[FFmpeg-cvslog] configure: Enable GCC vectorization on ≥4.9 on x86

Timothy Gu git at videolan.org
Tue Feb 9 01:25:58 CET 2016


ffmpeg | branch: master | Timothy Gu <timothygu99 at gmail.com> | Sat Jan 30 10:52:44 2016 -0800| [cb8646af24bd8e9627cc5e1c62b049a00fe0b07b] | committer: Timothy Gu

configure: Enable GCC vectorization on ≥4.9 on x86

4.9 was released precisely nine years after the first GCC version with
autovectorizer (4.0) and six years after the first GCC version with
`-ftree-vectorize` default to enabled on `-O3` (4.3). We've given GCC
enough time to fix those bugs.

FATE passes here on a x86-64 machine with both GCC 4.9.2 and 5.3.1.

Some optimization hotspots benefit greatly from this change, especially
those without handwritten assembly. For instance, the main function in
vf_phase is now 1.6x faster (1.2x overall) on my machine.

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=cb8646af24bd8e9627cc5e1c62b049a00fe0b07b
---

 configure |    6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/configure b/configure
index 1000cb1..2148f11 100755
--- a/configure
+++ b/configure
@@ -5935,7 +5935,11 @@ elif enabled ccc; then
     add_cflags -msg_disable nonstandcast
     add_cflags -msg_disable unsupieee
 elif enabled gcc; then
-    check_optflags -fno-tree-vectorize
+    case $gcc_basever in
+        4.9*) enabled x86 || check_optflags -fno-tree-vectorize ;;
+        4.*)                 check_optflags -fno-tree-vectorize ;;
+        *)    enabled x86 || check_optflags -fno-tree-vectorize ;;
+    esac
     check_cflags -Werror=format-security
     check_cflags -Werror=implicit-function-declaration
     check_cflags -Werror=missing-prototypes



More information about the ffmpeg-cvslog mailing list