FFmpeg merge diff after patches.

configure

Index: configure
diff --git a/configure b/configure
index 00f2ebe83a713243b25f5fb11b9a539f1c63dc85..d7f2e493c7bd93d842fca9fbf9288b92bcc12566 100755
--- a/configure
+++ b/configure
@@ -2152,7 +2152,7 @@ elif $cc -v 2>&1 | grep -qi ^gcc; then
         CC_DEPFLAGS='-MMD -MF $(@:.o=.d) -MT $@'
         AS_DEPFLAGS='-MMD -MF $(@:.o=.d) -MT $@'
     fi
-    speed_cflags='-O3'
+    speed_cflags='-O2'
     size_cflags='-Os'
 elif $cc --version 2>/dev/null | grep -q Intel; then
     cc_type=icc
@@ -2305,6 +2305,7 @@ elif $cc -v 2>&1 | grep -q Open64; then
     cc_ident=$($cc -v 2>&1 | head -n1 | tr -d :)
     CC_DEPFLAGS='-MMD -MF $(@:.o=.d) -MT $@'
     AS_DEPFLAGS='-MMD -MF $(@:.o=.d) -MT $@'
+# O2 produces smaller/faster code on P4 and same performance on Core/Atom.

[scherkus (not reviewing), 2012/02/06 17:46:29]

this might be doable via configure script and build_ffmpeg.sh

[DaleCurtis, 2012/02/06 18:59:01]

Sounds right, I can specify --extra-cflags=-O2 and it will override whatever
optimization level was set previously.

[scherkus (not reviewing), 2012/02/06 19:42:58]

another sanity check: does the -O2 we specify here match ffmpeg.gyp?

otherwise we're using different -O for win vs mac/linux and perhaps can drop
this entirely :)

[DaleCurtis, 2012/02/07 03:18:01]

Yes, we're using -O2 inside the gyp.

     speed_cflags='-O2'
     size_cflags='-Os'
     filter_cflags='filter_out -Wdisabled-optimization|-Wtype-limits|-fno-signed-zeros'
@@ -3098,6 +3099,10 @@ if ! disabled pthreads && ! enabled w32threads && ! enabled os2threads; then
         add_extralibs -pthreads
     elif check_func pthread_create -lpthreadGC2; then
         add_extralibs -lpthreadGC2
+# static pthreads improves startup time
+    elif check_func pthread_create -lpthreadGC2 -lws2_32; then
+        add_cflags -DPTW32_STATIC_LIB
+        add_extralibs -lpthreadGC2 -lws2_32

[scherkus (not reviewing), 2012/02/06 17:46:29]

this might be doable via configure script and build_ffmpeg.sh

[DaleCurtis, 2012/02/06 18:59:01]

Again, looks right. Here these would just be specified w/ win32.

     elif ! check_lib pthread.h pthread_create -lpthread; then
         disable pthreads
     fi
@@ -3380,7 +3385,8 @@ elif enabled ccc; then
 elif enabled gcc; then
     check_cflags -fno-tree-vectorize
     check_cflags -Werror=implicit-function-declaration
-    check_cflags -Werror=missing-prototypes
+# Removed to allow msys gcc 4.2.1-sjlj to compile
+#   check_cflags -Werror=missing-prototypes

[scherkus (not reviewing), 2012/02/06 17:46:29]

we should see whether this is still valid w/ our current windows toolchain

[DaleCurtis, 2012/02/06 18:59:01]

Will do, we're up to 4.5.0 on the current toolchain.

 elif enabled llvm_gcc; then
     check_cflags -mllvm -stack-alignment=16
 elif enabled clang; then