@@ -99,35 +99,33 @@ SECTION .text

     paddusw         m0, m1

 %endmacro

-; FIXME: HSUM_* saturates at 64k, while an 8x8 hadamard or dct block can get up to

+; FIXME: HSUM saturates at 64k, while an 8x8 hadamard or dct block can get up to

 ; about 100k on extreme inputs. But that's very unlikely to occur in natural video,

 ; and it's even more unlikely to not have any alternative mvs/modes with lower cost.

-%macro HSUM_MMX 3

-    mova            %2, %1

-    psrlq           %1, 32

+%macro HSUM 3

+%if cpuflag(sse2)

+    movhlps         %2, %1

     paddusw         %1, %2

-    mova            %2, %1

-    psrlq           %1, 16

+    pshuflw         %2, %1, 0xE

+    paddusw         %1, %2

+    pshuflw         %2, %1, 0x1

     paddusw         %1, %2

     movd            %3, %1

-%endmacro

-

-%macro HSUM_MMXEXT 3

+%elif cpuflag(mmxext)

     pshufw          %2, %1, 0xE

     paddusw         %1, %2

     pshufw          %2, %1, 0x1

     paddusw         %1, %2

     movd            %3, %1

-%endmacro

-

-%macro HSUM_SSE2 3

-    movhlps         %2, %1

-    paddusw         %1, %2

-    pshuflw         %2, %1, 0xE

+%elif cpuflag(mmx)

+    mova            %2, %1

+    psrlq           %1, 32

     paddusw         %1, %2

-    pshuflw         %2, %1, 0x1

+    mova            %2, %1

+    psrlq           %1, 16

     paddusw         %1, %2

     movd            %3, %1

+%endif

 %endmacro

 %macro STORE4 5

@@ -144,30 +142,30 @@ SECTION .text

     mova            %5, [%1+mmsize*3]

 %endmacro

-%macro hadamard8_16_wrapper 3

-cglobal hadamard8_diff_%1, 4, 4, %2

+%macro hadamard8_16_wrapper 2

+cglobal hadamard8_diff, 4, 4, %1

 %ifndef m8

-    %assign pad %3*mmsize-(4+stack_offset&(mmsize-1))

+    %assign pad %2*mmsize-(4+stack_offset&(mmsize-1))

     SUB            rsp, pad

 %endif

-    call hadamard8x8_diff_%1

+    call hadamard8x8_diff %+ SUFFIX

 %ifndef m8

     ADD            rsp, pad

 %endif

     RET

-cglobal hadamard8_diff16_%1, 5, 6, %2

+cglobal hadamard8_diff16, 5, 6, %1

 %ifndef m8

-    %assign pad %3*mmsize-(4+stack_offset&(mmsize-1))

+    %assign pad %2*mmsize-(4+stack_offset&(mmsize-1))

     SUB            rsp, pad

 %endif

-    call hadamard8x8_diff_%1

+    call hadamard8x8_diff %+ SUFFIX

     mov            r5d, eax

     add             r1, 8

     add             r2, 8

-    call hadamard8x8_diff_%1

+    call hadamard8x8_diff %+ SUFFIX

     add            r5d, eax

     cmp            r4d, 16

@@ -175,12 +173,12 @@ cglobal hadamard8_diff16_%1, 5, 6, %2

     lea             r1, [r1+r3*8-8]

     lea             r2, [r2+r3*8-8]

-    call hadamard8x8_diff_%1

+    call hadamard8x8_diff %+ SUFFIX

     add            r5d, eax

     add             r1, 8

     add             r2, 8

-    call hadamard8x8_diff_%1

+    call hadamard8x8_diff %+ SUFFIX

     add            r5d, eax

 .done:

@@ -191,7 +189,25 @@ cglobal hadamard8_diff16_%1, 5, 6, %2

     RET

 %endmacro

-%macro HADAMARD8_DIFF_MMX 1

+%macro HADAMARD8_DIFF 0-1

+%if cpuflag(sse2)

+hadamard8x8_diff %+ SUFFIX:

+    lea                          r0, [r3*3]

+    DIFF_PIXELS_8                r1, r2,  0, r3, r0, rsp+gprsize

+    HADAMARD8

+%if ARCH_X86_64

+    TRANSPOSE8x8W                 0,  1,  2,  3,  4,  5,  6,  7,  8

+%else

+    TRANSPOSE8x8W                 0,  1,  2,  3,  4,  5,  6,  7, [rsp+gprsize], [rsp+mmsize+gprsize]

+%endif

+    HADAMARD8

+    ABS_SUM_8x8         rsp+gprsize

+    HSUM                        m0, m1, eax

+    and                         eax, 0xFFFF

+    ret

+

+hadamard8_16_wrapper %1, 3

+%elif cpuflag(mmx)

 ALIGN 16

 ; int hadamard8_diff_##cpu(void *s, uint8_t *src1, uint8_t *src2,

 ;                          int stride, int h)

@@ -199,7 +215,7 @@ ALIGN 16

 ; note how r1, r2 and r3 are not clobbered in this function, so 16x16

 ; can simply call this 2x2x (and that's why we access rsp+gprsize

 ; everywhere, which is rsp of calling func

-hadamard8x8_diff_%1:

+hadamard8x8_diff %+ SUFFIX:

     lea                          r0, [r3*3]

     ; first 4x8 pixels

@@ -236,53 +252,35 @@ hadamard8x8_diff_%1:

     and                         rax, 0xFFFF

     ret

-hadamard8_16_wrapper %1, 0, 14

-%endmacro

-

-%macro HADAMARD8_DIFF_SSE2 2

-hadamard8x8_diff_%1:

-    lea                          r0, [r3*3]

-    DIFF_PIXELS_8                r1, r2,  0, r3, r0, rsp+gprsize

-    HADAMARD8

-%if ARCH_X86_64

-    TRANSPOSE8x8W                 0,  1,  2,  3,  4,  5,  6,  7,  8

-%else

-    TRANSPOSE8x8W                 0,  1,  2,  3,  4,  5,  6,  7, [rsp+gprsize], [rsp+mmsize+gprsize]

+hadamard8_16_wrapper 0, 14

 %endif

-    HADAMARD8

-    ABS_SUM_8x8         rsp+gprsize

-    HSUM_SSE2                    m0, m1, eax

-    and                         eax, 0xFFFF

-    ret

-

-hadamard8_16_wrapper %1, %2, 3

 %endmacro

-INIT_MMX

+INIT_MMX mmx

 %define ABS1 ABS1_MMX

-%define HSUM HSUM_MMX

-HADAMARD8_DIFF_MMX mmx

+HADAMARD8_DIFF

+INIT_MMX mmxext

 %define ABS1 ABS1_MMXEXT

-%define HSUM HSUM_MMXEXT

-HADAMARD8_DIFF_MMX mmxext

+HADAMARD8_DIFF

-INIT_XMM

+INIT_XMM sse2

 %define ABS2 ABS2_MMXEXT

 %if ARCH_X86_64

 %define ABS_SUM_8x8 ABS_SUM_8x8_64

 %else

 %define ABS_SUM_8x8 ABS_SUM_8x8_32

 %endif

-HADAMARD8_DIFF_SSE2 sse2, 10

+HADAMARD8_DIFF 10

+INIT_XMM ssse3

 %define ABS2        ABS2_SSSE3

 %define ABS_SUM_8x8 ABS_SUM_8x8_64

-HADAMARD8_DIFF_SSE2 ssse3, 9

+HADAMARD8_DIFF 9

-INIT_XMM

+INIT_XMM sse2

 ; sse16_sse2(void *v, uint8_t * pix1, uint8_t * pix2, int line_size, int h)

-cglobal sse16_sse2, 5, 5, 8

+cglobal sse16, 5, 5, 8

     shr      r4d, 1

     pxor      m0, m0         ; mm0 = 0

     pxor      m7, m7         ; mm7 holds the sum

@@ -28,7 +28,7 @@

 #define AVFORMAT_AIFF_H

 #include "avformat.h"

-#include "riff.h"

+#include "internal.h"

 static const AVCodecTag ff_codec_aiff_tags[] = {

     { AV_CODEC_ID_PCM_S16BE,    MKTAG('N','O','N','E') },

@@ -31,7 +31,6 @@

 #include "internal.h"

 #include "avio_internal.h"

 #include "pcm.h"

-#include "riff.h"

 /* if we don't know the size in advance */

 #define AU_UNKNOWN_SIZE ((uint32_t)(~0))

@@ -27,7 +27,6 @@

 #include "avformat.h"

 #include "internal.h"

-#include "riff.h"

 #include "isom.h"

 #include "mov_chan.h"

 #include "libavutil/intreadwrite.h"

@@ -27,7 +27,6 @@

 #include "avformat.h"

 #include "internal.h"

 #include "gxf.h"

-#include "riff.h"

 #include "audiointerleave.h"

 #define GXF_AUDIO_PACKET_SIZE 65536

@@ -355,6 +355,10 @@ void ff_compute_frame_duration(int *pnum, int *pden, AVStream *st,

 int ff_get_audio_frame_size(AVCodecContext *enc, int size, int mux);

+unsigned int ff_codec_get_tag(const AVCodecTag *tags, enum AVCodecID id);

+

+enum AVCodecID ff_codec_get_id(const AVCodecTag *tags, unsigned int tag);

+

 /**

  * Chooses a timebase for muxing the specified stream.

  *

@@ -26,7 +26,6 @@

 #include "avformat.h"

 #include "internal.h"

 #include "isom.h"

-#include "riff.h"

 #include "libavcodec/mpeg4audio.h"

 #include "libavcodec/mpegaudiodata.h"

@@ -23,7 +23,6 @@

 #include "libavcodec/bytestream.h"

 #include "avformat.h"

 #include "internal.h"

-#include "riff.h"

 #define LXF_MAX_PACKET_HEADER_SIZE 256

 #define LXF_HEADER_DATA_SIZE    120

@@ -23,7 +23,6 @@

 #include "libavutil/mathematics.h"

 #include "avformat.h"

 #include "internal.h"

-#include "riff.h"

 #include "libavutil/dict.h"

 #include "libavutil/intreadwrite.h"

@@ -149,6 +149,7 @@ const AVCodecTag ff_nut_video_tags[] = {

 static const AVCodecTag nut_audio_extra_tags[] = {

     { AV_CODEC_ID_PCM_ALAW,         MKTAG('A', 'L', 'A', 'W') },

     { AV_CODEC_ID_PCM_MULAW,        MKTAG('U', 'L', 'A', 'W') },

+    { AV_CODEC_ID_MP3,              MKTAG('M', 'P', '3', ' ') },

     { AV_CODEC_ID_NONE,             0                         }

 };

@@ -26,7 +26,7 @@

 //#include "libavutil/adler32.h"

 //#include "libavcodec/mpegaudio.h"

 #include "avformat.h"

-#include "riff.h"

+#include "internal.h"

 #include "metadata.h"

 #define      MAIN_STARTCODE (0x7A561F5F04ADULL + (((uint64_t)('N'<<8) + 'M')<<48))

@@ -28,6 +28,7 @@

 #include "libavutil/tree.h"

 #include "avio_internal.h"

 #include "nut.h"

+#include "riff.h"

 #define NUT_MAX_STREAMS 256    /* arbitrary sanity check value */

@@ -28,6 +28,7 @@

 #include "nut.h"

 #include "internal.h"

 #include "avio_internal.h"

+#include "riff.h"

 static int find_expected_header(AVCodecContext *c, int size, int key_frame,

                                 uint8_t out[64])

@@ -46,7 +46,6 @@

 #include "libavutil/des.h"

 #include "oma.h"

 #include "pcm.h"

-#include "riff.h"

 #include "id3v2.h"

@@ -389,7 +389,6 @@ const AVCodecTag ff_codec_wav_tags[] = {

     // for NuppelVideo (nuv.c)

     { AV_CODEC_ID_PCM_S16LE, MKTAG('R', 'A', 'W', 'A') },

     { AV_CODEC_ID_MP3,       MKTAG('L', 'A', 'M', 'E') },

-    { AV_CODEC_ID_MP3,       MKTAG('M', 'P', '3', ' ') },

     { AV_CODEC_ID_NONE,      0 },

 };

@@ -54,8 +54,6 @@ int ff_get_wav_header(AVIOContext *pb, AVCodecContext *codec, int size);

 extern const AVCodecTag ff_codec_bmp_tags[];

 extern const AVCodecTag ff_codec_wav_tags[];

-unsigned int ff_codec_get_tag(const AVCodecTag *tags, enum AVCodecID id);

-enum AVCodecID ff_codec_get_id(const AVCodecTag *tags, unsigned int tag);

 void ff_parse_specific_params(AVCodecContext *stream, int *au_rate, int *au_ssize, int *au_scale);

 typedef uint8_t ff_asf_guid[16];

@@ -26,7 +26,6 @@

 #include "libavutil/dict.h"

 #include "avformat.h"

 #include "internal.h"

-#include "riff.h"

 #include "rmsipr.h"

 #include "rm.h"

@@ -25,7 +25,6 @@

 #include "avformat.h"

 #include "internal.h"

 #include "pcm.h"

-#include "riff.h"

 #include "rso.h"

 static int rso_read_header(AVFormatContext *s)

@@ -26,7 +26,6 @@

 #include "avformat.h"

 #include "internal.h"

-#include "riff.h"

 #include "smjpeg.h"

 typedef struct SMJPEGMuxContext {

@@ -32,7 +32,7 @@

 #include "libavutil/fifo.h"

 #include "avformat.h"

 #include "avio.h"

-#include "riff.h"    /* for CodecTag */

+#include "internal.h"

 /* should have a generic way to indicate probable size */

 #define DUMMY_FILE_SIZE   (100 * 1024 * 1024)

@@ -23,7 +23,7 @@

 #define AVFORMAT_VOC_H

 #include "avformat.h"

-#include "riff.h"    /* for CodecTag */

+#include "internal.h"

 typedef struct voc_dec_context {

     int64_t remaining_size;

@@ -1,34 +1,34 @@

-FATE_VP3 += fate-vp31

+FATE_VP3-$(CONFIG_AVI_DEMUXER) += fate-vp31

 fate-vp31: CMD = framecrc -i $(SAMPLES)/vp3/vp31.avi

-FATE_VP3 += fate-vp3-coeff-level64

+FATE_VP3-$(CONFIG_MATROSKA_DEMUXER) += fate-vp3-coeff-level64

 fate-vp3-coeff-level64: CMD = framecrc -i $(SAMPLES)/vp3/coeff_level64.mkv

-FATE_SAMPLES_AVCONV += $(FATE_VP3)

-fate-vp3: $(FATE_VP3)

+FATE_SAMPLES_AVCONV-$(CONFIG_VP3_DECODER) += $(FATE_VP3-yes)

+fate-vp3: $(FATE_VP3-yes)

-FATE_SAMPLES_AVCONV += fate-vp5

+FATE_SAMPLES_AVCONV-$(call DEMDEC, AVI, VP5) += fate-vp5

 fate-vp5: CMD = framecrc -i $(SAMPLES)/vp5/potter512-400-partial.avi -an

-FATE_VP6 += fate-vp60

+FATE_VP6-$(call DEMDEC, EA, VP6) += fate-vp60

 fate-vp60: CMD = framecrc -i $(SAMPLES)/ea-vp6/g36.vp6

-FATE_VP6 += fate-vp61

+FATE_VP6-$(call DEMDEC, EA, VP6) += fate-vp61

 fate-vp61: CMD = framecrc -i $(SAMPLES)/ea-vp6/MovieSkirmishGondor.vp6 -t 4

-FATE_VP6 += fate-vp6a

+FATE_VP6-$(call DEMDEC, FLV, VP6A) += fate-vp6a

 fate-vp6a: CMD = framecrc -i $(SAMPLES)/flash-vp6/300x180-Scr-f8-056alpha.flv

-FATE_VP6 += fate-vp6f

+FATE_VP6-$(call DEMDEC, FLV, VP6F) += fate-vp6f

 fate-vp6f: CMD = framecrc -i $(SAMPLES)/flash-vp6/clip1024.flv

-FATE_SAMPLES_AVCONV += $(FATE_VP6)

-fate-vp6: $(FATE_VP6)

+FATE_SAMPLES_AVCONV += $(FATE_VP6-yes)

+fate-vp6: $(FATE_VP6-yes)

 VP8_SUITE = 001 002 003 004 005 006 007 008 009 010 011 012 013 014 015 016 017

 define FATE_VP8_SUITE

-FATE_VP8 += fate-vp8-test-vector$(2)-$(1)

+FATE_VP8-$(CONFIG_IVF_DEMUXER) += fate-vp8-test-vector$(2)-$(1)

 fate-vp8-test-vector$(2)-$(1): CMD = framemd5 $(3) -i $(SAMPLES)/vp8-test-vectors-r1/vp80-00-comprehensive-$(1).ivf

 fate-vp8-test-vector$(2)-$(1): REF = $(SRC_PATH)/tests/ref/fate/vp8-test-vector-$(1)

 endef

@@ -38,16 +38,16 @@ $(foreach N,$(VP8_SUITE),$(eval $(call FATE_VP8_SUITE,$(N),$(1),$(2))))

 # FIXME this file contains two frames with identical timestamps,

 # so ffmpeg drops one of them

-FATE_VP8 += fate-vp8-sign-bias$(1)

+FATE_VP8-$(CONFIG_IVF_DEMUXER) += fate-vp8-sign-bias$(1)

 fate-vp8-sign-bias$(1): CMD = framemd5 $(2) -i $(SAMPLES)/vp8/sintel-signbias.ivf

 fate-vp8-sign-bias$(1): REF = $(SRC_PATH)/tests/ref/fate/vp8-sign-bias

-FATE_VP8 += fate-vp8-size-change$(1)

+FATE_VP8-$(CONFIG_MATROSKA_DEMUXER) += fate-vp8-size-change$(1)

 fate-vp8-size-change$(1): CMD = framemd5 $(2) -flags +bitexact -i $(SAMPLES)/vp8/frame_size_change.webm -frames:v 30 -sws_flags bitexact+bilinear

 fate-vp8-size-change$(1): REF = $(SRC_PATH)/tests/ref/fate/vp8-size-change

 endef

 $(eval $(call FATE_VP8_FULL))

 $(eval $(call FATE_VP8_FULL,-emu-edge,-flags +emu_edge))

-FATE_SAMPLES_AVCONV += $(FATE_VP8)

-fate-vp8: $(FATE_VP8)

+FATE_SAMPLES_AVCONV-$(CONFIG_VP8_DECODER) += $(FATE_VP8-yes)

+fate-vp8: $(FATE_VP8-yes)

@@ -83,8 +83,10 @@ fate-wavpack-clipping: CMD = md5 -i $(SAMPLES)/wavpack/special/clipping.wv -f s1

 FATE_WAVPACK += fate-wavpack-falsestereo

 fate-wavpack-falsestereo: CMD = md5 -i $(SAMPLES)/wavpack/special/false_stereo.wv -f s16le

-FATE_WAVPACK += fate-wavpack-matroskamode

+FATE_WAVPACK-$(call DEMDEC, WV, WAVPACK) += $(FATE_WAVPACK)

+

+FATE_WAVPACK-$(call DEMDEC, MATROSKA, WAVPACK) += fate-wavpack-matroskamode

 fate-wavpack-matroskamode: CMD = md5 -i $(SAMPLES)/wavpack/special/matroska_mode.mka -f s16le

-FATE_SAMPLES_AVCONV-$(call DEMDEC, WV, WAVPACK) += $(FATE_WAVPACK)

-fate-wavpack: $(FATE_WAVPACK)

+FATE_SAMPLES_AVCONV += $(FATE_WAVPACK-yes)

+fate-wavpack: $(FATE_WAVPACK-yes)