and timed out.
diff --git a/demo/shared.c b/demo/shared.c
index 5ada460..7ef4756 100644
--- a/demo/shared.c
+++ b/demo/shared.c
@@ -3,7 +3,7 @@
void ndraw(const mp_int *a, const char *name)
{
char *buf;
- size_t size;
+ size_t size = 0;
mp_err err;
if ((err = mp_radix_size(a, 10, &size)) != MP_OKAY) {
diff --git a/makefile b/makefile
index 4045390..a04cd4d 100644
--- a/makefile
+++ b/makefile
@@ -77,7 +77,7 @@ profiled_single: pre_gen
rm -f *.o timing
$(CC) $(LTM_CFLAGS) -fbranch-probabilities -c pre_gen/tommath_amalgam.c -o tommath_amalgam.o
$(AR) $(ARFLAGS) $(LIBNAME) tommath_amalgam.o
- ranlib $(LIBNAME)
+ $(RANLIB) $(LIBNAME)
install: $(LIBNAME)
install -d $(DESTDIR)$(LIBPATH)
diff --git a/makefile_include.mk b/makefile_include.mk
index 4b5961c..aebbc8f 100644
--- a/makefile_include.mk
+++ b/makefile_include.mk
@@ -86,6 +86,12 @@ LTM_CFLAGS += -O3 -funroll-loops
LTM_CFLAGS += -fomit-frame-pointer
endif
+ifdef COMPILE_LTO
+LTM_CFLAGS += -flto
+AR = $(subst clang,llvm-ar,$(subst gcc,gcc-ar,$(CC)))
+RANLIB = $(subst clang,llvm-ranlib,$(subst gcc,gcc-ranlib,$(CC)))
+endif
+
endif # COMPILE_SIZE
ifneq ($(findstring clang,$(CC)),)
diff --git a/testme.sh b/testme.sh
index 6c71ac4..1a03f55 100755
--- a/testme.sh
+++ b/testme.sh
@@ -107,7 +107,7 @@ _make()
{
echo -ne " Compile $1 $2"
suffix=$(echo ${1}${2} | tr ' ' '_')
- CC="$1" CFLAGS="$2 $TEST_CFLAGS" make -j$MAKE_JOBS $3 $MAKE_OPTIONS > /dev/null 2>gcc_errors_${suffix}.log
+ CC="$1" CFLAGS="$2 $TEST_CFLAGS" make -j$MAKE_JOBS $3 $MAKE_OPTIONS 2>gcc_errors_${suffix}.log
errcnt=$(wc -l < gcc_errors_${suffix}.log)
if [[ ${errcnt} -gt 1 ]]; then
echo " failed"
@@ -354,7 +354,7 @@ _banner
if [[ "$TEST_VS_MTEST" != "" ]]
then
make clean > /dev/null
- _make "${compilers[0]} ${archflags[0]}" "$CFLAGS" "mtest_opponent"
+ _make "${compilers[0]}" "${archflags[0]} $CFLAGS" "mtest_opponent"
echo
_make "gcc" "$MTEST_RAND" "mtest"
echo
@@ -394,15 +394,15 @@ do
fi
if [[ "$VALGRIND_BIN" != "" ]]
then
- _runvalgrind "$i $a" "$CFLAGS"
+ _runvalgrind "$i" "$a $CFLAGS"
[ "$WITH_LOW_MP" != "1" ] && continue
- _runvalgrind "$i $a" "-DMP_16BIT $CFLAGS"
- _runvalgrind "$i $a" "-DMP_32BIT $CFLAGS"
+ _runvalgrind "$i" "$a -DMP_16BIT $CFLAGS"
+ _runvalgrind "$i" "$a -DMP_32BIT $CFLAGS"
else
- _runtest "$i $a" "$CFLAGS"
+ _runtest "$i" "$a $CFLAGS"
[ "$WITH_LOW_MP" != "1" ] && continue
- _runtest "$i $a" "-DMP_16BIT $CFLAGS"
- _runtest "$i $a" "-DMP_32BIT $CFLAGS"
+ _runtest "$i" "$a -DMP_16BIT $CFLAGS"
+ _runtest "$i" "$a -DMP_32BIT $CFLAGS"
fi
done
done