On Mon, 28 Jan 2013 11:32:52 -0800, "Ronald S. Bultje" <[email protected]> wrote: > From: "Ronald S. Bultje" <[email protected]> > diff --git a/libavcodec/sh4/Makefile b/libavcodec/sh4/Makefile > index aa17eab..6cdf9ca 100644 > --- a/libavcodec/sh4/Makefile > +++ b/libavcodec/sh4/Makefile > @@ -1,3 +1,5 @@ > OBJS += sh4/dsputil_align.o \ > sh4/dsputil_sh4.o \ > sh4/idct_sh4.o \ > + > +OBJS-$(CONFIG_HPELDSP) += sh4/hpeldsp_align.o > diff --git a/libavcodec/sh4/dsputil_align.c b/libavcodec/sh4/dsputil_align.c > index f723f76..bc9c403 100644 > --- a/libavcodec/sh4/dsputil_align.c > +++ b/libavcodec/sh4/dsputil_align.c > @@ -1,7 +1,8 @@ > /* > - * aligned/packed access motion > + * This is optimized for sh, which have post increment addressing (*p++). > + * Some CPU may be index (p[n]) faster than post increment (*p++). > * > - * Copyright (c) 2001-2003 BERO <[email protected]> > + * copyright (c) 2001-2003 BERO <[email protected]> > * > * This file is part of Libav. > * > @@ -20,35 +21,23 @@ > * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 > USA > */ >
This looks like it's moving a million things at once, namely merges qpel.c into dsputil_align.c, then splits off some parts to hpeldsp_align.c. Please split that, it's impossible to follow. Also hpeldsp_align.c #includes dsputil. Wasn't the whole point of this to make it stop depending on dsputil? -- Anton Khirnov _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
