Merge remote-tracking branch 'qatar/master'

* qatar/master: (22 commits)
  configure: enable memalign_hack automatically when needed
  swscale: unbreak the build on non-x86 systems.
  swscale: remove if(bitexact) branch from functions.
  swscale: remove if(canMMX2BeUsed) conditional.
  swscale: remove swScale_{c,MMX,MMX2} duplication.
  swscale: use emms_c().
  Move emms_c() from libavcodec to libavutil.
  tiff: set palette in the context when specified in TIFF_PAL tag
  rtsp: use strtoul to parse rtptime and seq values.
  pgssubdec: fix incorrect colors.
  dvdsubdec: fix incorrect colors.
  ape: Allow demuxing of files with metadata tags.
  swscale: remove dead macro WRITEBGR24OLD.
  swscale: remove AMD3DNOW "optimizations".
  swscale: remove duplicate code in ppc/ subdirectory.
  swscale: remove duplicated x86/ functions.
  swscale: force --enable-runtime-cpudetect and remove SWS_CPU_CAPS_*.
  vsrc_buffer.h: add file doxy
  vsrc_buffer: tweak error message in init()
  msmpeg4: reindent.
  ...
Merged-by: Michael Niedermayer <michaelni@gmx.at>

configure

@@ -2859,11 +2859,6 @@ check_header X11/extensions/XvMClib.h
 
 check_struct dxva2api.h DXVA_PictureParameters wDecodedPictureIndex
 
-if ! enabled_any memalign memalign_hack posix_memalign malloc_aligned &&
-     enabled_any $need_memalign ; then
-    die "Error, no aligned memory allocator but SSE enabled, disable it or use --enable-memalign-hack."
-fi
-
 disabled  zlib || check_lib   zlib.h      zlibVersion -lz   || disable  zlib
 disabled bzlib || check_lib2 bzlib.h BZ2_bzlibVersion -lbz2 || disable bzlib
 
@@ -3156,6 +3151,9 @@ check_deps $CONFIG_LIST       \
 
 enabled asm || { arch=c; disable $ARCH_LIST $ARCH_EXT_LIST; }
 
+! enabled_any memalign posix_memalign malloc_aligned &&
+    enabled_any $need_memalign && enable memalign_hack
+
 echo "install prefix            $prefix"
 echo "source path               $source_path"
 echo "C compiler                $cc"

doc/encoders.texi

@@ -433,3 +433,49 @@ For more information about libx264 and the supported options see:
 @url{http://www.videolan.org/developers/x264.html}
 
 @c man end VIDEO ENCODERS
+
+@subheading Floating-Point-Only AC-3 Encoding Options
+
+These options are only valid for the floating-point encoder and do not exist
+for the fixed-point encoder due to the corresponding features not being
+implemented in fixed-point.
+
+@table @option
+
+@item -channel_coupling @var{boolean}
+Enables/Disables use of channel coupling, which is an optional AC-3 feature
+that increases quality by combining high frequency information from multiple
+channels into a single channel. The per-channel high frequency information is
+sent with less accuracy in both the frequency and time domains. This allows
+more bits to be used for lower frequencies while preserving enough information
+to reconstruct the high frequencies. This option is enabled by default for the
+floating-point encoder and should generally be left as enabled except for
+testing purposes or to increase encoding speed.
+@table @option
+@item -1
+@itemx auto
+Selected by Encoder (default)
+@item 0
+@itemx off
+Disable Channel Coupling
+@item 1
+@itemx on
+Enable Channel Coupling
+@end table
+
+@item -cpl_start_band @var{number}
+Coupling Start Band. Sets the channel coupling start band, from 1 to 15. If a
+value higher than the bandwidth is used, it will be reduced to 1 less than the
+coupling end band. If @var{auto} is used, the start band will be determined by
+the encoder based on the bit rate, sample rate, and channel layout. This option
+has no effect if channel coupling is disabled.
+@table @option
+@item -1
+@itemx auto
+Selected by Encoder (default)
+@end table
+
+@end table
+
+@c man end ENCODERS
+

libavcodec/Makefile

@@ -269,8 +269,6 @@ OBJS-$(CONFIG_MPEG2VIDEO_ENCODER)      += mpeg12enc.o mpegvideo_enc.o \
                                           mpegvideo.o error_resilience.o
 OBJS-$(CONFIG_MPEG4_VAAPI_HWACCEL)     += vaapi_mpeg4.o
 OBJS-$(CONFIG_MSMPEG4V1_DECODER)       += msmpeg4.o msmpeg4data.o
-OBJS-$(CONFIG_MSMPEG4V1_ENCODER)       += msmpeg4.o msmpeg4data.o h263dec.o \
-                                          h263.o ituh263dec.o mpeg4videodec.o
 OBJS-$(CONFIG_MSMPEG4V2_DECODER)       += msmpeg4.o msmpeg4data.o h263dec.o \
                                           h263.o ituh263dec.o mpeg4videodec.o
 OBJS-$(CONFIG_MSMPEG4V2_ENCODER)       += msmpeg4.o msmpeg4data.o h263dec.o \

libavcodec/ac3.h

@@ -28,7 +28,8 @@
 #define AVCODEC_AC3_H
 
 #define AC3_MAX_CODED_FRAME_SIZE 3840 /* in bytes */
-#define AC3_MAX_CHANNELS 6 /* including LFE channel */
+#define AC3_MAX_CHANNELS 7            /**< maximum number of channels, including coupling channel */
+#define CPL_CH 0                      /**< coupling channel index */
 
 #define AC3_MAX_COEFS   256
 #define AC3_BLOCK_SIZE  256
@@ -158,7 +159,9 @@ typedef struct AC3EncOptions {
 
     /* other encoding options */
     int allow_per_frame_metadata;
-    int stereo_rematrixing;    
+    int stereo_rematrixing;
+    int channel_coupling;
+    int cpl_start;    
 } AC3EncOptions;
 
 

libavcodec/ac3dec.h

@@ -58,11 +58,6 @@
 #include "fft.h"
 #include "fmtconvert.h"
 
-/* override ac3.h to include coupling channel */
-#undef AC3_MAX_CHANNELS
-#define AC3_MAX_CHANNELS 7
-#define CPL_CH 0
-
 #define AC3_OUTPUT_LFEON  8
 
 #define SPX_MAX_BANDS    17

libavcodec/ac3dec_data.c

@@ -53,12 +53,6 @@ const uint8_t ff_eac3_hebap_tab[64] = {
     19, 19, 19, 19,
 };
 
-/**
- * Table E2.16 Default Coupling Banding Structure
- */
-const uint8_t ff_eac3_default_cpl_band_struct[18] =
-{ 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1 };
-
 /**
  * Table E2.15 Default Spectral Extension Banding Structure
  */

libavcodec/ac3dec_data.h

@@ -27,7 +27,6 @@
 extern const uint8_t ff_ac3_ungroup_3_in_5_bits_tab[32][3];
 
 extern const uint8_t ff_eac3_hebap_tab[64];
-extern const uint8_t ff_eac3_default_cpl_band_struct[18];
 extern const uint8_t ff_eac3_default_spx_band_struct[17];
 
 #endif /* AVCODEC_AC3DEC_DATA_H */

libavcodec/ac3enc_fixed.c

@@ -101,7 +101,7 @@ static void scale_coefficients(AC3EncodeContext *s)
 
     for (blk = 0; blk < AC3_MAX_BLOCKS; blk++) {
         AC3Block *block = &s->blocks[blk];
-        for (ch = 0; ch < s->channels; ch++) {
+        for (ch = 1; ch <= s->channels; ch++) {
             s->ac3dsp.ac3_rshift_int32(block->mdct_coef[ch], AC3_MAX_COEFS,
                                        block->coeff_shift[ch]);
         }

libavcodec/ac3enc_float.c

@@ -93,8 +93,10 @@ static int normalize_samples(AC3EncodeContext *s)
  */
 static void scale_coefficients(AC3EncodeContext *s)
 {
-    s->ac3dsp.float_to_fixed24(s->fixed_coef_buffer, s->mdct_coef_buffer,
-                               AC3_MAX_COEFS * AC3_MAX_BLOCKS * s->channels);
+    int chan_size = AC3_MAX_COEFS * AC3_MAX_BLOCKS;
+    s->ac3dsp.float_to_fixed24(s->fixed_coef_buffer + chan_size,
+                               s->mdct_coef_buffer  + chan_size,
+                               chan_size * s->channels);
 }
 
 

libavcodec/ac3tab.c

@@ -138,6 +138,13 @@ const uint16_t ff_ac3_bitrate_tab[19] = {
  */
 const uint8_t ff_ac3_rematrix_band_tab[5] = { 13, 25, 37, 61, 253 };
 
+/**
+ * Table E2.16 Default Coupling Banding Structure
+ */
+const uint8_t ff_eac3_default_cpl_band_struct[18] = {
+    0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 1, 0, 1, 1, 1, 1, 1
+};
+
 /* AC-3 MDCT window */
 
 /* MDCT window */

libavcodec/ac3tab.h

@@ -39,6 +39,7 @@ extern const uint8_t  ff_ac3_dec_channel_map[8][2][6];
 extern const uint16_t ff_ac3_sample_rate_tab[3];
 extern const uint16_t ff_ac3_bitrate_tab[19];
 extern const uint8_t  ff_ac3_rematrix_band_tab[5];
+extern const uint8_t  ff_eac3_default_cpl_band_struct[18];
 extern const int16_t  ff_ac3_window[AC3_WINDOW_SIZE/2];
 extern const uint8_t  ff_ac3_log_add_tab[260];
 extern const uint16_t ff_ac3_hearing_threshold_tab[AC3_CRITICAL_BANDS][3];

libavcodec/allcodecs.c

@@ -156,7 +156,7 @@ void avcodec_register_all(void)
     REGISTER_DECODER (MPEG1_VDPAU, mpeg1_vdpau);
     REGISTER_DECODER (MPEG2_CRYSTALHD, mpeg2_crystalhd);
     REGISTER_DECODER (MSMPEG4_CRYSTALHD, msmpeg4_crystalhd);
-    REGISTER_ENCDEC  (MSMPEG4V1, msmpeg4v1);
+    REGISTER_DECODER (MSMPEG4V1, msmpeg4v1);
     REGISTER_ENCDEC  (MSMPEG4V2, msmpeg4v2);
     REGISTER_ENCDEC  (MSMPEG4V3, msmpeg4v3);
     REGISTER_DECODER (MSRLE, msrle);

libavcodec/dsputil.h

@@ -628,13 +628,6 @@ static inline int get_penalty_factor(int lambda, int lambda2, int type){
     }
 }
 
-/**
- * Empty mmx state.
- * this must be called between any dsp function and float/double code.
- * for example sin(); dsp->idct_put(); emms_c(); cos()
- */
-#define emms_c()
-
 void dsputil_init_alpha(DSPContext* c, AVCodecContext *avctx);
 void dsputil_init_arm(DSPContext* c, AVCodecContext *avctx);
 void dsputil_init_bfin(DSPContext* c, AVCodecContext *avctx);
@@ -652,22 +645,9 @@ void ff_intrax8dsp_init(DSPContext* c, AVCodecContext *avctx);
 void ff_mlp_init(DSPContext* c, AVCodecContext *avctx);
 void ff_mlp_init_x86(DSPContext* c, AVCodecContext *avctx);
 
-#if HAVE_MMX
-
-#undef emms_c
 
-static inline void emms(void)
-{
-    __asm__ volatile ("emms;":::"memory");
-}
-
-#define emms_c() \
-{\
-    if(av_get_cpu_flags() & AV_CPU_FLAG_MMX)\
-        emms();\
-}
+#if ARCH_ARM
 
-#elif ARCH_ARM
 
 #if HAVE_NEON
 #   define STRIDE_ALIGN 16

libavcodec/mpegvideo_enc.c

@@ -638,15 +638,6 @@ av_cold int MPV_encode_init(AVCodecContext *avctx)
         s->low_delay= s->max_b_frames ? 0 : 1;
         avctx->delay= s->low_delay ? 0 : (s->max_b_frames + 1);
         break;
-    case CODEC_ID_MSMPEG4V1:
-        s->out_format = FMT_H263;
-        s->h263_msmpeg4 = 1;
-        s->h263_pred = 1;
-        s->unrestricted_mv = 1;
-        s->msmpeg4_version= 1;
-        avctx->delay=0;
-        s->low_delay=1;
-        break;
     case CODEC_ID_MSMPEG4V2:
         s->out_format = FMT_H263;
         s->h263_msmpeg4 = 1;
@@ -3807,18 +3798,6 @@ AVCodec ff_h263p_encoder = {
     .long_name= NULL_IF_CONFIG_SMALL("H.263+ / H.263-1998 / H.263 version 2"),
 };
 
-AVCodec ff_msmpeg4v1_encoder = {
-    "msmpeg4v1",
-    AVMEDIA_TYPE_VIDEO,
-    CODEC_ID_MSMPEG4V1,
-    sizeof(MpegEncContext),
-    MPV_encode_init,
-    MPV_encode_picture,
-    MPV_encode_end,
-    .pix_fmts= (const enum PixelFormat[]){PIX_FMT_YUV420P, PIX_FMT_NONE},
-    .long_name= NULL_IF_CONFIG_SMALL("MPEG-4 part 2 Microsoft variant version 1"),
-};
-
 AVCodec ff_msmpeg4v2_encoder = {
     "msmpeg4v2",
     AVMEDIA_TYPE_VIDEO,

libavcodec/msmpeg4.c

@@ -846,22 +846,14 @@ static void msmpeg4_encode_dc(MpegEncContext * s, int level, int n, int *dir_ptr
     int pred, extquant;
     int extrabits = 0;
 
-    if(s->msmpeg4_version==1){
-        int32_t *dc_val;
-        pred = msmpeg4v1_pred_dc(s, n, &dc_val);
-
-        /* update predictor */
-        *dc_val= level;
-    }else{
-        int16_t *dc_val;
-        pred = msmpeg4_pred_dc(s, n, &dc_val, dir_ptr);
+    int16_t *dc_val;
+    pred = msmpeg4_pred_dc(s, n, &dc_val, dir_ptr);
 
-        /* update predictor */
-        if (n < 4) {
-            *dc_val = level * s->y_dc_scale;
-        } else {
-            *dc_val = level * s->c_dc_scale;
-        }
+    /* update predictor */
+    if (n < 4) {
+        *dc_val = level * s->y_dc_scale;
+    } else {
+        *dc_val = level * s->c_dc_scale;
     }
 
     /* do the prediction */

libavcodec/msmpeg4.h

@@ -54,8 +54,7 @@ int ff_wmv2_decode_mb(MpegEncContext *s, DCTELEM block[6][64]);
                                 CONFIG_MSMPEG4V3_DECODER || \
                                 CONFIG_WMV2_DECODER      || \
                                 CONFIG_VC1_DECODER)
-#define CONFIG_MSMPEG4_ENCODER (CONFIG_MSMPEG4V1_ENCODER || \
-                                CONFIG_MSMPEG4V2_ENCODER || \
+#define CONFIG_MSMPEG4_ENCODER (CONFIG_MSMPEG4V2_ENCODER || \
                                 CONFIG_MSMPEG4V3_ENCODER || \
                                 CONFIG_WMV2_ENCODER)
 

libavcodec/tiff.c

@@ -39,6 +39,8 @@ typedef struct TiffContext {
 
     int width, height;
     unsigned int bpp, bppcount;
+    uint32_t palette[256];
+    int palette_is_set;
     int le;
     enum TiffCompr compr;
     int invert;
@@ -255,11 +257,15 @@ static int init_image(TiffContext *s)
         av_log(s->avctx, AV_LOG_ERROR, "get_buffer() failed\n");
         return ret;
     }
-    if (s->bpp == 8 && s->picture.data[1]){
-        /* make default grayscale pal */
-        pal = (uint32_t *) s->picture.data[1];
-        for (i = 0; i < 256; i++)
-            pal[i] = i * 0x010101;
+    if (s->avctx->pix_fmt == PIX_FMT_PAL8) {
+        if (s->palette_is_set) {
+            memcpy(s->picture.data[1], s->palette, sizeof(s->palette));
+        } else {
+            /* make default grayscale pal */
+            pal = (uint32_t *) s->picture.data[1];
+            for (i = 0; i < 256; i++)
+                pal[i] = i * 0x010101;
+        }
     }
     return 0;
 }
@@ -442,11 +448,7 @@ static int tiff_decode_tag(TiffContext *s, const uint8_t *start, const uint8_t *
         s->fill_order = value - 1;
         break;
     case TIFF_PAL:
-        if(s->avctx->pix_fmt != PIX_FMT_PAL8){
-            av_log(s->avctx, AV_LOG_ERROR, "Palette met but this is not palettized format\n");
-            return -1;
-        }
-        pal = (uint32_t *) s->picture.data[1];
+        pal = (uint32_t *) s->palette;
         off = type_sizes[type];
         rp = buf;
         gp = buf + count / 3 * off;
@@ -459,6 +461,7 @@ static int tiff_decode_tag(TiffContext *s, const uint8_t *start, const uint8_t *
             j |= tget(&bp, type, s->le) >> off;
             pal[i] = j;
         }
+        s->palette_is_set = 1;
         break;
     case TIFF_PLANAR:
         if(value == 2){

libavfilter/vf_mp.c

@@ -287,13 +287,8 @@ zrmjpeg
 CpuCaps gCpuCaps; //FIXME initialize this so optims work
 
 
-//exact copy from vf_scale.c
 int get_sws_cpuflags(void){
-    return
-          (gCpuCaps.hasMMX   ? SWS_CPU_CAPS_MMX   : 0)
-        | (gCpuCaps.hasMMX2  ? SWS_CPU_CAPS_MMX2  : 0)
-        | (gCpuCaps.has3DNow ? SWS_CPU_CAPS_3DNOW : 0)
-        | (gCpuCaps.hasAltiVec ? SWS_CPU_CAPS_ALTIVEC : 0);
+    return 0;
 }
 
 static void sws_getFlagsAndFilterFromCmdLine(int *flags, SwsFilter **srcFilterParam, SwsFilter **dstFilterParam)
@@ -348,7 +343,7 @@ struct SwsContext *sws_getContextFromCmdLine(int srcW, int srcH, int srcFormat,
         if (srcFormat == IMGFMT_RGB8 || srcFormat == IMGFMT_BGR8) sfmt = PIX_FMT_PAL8;
         sws_getFlagsAndFilterFromCmdLine(&flags, &srcFilterParam, &dstFilterParam);
 
-        return sws_getContext(srcW, srcH, sfmt, dstW, dstH, dfmt, flags | get_sws_cpuflags(), srcFilterParam, dstFilterParam, NULL);
+        return sws_getContext(srcW, srcH, sfmt, dstW, dstH, dfmt, flags , srcFilterParam, dstFilterParam, NULL);
 }
 
 typedef struct {

libavfilter/vsrc_buffer.h

@@ -27,6 +27,7 @@
  * memory buffer source API for video
  */
 
+
 #include "avfilter.h"
 
 /**

libavformat/ape.c

@@ -276,7 +276,7 @@ static int ape_read_header(AVFormatContext * s, AVFormatParameters * ap)
     ape->frames[0].nblocks = ape->blocksperframe;
     ape->frames[0].skip    = 0;
     for (i = 1; i < ape->totalframes; i++) {
-        ape->frames[i].pos      = ape->seektable[i] + ape->junklength; //ape->frames[i-1].pos + ape->blocksperframe;
+        ape->frames[i].pos      = ape->seektable[i] + ape->junklength;
         ape->frames[i].nblocks  = ape->blocksperframe;
         ape->frames[i - 1].size = ape->frames[i].pos - ape->frames[i - 1].pos;
         ape->frames[i].skip     = (ape->frames[i].pos - ape->frames[0].pos) & 3;

libavutil/internal.h

@@ -37,6 +37,7 @@
 #include "config.h"
 #include "attributes.h"
 #include "timer.h"
+#include "cpu.h"
 
 #ifndef attribute_align_arg
 #if ARCH_X86_32 && AV_GCC_VERSION_AT_LEAST(4,2)
@@ -222,4 +223,19 @@
 #   define ONLY_IF_THREADS_ENABLED(x) NULL
 #endif
 
+#if HAVE_MMX
+/**
+ * Empty mmx state.
+ * this must be called between any dsp function and float/double code.
+ * for example sin(); dsp->idct_put(); emms_c(); cos()
+ */
+static av_always_inline void emms_c(void)
+{
+    if(av_get_cpu_flags() & AV_CPU_FLAG_MMX)
+        __asm__ volatile ("emms" ::: "memory");
+}
+#else /* HAVE_MMX */
+#define emms_c()
+#endif /* HAVE_MMX */
+
 #endif /* AVUTIL_INTERNAL_H */

libswscale/bfin/swscale_bfin.c

@@ -79,15 +79,13 @@ static int yuyvtoyv12_unscaled(SwsContext *c, uint8_t* src[], int srcStride[], i
 void ff_bfin_get_unscaled_swscale(SwsContext *c)
 {
     SwsFunc swScale = c->swScale;
-    if (c->flags & SWS_CPU_CAPS_BFIN)
-        if (c->dstFormat == PIX_FMT_YUV420P)
-            if (c->srcFormat == PIX_FMT_UYVY422) {
-                av_log (NULL, AV_LOG_VERBOSE, "selecting Blackfin optimized uyvytoyv12_unscaled\n");
-                c->swScale = uyvytoyv12_unscaled;
-            }
-        if (c->dstFormat == PIX_FMT_YUV420P)
-            if (c->srcFormat == PIX_FMT_YUYV422) {
-                av_log (NULL, AV_LOG_VERBOSE, "selecting Blackfin optimized yuyvtoyv12_unscaled\n");
-                c->swScale = yuyvtoyv12_unscaled;
-            }
+
+    if (c->dstFormat == PIX_FMT_YUV420P && c->srcFormat == PIX_FMT_UYVY422) {
+        av_log (NULL, AV_LOG_VERBOSE, "selecting Blackfin optimized uyvytoyv12_unscaled\n");
+        c->swScale = uyvytoyv12_unscaled;
+    }
+    if (c->dstFormat == PIX_FMT_YUV420P && c->srcFormat == PIX_FMT_YUYV422) {
+        av_log (NULL, AV_LOG_VERBOSE, "selecting Blackfin optimized yuyvtoyv12_unscaled\n");
+        c->swScale = yuyvtoyv12_unscaled;
+    }
 }

libswscale/colorspace-test.c

@@ -33,31 +33,6 @@
 
 #define FUNC(s,d,n) {s,d,#n,n}
 
-static int cpu_caps;
-
-static char *args_parse(int argc, char *argv[])
-{
-    int o;
-
-    while ((o = getopt(argc, argv, "m23")) != -1) {
-        switch (o) {
-        case 'm':
-            cpu_caps |= SWS_CPU_CAPS_MMX;
-            break;
-        case '2':
-            cpu_caps |= SWS_CPU_CAPS_MMX2;
-            break;
-        case '3':
-            cpu_caps |= SWS_CPU_CAPS_3DNOW;
-            break;
-        default:
-            av_log(NULL, AV_LOG_ERROR, "Unknown option %c\n", o);
-        }
-    }
-
-    return argv[optind];
-}
-
 int main(int argc, char **argv)
 {
     int i, funcNum;
@@ -70,9 +45,7 @@ int main(int argc, char **argv)
         return -1;
 
     av_log(NULL, AV_LOG_INFO, "memory corruption test ...\n");
-    args_parse(argc, argv);
-    av_log(NULL, AV_LOG_INFO, "CPU capabilities forced to %x\n", cpu_caps);
-    sws_rgb2rgb_init(cpu_caps);
+    sws_rgb2rgb_init();
 
     for(funcNum=0; ; funcNum++) {
         struct func_info_s {

libswscale/options.c

@@ -48,12 +48,6 @@ static const AVOption options[] = {
     { "spline", "natural bicubic spline", 0, FF_OPT_TYPE_CONST, {.dbl = SWS_SPLINE }, INT_MIN, INT_MAX, VE, "sws_flags" },
     { "print_info", "print info", 0, FF_OPT_TYPE_CONST, {.dbl = SWS_PRINT_INFO }, INT_MIN, INT_MAX, VE, "sws_flags" },
     { "accurate_rnd", "accurate rounding", 0, FF_OPT_TYPE_CONST, {.dbl = SWS_ACCURATE_RND }, INT_MIN, INT_MAX, VE, "sws_flags" },
-    { "mmx", "MMX SIMD acceleration", 0, FF_OPT_TYPE_CONST, {.dbl = SWS_CPU_CAPS_MMX }, INT_MIN, INT_MAX, VE, "sws_flags" },
-    { "mmx2", "MMX2 SIMD acceleration", 0, FF_OPT_TYPE_CONST, {.dbl = SWS_CPU_CAPS_MMX2 }, INT_MIN, INT_MAX, VE, "sws_flags" },
-    { "sse2", "SSE2 SIMD acceleration", 0, FF_OPT_TYPE_CONST, {.dbl = SWS_CPU_CAPS_SSE2 }, INT_MIN, INT_MAX, VE, "sws_flags" },
-    { "3dnow", "3DNOW SIMD acceleration", 0, FF_OPT_TYPE_CONST, {.dbl = SWS_CPU_CAPS_3DNOW }, INT_MIN, INT_MAX, VE, "sws_flags" },
-    { "altivec", "AltiVec SIMD acceleration", 0, FF_OPT_TYPE_CONST, {.dbl = SWS_CPU_CAPS_ALTIVEC }, INT_MIN, INT_MAX, VE, "sws_flags" },
-    { "bfin", "Blackfin SIMD acceleration", 0, FF_OPT_TYPE_CONST, {.dbl = SWS_CPU_CAPS_BFIN }, INT_MIN, INT_MAX, VE, "sws_flags" },
     { "full_chroma_int", "full chroma interpolation", 0 , FF_OPT_TYPE_CONST, {.dbl = SWS_FULL_CHR_H_INT }, INT_MIN, INT_MAX, VE, "sws_flags" },
     { "full_chroma_inp", "full chroma input", 0 , FF_OPT_TYPE_CONST, {.dbl = SWS_FULL_CHR_H_INP }, INT_MIN, INT_MAX, VE, "sws_flags" },
     { "bitexact", "", 0 , FF_OPT_TYPE_CONST, {.dbl = SWS_BITEXACT }, INT_MIN, INT_MAX, VE, "sws_flags" },

libswscale/ppc/yuv2rgb_altivec.c

@@ -94,6 +94,7 @@ adjustment.
 #include "libswscale/rgb2rgb.h"
 #include "libswscale/swscale.h"
 #include "libswscale/swscale_internal.h"
+#include "libavutil/cpu.h"
 
 #undef PROFILE_THE_BEAST
 #undef INC_SCALING
@@ -692,7 +693,7 @@ static int altivec_uyvy_rgb32 (SwsContext *c,
 */
 SwsFunc ff_yuv2rgb_init_altivec(SwsContext *c)
 {
-    if (!(c->flags & SWS_CPU_CAPS_ALTIVEC))
+    if (!(av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC))
         return NULL;
 
     /*

libswscale/rgb2rgb.c

@@ -116,12 +116,11 @@ void (*yuyvtoyuv422)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const uint8_t
  32-bit C version, and and&add trick by Michael Niedermayer
 */
 
-void sws_rgb2rgb_init(int flags)
+void sws_rgb2rgb_init(void)
 {
     rgb2rgb_init_c();
-#if HAVE_MMX2 || HAVE_AMD3DNOW || HAVE_MMX
-    rgb2rgb_init_x86(flags);
-#endif /* HAVE_MMX2 || HAVE_AMD3DNOW || HAVE_MMX */
+    if (HAVE_MMX)
+        rgb2rgb_init_x86();
 }
 
 #if LIBSWSCALE_VERSION_MAJOR < 1

libswscale/rgb2rgb.h

@@ -166,8 +166,8 @@ extern void (*yuyvtoyuv422)(uint8_t *ydst, uint8_t *udst, uint8_t *vdst, const u
                             long width, long height,
                             long lumStride, long chromStride, long srcStride);
 
-void sws_rgb2rgb_init(int flags);
+void sws_rgb2rgb_init(void);
 
-void rgb2rgb_init_x86(int flags);
+void rgb2rgb_init_x86(void);
 
 #endif /* SWSCALE_RGB2RGB_H */

libswscale/rgb2rgb_template.c

@@ -278,25 +278,6 @@ static inline void rgb16tobgr24_c(const uint8_t *src, uint8_t *dst, long src_siz
     }
 }
 
-/*
- * mm0 = 00 B3 00 B2 00 B1 00 B0
- * mm1 = 00 G3 00 G2 00 G1 00 G0
- * mm2 = 00 R3 00 R2 00 R1 00 R0
- * mm6 = FF FF FF FF FF FF FF FF
- * mm7 = 00 00 00 00 00 00 00 00
- */
-#define PACK_RGB32 \
-    "packuswb   %%mm7, %%mm0    \n\t" /* 00 00 00 00 B3 B2 B1 B0 */ \
-    "packuswb   %%mm7, %%mm1    \n\t" /* 00 00 00 00 G3 G2 G1 G0 */ \
-    "packuswb   %%mm7, %%mm2    \n\t" /* 00 00 00 00 R3 R2 R1 R0 */ \
-    "punpcklbw  %%mm1, %%mm0    \n\t" /* G3 B3 G2 B2 G1 B1 G0 B0 */ \
-    "punpcklbw  %%mm6, %%mm2    \n\t" /* FF R3 FF R2 FF R1 FF R0 */ \
-    "movq       %%mm0, %%mm3    \n\t"                               \
-    "punpcklwd  %%mm2, %%mm0    \n\t" /* FF R1 G1 B1 FF R0 G0 B0 */ \
-    "punpckhwd  %%mm2, %%mm3    \n\t" /* FF R3 G3 B3 FF R2 G2 B2 */ \
-    MOVNTQ"     %%mm0,  %0      \n\t"                               \
-    MOVNTQ"     %%mm3, 8%0      \n\t"                               \
-
 static inline void rgb15to32_c(const uint8_t *src, uint8_t *dst, long src_size)
 {
     const uint16_t *end;

libswscale/swscale.c

@@ -63,6 +63,7 @@ untested special converters
 #include "libavutil/avassert.h"
 #include "libavutil/intreadwrite.h"
 #include "libavutil/x86_cpu.h"
+#include "libavutil/cpu.h"
 #include "libavutil/avutil.h"
 #include "libavutil/mathematics.h"
 #include "libavutil/bswap.h"
@@ -71,10 +72,6 @@ untested special converters
 #undef MOVNTQ
 #undef PAVGB
 
-//#undef HAVE_MMX2
-//#define HAVE_AMD3DNOW
-//#undef HAVE_MMX
-//#undef ARCH_X86
 #define DITHER1XBPP
 
 #define isPacked(x)         (       \
@@ -1262,57 +1259,13 @@ static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uin
 
 //Note: we have C, MMX, MMX2, 3DNOW versions, there is no 3DNOW+MMX2 one
 //Plain C versions
-#if CONFIG_RUNTIME_CPUDETECT
-#  define COMPILE_C 1
-#  if   ARCH_X86
-#    define COMPILE_MMX     1
-#    define COMPILE_MMX2    1
-#    define COMPILE_3DNOW   1
-#  elif ARCH_PPC
-#    define COMPILE_ALTIVEC HAVE_ALTIVEC
-#  endif
-#else /* CONFIG_RUNTIME_CPUDETECT */
-#  if   ARCH_X86
-#    if   HAVE_MMX2
-#      define COMPILE_MMX2  1
-#    elif HAVE_AMD3DNOW
-#      define COMPILE_3DNOW 1
-#    elif HAVE_MMX
-#      define COMPILE_MMX   1
-#    else
-#      define COMPILE_C     1
-#    endif
-#  elif ARCH_PPC && HAVE_ALTIVEC
-#    define COMPILE_ALTIVEC 1
-#  else
-#    define COMPILE_C       1
-#  endif
-#endif
-
-#ifndef COMPILE_C
-#  define COMPILE_C 0
-#endif
-#ifndef COMPILE_MMX
-#  define COMPILE_MMX 0
-#endif
-#ifndef COMPILE_MMX2
-#  define COMPILE_MMX2 0
-#endif
-#ifndef COMPILE_3DNOW
-#  define COMPILE_3DNOW 0
-#endif
-#ifndef COMPILE_ALTIVEC
-#  define COMPILE_ALTIVEC 0
-#endif
 
-#define COMPILE_TEMPLATE_MMX 0
 #define COMPILE_TEMPLATE_MMX2 0
-#define COMPILE_TEMPLATE_AMD3DNOW 0
 #define COMPILE_TEMPLATE_ALTIVEC 0
 
 #include "swscale_template.c"
 
-#if COMPILE_ALTIVEC
+#if HAVE_ALTIVEC
 #undef RENAME
 #undef COMPILE_TEMPLATE_ALTIVEC
 #define COMPILE_TEMPLATE_ALTIVEC 1
@@ -1320,90 +1273,42 @@ static inline void monoblack2Y(uint8_t *dst, const uint8_t *src, long width, uin
 #include "ppc/swscale_template.c"
 #endif
 
-#if ARCH_X86
-
 //MMX versions
-#if COMPILE_MMX
+#if HAVE_MMX
 #undef RENAME
-#undef COMPILE_TEMPLATE_MMX
 #undef COMPILE_TEMPLATE_MMX2
-#undef COMPILE_TEMPLATE_AMD3DNOW
-#define COMPILE_TEMPLATE_MMX 1
 #define COMPILE_TEMPLATE_MMX2 0
-#define COMPILE_TEMPLATE_AMD3DNOW 0
 #define RENAME(a) a ## _MMX
 #include "x86/swscale_template.c"
 #endif
 
 //MMX2 versions
-#if COMPILE_MMX2
+#if HAVE_MMX2
 #undef RENAME
-#undef COMPILE_TEMPLATE_MMX
 #undef COMPILE_TEMPLATE_MMX2
-#undef COMPILE_TEMPLATE_AMD3DNOW
-#define COMPILE_TEMPLATE_MMX 1
 #define COMPILE_TEMPLATE_MMX2 1
-#define COMPILE_TEMPLATE_AMD3DNOW 0
 #define RENAME(a) a ## _MMX2
 #include "x86/swscale_template.c"
 #endif
 
-//3DNOW versions
-#if COMPILE_3DNOW
-#undef RENAME
-#undef COMPILE_TEMPLATE_MMX
-#undef COMPILE_TEMPLATE_MMX2
-#undef COMPILE_TEMPLATE_AMD3DNOW
-#define COMPILE_TEMPLATE_MMX 1
-#define COMPILE_TEMPLATE_MMX2 0
-#define COMPILE_TEMPLATE_AMD3DNOW 1
-#define RENAME(a) a ## _3DNow
-#include "x86/swscale_template.c"
-#endif
-
-#endif //ARCH_X86
-
 SwsFunc ff_getSwsFunc(SwsContext *c)
 {
+    int cpu_flags = av_get_cpu_flags();
+
     sws_init_swScale_c(c);
 
-#if CONFIG_RUNTIME_CPUDETECT
-#if ARCH_X86
-    // ordered per speed fastest first
-    if (c->flags & SWS_CPU_CAPS_MMX2) {
-        sws_init_swScale_MMX2(c);
-        return swScale_MMX2;
-    } else if (c->flags & SWS_CPU_CAPS_3DNOW) {
-        sws_init_swScale_3DNow(c);
-        return swScale_3DNow;
-    } else if (c->flags & SWS_CPU_CAPS_MMX) {
+#if HAVE_MMX
+    if (cpu_flags & AV_CPU_FLAG_MMX)
         sws_init_swScale_MMX(c);
-        return swScale_MMX;
-    }
-
-#else
-#if COMPILE_ALTIVEC
-    if (c->flags & SWS_CPU_CAPS_ALTIVEC) {
-        sws_init_swScale_altivec(c);
-        return swScale_altivec;
-    }
 #endif
-#endif /* ARCH_X86 */
-#else //CONFIG_RUNTIME_CPUDETECT
-#if   COMPILE_TEMPLATE_MMX2
-    sws_init_swScale_MMX2(c);
-    return swScale_MMX2;
-#elif COMPILE_TEMPLATE_AMD3DNOW
-    sws_init_swScale_3DNow(c);
-    return swScale_3DNow;
-#elif COMPILE_TEMPLATE_MMX
-    sws_init_swScale_MMX(c);
-    return swScale_MMX;
-#elif COMPILE_TEMPLATE_ALTIVEC
-    sws_init_swScale_altivec(c);
-    return swScale_altivec;
+#if HAVE_MMX2
+    if (cpu_flags & AV_CPU_FLAG_MMX2)
+        sws_init_swScale_MMX2(c);
+#endif
+#if HAVE_ALTIVEC
+    if (cpu_flags & AV_CPU_FLAG_ALTIVEC)
+        sws_init_swScale_altivec(c);
 #endif
-#endif //!CONFIG_RUNTIME_CPUDETECT
 
     return swScale_c;
 }
@@ -1900,23 +1805,6 @@ static int planarCopyWrapper(SwsContext *c, const uint8_t* src[], int srcStride[
     return srcSliceH;
 }
 
-int ff_hardcodedcpuflags(void)
-{
-    int flags = 0;
-#if   COMPILE_TEMPLATE_MMX2
-    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_MMX2;
-#elif COMPILE_TEMPLATE_AMD3DNOW
-    flags |= SWS_CPU_CAPS_MMX|SWS_CPU_CAPS_3DNOW;
-#elif COMPILE_TEMPLATE_MMX
-    flags |= SWS_CPU_CAPS_MMX;
-#elif COMPILE_TEMPLATE_ALTIVEC
-    flags |= SWS_CPU_CAPS_ALTIVEC;
-#elif ARCH_BFIN
-    flags |= SWS_CPU_CAPS_BFIN;
-#endif
-    return flags;
-}
-
 void ff_get_unscaled_swscale(SwsContext *c)
 {
     const enum PixelFormat srcFormat = c->srcFormat;
@@ -2000,8 +1888,8 @@ void ff_get_unscaled_swscale(SwsContext *c)
     if(srcFormat == PIX_FMT_UYVY422 && dstFormat == PIX_FMT_YUV422P)
         c->swScale= uyvyToYuv422Wrapper;
 
-#if COMPILE_ALTIVEC
-    if ((c->flags & SWS_CPU_CAPS_ALTIVEC) &&
+#if HAVE_ALTIVEC
+    if ((av_get_cpu_flags() & AV_CPU_FLAG_ALTIVEC) &&
         !(c->flags & SWS_BITEXACT) &&
         srcFormat == PIX_FMT_YUV420P) {
         // unscaled YV12 -> packed YUV, we want speed
@@ -2031,8 +1919,7 @@ void ff_get_unscaled_swscale(SwsContext *c)
             c->swScale= planarCopyWrapper;
     }
 #if ARCH_BFIN
-    if (flags & SWS_CPU_CAPS_BFIN)
-        ff_bfin_get_unscaled_swscale (c);
+    ff_bfin_get_unscaled_swscale (c);
 #endif
 }
 

libswscale/swscale.h

@@ -95,13 +95,6 @@ const char *swscale_license(void);
 #define SWS_ACCURATE_RND      0x40000
 #define SWS_BITEXACT          0x80000
 
-#define SWS_CPU_CAPS_MMX      0x80000000
-#define SWS_CPU_CAPS_MMX2     0x20000000
-#define SWS_CPU_CAPS_3DNOW    0x40000000
-#define SWS_CPU_CAPS_ALTIVEC  0x10000000
-#define SWS_CPU_CAPS_BFIN     0x01000000
-#define SWS_CPU_CAPS_SSE2     0x02000000
-
 #define SWS_MAX_REDUCE_CUTOFF 0.002
 
 #define SWS_CS_ITU709         1

libswscale/swscale_internal.h

@@ -481,11 +481,6 @@ extern const AVClass sws_context_class;
  */
 void ff_get_unscaled_swscale(SwsContext *c);
 
-/**
- * Returns the SWS_CPU_CAPS for the optimized code compiled into swscale.
- */
-int ff_hardcodedcpuflags(void);
-
 /**
  * Returns function pointer to fastest main scaler path function depending
  * on architecture and available optimizations.

libswscale/x86/rgb2rgb.c

@@ -27,6 +27,7 @@
 
 #include "config.h"
 #include "libavutil/x86_cpu.h"
+#include "libavutil/cpu.h"
 #include "libavutil/bswap.h"
 #include "libswscale/rgb2rgb.h"
 #include "libswscale/swscale.h"
@@ -122,16 +123,16 @@ DECLARE_ASM_CONST(8, uint64_t, blue_15mask)  = 0x0000001f0000001fULL;
  32-bit C version, and and&add trick by Michael Niedermayer
 */
 
-void rgb2rgb_init_x86(int flags)
+void rgb2rgb_init_x86(void)
 {
-#if HAVE_MMX2 || HAVE_AMD3DNOW || HAVE_MMX
-    if (flags & SWS_CPU_CAPS_SSE2)
-        rgb2rgb_init_SSE2();
-    else if (flags & SWS_CPU_CAPS_MMX2)
-        rgb2rgb_init_MMX2();
-    else if (flags & SWS_CPU_CAPS_3DNOW)
-        rgb2rgb_init_3DNOW();
-    else if (flags & SWS_CPU_CAPS_MMX)
+    int cpu_flags = av_get_cpu_flags();
+
+    if (HAVE_MMX      && cpu_flags & AV_CPU_FLAG_MMX)
         rgb2rgb_init_MMX();
-#endif /* HAVE_MMX2 || HAVE_AMD3DNOW || HAVE_MMX */
+    if (HAVE_AMD3DNOW && cpu_flags & AV_CPU_FLAG_3DNOW)
+        rgb2rgb_init_3DNOW();
+    if (HAVE_MMX2     && cpu_flags & AV_CPU_FLAG_MMX2)
+        rgb2rgb_init_MMX2();
+    if (HAVE_SSE      && cpu_flags & AV_CPU_FLAG_SSE2)
+        rgb2rgb_init_SSE2();
 }

libswscale/x86/yuv2rgb_mmx.c

@@ -34,6 +34,7 @@
 #include "libswscale/swscale.h"
 #include "libswscale/swscale_internal.h"
 #include "libavutil/x86_cpu.h"
+#include "libavutil/cpu.h"
 
 #define DITHER1XBPP // only for MMX
 
@@ -46,57 +47,58 @@ DECLARE_ASM_CONST(8, uint64_t, pb_03) = 0x0303030303030303ULL;
 DECLARE_ASM_CONST(8, uint64_t, pb_07) = 0x0707070707070707ULL;
 
 //MMX versions
+#if HAVE_MMX
 #undef RENAME
-#undef HAVE_MMX2
-#undef HAVE_AMD3DNOW
-#define HAVE_MMX2 0
-#define HAVE_AMD3DNOW 0
+#undef COMPILE_TEMPLATE_MMX2
+#define COMPILE_TEMPLATE_MMX2 0
 #define RENAME(a) a ## _MMX
 #include "yuv2rgb_template.c"
+#endif /* HAVE_MMX */
 
 //MMX2 versions
+#if HAVE_MMX2
 #undef RENAME
-#undef HAVE_MMX2
-#define HAVE_MMX2 1
+#undef COMPILE_TEMPLATE_MMX2
+#define COMPILE_TEMPLATE_MMX2 1
 #define RENAME(a) a ## _MMX2
 #include "yuv2rgb_template.c"
+#endif /* HAVE_MMX2 */
 
 SwsFunc ff_yuv2rgb_init_mmx(SwsContext *c)
 {
-    if (c->flags & SWS_CPU_CAPS_MMX2) {
+    int cpu_flags = av_get_cpu_flags();
+
+    if (c->srcFormat != PIX_FMT_YUV420P &&
+        c->srcFormat != PIX_FMT_YUVA420P)
+        return NULL;
+
+    if (HAVE_MMX2 && cpu_flags & AV_CPU_FLAG_MMX2) {
         switch (c->dstFormat) {
-        case PIX_FMT_RGB32:
-            if (CONFIG_SWSCALE_ALPHA && c->srcFormat == PIX_FMT_YUVA420P) {
-                if (HAVE_7REGS) return yuva420_rgb32_MMX2;
-                break;
-            } else return yuv420_rgb32_MMX2;
-        case PIX_FMT_BGR32:
-            if (CONFIG_SWSCALE_ALPHA && c->srcFormat == PIX_FMT_YUVA420P) {
-                if (HAVE_7REGS) return yuva420_bgr32_MMX2;
-                break;
-            } else return yuv420_bgr32_MMX2;
         case PIX_FMT_RGB24:  return yuv420_rgb24_MMX2;
         case PIX_FMT_BGR24:  return yuv420_bgr24_MMX2;
-        case PIX_FMT_RGB565: return yuv420_rgb16_MMX2;
-        case PIX_FMT_RGB555: return yuv420_rgb15_MMX2;
         }
     }
-    if (c->flags & SWS_CPU_CAPS_MMX) {
+
+    if (HAVE_MMX && cpu_flags & AV_CPU_FLAG_MMX) {
         switch (c->dstFormat) {
-        case PIX_FMT_RGB32:
-            if (CONFIG_SWSCALE_ALPHA && c->srcFormat == PIX_FMT_YUVA420P) {
-                if (HAVE_7REGS) return yuva420_rgb32_MMX;
-                break;
-            } else return yuv420_rgb32_MMX;
-        case PIX_FMT_BGR32:
-            if (CONFIG_SWSCALE_ALPHA && c->srcFormat == PIX_FMT_YUVA420P) {
-                if (HAVE_7REGS) return yuva420_bgr32_MMX;
-                break;
-            } else return yuv420_bgr32_MMX;
-        case PIX_FMT_RGB24:  return yuv420_rgb24_MMX;
-        case PIX_FMT_BGR24:  return yuv420_bgr24_MMX;
-        case PIX_FMT_RGB565: return yuv420_rgb16_MMX;
-        case PIX_FMT_RGB555: return yuv420_rgb15_MMX;
+            case PIX_FMT_RGB32:
+                if (CONFIG_SWSCALE_ALPHA && c->srcFormat == PIX_FMT_YUVA420P) {
+#if HAVE_7REGS
+                    return yuva420_rgb32_MMX;
+#endif
+                    break;
+                } else return yuv420_rgb32_MMX;
+            case PIX_FMT_BGR32:
+                if (CONFIG_SWSCALE_ALPHA && c->srcFormat == PIX_FMT_YUVA420P) {
+#if HAVE_7REGS
+                    return yuva420_bgr32_MMX;
+#endif
+                    break;
+                } else return yuv420_bgr32_MMX;
+            case PIX_FMT_RGB24:  return yuv420_rgb24_MMX;
+            case PIX_FMT_BGR24:  return yuv420_bgr24_MMX;
+            case PIX_FMT_RGB565: return yuv420_rgb16_MMX;
+            case PIX_FMT_RGB555: return yuv420_rgb15_MMX;
         }
     }
 

libswscale/x86/yuv2rgb_template.c

@@ -25,14 +25,7 @@
 #undef EMMS
 #undef SFENCE
 
-#if HAVE_AMD3DNOW
-/* On K6 femms is faster than emms. On K7 femms is directly mapped to emms. */
-#define EMMS   "femms"
-#else
-#define EMMS   "emms"
-#endif
-
-#if HAVE_MMX2
+#if COMPILE_TEMPLATE_MMX2
 #define MOVNTQ "movntq"
 #define SFENCE "sfence"
 #else
@@ -159,7 +152,8 @@
     }                                                             \
 
 #define YUV2RGB_ENDFUNC                          \
-    __asm__ volatile (SFENCE"\n\t"EMMS);         \
+    __asm__ volatile (SFENCE"\n\t"               \
+                    "emms    \n\t");             \
     return srcSliceH;                            \
 
 #define IF0(x)
@@ -188,6 +182,7 @@
     "paddusb "GREEN_DITHER"(%4), %%mm2\n\t"      \
     "paddusb "RED_DITHER"(%4),   %%mm1\n\t"      \
 
+#if !COMPILE_TEMPLATE_MMX2
 static inline int RENAME(yuv420_rgb15)(SwsContext *c, const uint8_t *src[],
                                        int srcStride[],
                                        int srcSliceY, int srcSliceH,
@@ -243,6 +238,7 @@ static inline int RENAME(yuv420_rgb16)(SwsContext *c, const uint8_t *src[],
     YUV2RGB_OPERANDS
     YUV2RGB_ENDFUNC
 }
+#endif /* !COMPILE_TEMPLATE_MMX2 */
 
 #define RGB_PACK24(blue, red)\
     "packuswb  %%mm3,      %%mm0 \n" /* R0 R2 R4 R6 R1 R3 R5 R7 */\
@@ -259,7 +255,7 @@ static inline int RENAME(yuv420_rgb16)(SwsContext *c, const uint8_t *src[],
     "punpckhwd %%mm6,      %%mm5 \n" /* R4 G4 B4 R5 R6 G6 B6 R7 */\
     RGB_PACK24_B
 
-#if HAVE_MMX2
+#if COMPILE_TEMPLATE_MMX2
 DECLARE_ASM_CONST(8, int16_t, mask1101[4]) = {-1,-1, 0,-1};
 DECLARE_ASM_CONST(8, int16_t, mask0010[4]) = { 0, 0,-1, 0};
 DECLARE_ASM_CONST(8, int16_t, mask0110[4]) = { 0,-1,-1, 0};
@@ -366,6 +362,7 @@ static inline int RENAME(yuv420_bgr24)(SwsContext *c, const uint8_t *src[],
     MOVNTQ "   %%mm5,       16(%1)\n\t"      \
     MOVNTQ "   %%mm"alpha", 24(%1)\n\t"      \
 
+#if !COMPILE_TEMPLATE_MMX2
 static inline int RENAME(yuv420_rgb32)(SwsContext *c, const uint8_t *src[],
                                        int srcStride[],
                                        int srcSliceY, int srcSliceH,
@@ -386,12 +383,12 @@ static inline int RENAME(yuv420_rgb32)(SwsContext *c, const uint8_t *src[],
     YUV2RGB_ENDFUNC
 }
 
+#if HAVE_7REGS && CONFIG_SWSCALE_ALPHA
 static inline int RENAME(yuva420_rgb32)(SwsContext *c, const uint8_t *src[],
                                         int srcStride[],
                                         int srcSliceY, int srcSliceH,
                                         uint8_t *dst[], int dstStride[])
 {
-#if HAVE_7REGS
     int y, h_size;
 
     YUV2RGB_LOOP(4)
@@ -406,9 +403,8 @@ static inline int RENAME(yuva420_rgb32)(SwsContext *c, const uint8_t *src[],
     YUV2RGB_ENDLOOP(4)
     YUV2RGB_OPERANDS_ALPHA
     YUV2RGB_ENDFUNC
-#endif
-    return 0;
 }
+#endif
 
 static inline int RENAME(yuv420_bgr32)(SwsContext *c, const uint8_t *src[],
                                        int srcStride[],
@@ -430,12 +426,12 @@ static inline int RENAME(yuv420_bgr32)(SwsContext *c, const uint8_t *src[],
     YUV2RGB_ENDFUNC
 }
 
+#if HAVE_7REGS && CONFIG_SWSCALE_ALPHA
 static inline int RENAME(yuva420_bgr32)(SwsContext *c, const uint8_t *src[],
                                         int srcStride[],
                                         int srcSliceY, int srcSliceH,
                                         uint8_t *dst[], int dstStride[])
 {
-#if HAVE_7REGS
     int y, h_size;
 
     YUV2RGB_LOOP(4)
@@ -450,6 +446,7 @@ static inline int RENAME(yuva420_bgr32)(SwsContext *c, const uint8_t *src[],
     YUV2RGB_ENDLOOP(4)
     YUV2RGB_OPERANDS_ALPHA
     YUV2RGB_ENDFUNC
-#endif
-    return 0;
 }
+#endif
+
+#endif /* !COMPILE_TEMPLATE_MMX2 */

libswscale/yuv2rgb.c

@@ -32,7 +32,7 @@
 #include "rgb2rgb.h"
 #include "swscale.h"
 #include "swscale_internal.h"
-#include "libavutil/x86_cpu.h"
+#include "libavutil/cpu.h"
 #include "libavutil/bswap.h"
 
 extern const uint8_t dither_4x4_16[4][8];
@@ -579,24 +579,18 @@ CLOSEYUV2RGBFUNC(1)
 SwsFunc ff_yuv2rgb_get_func_ptr(SwsContext *c)
 {
     SwsFunc t = NULL;
-#if HAVE_MMX
-     t = ff_yuv2rgb_init_mmx(c);
-#endif
-#if HAVE_VIS
-    t = ff_yuv2rgb_init_vis(c);
-#endif
-#if CONFIG_MLIB
-    t = ff_yuv2rgb_init_mlib(c);
-#endif
-#if HAVE_ALTIVEC
-    if (c->flags & SWS_CPU_CAPS_ALTIVEC)
-        t = ff_yuv2rgb_init_altivec(c);
-#endif
 
-#if ARCH_BFIN
-    if (c->flags & SWS_CPU_CAPS_BFIN)
+    if (HAVE_MMX) {
+        t = ff_yuv2rgb_init_mmx(c);
+    } else if (HAVE_VIS) {
+        t = ff_yuv2rgb_init_vis(c);
+    } else if (CONFIG_MLIB) {
+        t = ff_yuv2rgb_init_mlib(c);
+    } else if (HAVE_ALTIVEC) {
+        t = ff_yuv2rgb_init_altivec(c);
+    } else if (ARCH_BFIN) {
         t = ff_yuv2rgb_get_func_ptr_bfin(c);
-#endif
+    }
 
     if (t)
         return t;