Merge commit 'bfe5454cd238b16e7977085f880205229103eccb'

* commit 'bfe5454cd238b16e7977085f880205229103eccb': lavf: move ff_codec_get_tag() and ff_codec_get_id() definitions to internal.h lavf: move "MP3 " fourcc from riff to nut fate: vpx: Add dependencies fate: Fix wavpack-matroskamode test dependencies x86: dsputilenc: port to cpuflags Conflicts: libavformat/internal.h libavformat/nut.c tests/fate/vpx.mak Merged-by: Michael Niedermayer <[email protected]>
SaschaVasarrhelyi · Nov 29, 2012 · 076300b · 076300b
2 parents 7dc0ed8 + bfe5454
commit 076300b
Show file tree

Hide file tree

Showing 23 changed files with 87 additions and 93 deletions.
diff --git a/libavcodec/x86/dsputilenc.asm b/libavcodec/x86/dsputilenc.asm
@@ -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
@@ -144,43 +142,43 @@ 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
     jne .done
 
     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:
@@ -191,15 +189,33 @@ 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)
 ; r0 = void *s = unused, int h = unused (always 8)
 ; 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

diff --git a/libavformat/aiff.h b/libavformat/aiff.h
@@ -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') },

diff --git a/libavformat/au.c b/libavformat/au.c
@@ -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))

diff --git a/libavformat/cafdec.c b/libavformat/cafdec.c
@@ -27,7 +27,6 @@
 
 #include "avformat.h"
 #include "internal.h"
-#include "riff.h"
 #include "isom.h"
 #include "mov_chan.h"
 #include "libavutil/intreadwrite.h"

diff --git a/libavformat/gxfenc.c b/libavformat/gxfenc.c
@@ -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

diff --git a/libavformat/internal.h b/libavformat/internal.h
@@ -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.
  *

diff --git a/libavformat/isom.c b/libavformat/isom.c
@@ -26,7 +26,6 @@
 #include "avformat.h"
 #include "internal.h"
 #include "isom.h"
-#include "riff.h"
 #include "libavcodec/mpeg4audio.h"
 #include "libavcodec/mpegaudiodata.h"
 

diff --git a/libavformat/lxfdec.c b/libavformat/lxfdec.c
@@ -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

diff --git a/libavformat/nsvdec.c b/libavformat/nsvdec.c
@@ -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"
 

diff --git a/libavformat/nut.c b/libavformat/nut.c
@@ -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                         }
 };
 

diff --git a/libavformat/nut.h b/libavformat/nut.h
@@ -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))

diff --git a/libavformat/nutdec.c b/libavformat/nutdec.c
@@ -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 */
 

diff --git a/libavformat/nutenc.c b/libavformat/nutenc.c
@@ -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])

diff --git a/libavformat/omadec.c b/libavformat/omadec.c
@@ -46,7 +46,6 @@
 #include "libavutil/des.h"
 #include "oma.h"
 #include "pcm.h"
-#include "riff.h"
 #include "id3v2.h"
 
 

diff --git a/libavformat/riff.c b/libavformat/riff.c
@@ -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 },
 };
 

diff --git a/libavformat/riff.h b/libavformat/riff.h
@@ -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];

diff --git a/libavformat/rmdec.c b/libavformat/rmdec.c
@@ -26,7 +26,6 @@
 #include "libavutil/dict.h"
 #include "avformat.h"
 #include "internal.h"
-#include "riff.h"
 #include "rmsipr.h"
 #include "rm.h"
 

diff --git a/libavformat/rsodec.c b/libavformat/rsodec.c
@@ -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)

diff --git a/libavformat/smjpegenc.c b/libavformat/smjpegenc.c
@@ -26,7 +26,6 @@
 
 #include "avformat.h"
 #include "internal.h"
-#include "riff.h"
 #include "smjpeg.h"
 
 typedef struct SMJPEGMuxContext {

diff --git a/libavformat/swf.h b/libavformat/swf.h
@@ -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)

diff --git a/libavformat/voc.h b/libavformat/voc.h
@@ -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;