diff --git a/Makefile.pre.in b/Makefile.pre.in index cfa6a5f4ef5..17c0b5785d0 100644 --- a/Makefile.pre.in +++ b/Makefile.pre.in @@ -504,7 +504,7 @@ profile-opt: $(MAKE) profile-removal build_all_generate_profile: - $(MAKE) all CFLAGS_NODIST="$(CFLAGS) $(PGO_PROF_GEN_FLAG)" LDFLAGS="$(LDFLAGS) $(PGO_PROF_GEN_FLAG)" LIBS="$(LIBS)" + $(MAKE) all CFLAGS_NODIST="$(CFLAGS) $(PGO_PROF_GEN_FLAG) @LTOFLAGS@" LDFLAGS="$(LDFLAGS) $(PGO_PROF_GEN_FLAG) @LTOFLAGS@" LIBS="$(LIBS)" run_profile_task: : # FIXME: can't run for a cross build @@ -514,7 +514,7 @@ build_all_merge_profile: $(LLVM_PROF_MERGER) build_all_use_profile: - $(MAKE) all CFLAGS_NODIST="$(CFLAGS) $(PGO_PROF_USE_FLAG)" + $(MAKE) all CFLAGS_NODIST="$(CFLAGS) $(PGO_PROF_USE_FLAG) @LTOFLAGS@" LDFLAGS="$(LDFLAGS) @LTOFLAGS@" # Compile and run with gcov .PHONY=coverage coverage-lcov coverage-report diff --git a/Misc/NEWS b/Misc/NEWS index 124866423d8..8bf583f4d04 100644 --- a/Misc/NEWS +++ b/Misc/NEWS @@ -388,6 +388,12 @@ Tests Build ----- +- Issue #25702: A --with-lto configure option has been added that will + enable link time optimizations at build time during a make profile-opt. + Some compilers and toolchains are known to not produce stable code when + using LTO, be sure to test things thoroughly before relying on it. + It can provide a few % speed up over profile-opt alone. + - Issue #26624: Adds validation of ucrtbase[d].dll version with warning for old versions. diff --git a/configure b/configure index 0eff26d0054..68970848702 100755 --- a/configure +++ b/configure @@ -673,6 +673,7 @@ LLVM_PROF_FILE LLVM_PROF_MERGER PGO_PROF_USE_FLAG PGO_PROF_GEN_FLAG +LTOFLAGS ABIFLAGS LN MKDIR_P @@ -807,6 +808,7 @@ with_suffix enable_shared enable_profiling with_pydebug +with_lto with_hash_algorithm with_address_sanitizer with_libs @@ -1487,6 +1489,8 @@ Optional Packages: compiler --with-suffix=.exe set executable suffix --with-pydebug build with Py_DEBUG defined + --with-lto Enable Link Time Optimization in PGO builds. + Disabled by default. --with-hash-algorithm=[fnv|siphash24] select hash algorithm --with-address-sanitizer @@ -6560,6 +6564,48 @@ $as_echo "no" >&6; } fi +# Enable LTO flags + +{ $as_echo "$as_me:${as_lineno-$LINENO}: checking for --with-lto" >&5 +$as_echo_n "checking for --with-lto... " >&6; } + +# Check whether --with-lto was given. +if test "${with_lto+set}" = set; then : + withval=$with_lto; +if test "$withval" != no +then + Py_LTO='true' + { $as_echo "$as_me:${as_lineno-$LINENO}: result: yes" >&5 +$as_echo "yes" >&6; }; +else + Py_LTO='false' + { $as_echo "$as_me:${as_lineno-$LINENO}: result: no" >&5 +$as_echo "no" >&6; }; +fi +else + { $as_echo "$as_me:${as_lineno-$LINENO}: result: no" >&5 +$as_echo "no" >&6; } +fi + +if test "$Py_LTO" = 'true' ; then + case $CC in + *clang*) + # Any changes made here should be reflected in the GCC+Darwin case below + LTOFLAGS="-flto" + ;; + *gcc*) + case $ac_sys_system in + Darwin*) + LTOFLAGS="-flto" + ;; + *) + LTOFLAGS="-flto -fuse-linker-plugin -ffat-lto-objects -flto-partition=none" + ;; + esac + ;; + esac +fi + # Enable PGO flags. diff --git a/configure.ac b/configure.ac index 71498300141..b13cf059a75 100644 --- a/configure.ac +++ b/configure.ac @@ -1235,6 +1235,39 @@ else AC_MSG_RESULT(no); Py_DEBUG='false' fi], [AC_MSG_RESULT(no)]) +# Enable LTO flags +AC_SUBST(LTOFLAGS) +AC_MSG_CHECKING(for --with-lto) +AC_ARG_WITH(lto, AS_HELP_STRING([--with-lto], [Enable Link Time Optimization in PGO builds. Disabled by default.]), +[ +if test "$withval" != no +then + Py_LTO='true' + AC_MSG_RESULT(yes); +else + Py_LTO='false' + AC_MSG_RESULT(no); +fi], +[AC_MSG_RESULT(no)]) +if test "$Py_LTO" = 'true' ; then + case $CC in + *clang*) + # Any changes made here should be reflected in the GCC+Darwin case below + LTOFLAGS="-flto" + ;; + *gcc*) + case $ac_sys_system in + Darwin*) + LTOFLAGS="-flto" + ;; + *) + LTOFLAGS="-flto -fuse-linker-plugin -ffat-lto-objects -flto-partition=none" + ;; + esac + ;; + esac +fi + # Enable PGO flags. AC_SUBST(PGO_PROF_GEN_FLAG) AC_SUBST(PGO_PROF_USE_FLAG)