Merge remote-tracking branch 'qatar/release/0.8' into release/0.10

* qatar/release/0.8: (154 commits)
  Update Changelog for the 0.8.1 Release
  dca: include libavutil/mathematics.h for possibly missing M_SQRT1_2
  dca: don't use av_clip_uintp2().
  snow: check reference frame indices.
  snow: reject unsupported chroma shifts.
  xa_adpcm: limit filter to prevent xa_adpcm_table[] array bounds overruns.
  h264: increase reference poc list from 16 to 32.
  h264: stricter reference limit enforcement.
  h264: improve parsing of broken AVC SPS
  Replace computations of remaining bits with calls to get_bits_left().
  png: convert to bytestream2 API.
  roqvideo: convert to bytestream2 API.
  smc: port to bytestream2 API.
  tgq: convert to bytestream2 API.
  algmm: convert to bytestream2 API.
  jvdec: unbreak video decoding
  h264: Fix invalid interlaced/progressive MB combinations for direct mode prediction.
  libx264: add 'stats' private option for setting 2pass stats filename.
  libx264: fix help text for slice-max-size option.
  avconv: reindent
  ...

Conflicts:
	Changelog
	RELEASE
	avconv.c
	doc/APIchanges
	ffplay.c
	libavcodec/Makefile
	libavcodec/aacdec.c
	libavcodec/alsdec.c
	libavcodec/atrac3.c
	libavcodec/avcodec.h
	libavcodec/dvdata.c
	libavcodec/fraps.c
	libavcodec/golomb.h
	libavcodec/h264.c
	libavcodec/h264.h
	libavcodec/h264_cabac.c
	libavcodec/h264_cavlc.c
	libavcodec/h264_direct.c
	libavcodec/h264_parser.c
	libavcodec/h264_ps.c
	libavcodec/h264idct_template.c
	libavcodec/indeo3.c
	libavcodec/kgv1dec.c
	libavcodec/kmvc.c
	libavcodec/mjpegbdec.c
	libavcodec/mmvideo.c
	libavcodec/mpegaudiodec.c
	libavcodec/mpegvideo.h
	libavcodec/options.c
	libavcodec/pngdec.c
	libavcodec/roqvideodec.c
	libavcodec/shorten.c
	libavcodec/svq3.c
	libavcodec/utils.c
	libavcodec/version.h
	libavcodec/wmadec.c
	libavcodec/xxan.c
	libavformat/Makefile
	libavformat/asfdec.c
	libavformat/dv.c
	libavformat/mov.c
	libavformat/nsvdec.c
	libavformat/utils.c
	libavformat/version.h
	libavutil/avutil.h
	libavutil/error.c
	libavutil/error.h
	libswscale/swscale.c
	libswscale/utils.c
	libswscale/x86/swscale_template.c
	tests/ref/acodec/g722

Merged-by: Michael Niedermayer <michaelni@gmx.at>

3 files changed, 72 insertions, 31 deletions

diff --git a/libswscale/x86/scale.asm b/libswscale/x86/scale.asm
index 09313b926f..692a88fe39 100644
--- a/libswscale/x86/scale.asm
+++ b/libswscale/x86/scale.asm
@@ -38,7 +38,7 @@ SECTION .text
 ;                               (SwsContext *c, int{16,32}_t *dst,
 ;                                int dstW, const uint{8,16}_t *src,
 ;                                const int16_t *filter,
-;                                const int16_t *filterPos, int filterSize);
+;                                const int32_t *filterPos, int filterSize);
 ;
 ; Scale one horizontal line. Input is either 8-bits width or 16-bits width
 ; ($source_width can be either 8, 9, 10 or 16, difference is whether we have to
@@ -53,6 +53,9 @@ SECTION .text
 cglobal hscale%1to%2_%4_%5, %6, 7, %7
 %ifdef ARCH_X86_64
     movsxd        r2, r2d
+%define mov32 movsxd
+%else ; x86-32
+%define mov32 mov
 %endif ; x86-64
 %if %2 == 19
 %if mmsize == 8 ; mmx
@@ -95,14 +98,14 @@ cglobal hscale%1to%2_%4_%5, %6, 7, %7
 %else ; %2 == 19
     lea           r1, [r1+r2*(4>>r2shr)]
 %endif ; %2 == 15/19
-    lea           r5, [r5+r2*(2>>r2shr)]
+    lea           r5, [r5+r2*(4>>r2shr)]
     neg           r2
 
 .loop:
 %if %3 == 4 ; filterSize == 4 scaling
     ; load 2x4 or 4x4 source pixels into m0/m1
-    movsx         r0, word [r5+r2*2+0]   ; filterPos[0]
-    movsx         r6, word [r5+r2*2+2]   ; filterPos[1]
+    mov32         r0, dword [r5+r2*4+0]  ; filterPos[0]
+    mov32         r6, dword [r5+r2*4+4]  ; filterPos[1]
     movlh         m0, [r3+r0*srcmul]     ; src[filterPos[0] + {0,1,2,3}]
 %if mmsize == 8
     movlh         m1, [r3+r6*srcmul]     ; src[filterPos[1] + {0,1,2,3}]
@@ -112,8 +115,8 @@ cglobal hscale%1to%2_%4_%5, %6, 7, %7
 %else ; %1 == 8
     movd          m4, [r3+r6*srcmul]     ; src[filterPos[1] + {0,1,2,3}]
 %endif
-    movsx         r0, word [r5+r2*2+4]   ; filterPos[2]
-    movsx         r6, word [r5+r2*2+6]   ; filterPos[3]
+    mov32         r0, dword [r5+r2*4+8]  ; filterPos[2]
+    mov32         r6, dword [r5+r2*4+12] ; filterPos[3]
     movlh         m1, [r3+r0*srcmul]     ; src[filterPos[2] + {0,1,2,3}]
 %if %1 > 8
     movhps        m1, [r3+r6*srcmul]     ; src[filterPos[3] + {0,1,2,3}]
@@ -156,8 +159,8 @@ cglobal hscale%1to%2_%4_%5, %6, 7, %7
 %endif ; mmx/sse2/ssse3/sse4
 %else ; %3 == 8, i.e. filterSize == 8 scaling
     ; load 2x8 or 4x8 source pixels into m0, m1, m4 and m5
-    movsx         r0, word [r5+r2*1+0]   ; filterPos[0]
-    movsx         r6, word [r5+r2*1+2]   ; filterPos[1]
+    mov32         r0, dword [r5+r2*2+0]  ; filterPos[0]
+    mov32         r6, dword [r5+r2*2+4]  ; filterPos[1]
     movbh         m0, [r3+ r0   *srcmul] ; src[filterPos[0] + {0,1,2,3,4,5,6,7}]
 %if mmsize == 8
     movbh         m1, [r3+(r0+4)*srcmul] ; src[filterPos[0] + {4,5,6,7}]
@@ -165,8 +168,8 @@ cglobal hscale%1to%2_%4_%5, %6, 7, %7
     movbh         m5, [r3+(r6+4)*srcmul] ; src[filterPos[1] + {4,5,6,7}]
 %else ; mmsize == 16
     movbh         m1, [r3+ r6   *srcmul] ; src[filterPos[1] + {0,1,2,3,4,5,6,7}]
-    movsx         r0, word [r5+r2*1+4]   ; filterPos[2]
-    movsx         r6, word [r5+r2*1+6]   ; filterPos[3]
+    mov32         r0, dword [r5+r2*2+8]  ; filterPos[2]
+    mov32         r6, dword [r5+r2*2+12] ; filterPos[3]
     movbh         m4, [r3+ r0   *srcmul] ; src[filterPos[2] + {0,1,2,3,4,5,6,7}]
     movbh         m5, [r3+ r6   *srcmul] ; src[filterPos[3] + {0,1,2,3,4,5,6,7}]
 %endif ; mmsize == 8/16
@@ -251,7 +254,7 @@ cglobal hscale%1to%2_%4_%5, %6, 7, %7
 %define r1x     r1
 %define filter2 r6m
 %endif ; x86-32/64
-    lea           r5, [r5+r2*2]
+    lea           r5, [r5+r2*4]
 %if %2 == 15
     lea           r1, [r1+r2*2]
 %else ; %2 == 19
@@ -261,8 +264,8 @@ cglobal hscale%1to%2_%4_%5, %6, 7, %7
     neg           r2
 
 .loop:
-    movsx         r0, word [r5+r2*2+0]   ; filterPos[0]
-    movsx        r1x, word [r5+r2*2+2]   ; filterPos[1]
+    mov32         r0, dword [r5+r2*4+0]  ; filterPos[0]
+    mov32        r1x, dword [r5+r2*4+4]  ; filterPos[1]
     ; FIXME maybe do 4px/iteration on x86-64 (x86-32 wouldn't have enough regs)?
     pxor          m4, m4
     pxor          m5, m5
@@ -293,7 +296,7 @@ cglobal hscale%1to%2_%4_%5, %6, 7, %7
     jl .innerloop
 
 %ifidn %4, X4
-    movsx        r1x, word [r5+r2*2+2]   ; filterPos[1]
+    mov32        r1x, dword [r5+r2*4+4]  ; filterPos[1]
     movlh         m0, [src_reg+r0 *srcmul] ; split last 4 srcpx of dstpx[0]
     sub          r1x, r6                   ; and first 4 srcpx of dstpx[1]
 %if %1 > 8
diff --git a/libswscale/x86/swscale_mmx.c b/libswscale/x86/swscale_mmx.c
index d0bb861618..fdfd1f215a 100644
--- a/libswscale/x86/swscale_mmx.c
+++ b/libswscale/x86/swscale_mmx.c
@@ -108,8 +108,8 @@ void updateMMXDitherTables(SwsContext *c, int dstY, int lumBufIndex, int chrBufI
     int16_t **alpPixBuf= c->alpPixBuf;
     const int vLumBufSize= c->vLumBufSize;
     const int vChrBufSize= c->vChrBufSize;
-    int16_t *vLumFilterPos= c->vLumFilterPos;
-    int16_t *vChrFilterPos= c->vChrFilterPos;
+    int32_t *vLumFilterPos= c->vLumFilterPos;
+    int32_t *vChrFilterPos= c->vChrFilterPos;
     int16_t *vLumFilter= c->vLumFilter;
     int16_t *vChrFilter= c->vChrFilter;
     int32_t *lumMmxFilter= c->lumMmxFilter;
@@ -132,6 +132,44 @@ void updateMMXDitherTables(SwsContext *c, int dstY, int lumBufIndex, int chrBufI
         const int16_t **chrUSrcPtr= (const int16_t **)(void*) chrUPixBuf + chrBufIndex + firstChrSrcY - lastInChrBuf + vChrBufSize;
         const int16_t **alpSrcPtr= (CONFIG_SWSCALE_ALPHA && alpPixBuf) ? (const int16_t **)(void*) alpPixBuf + lumBufIndex + firstLumSrcY - lastInLumBuf + vLumBufSize : NULL;
         int i;
+
+        if (firstLumSrcY < 0 || firstLumSrcY + vLumFilterSize > c->srcH) {
+            const int16_t **tmpY = (const int16_t **) lumPixBuf + 2 * vLumBufSize;
+            int neg = -firstLumSrcY, i, end = FFMIN(c->srcH - firstLumSrcY, vLumFilterSize);
+            for (i = 0; i < neg;            i++)
+                tmpY[i] = lumSrcPtr[neg];
+            for (     ; i < end;            i++)
+                tmpY[i] = lumSrcPtr[i];
+            for (     ; i < vLumFilterSize; i++)
+                tmpY[i] = tmpY[i-1];
+            lumSrcPtr = tmpY;
+
+            if (alpSrcPtr) {
+                const int16_t **tmpA = (const int16_t **) alpPixBuf + 2 * vLumBufSize;
+                for (i = 0; i < neg;            i++)
+                    tmpA[i] = alpSrcPtr[neg];
+                for (     ; i < end;            i++)
+                    tmpA[i] = alpSrcPtr[i];
+                for (     ; i < vLumFilterSize; i++)
+                    tmpA[i] = tmpA[i - 1];
+                alpSrcPtr = tmpA;
+            }
+        }
+        if (firstChrSrcY < 0 || firstChrSrcY + vChrFilterSize > c->chrSrcH) {
+            const int16_t **tmpU = (const int16_t **) chrUPixBuf + 2 * vChrBufSize;
+            int neg = -firstChrSrcY, i, end = FFMIN(c->chrSrcH - firstChrSrcY, vChrFilterSize);
+            for (i = 0; i < neg;            i++) {
+                tmpU[i] = chrUSrcPtr[neg];
+            }
+            for (     ; i < end;            i++) {
+                tmpU[i] = chrUSrcPtr[i];
+            }
+            for (     ; i < vChrFilterSize; i++) {
+                tmpU[i] = tmpU[i - 1];
+            }
+            chrUSrcPtr = tmpU;
+        }
+
         if (flags & SWS_ACCURATE_RND) {
             int s= APCK_SIZE / 8;
             for (i=0; i<vLumFilterSize; i+=2) {
@@ -242,7 +280,7 @@ extern void ff_hscale ## from_bpc ## to ## to_bpc ## _ ## filter_n ## _ ## opt(
                                                 SwsContext *c, int16_t *data, \
                                                 int dstW, const uint8_t *src, \
                                                 const int16_t *filter, \
-                                                const int16_t *filterPos, int filterSize)
+                                                const int32_t *filterPos, int filterSize)
 
 #define SCALE_FUNCS(filter_n, opt) \
     SCALE_FUNC(filter_n,  8, 15, opt); \
diff --git a/libswscale/x86/swscale_template.c b/libswscale/x86/swscale_template.c
index e92d927440..61ee3ebae0 100644
--- a/libswscale/x86/swscale_template.c
+++ b/libswscale/x86/swscale_template.c
@@ -762,10 +762,10 @@ static void RENAME(yuv2yuyv422_X)(SwsContext *c, const int16_t *lumFilter,
     "1:                                 \n\t"\
     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
-    "add           "UV_OFFx2"("#c"), "#index"  \n\t" \
+    "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
-    "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \
+    "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
     "psubw             %%mm3, %%mm2     \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
     "psubw             %%mm4, %%mm5     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0    \n\t"\
@@ -993,10 +993,10 @@ static void RENAME(yuv2rgb565_2)(SwsContext *c, const int16_t *buf[2],
     "1:                                 \n\t"\
     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
-    "add           "UV_OFFx2"("#c"), "#index"  \n\t" \
+    "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
-    "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \
+    "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
     "psubw             %%mm3, %%mm2     \n\t" /* uvbuf0[eax] - uvbuf1[eax]*/\
     "psubw             %%mm4, %%mm5     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048]*/\
     "movq "CHR_MMX_FILTER_OFFSET"+8("#c"), %%mm0    \n\t"\
@@ -1048,9 +1048,9 @@ static void RENAME(yuv2yuyv422_2)(SwsContext *c, const int16_t *buf[2],
     ".p2align              4            \n\t"\
     "1:                                 \n\t"\
     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\
-    "add           "UV_OFFx2"("#c"), "#index"  \n\t" \
+    "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
     "movq     (%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\
-    "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \
+    "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
     "psraw                $4, %%mm3     \n\t" /* uvbuf0[eax] - uvbuf1[eax] >>4*/\
     "psraw                $4, %%mm4     \n\t" /* uvbuf0[eax+2048] - uvbuf1[eax+2048] >>4*/\
     "psubw  "U_OFFSET"("#c"), %%mm3     \n\t" /* (U-128)8*/\
@@ -1101,10 +1101,10 @@ static void RENAME(yuv2yuyv422_2)(SwsContext *c, const int16_t *buf[2],
     "1:                                 \n\t"\
     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
-    "add           "UV_OFFx2"("#c"), "#index"  \n\t" \
+    "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
-    "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \
+    "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
     "psrlw                $5, %%mm3     \n\t" /*FIXME might overflow*/\
@@ -1368,9 +1368,9 @@ static void RENAME(yuv2rgb565_1)(SwsContext *c, const int16_t *buf0,
     ".p2align              4            \n\t"\
     "1:                                 \n\t"\
     "movq     (%2, "#index"), %%mm3     \n\t" /* uvbuf0[eax]*/\
-    "add           "UV_OFFx2"("#c"), "#index"  \n\t" \
+    "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
     "movq     (%2, "#index"), %%mm4     \n\t" /* uvbuf0[eax+2048]*/\
-    "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \
+    "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
     "psraw                $7, %%mm3     \n\t" \
     "psraw                $7, %%mm4     \n\t" \
     "movq  (%0, "#index", 2), %%mm1     \n\t" /*buf0[eax]*/\
@@ -1386,10 +1386,10 @@ static void RENAME(yuv2rgb565_1)(SwsContext *c, const int16_t *buf0,
     "1:                                 \n\t"\
     "movq     (%2, "#index"), %%mm2     \n\t" /* uvbuf0[eax]*/\
     "movq     (%3, "#index"), %%mm3     \n\t" /* uvbuf1[eax]*/\
-    "add           "UV_OFFx2"("#c"), "#index"  \n\t" \
+    "add "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
     "movq     (%2, "#index"), %%mm5     \n\t" /* uvbuf0[eax+2048]*/\
     "movq     (%3, "#index"), %%mm4     \n\t" /* uvbuf1[eax+2048]*/\
-    "sub           "UV_OFFx2"("#c"), "#index"  \n\t" \
+    "sub "UV_OFF_BYTE"("#c"), "#index"  \n\t" \
     "paddw             %%mm2, %%mm3     \n\t" /* uvbuf0[eax] + uvbuf1[eax]*/\
     "paddw             %%mm5, %%mm4     \n\t" /* uvbuf0[eax+2048] + uvbuf1[eax+2048]*/\
     "psrlw                $8, %%mm3     \n\t" \
@@ -1579,7 +1579,7 @@ static void RENAME(hyscale_fast)(SwsContext *c, int16_t *dst,
                                  int dstWidth, const uint8_t *src,
                                  int srcW, int xInc)
 {
-    int16_t *filterPos = c->hLumFilterPos;
+    int32_t *filterPos = c->hLumFilterPos;
     int16_t *filter    = c->hLumFilter;
     void    *mmx2FilterCode= c->lumMmx2FilterCode;
     int i;
@@ -1675,7 +1675,7 @@ static void RENAME(hcscale_fast)(SwsContext *c, int16_t *dst1, int16_t *dst2,
                                  int dstWidth, const uint8_t *src1,
                                  const uint8_t *src2, int srcW, int xInc)
 {
-    int16_t *filterPos = c->hChrFilterPos;
+    int32_t *filterPos = c->hChrFilterPos;
     int16_t *filter    = c->hChrFilter;
     void    *mmx2FilterCode= c->chrMmx2FilterCode;
     int i;