Merge remote branch 'qatar/master'

* qatar/master: (23 commits)
  ac3enc: correct the flipped sign in the ac3_fixed encoder
  Eliminate pointless '#if 1' statements without matching '#else'.
  Add AVX FFT implementation.
  Increase alignment of av_malloc() as needed by AVX ASM.
  Update x86inc.asm from x264 to allow AVX emulation using SSE and MMX.
  mjpeg: Detect overreads in mjpeg_decode_scan() and error out.
  documentation: extend documentation for ffmpeg -aspect option
  APIChanges: update commit hashes for recent additions.
  lavc: deprecate FF_*_TYPE macros in favor of AV_PICTURE_TYPE_* enums
  aac: add headers needed for log2f()
  lavc: remove FF_API_MB_Q cruft
  lavc: remove FF_API_RATE_EMU cruft
  lavc: remove FF_API_HURRY_UP cruft
  pad: make the filter parametric
  vsrc_movie: add key_frame and pict_type.
  vsrc_movie: fix leak in request_frame()
  lavfi: add key_frame and pict_type to AVFilterBufferRefVideo.
  vsrc_buffer: add sample_aspect_ratio fields to arguments.
  lavfi: add fieldorder filter
  scale: make the filter parametric
  ...

Conflicts:
	Changelog
	doc/filters.texi
	ffmpeg.c
	libavcodec/ac3dec.h
	libavcodec/dsputil.c
	libavfilter/avfilter.h
	libavfilter/vf_scale.c
	libavfilter/vf_yadif.c
	libavfilter/vsrc_buffer.c

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

48 files changed, 735 insertions, 393 deletions

diff --git a/libavcodec/aac.h b/libavcodec/aac.h
index 73bc40843f..54aab6e0bc 100644
--- a/libavcodec/aac.h
+++ b/libavcodec/aac.h
@@ -223,9 +223,9 @@ typedef struct {
     float sf[120];                                  ///< scalefactors
     int sf_idx[128];                                ///< scalefactor indices (used by encoder)
     uint8_t zeroes[128];                            ///< band is not coded (used by encoder)
-    DECLARE_ALIGNED(16, float,   coeffs)[1024];     ///< coefficients for IMDCT
-    DECLARE_ALIGNED(16, float,   saved)[1024];      ///< overlap
-    DECLARE_ALIGNED(16, float,   ret)[2048];        ///< PCM output
+    DECLARE_ALIGNED(32, float,   coeffs)[1024];     ///< coefficients for IMDCT
+    DECLARE_ALIGNED(32, float,   saved)[1024];      ///< overlap
+    DECLARE_ALIGNED(32, float,   ret)[2048];        ///< PCM output
     DECLARE_ALIGNED(16, int16_t, ltp_state)[3072];  ///< time signal for LTP
     PredictorState predictor_state[MAX_PREDICTORS];
 } SingleChannelElement;
@@ -272,7 +272,7 @@ typedef struct {
      * @defgroup temporary aligned temporary buffers (We do not want to have these on the stack.)
      * @{
      */
-    DECLARE_ALIGNED(16, float, buf_mdct)[1024];
+    DECLARE_ALIGNED(32, float, buf_mdct)[1024];
     /** @} */
 
     /**
@@ -296,7 +296,7 @@ typedef struct {
     int sf_offset;                                    ///< offset into pow2sf_tab as appropriate for dsp.float_to_int16
     /** @} */
 
-    DECLARE_ALIGNED(16, float, temp)[128];
+    DECLARE_ALIGNED(32, float, temp)[128];
 
     enum OCStatus output_configured;
 } AACContext;
diff --git a/libavcodec/aaccoder.c b/libavcodec/aaccoder.c
index 9748fe1e1a..35b31c708a 100644
--- a/libavcodec/aaccoder.c
+++ b/libavcodec/aaccoder.c
@@ -37,6 +37,7 @@
 #include "aac.h"
 #include "aacenc.h"
 #include "aactab.h"
+#include "libavutil/libm.h"
 
 /** bits needed to code codebook run value for long windows */
 static const uint8_t run_value_bits_long[64] = {
diff --git a/libavcodec/aacenc.h b/libavcodec/aacenc.h
index 1c8467990c..c12475a187 100644
--- a/libavcodec/aacenc.h
+++ b/libavcodec/aacenc.h
@@ -64,7 +64,7 @@ typedef struct AACEncContext {
     int last_frame;
     float lambda;
     DECLARE_ALIGNED(16, int,   qcoefs)[96];      ///< quantized coefficients
-    DECLARE_ALIGNED(16, float, scoefs)[1024];    ///< scaled coefficients
+    DECLARE_ALIGNED(32, float, scoefs)[1024];    ///< scaled coefficients
 } AACEncContext;
 
 #endif /* AVCODEC_AACENC_H */
diff --git a/libavcodec/aacsbr.c b/libavcodec/aacsbr.c
index 6ac2cbc06a..afff6931fb 100644
--- a/libavcodec/aacsbr.c
+++ b/libavcodec/aacsbr.c
@@ -32,6 +32,7 @@
 #include "aacsbrdata.h"
 #include "fft.h"
 #include "aacps.h"
+#include "libavutil/libm.h"
 
 #include <stdint.h>
 #include <float.h>
diff --git a/libavcodec/ac3dec.h b/libavcodec/ac3dec.h
index 91008e5c8f..d3fe3e782b 100644
--- a/libavcodec/ac3dec.h
+++ b/libavcodec/ac3dec.h
@@ -201,13 +201,13 @@ typedef struct {
 ///@}
 
 ///@defgroup arrays aligned arrays
-    DECLARE_ALIGNED(16, int,   fixed_coeffs)[AC3_MAX_CHANNELS][AC3_MAX_COEFS];       ///< fixed-point transform coefficients
-    DECLARE_ALIGNED(16, float, transform_coeffs)[AC3_MAX_CHANNELS][AC3_MAX_COEFS];   ///< transform coefficients
-    DECLARE_ALIGNED(16, float, delay)[AC3_MAX_CHANNELS][AC3_BLOCK_SIZE];             ///< delay - added to the next block
-    DECLARE_ALIGNED(16, float, window)[AC3_BLOCK_SIZE];                              ///< window coefficients
-    DECLARE_ALIGNED(16, float, tmp_output)[AC3_BLOCK_SIZE];                          ///< temporary storage for output before windowing
-    DECLARE_ALIGNED(16, float, output)[AC3_MAX_CHANNELS][AC3_BLOCK_SIZE];            ///< output after imdct transform and windowing
-    DECLARE_ALIGNED(16, uint8_t, input_buffer)[AC3_FRAME_BUFFER_SIZE + FF_INPUT_BUFFER_PADDING_SIZE]; ///< temp buffer to prevent overread
+    DECLARE_ALIGNED(16, int,   fixed_coeffs)[AC3_MAX_CHANNELS][AC3_MAX_COEFS];       ///> fixed-point transform coefficients
+    DECLARE_ALIGNED(32, float, transform_coeffs)[AC3_MAX_CHANNELS][AC3_MAX_COEFS];   ///< transform coefficients
+    DECLARE_ALIGNED(32, float, delay)[AC3_MAX_CHANNELS][AC3_BLOCK_SIZE];             ///< delay - added to the next block
+    DECLARE_ALIGNED(32, float, window)[AC3_BLOCK_SIZE];                              ///< window coefficients
+    DECLARE_ALIGNED(32, float, tmp_output)[AC3_BLOCK_SIZE];                          ///< temporary storage for output before windowing
+    DECLARE_ALIGNED(32, float, output)[AC3_MAX_CHANNELS][AC3_BLOCK_SIZE];            ///< output after imdct transform and windowing
+    DECLARE_ALIGNED(32, uint8_t, input_buffer)[AC3_FRAME_BUFFER_SIZE + FF_INPUT_BUFFER_PADDING_SIZE]; ///< temp buffer to prevent overread
 ///@}
 } AC3DecodeContext;
 
diff --git a/libavcodec/ac3enc.c b/libavcodec/ac3enc.c
index 77ebcdc9e2..b0b4075a36 100644
--- a/libavcodec/ac3enc.c
+++ b/libavcodec/ac3enc.c
@@ -171,7 +171,7 @@ typedef struct AC3EncodeContext {
 
     uint8_t exp_strategy[AC3_MAX_CHANNELS][AC3_MAX_BLOCKS]; ///< exponent strategies
 
-    DECLARE_ALIGNED(16, SampleType, windowed_samples)[AC3_WINDOW_SIZE];
+    DECLARE_ALIGNED(32, SampleType, windowed_samples)[AC3_WINDOW_SIZE];
 } AC3EncodeContext;
 
 typedef struct AC3Mant {
diff --git a/libavcodec/ac3enc_fixed.c b/libavcodec/ac3enc_fixed.c
index 58590ac958..24c11680a6 100644
--- a/libavcodec/ac3enc_fixed.c
+++ b/libavcodec/ac3enc_fixed.c
@@ -47,7 +47,7 @@ static av_cold void mdct_end(AC3MDCTContext *mdct)
 static av_cold int mdct_init(AVCodecContext *avctx, AC3MDCTContext *mdct,
                              int nbits)
 {
-    int ret = ff_mdct_init(&mdct->fft, nbits, 0, 1.0);
+    int ret = ff_mdct_init(&mdct->fft, nbits, 0, -1.0);
     mdct->window = ff_ac3_window;
     return ret;
 }
diff --git a/libavcodec/atrac1.c b/libavcodec/atrac1.c
index 8d8267556a..1f4c36c141 100644
--- a/libavcodec/atrac1.c
+++ b/libavcodec/atrac1.c
@@ -60,11 +60,11 @@ typedef struct {
     int                 log2_block_count[AT1_QMF_BANDS];    ///< log2 number of blocks in a band
     int                 num_bfus;                           ///< number of Block Floating Units
     float*              spectrum[2];
-    DECLARE_ALIGNED(16, float, spec1)[AT1_SU_SAMPLES];     ///< mdct buffer
-    DECLARE_ALIGNED(16, float, spec2)[AT1_SU_SAMPLES];     ///< mdct buffer
-    DECLARE_ALIGNED(16, float, fst_qmf_delay)[46];         ///< delay line for the 1st stacked QMF filter
-    DECLARE_ALIGNED(16, float, snd_qmf_delay)[46];         ///< delay line for the 2nd stacked QMF filter
-    DECLARE_ALIGNED(16, float, last_qmf_delay)[256+23];    ///< delay line for the last stacked QMF filter
+    DECLARE_ALIGNED(32, float, spec1)[AT1_SU_SAMPLES];     ///< mdct buffer
+    DECLARE_ALIGNED(32, float, spec2)[AT1_SU_SAMPLES];     ///< mdct buffer
+    DECLARE_ALIGNED(32, float, fst_qmf_delay)[46];         ///< delay line for the 1st stacked QMF filter
+    DECLARE_ALIGNED(32, float, snd_qmf_delay)[46];         ///< delay line for the 2nd stacked QMF filter
+    DECLARE_ALIGNED(32, float, last_qmf_delay)[256+23];    ///< delay line for the last stacked QMF filter
 } AT1SUCtx;
 
 /**
@@ -72,13 +72,13 @@ typedef struct {
  */
 typedef struct {
     AT1SUCtx            SUs[AT1_MAX_CHANNELS];              ///< channel sound unit
-    DECLARE_ALIGNED(16, float, spec)[AT1_SU_SAMPLES];      ///< the mdct spectrum buffer
+    DECLARE_ALIGNED(32, float, spec)[AT1_SU_SAMPLES];      ///< the mdct spectrum buffer
 
-    DECLARE_ALIGNED(16, float,  low)[256];
-    DECLARE_ALIGNED(16, float,  mid)[256];
-    DECLARE_ALIGNED(16, float, high)[512];
+    DECLARE_ALIGNED(32, float,  low)[256];
+    DECLARE_ALIGNED(32, float,  mid)[256];
+    DECLARE_ALIGNED(32, float, high)[512];
     float*              bands[3];
-    DECLARE_ALIGNED(16, float, out_samples)[AT1_MAX_CHANNELS][AT1_SU_SAMPLES];
+    DECLARE_ALIGNED(32, float, out_samples)[AT1_MAX_CHANNELS][AT1_SU_SAMPLES];
     FFTContext          mdct_ctx[3];
     int                 channels;
     DSPContext          dsp;
diff --git a/libavcodec/atrac3.c b/libavcodec/atrac3.c
index bd49169465..20ab75dfd7 100644
--- a/libavcodec/atrac3.c
+++ b/libavcodec/atrac3.c
@@ -74,8 +74,8 @@ typedef struct {
     int               gcBlkSwitch;
     gain_block        gainBlock[2];
 
-    DECLARE_ALIGNED(16, float, spectrum)[1024];
-    DECLARE_ALIGNED(16, float, IMDCT_buf)[1024];
+    DECLARE_ALIGNED(32, float, spectrum)[1024];
+    DECLARE_ALIGNED(32, float, IMDCT_buf)[1024];
 
     float             delayBuf1[46]; ///<qmf delay buffers
     float             delayBuf2[46];
@@ -122,7 +122,7 @@ typedef struct {
     FFTContext          mdct_ctx;
 } ATRAC3Context;
 
-static DECLARE_ALIGNED(16, float,mdct_window)[512];
+static DECLARE_ALIGNED(32, float, mdct_window)[512];
 static VLC              spectral_coeff_tab[7];
 static float            gain_tab1[16];
 static float            gain_tab2[31];
diff --git a/libavcodec/avcodec.h b/libavcodec/avcodec.h
index 58a38fa88c..b69230f6fe 100644
--- a/libavcodec/avcodec.h
+++ b/libavcodec/avcodec.h
@@ -766,7 +766,7 @@ typedef struct AVPanScan{
      * - encoding: Set by libavcodec. for coded_picture (and set by user for input).\
      * - decoding: Set by libavcodec.\
      */\
-    int pict_type;\
+    enum AVPictureType pict_type;\
 \
     /**\
      * presentation timestamp in time_base units (time when frame should be shown to user)\
@@ -1016,14 +1016,16 @@ typedef struct AVPanScan{
 #define FF_BUFFER_TYPE_SHARED   4 ///< Buffer from somewhere else; don't deallocate image (data/base), all other tables are not shared.
 #define FF_BUFFER_TYPE_COPY     8 ///< Just a (modified) copy of some other buffer, don't deallocate anything.
 
-
-#define FF_I_TYPE  1 ///< Intra
-#define FF_P_TYPE  2 ///< Predicted
-#define FF_B_TYPE  3 ///< Bi-dir predicted
-#define FF_S_TYPE  4 ///< S(GMC)-VOP MPEG4
-#define FF_SI_TYPE 5 ///< Switching Intra
-#define FF_SP_TYPE 6 ///< Switching Predicted
-#define FF_BI_TYPE 7
+#if FF_API_OLD_FF_PICT_TYPES
+/* DEPRECATED, directly use the AV_PICTURE_TYPE_* enum values */
+#define FF_I_TYPE  AV_PICTURE_TYPE_I  ///< Intra
+#define FF_P_TYPE  AV_PICTURE_TYPE_P  ///< Predicted
+#define FF_B_TYPE  AV_PICTURE_TYPE_B  ///< Bi-dir predicted
+#define FF_S_TYPE  AV_PICTURE_TYPE_S  ///< S(GMC)-VOP MPEG4
+#define FF_SI_TYPE AV_PICTURE_TYPE_SI ///< Switching Intra
+#define FF_SP_TYPE AV_PICTURE_TYPE_SP ///< Switching Predicted
+#define FF_BI_TYPE AV_PICTURE_TYPE_BI
+#endif
 
 #define FF_BUFFER_HINTS_VALID    0x01 // Buffer hints value is meaningful (if 0 ignore).
 #define FF_BUFFER_HINTS_READABLE 0x02 // Codec will read from buffer.
@@ -1215,16 +1217,6 @@ typedef struct AVCodecContext {
      */
     enum PixelFormat pix_fmt;
 
-#if FF_API_RATE_EMU
-    /**
-     * Frame rate emulation. If not zero, the lower layer (i.e. format handler)
-     * has to read frames at native frame rate.
-     * - encoding: Set by user.
-     * - decoding: unused
-     */
-    attribute_deprecated int rate_emu;
-#endif
-
     /**
      * If non NULL, 'draw_horiz_band' is called by the libavcodec
      * decoder to draw a horizontal band. It improves cache usage. Not
@@ -1326,16 +1318,6 @@ typedef struct AVCodecContext {
 
     int b_frame_strategy;
 
-#if FF_API_HURRY_UP
-    /**
-     * hurry up amount
-     * - encoding: unused
-     * - decoding: Set by user. 1-> Skip B-frames, 2-> Skip IDCT/dequant too, 5-> Skip everything except header
-     * @deprecated Deprecated in favor of skip_idct and skip_frame.
-     */
-    attribute_deprecated int hurry_up;
-#endif
-
     struct AVCodec *codec;
 
     void *priv_data;
@@ -1800,22 +1782,6 @@ typedef struct AVCodecContext {
      */
     uint64_t error[4];
 
-#if FF_API_MB_Q
-    /**
-     * minimum MB quantizer
-     * - encoding: unused
-     * - decoding: unused
-     */
-    attribute_deprecated int mb_qmin;
-
-    /**
-     * maximum MB quantizer
-     * - encoding: unused
-     * - decoding: unused
-     */
-    attribute_deprecated int mb_qmax;
-#endif
-
     /**
      * motion estimation comparison function
      * - encoding: Set by user.
@@ -3866,13 +3832,17 @@ void avcodec_default_free_buffers(AVCodecContext *s);
 
 /* misc useful functions */
 
+#if FF_API_OLD_FF_PICT_TYPES
 /**
  * Return a single letter to describe the given picture type pict_type.
  *
  * @param[in] pict_type the picture type
  * @return A single character representing the picture type.
+ * @deprecated Use av_get_picture_type_char() instead.
  */
+attribute_deprecated
 char av_get_pict_type_char(int pict_type);
+#endif
 
 /**
  * Return codec bits per sample.
diff --git a/libavcodec/binkaudio.c b/libavcodec/binkaudio.c
index d879efc2e0..bf1d412ed1 100644
--- a/libavcodec/binkaudio.c
+++ b/libavcodec/binkaudio.c
@@ -55,7 +55,7 @@ typedef struct {
     int num_bands;
     unsigned int *bands;
     float root;
-    DECLARE_ALIGNED(16, FFTSample, coeffs)[BINK_BLOCK_MAX_SIZE];
+    DECLARE_ALIGNED(32, FFTSample, coeffs)[BINK_BLOCK_MAX_SIZE];
     DECLARE_ALIGNED(16, short, previous)[BINK_BLOCK_MAX_SIZE / 16];  ///< coeffs from previous audio block
     float *coeffs_ptr[MAX_CHANNELS]; ///< pointers to the coeffs arrays for float_to_int16_interleave
     union {
diff --git a/libavcodec/cook.c b/libavcodec/cook.c
index 286ecd42e5..249b966eaa 100644
--- a/libavcodec/cook.c
+++ b/libavcodec/cook.c
@@ -153,7 +153,7 @@ typedef struct cook {
     /* data buffers */
 
     uint8_t*            decoded_bytes_buffer;
-    DECLARE_ALIGNED(16, float,mono_mdct_output)[2048];
+    DECLARE_ALIGNED(32, float, mono_mdct_output)[2048];
     float               decode_buffer_1[1024];
     float               decode_buffer_2[1024];
     float               decode_buffer_0[1060]; /* static allocation for joint decode */
diff --git a/libavcodec/dca.c b/libavcodec/dca.c
index a615d5d18b..1e26eedc3c 100644
--- a/libavcodec/dca.c
+++ b/libavcodec/dca.c
@@ -321,16 +321,16 @@ typedef struct {
 
     /* Subband samples history (for ADPCM) */
     float subband_samples_hist[DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][4];
-    DECLARE_ALIGNED(16, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
-    DECLARE_ALIGNED(16, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
+    DECLARE_ALIGNED(32, float, subband_fir_hist)[DCA_PRIM_CHANNELS_MAX][512];
+    DECLARE_ALIGNED(32, float, subband_fir_noidea)[DCA_PRIM_CHANNELS_MAX][32];
     int hist_index[DCA_PRIM_CHANNELS_MAX];
-    DECLARE_ALIGNED(16, float, raXin)[32];
+    DECLARE_ALIGNED(32, float, raXin)[32];
 
     int output;                 ///< type of output
     float scale_bias;           ///< output scale
 
-    DECLARE_ALIGNED(16, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
-    DECLARE_ALIGNED(16, float, samples)[(DCA_PRIM_CHANNELS_MAX+1)*256];
+    DECLARE_ALIGNED(32, float, subband_samples)[DCA_BLOCKS_MAX][DCA_PRIM_CHANNELS_MAX][DCA_SUBBANDS][8];
+    DECLARE_ALIGNED(32, float, samples)[(DCA_PRIM_CHANNELS_MAX+1)*256];
     const float *samples_chanptr[DCA_PRIM_CHANNELS_MAX+1];
 
     uint8_t dca_buffer[DCA_MAX_FRAME_SIZE + DCA_MAX_EXSS_HEADER_SIZE + DCA_BUFFER_PADDING_SIZE];
diff --git a/libavcodec/dct-test.c b/libavcodec/dct-test.c
index b45f23ed27..f8e981be9b 100644
--- a/libavcodec/dct-test.c
+++ b/libavcodec/dct-test.c
@@ -312,18 +312,16 @@ static void dct_error(const char *name, int is_idct,
     }
     for(i=0; i<64; i++) sysErrMax= FFMAX(sysErrMax, FFABS(sysErr[i]));
 
-#if 1 // dump systematic errors
     for(i=0; i<64; i++){
         if(i%8==0) printf("\n");
         printf("%7d ", (int)sysErr[i]);
     }
     printf("\n");
-#endif
 
     printf("%s %s: err_inf=%d err2=%0.8f syserr=%0.8f maxout=%d blockSumErr=%d\n",
            is_idct ? "IDCT" : "DCT",
            name, err_inf, (double)err2 / NB_ITS / 64.0, (double)sysErrMax / NB_ITS, maxout, blockSumErrMax);
-#if 1 //Speed test
+
     /* speed test */
     for(i=0;i<64;i++)
         block1[i] = 0;
@@ -376,7 +374,6 @@ static void dct_error(const char *name, int is_idct,
     printf("%s %s: %0.1f kdct/s\n",
            is_idct ? "IDCT" : "DCT",
            name, (double)it1 * 1000.0 / (double)ti1);
-#endif
 }
 
 DECLARE_ALIGNED(8, static uint8_t, img_dest)[64];
diff --git a/libavcodec/error_resilience.c b/libavcodec/error_resilience.c
index dc015b9f6a..ed31582199 100644
--- a/libavcodec/error_resilience.c
+++ b/libavcodec/error_resilience.c
@@ -786,7 +786,6 @@ void ff_er_frame_end(MpegEncContext *s){
         }
     }
 
-#if 1
     /* handle overlapping slices */
     for(error_type=1; error_type<=3; error_type++){
         int end_ok=0;
@@ -807,8 +806,7 @@ void ff_er_frame_end(MpegEncContext *s){
                 end_ok=0;
         }
     }
-#endif
-#if 1
+
     /* handle slices with partitions of different length */
     if(s->partitioned_frame){
         int end_ok=0;
@@ -829,7 +827,7 @@ void ff_er_frame_end(MpegEncContext *s){
                 end_ok=0;
         }
     }
-#endif
+
     /* handle missing slices */
     if(s->error_recognition>=4){
         int end_ok=1;
@@ -853,7 +851,6 @@ void ff_er_frame_end(MpegEncContext *s){
         }
     }
 
-#if 1
     /* backward mark errors */
     distance=9999999;
     for(error_type=1; error_type<=3; error_type++){
@@ -878,7 +875,6 @@ void ff_er_frame_end(MpegEncContext *s){
                 distance= 9999999;
         }
     }
-#endif
 
     /* forward mark errors */
     error=0;
@@ -893,7 +889,7 @@ void ff_er_frame_end(MpegEncContext *s){
             s->error_status_table[mb_xy]|= error;
         }
     }
-#if 1
+
     /* handle not partitioned case */
     if(!s->partitioned_frame){
         for(i=0; i<s->mb_num; i++){
@@ -904,7 +900,6 @@ void ff_er_frame_end(MpegEncContext *s){
             s->error_status_table[mb_xy]= error;
         }
     }
-#endif
 
     dc_error= ac_error= mv_error=0;
     for(i=0; i<s->mb_num; i++){
@@ -1065,16 +1060,15 @@ void ff_er_frame_end(MpegEncContext *s){
             s->dc_val[2][mb_x + mb_y*s->mb_stride]= (dcv+4)>>3;
         }
     }
-#if 1
+
     /* guess DC for damaged blocks */
     guess_dc(s, s->dc_val[0], s->mb_width*2, s->mb_height*2, s->b8_stride, 1);
     guess_dc(s, s->dc_val[1], s->mb_width  , s->mb_height  , s->mb_stride, 0);
     guess_dc(s, s->dc_val[2], s->mb_width  , s->mb_height  , s->mb_stride, 0);
-#endif
+
     /* filter luma DC */
     filter181(s->dc_val[0], s->mb_width*2, s->mb_height*2, s->b8_stride);
 
-#if 1
     /* render DC only intra */
     for(mb_y=0; mb_y<s->mb_height; mb_y++){
         for(mb_x=0; mb_x<s->mb_width; mb_x++){
@@ -1094,7 +1088,6 @@ void ff_er_frame_end(MpegEncContext *s){
             put_dc(s, dest_y, dest_cb, dest_cr, mb_x, mb_y);
         }
     }
-#endif
 
     if(s->avctx->error_concealment&FF_EC_DEBLOCK){
         /* filter horizontal block boundaries */
diff --git a/libavcodec/fft.c b/libavcodec/fft.c
index 58484e047b..4eb46f14c8 100644
--- a/libavcodec/fft.c
+++ b/libavcodec/fft.c
@@ -93,6 +93,44 @@ av_cold void ff_init_ff_cos_tabs(int index)
 #endif
 }
 
+static const int avx_tab[] = {
+    0, 4, 1, 5, 8, 12, 9, 13, 2, 6, 3, 7, 10, 14, 11, 15
+};
+
+static int is_second_half_of_fft32(int i, int n)
+{
+    if (n <= 32)
+        return i >= 16;
+    else if (i < n/2)
+        return is_second_half_of_fft32(i, n/2);
+    else if (i < 3*n/4)
+        return is_second_half_of_fft32(i - n/2, n/4);
+    else
+        return is_second_half_of_fft32(i - 3*n/4, n/4);
+}
+
+static av_cold void fft_perm_avx(FFTContext *s)
+{
+    int i;
+    int n = 1 << s->nbits;
+
+    for (i = 0; i < n; i += 16) {
+        int k;
+        if (is_second_half_of_fft32(i, n)) {
+            for (k = 0; k < 16; k++)
+                s->revtab[-split_radix_permutation(i + k, n, s->inverse) & (n - 1)] =
+                    i + avx_tab[k];
+
+        } else {
+            for (k = 0; k < 16; k++) {
+                int j = i + k;
+                j = (j & ~7) | ((j >> 1) & 3) | ((j << 2) & 4);
+                s->revtab[-split_radix_permutation(i + k, n, s->inverse) & (n - 1)] = j;
+            }
+        }
+    }
+}
+
 av_cold int ff_fft_init(FFTContext *s, int nbits, int inverse)
 {
     int i, j, n;
@@ -132,11 +170,16 @@ av_cold int ff_fft_init(FFTContext *s, int nbits, int inverse)
     for(j=4; j<=nbits; j++) {
         ff_init_ff_cos_tabs(j);
     }
-    for(i=0; i<n; i++) {
-        int j = i;
-        if (s->fft_permutation == FF_FFT_PERM_SWAP_LSBS)
-            j = (j&~3) | ((j>>1)&1) | ((j<<1)&2);
-        s->revtab[-split_radix_permutation(i, n, s->inverse) & (n-1)] = j;
+
+    if (s->fft_permutation == FF_FFT_PERM_AVX) {
+        fft_perm_avx(s);
+    } else {
+        for(i=0; i<n; i++) {
+            int j = i;
+            if (s->fft_permutation == FF_FFT_PERM_SWAP_LSBS)
+                j = (j&~3) | ((j>>1)&1) | ((j<<1)&2);
+            s->revtab[-split_radix_permutation(i, n, s->inverse) & (n-1)] = j;
+        }
     }
 
     return 0;
diff --git a/libavcodec/fft.h b/libavcodec/fft.h
index 8280a361cc..24db7e3d24 100644
--- a/libavcodec/fft.h
+++ b/libavcodec/fft.h
@@ -85,6 +85,7 @@ struct FFTContext {
     int fft_permutation;
 #define FF_FFT_PERM_DEFAULT   0
 #define FF_FFT_PERM_SWAP_LSBS 1
+#define FF_FFT_PERM_AVX       2
     int mdct_permutation;
 #define FF_MDCT_PERM_NONE       0
 #define FF_MDCT_PERM_INTERLEAVE 1
@@ -97,7 +98,7 @@ struct FFTContext {
 #endif
 
 #define COSTABLE(size) \
-    COSTABLE_CONST DECLARE_ALIGNED(16, FFTSample, FFT_NAME(ff_cos_##size))[size/2]
+    COSTABLE_CONST DECLARE_ALIGNED(32, FFTSample, FFT_NAME(ff_cos_##size))[size/2]
 
 extern COSTABLE(16);
 extern COSTABLE(32);
diff --git a/libavcodec/h261dec.c b/libavcodec/h261dec.c
index 060aef6173..d732123d3e 100644
--- a/libavcodec/h261dec.c
+++ b/libavcodec/h261dec.c
@@ -599,10 +599,6 @@ retry:
     s->current_picture.pict_type= s->pict_type;
     s->current_picture.key_frame= s->pict_type == FF_I_TYPE;
 
-#if FF_API_HURRY_UP
-    /* skip everything if we are in a hurry>=5 */
-    if(avctx->hurry_up>=5) return get_consumed_bytes(s, buf_size);
-#endif
     if(  (avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type==FF_B_TYPE)
        ||(avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type!=FF_I_TYPE)
        || avctx->skip_frame >= AVDISCARD_ALL)
diff --git a/libavcodec/h263dec.c b/libavcodec/h263dec.c
index 4151f4b012..72622e693f 100644
--- a/libavcodec/h263dec.c
+++ b/libavcodec/h263dec.c
@@ -612,18 +612,10 @@ retry:
 
     /* skip B-frames if we don't have reference frames */
     if(s->last_picture_ptr==NULL && (s->pict_type==FF_B_TYPE || s->dropable)) return get_consumed_bytes(s, buf_size);
-#if FF_API_HURRY_UP
-    /* skip b frames if we are in a hurry */
-    if(avctx->hurry_up && s->pict_type==FF_B_TYPE) return get_consumed_bytes(s, buf_size);
-#endif
     if(   (avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type==FF_B_TYPE)
        || (avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type!=FF_I_TYPE)
        ||  avctx->skip_frame >= AVDISCARD_ALL)
         return get_consumed_bytes(s, buf_size);
-#if FF_API_HURRY_UP
-    /* skip everything if we are in a hurry>=5 */
-    if(avctx->hurry_up>=5) return get_consumed_bytes(s, buf_size);
-#endif
 
     if(s->next_p_frame_damaged){
         if(s->pict_type==FF_B_TYPE)
diff --git a/libavcodec/h264.c b/libavcodec/h264.c
index 22a57866d4..285dbeadd6 100644
--- a/libavcodec/h264.c
+++ b/libavcodec/h264.c
@@ -2966,11 +2966,7 @@ static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size){
         buf_index += consumed;
 
         //FIXME do not discard SEI id
-        if(
-#if FF_API_HURRY_UP
-           (s->hurry_up == 1 && h->nal_ref_idc  == 0) ||
-#endif
-           (avctx->skip_frame >= AVDISCARD_NONREF && h->nal_ref_idc  == 0))
+        if(avctx->skip_frame >= AVDISCARD_NONREF && h->nal_ref_idc  == 0)
             continue;
 
       again:
@@ -3007,9 +3003,6 @@ static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size){
             }
 
             if(hx->redundant_pic_count==0
-#if FF_API_HURRY_UP
-               && hx->s.hurry_up < 5
-#endif
                && (avctx->skip_frame < AVDISCARD_NONREF || hx->nal_ref_idc)
                && (avctx->skip_frame < AVDISCARD_BIDIR  || hx->slice_type_nos!=FF_B_TYPE)
                && (avctx->skip_frame < AVDISCARD_NONKEY || hx->slice_type_nos==FF_I_TYPE)
@@ -3047,9 +3040,6 @@ static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size){
 
             if(hx->redundant_pic_count==0 && hx->intra_gb_ptr && hx->s.data_partitioning
                && s->context_initialized
-#if FF_API_HURRY_UP
-               && s->hurry_up < 5
-#endif
                && (avctx->skip_frame < AVDISCARD_NONREF || hx->nal_ref_idc)
                && (avctx->skip_frame < AVDISCARD_BIDIR  || hx->slice_type_nos!=FF_B_TYPE)
                && (avctx->skip_frame < AVDISCARD_NONKEY || hx->slice_type_nos==FF_I_TYPE)
@@ -3186,11 +3176,7 @@ static int decode_frame(AVCodecContext *avctx,
     }
 
     if(!(s->flags2 & CODEC_FLAG2_CHUNKS) && !s->current_picture_ptr){
-        if (avctx->skip_frame >= AVDISCARD_NONREF
-#if FF_API_HURRY_UP
-                || s->hurry_up
-#endif
-           )
+        if (avctx->skip_frame >= AVDISCARD_NONREF)
             return 0;
         av_log(avctx, AV_LOG_ERROR, "no frame!\n");
         return -1;
diff --git a/libavcodec/h264.h b/libavcodec/h264.h
index ab02de1dfb..a0f688d180 100644
--- a/libavcodec/h264.h
+++ b/libavcodec/h264.h
@@ -1007,7 +1007,6 @@ static void fill_decode_caches(H264Context *h, int mb_type){
     }
     }
 
-#if 1
     if(IS_INTER(mb_type) || (IS_DIRECT(mb_type) && h->direct_spatial_mv_pred)){
         int list;
         for(list=0; list<h->list_count; list++){
@@ -1182,7 +1181,6 @@ static void fill_decode_caches(H264Context *h, int mb_type){
             }
         }
     }
-#endif
 
         h->neighbor_transform_size= !!IS_8x8DCT(top_type) + !!IS_8x8DCT(left_type[0]);
 }
diff --git a/libavcodec/imc.c b/libavcodec/imc.c
index cb0c7102d8..f08efe46ec 100644
--- a/libavcodec/imc.c
+++ b/libavcodec/imc.c
@@ -88,7 +88,7 @@ typedef struct {
 
     DSPContext dsp;
     FFTContext fft;
-    DECLARE_ALIGNED(16, FFTComplex, samples)[COEFFS/2];
+    DECLARE_ALIGNED(32, FFTComplex, samples)[COEFFS/2];
     float *out_samples;
 } IMCContext;
 
diff --git a/libavcodec/motion_est_template.c b/libavcodec/motion_est_template.c
index 8f730ef14f..3a37bde063 100644
--- a/libavcodec/motion_est_template.c
+++ b/libavcodec/motion_est_template.c
@@ -158,7 +158,6 @@ static int hpel_motion_search(MpegEncContext * s,
         const int b= score_map[(index+(1<<ME_MAP_SHIFT))&(ME_MAP_SIZE-1)]
                      + (mv_penalty[bx   - pred_x] + mv_penalty[by+2 - pred_y])*c->penalty_factor;
 
-#if 1
         int key;
         int map_generation= c->map_generation;
 #ifndef NDEBUG
@@ -172,7 +171,6 @@ static int hpel_motion_search(MpegEncContext * s,
         assert(map[(index+1)&(ME_MAP_SIZE-1)] == key);
         key= ((my)<<ME_MAP_MV_BITS) + (mx-1) + map_generation;
         assert(map[(index-1)&(ME_MAP_SIZE-1)] == key);
-#endif
         if(t<=b){
             CHECK_HALF_MV(0, 1, mx  ,my-1)
             if(l<=r){
diff --git a/libavcodec/mpeg12.c b/libavcodec/mpeg12.c
index 192ecdd478..510a2b0809 100644
--- a/libavcodec/mpeg12.c
+++ b/libavcodec/mpeg12.c
@@ -2476,18 +2476,10 @@ static int decode_chunks(AVCodecContext *avctx,
                 /* Skip P-frames if we do not have a reference frame or we have an invalid header. */
                     if(s2->pict_type==FF_P_TYPE && !s->sync) break;
                 }
-#if FF_API_HURRY_UP
-                /* Skip B-frames if we are in a hurry. */
-                if(avctx->hurry_up && s2->pict_type==FF_B_TYPE) break;
-#endif
                 if(  (avctx->skip_frame >= AVDISCARD_NONREF && s2->pict_type==FF_B_TYPE)
                     ||(avctx->skip_frame >= AVDISCARD_NONKEY && s2->pict_type!=FF_I_TYPE)
                     || avctx->skip_frame >= AVDISCARD_ALL)
                     break;
-#if FF_API_HURRY_UP
-                /* Skip everything if we are in a hurry>=5. */
-                if(avctx->hurry_up>=5) break;
-#endif
 
                 if (!s->mpeg_enc_ctx_allocated) break;
 
diff --git a/libavcodec/mpegvideo.c b/libavcodec/mpegvideo.c
index 819f58dc95..73305d1324 100644
--- a/libavcodec/mpegvideo.c
+++ b/libavcodec/mpegvideo.c
@@ -1131,9 +1131,6 @@ int MPV_frame_start(MpegEncContext *s, AVCodecContext *avctx)
         }
     }
 
-#if FF_API_HURRY_UP
-    s->hurry_up= s->avctx->hurry_up;
-#endif
     s->error_recognition= avctx->error_recognition;
 
     /* set dequantizer, we can't do it during init as it might change for mpeg4
@@ -2125,9 +2122,6 @@ void MPV_decode_mb_internal(MpegEncContext *s, DCTELEM block[12][64],
             }
 
             /* skip dequant / idct if we are really late ;) */
-#if FF_API_HURRY_UP
-            if(s->hurry_up>1) goto skip_idct;
-#endif
             if(s->avctx->skip_idct){
                 if(  (s->avctx->skip_idct >= AVDISCARD_NONREF && s->pict_type == FF_B_TYPE)
                    ||(s->avctx->skip_idct >= AVDISCARD_NONKEY && s->pict_type != FF_I_TYPE)
diff --git a/libavcodec/mpegvideo.h b/libavcodec/mpegvideo.h
index 3836a6ef38..c90912f853 100644
--- a/libavcodec/mpegvideo.h
+++ b/libavcodec/mpegvideo.h
@@ -391,11 +391,6 @@ typedef struct MpegEncContext {
     int no_rounding;  /**< apply no rounding to motion compensation (MPEG4, msmpeg4, ...)
                         for b-frames rounding mode is always 0 */
 
-#if FF_API_HURRY_UP
-    int hurry_up;     /**< when set to 1 during decoding, b frames will be skipped
-                         when set to 2 idct/dequant will be skipped too */
-#endif
-
     /* macroblock layer */
     int mb_x, mb_y;
     int mb_skip_run;
diff --git a/libavcodec/msmpeg4.c b/libavcodec/msmpeg4.c
index 7b179b6d5e..9cbe2b8bef 100644
--- a/libavcodec/msmpeg4.c
+++ b/libavcodec/msmpeg4.c
@@ -985,10 +985,9 @@ void ff_msmpeg4_encode_block(MpegEncContext * s, DCTELEM * block, int n)
             if(level<=MAX_LEVEL && run<=MAX_RUN){
                 s->ac_stats[s->mb_intra][n>3][level][run][last]++;
             }
-#if 0
-else
-    s->ac_stats[s->mb_intra][n>3][40][63][0]++; //esc3 like
-#endif
+
+            s->ac_stats[s->mb_intra][n > 3][40][63][0]++; //esc3 like
+
             code = get_rl_index(rl, last, run, level);
             put_bits(&s->pb, rl->table_vlc[code][1], rl->table_vlc[code][0]);
             if (code == rl->n) {
diff --git a/libavcodec/nellymoserdec.c b/libavcodec/nellymoserdec.c
index 5ad49abceb..59c1b3bdd8 100644
--- a/libavcodec/nellymoserdec.c
+++ b/libavcodec/nellymoserdec.c
@@ -47,7 +47,7 @@
 
 typedef struct NellyMoserDecodeContext {
     AVCodecContext* avctx;
-    DECLARE_ALIGNED(16, float,float_buf)[NELLY_SAMPLES];
+    DECLARE_ALIGNED(32, float, float_buf)[NELLY_SAMPLES];
     float           state[128];
     AVLFG           random_state;
     GetBitContext   gb;
@@ -55,7 +55,7 @@ typedef struct NellyMoserDecodeContext {
     DSPContext      dsp;
     FFTContext      imdct_ctx;
     FmtConvertContext fmt_conv;
-    DECLARE_ALIGNED(16, float,imdct_out)[NELLY_BUF_LEN * 2];
+    DECLARE_ALIGNED(32, float, imdct_out)[NELLY_BUF_LEN * 2];
 } NellyMoserDecodeContext;
 
 static void overlap_and_window(NellyMoserDecodeContext *s, float *state, float *audio, float *a_in)
diff --git a/libavcodec/nellymoserenc.c b/libavcodec/nellymoserenc.c
index 81a1d43e4e..1d35cda9a1 100644
--- a/libavcodec/nellymoserenc.c
+++ b/libavcodec/nellymoserenc.c
@@ -55,9 +55,9 @@ typedef struct NellyMoserEncodeContext {
     int             have_saved;
     DSPContext      dsp;
     FFTContext      mdct_ctx;
-    DECLARE_ALIGNED(16, float, mdct_out)[NELLY_SAMPLES];
-    DECLARE_ALIGNED(16, float, in_buff)[NELLY_SAMPLES];
-    DECLARE_ALIGNED(16, float, buf)[2][3 * NELLY_BUF_LEN];     ///< sample buffer
+    DECLARE_ALIGNED(32, float, mdct_out)[NELLY_SAMPLES];
+    DECLARE_ALIGNED(32, float, in_buff)[NELLY_SAMPLES];
+    DECLARE_ALIGNED(32, float, buf)[2][3 * NELLY_BUF_LEN];     ///< sample buffer
     float           (*opt )[NELLY_BANDS];
     uint8_t         (*path)[NELLY_BANDS];
 } NellyMoserEncodeContext;
diff --git a/libavcodec/options.c b/libavcodec/options.c
index ebe228ee2f..2a70b1b21b 100644
--- a/libavcodec/options.c
+++ b/libavcodec/options.c
@@ -105,9 +105,6 @@ static const AVOption options[]={
 {"extradata_size", NULL, OFFSET(extradata_size), FF_OPT_TYPE_INT, DEFAULT, INT_MIN, INT_MAX},
 {"time_base", NULL, OFFSET(time_base), FF_OPT_TYPE_RATIONAL, DEFAULT, INT_MIN, INT_MAX},
 {"g", "set the group of picture size", OFFSET(gop_size), FF_OPT_TYPE_INT, 12, INT_MIN, INT_MAX, V|E},
-#if FF_API_RATE_EMU
-{"rate_emu", "frame rate emulation", OFFSET(rate_emu), FF_OPT_TYPE_INT, DEFAULT, INT_MIN, INT_MAX},
-#endif
 {"ar", "set audio sampling rate (in Hz)", OFFSET(sample_rate), FF_OPT_TYPE_INT, DEFAULT, INT_MIN, INT_MAX},
 {"ac", "set number of audio channels", OFFSET(channels), FF_OPT_TYPE_INT, DEFAULT, INT_MIN, INT_MAX},
 {"cutoff", "set cutoff bandwidth", OFFSET(cutoff), FF_OPT_TYPE_INT, DEFAULT, INT_MIN, INT_MAX, A|E},
@@ -124,9 +121,6 @@ static const AVOption options[]={
 {"rc_strategy", "ratecontrol method", OFFSET(rc_strategy), FF_OPT_TYPE_INT, DEFAULT, INT_MIN, INT_MAX, V|E},
 {"b_strategy", "strategy to choose between I/P/B-frames", OFFSET(b_frame_strategy), FF_OPT_TYPE_INT, 0, INT_MIN, INT_MAX, V|E},
 {"wpredp", "weighted prediction analysis method", OFFSET(weighted_p_pred), FF_OPT_TYPE_INT, 0, INT_MIN, INT_MAX, V|E},
-#if FF_API_HURRY_UP
-{"hurry_up", "deprecated, use skip_idct/skip_frame instead", OFFSET(hurry_up), FF_OPT_TYPE_INT, DEFAULT, INT_MIN, INT_MAX, V|D},
-#endif
 {"ps", "rtp payload size in bytes", OFFSET(rtp_payload_size), FF_OPT_TYPE_INT, DEFAULT, INT_MIN, INT_MAX, V|E},
 {"mv_bits", NULL, OFFSET(mv_bits), FF_OPT_TYPE_INT, DEFAULT, INT_MIN, INT_MAX},
 {"header_bits", NULL, OFFSET(header_bits), FF_OPT_TYPE_INT, DEFAULT, INT_MIN, INT_MAX},
@@ -253,10 +247,6 @@ static const AVOption options[]={
 {"pf", "forward predicted MVs of P-frames", 0, FF_OPT_TYPE_CONST, FF_DEBUG_VIS_MV_P_FOR, INT_MIN, INT_MAX, V|D, "debug_mv"},
 {"bf", "forward predicted MVs of B-frames", 0, FF_OPT_TYPE_CONST, FF_DEBUG_VIS_MV_B_FOR, INT_MIN, INT_MAX, V|D, "debug_mv"},
 {"bb", "backward predicted MVs of B-frames", 0, FF_OPT_TYPE_CONST, FF_DEBUG_VIS_MV_B_BACK, INT_MIN, INT_MAX, V|D, "debug_mv"},
-#if FF_API_MB_Q
-{"mb_qmin", "obsolete, use qmin", OFFSET(mb_qmin), FF_OPT_TYPE_INT, DEFAULT, INT_MIN, INT_MAX, V|E},
-{"mb_qmax", "obsolete, use qmax", OFFSET(mb_qmax), FF_OPT_TYPE_INT, DEFAULT, INT_MIN, INT_MAX, V|E},
-#endif
 {"cmp", "full pel me compare function", OFFSET(me_cmp), FF_OPT_TYPE_INT, DEFAULT, INT_MIN, INT_MAX, V|E, "cmp_func"},
 {"subcmp", "sub pel me compare function", OFFSET(me_sub_cmp), FF_OPT_TYPE_INT, DEFAULT, INT_MIN, INT_MAX, V|E, "cmp_func"},
 {"mbcmp", "macroblock compare function", OFFSET(mb_cmp), FF_OPT_TYPE_INT, DEFAULT, INT_MIN, INT_MAX, V|E, "cmp_func"},
diff --git a/libavcodec/pthread.c b/libavcodec/pthread.c
index e0eea4b1f1..6d4ee549cb 100644
--- a/libavcodec/pthread.c
+++ b/libavcodec/pthread.c
@@ -380,9 +380,6 @@ static void update_context_from_user(AVCodecContext *dst, AVCodecContext *src)
     dst->release_buffer = src->release_buffer;
 
     dst->opaque   = src->opaque;
-#if FF_API_HURRY_UP
-    dst->hurry_up = src->hurry_up;
-#endif
     dst->dsp_mask = src->dsp_mask;
     dst->debug    = src->debug;
     dst->debug_mv = src->debug_mv;
diff --git a/libavcodec/qdm2.c b/libavcodec/qdm2.c
index 0d5eeac0ce..a64870a3f9 100644
--- a/libavcodec/qdm2.c
+++ b/libavcodec/qdm2.c
@@ -120,7 +120,7 @@ typedef struct {
 } FFTCoefficient;
 
 typedef struct {
-    DECLARE_ALIGNED(16, QDM2Complex, complex)[MPA_MAX_CHANNELS][256];
+    DECLARE_ALIGNED(32, QDM2Complex, complex)[MPA_MAX_CHANNELS][256];
 } QDM2FFT;
 
 /**
diff --git a/libavcodec/rv34.c b/libavcodec/rv34.c
index 32d8198eeb..b11e84ea0b 100644
--- a/libavcodec/rv34.c
+++ b/libavcodec/rv34.c
@@ -1454,19 +1454,10 @@ int ff_rv34_decode_frame(AVCodecContext *avctx,
     }
     if((!s->last_picture_ptr || !s->last_picture_ptr->data[0]) && si.type == FF_B_TYPE)
         return -1;
-#if FF_API_HURRY_UP
-    /* skip b frames if we are in a hurry */
-    if(avctx->hurry_up && si.type==FF_B_TYPE) return buf_size;
-#endif
     if(   (avctx->skip_frame >= AVDISCARD_NONREF && si.type==FF_B_TYPE)
        || (avctx->skip_frame >= AVDISCARD_NONKEY && si.type!=FF_I_TYPE)
        ||  avctx->skip_frame >= AVDISCARD_ALL)
         return buf_size;
-#if FF_API_HURRY_UP
-    /* skip everything if we are in a hurry>=5 */
-    if(avctx->hurry_up>=5)
-        return buf_size;
-#endif
 
     for(i=0; i<slice_count; i++){
         int offset= get_slice_offset(avctx, slices_hdr, i);
diff --git a/libavcodec/sh4/qpel.c b/libavcodec/sh4/qpel.c
index 917068d0c6..2e30ae1137 100644
--- a/libavcodec/sh4/qpel.c
+++ b/libavcodec/sh4/qpel.c
@@ -897,7 +897,6 @@ QPEL_MC(0, avg_       , _       , op_avg)
 #undef op_put
 #undef op_put_no_rnd
 
-#if 1
 #define H264_LOWPASS(OPNAME, OP, OP2) \
 static inline void OPNAME ## h264_qpel_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,int w,int h){\
     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
@@ -1298,7 +1297,6 @@ H264_MC(avg_, 16)
 #undef op_put
 #undef op2_avg
 #undef op2_put
-#endif
 
 static void wmv2_mspel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){
     uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;
diff --git a/libavcodec/snow.c b/libavcodec/snow.c
index f330fe32e6..0668f67b57 100644
--- a/libavcodec/snow.c
+++ b/libavcodec/snow.c
@@ -3293,10 +3293,8 @@ static void iterative_me(SnowContext *s){
                 }
                 best_rd= ref_rd;
                 *block= ref_b;
-#if 1
                 check_block(s, mb_x, mb_y, color, 1, *obmc_edged, &best_rd);
                 //FIXME RD style color selection
-#endif
                 if(!same_block(block, &backup)){
                     if(tb ) tb ->type &= ~BLOCK_OPT;
                     if(lb ) lb ->type &= ~BLOCK_OPT;
diff --git a/libavcodec/svq1dec.c b/libavcodec/svq1dec.c
index cb84ea98a3..dca1ea0c9a 100644
--- a/libavcodec/svq1dec.c
+++ b/libavcodec/svq1dec.c
@@ -684,9 +684,6 @@ static int svq1_decode_frame(AVCodecContext *avctx,
   //this should be removed after libavcodec can handle more flexible picture types & ordering
   if(s->pict_type==FF_B_TYPE && s->last_picture_ptr==NULL) return buf_size;
 
-#if FF_API_HURRY_UP
-  if(avctx->hurry_up && s->pict_type==FF_B_TYPE) return buf_size;
-#endif
   if(  (avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type==FF_B_TYPE)
      ||(avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type!=FF_I_TYPE)
      || avctx->skip_frame >= AVDISCARD_ALL)
diff --git a/libavcodec/svq3.c b/libavcodec/svq3.c
index c7119b975b..ece514dc2e 100644
--- a/libavcodec/svq3.c
+++ b/libavcodec/svq3.c
@@ -952,14 +952,6 @@ static int svq3_decode_frame(AVCodecContext *avctx,
     /* Skip B-frames if we do not have reference frames. */
     if (s->last_picture_ptr == NULL && s->pict_type == FF_B_TYPE)
         return 0;
-#if FF_API_HURRY_UP
-    /* Skip B-frames if we are in a hurry. */
-    if (avctx->hurry_up && s->pict_type == FF_B_TYPE)
-        return 0;
-    /* Skip everything if we are in a hurry >= 5. */
-    if (avctx->hurry_up >= 5)
-        return 0;
-#endif
     if (  (avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type == FF_B_TYPE)
         ||(avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type != FF_I_TYPE)
         || avctx->skip_frame >= AVDISCARD_ALL)
diff --git a/libavcodec/utils.c b/libavcodec/utils.c
index d6b71b42db..6cdbb43e48 100644
--- a/libavcodec/utils.c
+++ b/libavcodec/utils.c
@@ -1105,18 +1105,11 @@ void avcodec_default_free_buffers(AVCodecContext *s){
     s->internal_buffer_count=0;
 }
 
+#if FF_API_OLD_FF_PICT_TYPES
 char av_get_pict_type_char(int pict_type){
-    switch(pict_type){
-    case FF_I_TYPE: return 'I';
-    case FF_P_TYPE: return 'P';
-    case FF_B_TYPE: return 'B';
-    case FF_S_TYPE: return 'S';
-    case FF_SI_TYPE:return 'i';
-    case FF_SP_TYPE:return 'p';
-    case FF_BI_TYPE:return 'b';
-    default:        return '?';
-    }
+    return av_get_picture_type_char(pict_type);
 }
+#endif
 
 int av_get_bits_per_sample(enum CodecID codec_id){
     switch(codec_id){
diff --git a/libavcodec/vc1dec.c b/libavcodec/vc1dec.c
index 27695f184a..ddd59700fa 100644
--- a/libavcodec/vc1dec.c
+++ b/libavcodec/vc1dec.c
@@ -3519,21 +3519,11 @@ static int vc1_decode_frame(AVCodecContext *avctx,
     if(s->last_picture_ptr==NULL && (s->pict_type==FF_B_TYPE || s->dropable)){
         goto err;
     }
-#if FF_API_HURRY_UP
-    /* skip b frames if we are in a hurry */
-    if(avctx->hurry_up && s->pict_type==FF_B_TYPE) return -1;//buf_size;
-#endif
     if(   (avctx->skip_frame >= AVDISCARD_NONREF && s->pict_type==FF_B_TYPE)
        || (avctx->skip_frame >= AVDISCARD_NONKEY && s->pict_type!=FF_I_TYPE)
        ||  avctx->skip_frame >= AVDISCARD_ALL) {
         goto end;
     }
-#if FF_API_HURRY_UP
-    /* skip everything if we are in a hurry>=5 */
-    if(avctx->hurry_up>=5) {
-        goto err;
-    }
-#endif
 
     if(s->next_p_frame_damaged){
         if(s->pict_type==FF_B_TYPE)
diff --git a/libavcodec/version.h b/libavcodec/version.h
index e5ca102814..10de67c85a 100644
--- a/libavcodec/version.h
+++ b/libavcodec/version.h
@@ -22,7 +22,7 @@
 
 #define LIBAVCODEC_VERSION_MAJOR 53
 #define LIBAVCODEC_VERSION_MINOR  1
-#define LIBAVCODEC_VERSION_MICRO  0
+#define LIBAVCODEC_VERSION_MICRO  1
 
 #define LIBAVCODEC_VERSION_INT  AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \
                                                LIBAVCODEC_VERSION_MINOR, \
@@ -47,15 +47,6 @@
 #ifndef FF_API_OLD_AUDIOCONVERT
 #define FF_API_OLD_AUDIOCONVERT (LIBAVCODEC_VERSION_MAJOR < 54)
 #endif
-#ifndef FF_API_HURRY_UP
-#define FF_API_HURRY_UP         (LIBAVCODEC_VERSION_MAJOR < 53)
-#endif
-#ifndef FF_API_RATE_EMU
-#define FF_API_RATE_EMU         (LIBAVCODEC_VERSION_MAJOR < 53)
-#endif
-#ifndef FF_API_MB_Q
-#define FF_API_MB_Q             (LIBAVCODEC_VERSION_MAJOR < 53)
-#endif
 #ifndef FF_API_ANTIALIAS_ALGO
 #define FF_API_ANTIALIAS_ALGO   (LIBAVCODEC_VERSION_MAJOR < 54)
 #endif
@@ -68,5 +59,8 @@
 #ifndef FF_API_THREAD_INIT
 #define FF_API_THREAD_INIT      (LIBAVCODEC_VERSION_MAJOR < 54)
 #endif
+#ifndef FF_API_OLD_FF_PICT_TYPES
+#define FF_API_OLD_FF_PICT_TYPES (LIBAVCODEC_VERSION_MAJOR < 54)
+#endif
 
 #endif /* AVCODEC_VERSION_H */
diff --git a/libavcodec/wma.h b/libavcodec/wma.h
index a51b3e83cf..fd9d5d8d3b 100644
--- a/libavcodec/wma.h
+++ b/libavcodec/wma.h
@@ -113,15 +113,15 @@ typedef struct WMACodecContext {
     uint8_t ms_stereo;                      ///< true if mid/side stereo mode
     uint8_t channel_coded[MAX_CHANNELS];    ///< true if channel is coded
     int exponents_bsize[MAX_CHANNELS];      ///< log2 ratio frame/exp. length
-    DECLARE_ALIGNED(16, float, exponents)[MAX_CHANNELS][BLOCK_MAX_SIZE];
+    DECLARE_ALIGNED(32, float, exponents)[MAX_CHANNELS][BLOCK_MAX_SIZE];
     float max_exponent[MAX_CHANNELS];
     WMACoef coefs1[MAX_CHANNELS][BLOCK_MAX_SIZE];
-    DECLARE_ALIGNED(16, float, coefs)[MAX_CHANNELS][BLOCK_MAX_SIZE];
-    DECLARE_ALIGNED(16, FFTSample, output)[BLOCK_MAX_SIZE * 2];
+    DECLARE_ALIGNED(32, float, coefs)[MAX_CHANNELS][BLOCK_MAX_SIZE];
+    DECLARE_ALIGNED(32, FFTSample, output)[BLOCK_MAX_SIZE * 2];
     FFTContext mdct_ctx[BLOCK_NB_SIZES];
     float *windows[BLOCK_NB_SIZES];
     /* output buffer for one frame and the last for IMDCT windowing */
-    DECLARE_ALIGNED(16, float, frame_out)[MAX_CHANNELS][BLOCK_MAX_SIZE * 2];
+    DECLARE_ALIGNED(32, float, frame_out)[MAX_CHANNELS][BLOCK_MAX_SIZE * 2];
     /* last frame info */
     uint8_t last_superframe[MAX_CODED_SUPERFRAME_SIZE + 4]; /* padding added */
     int last_bitoffset;
diff --git a/libavcodec/wmaprodec.c b/libavcodec/wmaprodec.c
index b207e18b2f..c0492c36f5 100644
--- a/libavcodec/wmaprodec.c
+++ b/libavcodec/wmaprodec.c
@@ -145,7 +145,7 @@ typedef struct {
     uint8_t  table_idx;                               ///< index in sf_offsets for the scale factor reference block
     float*   coeffs;                                  ///< pointer to the subframe decode buffer
     uint16_t num_vec_coeffs;                          ///< number of vector coded coefficients
-    DECLARE_ALIGNED(16, float, out)[WMAPRO_BLOCK_MAX_SIZE + WMAPRO_BLOCK_MAX_SIZE / 2]; ///< output buffer
+    DECLARE_ALIGNED(32, float, out)[WMAPRO_BLOCK_MAX_SIZE + WMAPRO_BLOCK_MAX_SIZE / 2]; ///< output buffer
 } WMAProChannelCtx;
 
 /**
@@ -170,7 +170,7 @@ typedef struct WMAProDecodeCtx {
                       FF_INPUT_BUFFER_PADDING_SIZE];///< compressed frame data
     PutBitContext    pb;                            ///< context for filling the frame_data buffer
     FFTContext       mdct_ctx[WMAPRO_BLOCK_SIZES];  ///< MDCT context per block size
-    DECLARE_ALIGNED(16, float, tmp)[WMAPRO_BLOCK_MAX_SIZE]; ///< IMDCT output buffer
+    DECLARE_ALIGNED(32, float, tmp)[WMAPRO_BLOCK_MAX_SIZE]; ///< IMDCT output buffer
     float*           windows[WMAPRO_BLOCK_SIZES];   ///< windows for the different block sizes
 
     /* frame size dependent frame information (set during initialization) */
diff --git a/libavcodec/wmavoice.c b/libavcodec/wmavoice.c
index dfdfabbbb3..cd640b01fc 100644
--- a/libavcodec/wmavoice.c
+++ b/libavcodec/wmavoice.c
@@ -275,11 +275,11 @@ typedef struct {
                                   ///< by postfilter
     float denoise_filter_cache[MAX_FRAMESIZE];
     int   denoise_filter_cache_size; ///< samples in #denoise_filter_cache
-    DECLARE_ALIGNED(16, float, tilted_lpcs_pf)[0x80];
+    DECLARE_ALIGNED(32, float, tilted_lpcs_pf)[0x80];
                                   ///< aligned buffer for LPC tilting
-    DECLARE_ALIGNED(16, float, denoise_coeffs_pf)[0x80];
+    DECLARE_ALIGNED(32, float, denoise_coeffs_pf)[0x80];
                                   ///< aligned buffer for denoise coefficients
-    DECLARE_ALIGNED(16, float, synth_filter_out_buf)[0x80 + MAX_LSPS_ALIGN16];
+    DECLARE_ALIGNED(32, float, synth_filter_out_buf)[0x80 + MAX_LSPS_ALIGN16];
                                   ///< aligned buffer for postfilter speech
                                   ///< synthesis
     /**
diff --git a/libavcodec/x86/fft.c b/libavcodec/x86/fft.c
index 37081d67fc..13f7ea6ec2 100644
--- a/libavcodec/x86/fft.c
+++ b/libavcodec/x86/fft.c
@@ -25,7 +25,14 @@ av_cold void ff_fft_init_mmx(FFTContext *s)
 {
 #if HAVE_YASM
     int has_vectors = av_get_cpu_flags();
-    if (has_vectors & AV_CPU_FLAG_SSE && HAVE_SSE) {
+    if (has_vectors & AV_CPU_FLAG_AVX && HAVE_AVX && s->nbits >= 5) {
+        /* AVX for SB */
+        s->imdct_calc      = ff_imdct_calc_sse;
+        s->imdct_half      = ff_imdct_half_avx;
+        s->fft_permute     = ff_fft_permute_sse;
+        s->fft_calc        = ff_fft_calc_avx;
+        s->fft_permutation = FF_FFT_PERM_AVX;
+    } else if (has_vectors & AV_CPU_FLAG_SSE && HAVE_SSE) {
         /* SSE for P3/P4/K8 */
         s->imdct_calc  = ff_imdct_calc_sse;
         s->imdct_half  = ff_imdct_half_sse;
diff --git a/libavcodec/x86/fft.h b/libavcodec/x86/fft.h
index 2affacb62b..611db8508f 100644
--- a/libavcodec/x86/fft.h
+++ b/libavcodec/x86/fft.h
@@ -22,6 +22,7 @@
 #include "libavcodec/fft.h"
 
 void ff_fft_permute_sse(FFTContext *s, FFTComplex *z);
+void ff_fft_calc_avx(FFTContext *s, FFTComplex *z);
 void ff_fft_calc_sse(FFTContext *s, FFTComplex *z);
 void ff_fft_calc_3dn(FFTContext *s, FFTComplex *z);
 void ff_fft_calc_3dn2(FFTContext *s, FFTComplex *z);
@@ -32,6 +33,7 @@ void ff_imdct_calc_3dn2(FFTContext *s, FFTSample *output, const FFTSample *input
 void ff_imdct_half_3dn2(FFTContext *s, FFTSample *output, const FFTSample *input);
 void ff_imdct_calc_sse(FFTContext *s, FFTSample *output, const FFTSample *input);
 void ff_imdct_half_sse(FFTContext *s, FFTSample *output, const FFTSample *input);
+void ff_imdct_half_avx(FFTContext *s, FFTSample *output, const FFTSample *input);
 void ff_dct32_float_sse(FFTSample *out, const FFTSample *in);
 
 #endif
diff --git a/libavcodec/x86/fft_mmx.asm b/libavcodec/x86/fft_mmx.asm
index 68e20df3b0..b923f1f9a3 100644
--- a/libavcodec/x86/fft_mmx.asm
+++ b/libavcodec/x86/fft_mmx.asm
@@ -1,6 +1,7 @@
 ;******************************************************************************
 ;* FFT transform with SSE/3DNow optimizations
 ;* Copyright (c) 2008 Loren Merritt
+;* Copyright (c) 2011 Vitor Sessak
 ;*
 ;* This algorithm (though not any of the implementation details) is
 ;* based on libdjbfft by D. J. Bernstein.
@@ -49,9 +50,21 @@ endstruc
 SECTION_RODATA
 
 %define M_SQRT1_2 0.70710678118654752440
-ps_root2: times 4 dd M_SQRT1_2
-ps_root2mppm: dd -M_SQRT1_2, M_SQRT1_2, M_SQRT1_2, -M_SQRT1_2
-ps_p1p1m1p1: dd 0, 0, 1<<31, 0
+%define M_COS_PI_1_8 0.923879532511287
+%define M_COS_PI_3_8 0.38268343236509
+
+align 32
+ps_cos16_1: dd 1.0, M_COS_PI_1_8, M_SQRT1_2, M_COS_PI_3_8, 1.0, M_COS_PI_1_8, M_SQRT1_2, M_COS_PI_3_8
+ps_cos16_2: dd 0, M_COS_PI_3_8, M_SQRT1_2, M_COS_PI_1_8, 0, -M_COS_PI_3_8, -M_SQRT1_2, -M_COS_PI_1_8
+
+ps_root2: times 8 dd M_SQRT1_2
+ps_root2mppm: dd -M_SQRT1_2, M_SQRT1_2, M_SQRT1_2, -M_SQRT1_2, -M_SQRT1_2, M_SQRT1_2, M_SQRT1_2, -M_SQRT1_2
+ps_p1p1m1p1: dd 0, 0, 1<<31, 0, 0, 0, 1<<31, 0
+
+perm1: dd 0x00, 0x02, 0x03, 0x01, 0x03, 0x00, 0x02, 0x01
+perm2: dd 0x00, 0x01, 0x02, 0x03, 0x01, 0x00, 0x02, 0x03
+ps_p1p1m1p1root2: dd 1.0, 1.0, -1.0, 1.0, M_SQRT1_2, M_SQRT1_2, M_SQRT1_2, M_SQRT1_2
+ps_m1m1p1m1p1m1m1m1: dd 1<<31, 1<<31, 0, 1<<31, 0, 1<<31, 1<<31, 1<<31
 ps_m1p1: dd 1<<31, 0
 
 %assign i 16
@@ -96,51 +109,80 @@ section .text align=16
     SWAP     %3, %6
 %endmacro
 
+;  in: %1 = {r0,i0,r2,i2,r4,i4,r6,i6}
+;      %2 = {r1,i1,r3,i3,r5,i5,r7,i7}
+;      %3, %4, %5 tmp
+; out: %1 = {r0,r1,r2,r3,i0,i1,i2,i3}
+;      %2 = {r4,r5,r6,r7,i4,i5,i6,i7}
+%macro T8_AVX 5
+    vsubps     %5, %1, %2       ; v  = %1 - %2
+    vaddps     %3, %1, %2       ; w  = %1 + %2
+    vmulps     %2, %5, [ps_p1p1m1p1root2]  ; v *= vals1
+    vpermilps  %2, %2, [perm1]
+    vblendps   %1, %2, %3, 0x33 ; q = {w1,w2,v4,v2,w5,w6,v7,v6}
+    vshufps    %5, %3, %2, 0x4e ; r = {w3,w4,v1,v3,w7,w8,v8,v5}
+    vsubps     %4, %5, %1       ; s = r - q
+    vaddps     %1, %5, %1       ; u = r + q
+    vpermilps  %1, %1, [perm2]  ; k  = {u1,u2,u3,u4,u6,u5,u7,u8}
+    vshufps    %5, %4, %1, 0xbb
+    vshufps    %3, %4, %1, 0xee
+    vperm2f128 %3, %3, %5, 0x13
+    vxorps     %4, %4, [ps_m1m1p1m1p1m1m1m1]  ; s *= {1,1,-1,-1,1,-1,-1,-1}
+    vshufps    %2, %1, %4, 0xdd
+    vshufps    %1, %1, %4, 0x88
+    vperm2f128 %4, %2, %1, 0x02 ; v  = {k1,k3,s1,s3,k2,k4,s2,s4}
+    vperm2f128 %1, %1, %2, 0x13 ; w  = {k6,k8,s6,s8,k5,k7,s5,s7}
+    vsubps     %5, %1, %3
+    vblendps   %1, %5, %1, 0x55 ; w -= {0,s7,0,k7,0,s8,0,k8}
+    vsubps     %2, %4, %1       ; %2 = v - w
+    vaddps     %1, %4, %1       ; %1 = v + w
+%endmacro
+
+; In SSE mode do one fft4 transforms
 ; in:  %1={r0,i0,r2,i2} %2={r1,i1,r3,i3}
 ; out: %1={r0,r1,r2,r3} %2={i0,i1,i2,i3}
+;
+; In AVX mode do two fft4 transforms
+; in:  %1={r0,i0,r2,i2,r4,i4,r6,i6} %2={r1,i1,r3,i3,r5,i5,r7,i7}
+; out: %1={r0,r1,r2,r3,r4,r5,r6,r7} %2={i0,i1,i2,i3,i4,i5,i6,i7}
 %macro T4_SSE 3
-    mova     %3, %1
-    addps    %1, %2       ; {t1,t2,t6,t5}
-    subps    %3, %2       ; {t3,t4,-t8,t7}
-    xorps    %3, [ps_p1p1m1p1]
-    mova     %2, %1
-    shufps   %1, %3, 0x44 ; {t1,t2,t3,t4}
-    shufps   %2, %3, 0xbe ; {t6,t5,t7,t8}
-    mova     %3, %1
-    addps    %1, %2       ; {r0,i0,r1,i1}
-    subps    %3, %2       ; {r2,i2,r3,i3}
-    mova     %2, %1
-    shufps   %1, %3, 0x88 ; {r0,r1,r2,r3}
-    shufps   %2, %3, 0xdd ; {i0,i1,i2,i3}
+    subps    %3, %1, %2       ; {t3,t4,-t8,t7}
+    addps    %1, %1, %2       ; {t1,t2,t6,t5}
+    xorps    %3, %3, [ps_p1p1m1p1]
+    shufps   %2, %1, %3, 0xbe ; {t6,t5,t7,t8}
+    shufps   %1, %1, %3, 0x44 ; {t1,t2,t3,t4}
+    subps    %3, %1, %2       ; {r2,i2,r3,i3}
+    addps    %1, %1, %2       ; {r0,i0,r1,i1}
+    shufps   %2, %1, %3, 0xdd ; {i0,i1,i2,i3}
+    shufps   %1, %1, %3, 0x88 ; {r0,r1,r2,r3}
 %endmacro
 
+; In SSE mode do one FFT8
 ; in:  %1={r0,r1,r2,r3} %2={i0,i1,i2,i3} %3={r4,i4,r6,i6} %4={r5,i5,r7,i7}
 ; out: %1={r0,r1,r2,r3} %2={i0,i1,i2,i3} %1={r4,r5,r6,r7} %2={i4,i5,i6,i7}
+;
+; In AVX mode do two FFT8
+; in:  %1={r0,i0,r2,i2,r8, i8, r10,i10} %2={r1,i1,r3,i3,r9, i9, r11,i11}
+;      %3={r4,i4,r6,i6,r12,i12,r14,i14} %4={r5,i5,r7,i7,r13,i13,r15,i15}
+; out: %1={r0,r1,r2,r3,r8, r9, r10,r11} %2={i0,i1,i2,i3,i8, i9, i10,i11}
+;      %3={r4,r5,r6,r7,r12,r13,r14,r15} %4={i4,i5,i6,i7,i12,i13,i14,i15}
 %macro T8_SSE 6
-    mova     %6, %3
-    subps    %3, %4       ; {r5,i5,r7,i7}
-    addps    %6, %4       ; {t1,t2,t3,t4}
-    mova     %4, %3
-    shufps   %4, %4, 0xb1 ; {i5,r5,i7,r7}
-    mulps    %3, [ps_root2mppm] ; {-r5,i5,r7,-i7}
-    mulps    %4, [ps_root2]
-    addps    %3, %4       ; {t8,t7,ta,t9}
-    mova     %4, %6
-    shufps   %6, %3, 0x36 ; {t3,t2,t9,t8}
-    shufps   %4, %3, 0x9c ; {t1,t4,t7,ta}
-    mova     %3, %6
-    addps    %6, %4       ; {t1,t2,t9,ta}
-    subps    %3, %4       ; {t6,t5,tc,tb}
-    mova     %4, %6
-    shufps   %6, %3, 0xd8 ; {t1,t9,t5,tb}
-    shufps   %4, %3, 0x8d ; {t2,ta,t6,tc}
-    mova     %3, %1
-    mova     %5, %2
-    addps    %1, %6       ; {r0,r1,r2,r3}
-    addps    %2, %4       ; {i0,i1,i2,i3}
-    subps    %3, %6       ; {r4,r5,r6,r7}
-    subps    %5, %4       ; {i4,i5,i6,i7}
-    SWAP     %4, %5
+    addps    %6, %3, %4       ; {t1,t2,t3,t4}
+    subps    %3, %3, %4       ; {r5,i5,r7,i7}
+    shufps   %4, %3, %3, 0xb1 ; {i5,r5,i7,r7}
+    mulps    %3, %3, [ps_root2mppm] ; {-r5,i5,r7,-i7}
+    mulps    %4, %4, [ps_root2]
+    addps    %3, %3, %4       ; {t8,t7,ta,t9}
+    shufps   %4, %6, %3, 0x9c ; {t1,t4,t7,ta}
+    shufps   %6, %6, %3, 0x36 ; {t3,t2,t9,t8}
+    subps    %3, %6, %4       ; {t6,t5,tc,tb}
+    addps    %6, %6, %4       ; {t1,t2,t9,ta}
+    shufps   %5, %6, %3, 0x8d ; {t2,ta,t6,tc}
+    shufps   %6, %6, %3, 0xd8 ; {t1,t9,t5,tb}
+    subps    %3, %1, %6       ; {r4,r5,r6,r7}
+    addps    %1, %1, %6       ; {r0,r1,r2,r3}
+    subps    %4, %2, %5       ; {i4,i5,i6,i7}
+    addps    %2, %2, %5       ; {i0,i1,i2,i3}
 %endmacro
 
 ; scheduled for cpu-bound sizes
@@ -148,52 +190,44 @@ section .text align=16
 IF%1 mova    m4, Z(4)
 IF%1 mova    m5, Z(5)
     mova     m0, %2 ; wre
-    mova     m2, m4
     mova     m1, %3 ; wim
-    mova     m3, m5
-    mulps    m2, m0 ; r2*wre
+    mulps    m2, m4, m0 ; r2*wre
 IF%1 mova    m6, Z2(6)
-    mulps    m3, m1 ; i2*wim
+    mulps    m3, m5, m1 ; i2*wim
 IF%1 mova    m7, Z2(7)
-    mulps    m4, m1 ; r2*wim
-    mulps    m5, m0 ; i2*wre
-    addps    m2, m3 ; r2*wre + i2*wim
-    mova     m3, m1
-    mulps    m1, m6 ; r3*wim
-    subps    m5, m4 ; i2*wre - r2*wim
-    mova     m4, m0
-    mulps    m3, m7 ; i3*wim
-    mulps    m4, m6 ; r3*wre
-    mulps    m0, m7 ; i3*wre
-    subps    m4, m3 ; r3*wre - i3*wim
+    mulps    m4, m4, m1 ; r2*wim
+    mulps    m5, m5, m0 ; i2*wre
+    addps    m2, m2, m3 ; r2*wre + i2*wim
+    mulps    m3, m1, m7 ; i3*wim
+    subps    m5, m5, m4 ; i2*wre - r2*wim
+    mulps    m1, m1, m6 ; r3*wim
+    mulps    m4, m0, m6 ; r3*wre
+    mulps    m0, m0, m7 ; i3*wre
+    subps    m4, m4, m3 ; r3*wre - i3*wim
     mova     m3, Z(0)
-    addps    m0, m1 ; i3*wre + r3*wim
-    mova     m1, m4
-    addps    m4, m2 ; t5
-    subps    m1, m2 ; t3
-    subps    m3, m4 ; r2
-    addps    m4, Z(0) ; r0
+    addps    m0, m0, m1 ; i3*wre + r3*wim
+    subps    m1, m4, m2 ; t3
+    addps    m4, m4, m2 ; t5
+    subps    m3, m3, m4 ; r2
+    addps    m4, m4, Z(0) ; r0
     mova     m6, Z(2)
     mova   Z(4), m3
     mova   Z(0), m4
-    mova     m3, m5
-    subps    m5, m0 ; t4
-    mova     m4, m6
-    subps    m6, m5 ; r3
-    addps    m5, m4 ; r1
-    mova  Z2(6), m6
-    mova   Z(2), m5
+    subps    m3, m5, m0 ; t4
+    subps    m4, m6, m3 ; r3
+    addps    m3, m3, m6 ; r1
+    mova  Z2(6), m4
+    mova   Z(2), m3
     mova     m2, Z(3)
-    addps    m3, m0 ; t6
-    subps    m2, m1 ; i3
+    addps    m3, m5, m0 ; t6
+    subps    m2, m2, m1 ; i3
     mova     m7, Z(1)
-    addps    m1, Z(3) ; i1
+    addps    m1, m1, Z(3) ; i1
     mova  Z2(7), m2
     mova   Z(3), m1
-    mova     m4, m7
-    subps    m7, m3 ; i2
-    addps    m3, m4 ; i0
-    mova   Z(5), m7
+    subps    m4, m7, m3 ; i2
+    addps    m3, m3, m7 ; i0
+    mova   Z(5), m4
     mova   Z(1), m3
 %endmacro
 
@@ -201,77 +235,55 @@ IF%1 mova    m7, Z2(7)
 %macro PASS_BIG 1 ; (!interleave)
     mova     m4, Z(4) ; r2
     mova     m5, Z(5) ; i2
-    mova     m2, m4
     mova     m0, [wq] ; wre
-    mova     m3, m5
     mova     m1, [wq+o1q] ; wim
-    mulps    m2, m0 ; r2*wre
+    mulps    m2, m4, m0 ; r2*wre
     mova     m6, Z2(6) ; r3
-    mulps    m3, m1 ; i2*wim
+    mulps    m3, m5, m1 ; i2*wim
     mova     m7, Z2(7) ; i3
-    mulps    m4, m1 ; r2*wim
-    mulps    m5, m0 ; i2*wre
-    addps    m2, m3 ; r2*wre + i2*wim
-    mova     m3, m1
-    mulps    m1, m6 ; r3*wim
-    subps    m5, m4 ; i2*wre - r2*wim
-    mova     m4, m0
-    mulps    m3, m7 ; i3*wim
-    mulps    m4, m6 ; r3*wre
-    mulps    m0, m7 ; i3*wre
-    subps    m4, m3 ; r3*wre - i3*wim
+    mulps    m4, m4, m1 ; r2*wim
+    mulps    m5, m5, m0 ; i2*wre
+    addps    m2, m2, m3 ; r2*wre + i2*wim
+    mulps    m3, m1, m7 ; i3*wim
+    mulps    m1, m1, m6 ; r3*wim
+    subps    m5, m5, m4 ; i2*wre - r2*wim
+    mulps    m4, m0, m6 ; r3*wre
+    mulps    m0, m0, m7 ; i3*wre
+    subps    m4, m4, m3 ; r3*wre - i3*wim
     mova     m3, Z(0)
-    addps    m0, m1 ; i3*wre + r3*wim
-    mova     m1, m4
-    addps    m4, m2 ; t5
-    subps    m1, m2 ; t3
-    subps    m3, m4 ; r2
-    addps    m4, Z(0) ; r0
+    addps    m0, m0, m1 ; i3*wre + r3*wim
+    subps    m1, m4, m2 ; t3
+    addps    m4, m4, m2 ; t5
+    subps    m3, m3, m4 ; r2
+    addps    m4, m4, Z(0) ; r0
     mova     m6, Z(2)
     mova   Z(4), m3
     mova   Z(0), m4
-    mova     m3, m5
-    subps    m5, m0 ; t4
-    mova     m4, m6
-    subps    m6, m5 ; r3
-    addps    m5, m4 ; r1
-IF%1 mova Z2(6), m6
-IF%1 mova  Z(2), m5
+    subps    m3, m5, m0 ; t4
+    subps    m4, m6, m3 ; r3
+    addps    m3, m3, m6 ; r1
+IF%1 mova Z2(6), m4
+IF%1 mova  Z(2), m3
     mova     m2, Z(3)
-    addps    m3, m0 ; t6
-    subps    m2, m1 ; i3
+    addps    m5, m5, m0 ; t6
+    subps    m2, m2, m1 ; i3
     mova     m7, Z(1)
-    addps    m1, Z(3) ; i1
+    addps    m1, m1, Z(3) ; i1
 IF%1 mova Z2(7), m2
 IF%1 mova  Z(3), m1
-    mova     m4, m7
-    subps    m7, m3 ; i2
-    addps    m3, m4 ; i0
-IF%1 mova  Z(5), m7
-IF%1 mova  Z(1), m3
+    subps    m6, m7, m5 ; i2
+    addps    m5, m5, m7 ; i0
+IF%1 mova  Z(5), m6
+IF%1 mova  Z(1), m5
 %if %1==0
-    mova     m4, m5 ; r1
-    mova     m0, m6 ; r3
-    unpcklps m5, m1
-    unpckhps m4, m1
-    unpcklps m6, m2
-    unpckhps m0, m2
+    INTERL m1, m3, m7, Z, 2
+    INTERL m2, m4, m0, Z2, 6
+
     mova     m1, Z(0)
     mova     m2, Z(4)
-    mova   Z(2), m5
-    mova   Z(3), m4
-    mova  Z2(6), m6
-    mova  Z2(7), m0
-    mova     m5, m1 ; r0
-    mova     m4, m2 ; r2
-    unpcklps m1, m3
-    unpckhps m5, m3
-    unpcklps m2, m7
-    unpckhps m4, m7
-    mova   Z(0), m1
-    mova   Z(1), m5
-    mova   Z(4), m2
-    mova   Z(5), m4
+
+    INTERL m5, m1, m3, Z, 0
+    INTERL m6, m2, m7, Z, 4
 %endif
 %endmacro
 
@@ -281,13 +293,106 @@ IF%1 mova  Z(1), m3
     punpckhdq %3, %2
 %endmacro
 
-INIT_XMM
-%define mova movaps
-
 %define Z(x) [r0+mmsize*x]
 %define Z2(x) [r0+mmsize*x]
+%define ZH(x) [r0+mmsize*x+mmsize/2]
+
+INIT_YMM
+
+align 16
+fft8_avx:
+    mova      m0, Z(0)
+    mova      m1, Z(1)
+    T8_AVX    m0, m1, m2, m3, m4
+    mova      Z(0), m0
+    mova      Z(1), m1
+    ret
+
+
+align 16
+fft16_avx:
+    mova       m2, Z(2)
+    mova       m3, Z(3)
+    T4_SSE     m2, m3, m7
+
+    mova       m0, Z(0)
+    mova       m1, Z(1)
+    T8_AVX     m0, m1, m4, m5, m7
+
+    mova       m4, [ps_cos16_1]
+    mova       m5, [ps_cos16_2]
+    vmulps     m6, m2, m4
+    vmulps     m7, m3, m5
+    vaddps     m7, m7, m6
+    vmulps     m2, m2, m5
+    vmulps     m3, m3, m4
+    vsubps     m3, m3, m2
+    vblendps   m2, m7, m3, 0xf0
+    vperm2f128 m3, m7, m3, 0x21
+    vaddps     m4, m2, m3
+    vsubps     m2, m3, m2
+    vperm2f128 m2, m2, m2, 0x01
+    vsubps     m3, m1, m2
+    vaddps     m1, m1, m2
+    vsubps     m5, m0, m4
+    vaddps     m0, m0, m4
+    vextractf128   Z(0), m0, 0
+    vextractf128  ZH(0), m1, 0
+    vextractf128   Z(1), m0, 1
+    vextractf128  ZH(1), m1, 1
+    vextractf128   Z(2), m5, 0
+    vextractf128  ZH(2), m3, 0
+    vextractf128   Z(3), m5, 1
+    vextractf128  ZH(3), m3, 1
+    ret
+
+align 16
+fft32_avx:
+    call fft16_avx
+
+    mova m0, Z(4)
+    mova m1, Z(5)
+
+    T4_SSE      m0, m1, m4
+
+    mova m2, Z(6)
+    mova m3, Z(7)
+
+    T8_SSE      m0, m1, m2, m3, m4, m6
+    ; m0={r0,r1,r2,r3,r8, r9, r10,r11} m1={i0,i1,i2,i3,i8, i9, i10,i11}
+    ; m2={r4,r5,r6,r7,r12,r13,r14,r15} m3={i4,i5,i6,i7,i12,i13,i14,i15}
+
+    vperm2f128  m4, m0, m2, 0x20
+    vperm2f128  m5, m1, m3, 0x20
+    vperm2f128  m6, m0, m2, 0x31
+    vperm2f128  m7, m1, m3, 0x31
+
+    PASS_SMALL 0, [cos_32], [cos_32+32]
+
+    ret
+
+fft32_interleave_avx:
+    call fft32_avx
+    mov r2d, 32
+.deint_loop:
+    mova     m2, Z(0)
+    mova     m3, Z(1)
+    vunpcklps      m0, m2, m3
+    vunpckhps      m1, m2, m3
+    vextractf128   Z(0), m0, 0
+    vextractf128  ZH(0), m1, 0
+    vextractf128   Z(1), m0, 1
+    vextractf128  ZH(1), m1, 1
+    add r0, mmsize*2
+    sub r2d, mmsize/4
+    jg .deint_loop
+    ret
+
+INIT_XMM
+%define movdqa  movaps
 
 align 16
+fft4_avx:
 fft4_sse:
     mova     m0, Z(0)
     mova     m1, Z(1)
@@ -406,6 +511,8 @@ FFT48_3DN _3dn
 
 %define Z(x) [zq + o1q*(x&6) + mmsize*(x&1)]
 %define Z2(x) [zq + o3q + mmsize*(x&1)]
+%define ZH(x) [zq + o1q*(x&6) + mmsize*(x&1) + mmsize/2]
+%define Z2H(x) [zq + o3q + mmsize*(x&1) + mmsize/2]
 
 %macro DECL_PASS 2+ ; name, payload
 align 16
@@ -423,8 +530,34 @@ DEFINE_ARGS z, w, n, o1, o3
     rep ret
 %endmacro
 
+INIT_YMM
+
+%macro INTERL_AVX 5
+    vunpckhps      %3, %2, %1
+    vunpcklps      %2, %2, %1
+    vextractf128   %4(%5), %2, 0
+    vextractf128  %4 %+ H(%5), %3, 0
+    vextractf128   %4(%5 + 1), %2, 1
+    vextractf128  %4 %+ H(%5 + 1), %3, 1
+%endmacro
+
+%define INTERL INTERL_AVX
+
+DECL_PASS pass_avx, PASS_BIG 1
+DECL_PASS pass_interleave_avx, PASS_BIG 0
+
 INIT_XMM
-%define mova movaps
+
+%macro INTERL_SSE 5
+    mova     %3, %2
+    unpcklps %2, %1
+    unpckhps %3, %1
+    mova  %4(%5), %2
+    mova  %4(%5+1), %3
+%endmacro
+
+%define INTERL INTERL_SSE
+
 DECL_PASS pass_sse, PASS_BIG 1
 DECL_PASS pass_interleave_sse, PASS_BIG 0
 
@@ -457,9 +590,12 @@ DECL_PASS pass_interleave_3dn, PASS_BIG 0
 
 %macro DECL_FFT 2-3 ; nbits, cpu, suffix
 %xdefine list_of_fft fft4%2 SECTION_REL, fft8%2 SECTION_REL
-%if %1==5
+%if %1>=5
 %xdefine list_of_fft list_of_fft, fft16%2 SECTION_REL
 %endif
+%if %1>=6
+%xdefine list_of_fft list_of_fft, fft32%3%2 SECTION_REL
+%endif
 
 %assign n 1<<%1
 %rep 17-%1
@@ -492,9 +628,14 @@ section .text
 ; The others pass args in registers and don't spill anything.
 cglobal fft_dispatch%3%2, 2,5,8, z, nbits
     FFT_DISPATCH %3%2, nbits
+%ifidn %2, _avx
+    vzeroupper
+%endif
     RET
 %endmacro ; DECL_FFT
 
+DECL_FFT 6, _avx
+DECL_FFT 6, _avx, _interleave
 DECL_FFT 5, _sse
 DECL_FFT 5, _sse, _interleave
 DECL_FFT 4, _3dn
@@ -533,21 +674,53 @@ INIT_XMM
 %endmacro
 
 %macro CMUL 6 ;j, xmm0, xmm1, 3, 4, 5
-    movaps   xmm6, [%4+%1*2]
-    movaps   %2,   [%4+%1*2+0x10]
-    movaps   %3,   xmm6
-    movaps   xmm7, %2
-    mulps    xmm6, [%5+%1]
-    mulps    %2,   [%6+%1]
-    mulps    %3,   [%6+%1]
-    mulps    xmm7, [%5+%1]
-    subps    %2,   xmm6
-    addps    %3,   xmm7
+    mulps      m6, %3, [%5+%1]
+    mulps      m7, %2, [%5+%1]
+    mulps      %2, %2, [%6+%1]
+    mulps      %3, %3, [%6+%1]
+    subps      %2, %2, m6
+    addps      %3, %3, m7
+%endmacro
+
+%macro POSROTATESHUF_AVX 5 ;j, k, z+n8, tcos+n8, tsin+n8
+.post:
+    vmovaps      ymm1,   [%3+%1*2]
+    vmovaps      ymm0,   [%3+%1*2+0x20]
+    vmovaps      ymm3,   [%3+%2*2]
+    vmovaps      ymm2,   [%3+%2*2+0x20]
+
+    CMUL         %1, ymm0, ymm1, %3, %4, %5
+    CMUL         %2, ymm2, ymm3, %3, %4, %5
+    vshufps      ymm1, ymm1, ymm1, 0x1b
+    vshufps      ymm3, ymm3, ymm3, 0x1b
+    vperm2f128   ymm1, ymm1, ymm1, 0x01
+    vperm2f128   ymm3, ymm3, ymm3, 0x01
+    vunpcklps    ymm6, ymm2, ymm1
+    vunpckhps    ymm4, ymm2, ymm1
+    vunpcklps    ymm7, ymm0, ymm3
+    vunpckhps    ymm5, ymm0, ymm3
+
+    vextractf128 [%3+%1*2],      ymm7, 0
+    vextractf128 [%3+%1*2+0x10], ymm5, 0
+    vextractf128 [%3+%1*2+0x20], ymm7, 1
+    vextractf128 [%3+%1*2+0x30], ymm5, 1
+
+    vextractf128 [%3+%2*2],      ymm6, 0
+    vextractf128 [%3+%2*2+0x10], ymm4, 0
+    vextractf128 [%3+%2*2+0x20], ymm6, 1
+    vextractf128 [%3+%2*2+0x30], ymm4, 1
+    sub      %2,   0x20
+    add      %1,   0x20
+    jl       .post
 %endmacro
 
 %macro POSROTATESHUF 5 ;j, k, z+n8, tcos+n8, tsin+n8
 .post:
+    movaps   xmm1, [%3+%1*2]
+    movaps   xmm0, [%3+%1*2+0x10]
     CMUL     %1,   xmm0, xmm1, %3, %4, %5
+    movaps   xmm5, [%3+%2*2]
+    movaps   xmm4, [%3+%2*2+0x10]
     CMUL     %2,   xmm4, xmm5, %3, %4, %5
     shufps   xmm1, xmm1, 0x1b
     shufps   xmm5, xmm5, 0x1b
@@ -566,7 +739,8 @@ INIT_XMM
     jl       .post
 %endmacro
 
-cglobal imdct_half_sse, 3,7,8; FFTContext *s, FFTSample *output, const FFTSample *input
+%macro DECL_IMDCT 2
+cglobal imdct_half%1, 3,7,8; FFTContext *s, FFTSample *output, const FFTSample *input
 %ifdef ARCH_X86_64
 %define rrevtab r10
 %define rtcos   r11
@@ -641,7 +815,7 @@ cglobal imdct_half_sse, 3,7,8; FFTContext *s, FFTSample *output, const FFTSample
     mov  r0, r1
     mov  r1d, [r5+FFTContext.nbits]
 
-    FFT_DISPATCH _sse, r1
+    FFT_DISPATCH %1, r1
 
     mov  r0d, [r5+FFTContext.mdctsize]
     add  r6, r0
@@ -653,9 +827,9 @@ cglobal imdct_half_sse, 3,7,8; FFTContext *s, FFTSample *output, const FFTSample
     mov  rtsin, [esp+4]
 %endif
     neg  r0
-    mov  r1, -16
+    mov  r1, -mmsize
     sub  r1, r0
-    POSROTATESHUF r0, r1, r6, rtcos, rtsin
+    %2 r0, r1, r6, rtcos, rtsin
 %ifdef ARCH_X86_64
     pop  r14
     pop  r13
@@ -663,4 +837,14 @@ cglobal imdct_half_sse, 3,7,8; FFTContext *s, FFTSample *output, const FFTSample
 %else
     add esp, 12
 %endif
+%ifidn avx_enabled, 1
+    vzeroupper
+%endif
     RET
+%endmacro
+
+DECL_IMDCT _sse, POSROTATESHUF
+
+INIT_YMM
+
+DECL_IMDCT _avx, POSROTATESHUF_AVX
diff --git a/libavcodec/x86/fft_sse.c b/libavcodec/x86/fft_sse.c
index 0d151bc446..801dc1bc8a 100644
--- a/libavcodec/x86/fft_sse.c
+++ b/libavcodec/x86/fft_sse.c
@@ -28,6 +28,12 @@ DECLARE_ASM_CONST(16, int, ff_m1m1m1m1)[4] =
 
 void ff_fft_dispatch_sse(FFTComplex *z, int nbits);
 void ff_fft_dispatch_interleave_sse(FFTComplex *z, int nbits);
+void ff_fft_dispatch_interleave_avx(FFTComplex *z, int nbits);
+
+void ff_fft_calc_avx(FFTContext *s, FFTComplex *z)
+{
+    ff_fft_dispatch_interleave_avx(z, s->nbits);
+}
 
 void ff_fft_calc_sse(FFTContext *s, FFTComplex *z)
 {
@@ -77,7 +83,7 @@ void ff_imdct_calc_sse(FFTContext *s, FFTSample *output, const FFTSample *input)
     long n = s->mdct_size;
     long n4 = n >> 2;
 
-    ff_imdct_half_sse(s, output+n4, input);
+    s->imdct_half(s, output + n4, input);
 
     j = -n;
     k = n-16;
diff --git a/libavcodec/x86/x86inc.asm b/libavcodec/x86/x86inc.asm
index b7d17742e4..53091c14c9 100644
--- a/libavcodec/x86/x86inc.asm
+++ b/libavcodec/x86/x86inc.asm
@@ -1,10 +1,11 @@
 ;*****************************************************************************
 ;* x86inc.asm
 ;*****************************************************************************
-;* Copyright (C) 2005-2008 x264 project
+;* Copyright (C) 2005-2011 x264 project
 ;*
 ;* Authors: Loren Merritt <lorenm@u.washington.edu>
 ;*          Anton Mitrofanov <BugMaster@narod.ru>
+;*          Jason Garrett-Glaser <darkshikari@gmail.com>
 ;*
 ;* Permission to use, copy, modify, and/or distribute this software for any
 ;* purpose with or without fee is hereby granted, provided that the above
@@ -499,6 +500,7 @@ SECTION .note.GNU-stack noalloc noexec nowrite progbits
 %endmacro
 
 %macro INIT_MMX 0
+    %assign avx_enabled 0
     %define RESET_MM_PERMUTATION INIT_MMX
     %define mmsize 8
     %define num_mmregs 8
@@ -520,6 +522,7 @@ SECTION .note.GNU-stack noalloc noexec nowrite progbits
 %endmacro
 
 %macro INIT_XMM 0
+    %assign avx_enabled 0
     %define RESET_MM_PERMUTATION INIT_XMM
     %define mmsize 16
     %define num_mmregs 8
@@ -538,6 +541,31 @@ SECTION .note.GNU-stack noalloc noexec nowrite progbits
     %endrep
 %endmacro
 
+%macro INIT_AVX 0
+    INIT_XMM
+    %assign avx_enabled 1
+    %define PALIGNR PALIGNR_SSSE3
+    %define RESET_MM_PERMUTATION INIT_AVX
+%endmacro
+
+%macro INIT_YMM 0
+    %assign avx_enabled 1
+    %define RESET_MM_PERMUTATION INIT_YMM
+    %define mmsize 32
+    %define num_mmregs 8
+    %ifdef ARCH_X86_64
+    %define num_mmregs 16
+    %endif
+    %define mova vmovaps
+    %define movu vmovups
+    %assign %%i 0
+    %rep num_mmregs
+    CAT_XDEFINE m, %%i, ymm %+ %%i
+    CAT_XDEFINE nymm, %%i, %%i
+    %assign %%i %%i+1
+    %endrep
+%endmacro
+
 INIT_MMX
 
 ; I often want to use macros that permute their arguments. e.g. there's no
@@ -645,3 +673,222 @@ INIT_MMX
         sub %1, %2
     %endif
 %endmacro
+
+;=============================================================================
+; AVX abstraction layer
+;=============================================================================
+
+%assign i 0
+%rep 16
+    %if i < 8
+        CAT_XDEFINE sizeofmm, i, 8
+    %endif
+    CAT_XDEFINE sizeofxmm, i, 16
+    CAT_XDEFINE sizeofymm, i, 32
+%assign i i+1
+%endrep
+%undef i
+
+;%1 == instruction
+;%2 == 1 if float, 0 if int
+;%3 == 0 if 3-operand (xmm, xmm, xmm), 1 if 4-operand (xmm, xmm, xmm, imm)
+;%4 == number of operands given
+;%5+: operands
+%macro RUN_AVX_INSTR 6-7+
+    %if sizeof%5==32
+        v%1 %5, %6, %7
+    %else
+        %if sizeof%5==8
+            %define %%regmov movq
+        %elif %2
+            %define %%regmov movaps
+        %else
+            %define %%regmov movdqa
+        %endif
+
+        %if %4>=3+%3
+            %ifnidn %5, %6
+                %if avx_enabled && sizeof%5==16
+                    v%1 %5, %6, %7
+                %else
+                    %%regmov %5, %6
+                    %1 %5, %7
+                %endif
+            %else
+                %1 %5, %7
+            %endif
+        %elif %3
+            %1 %5, %6, %7
+        %else
+            %1 %5, %6
+        %endif
+    %endif
+%endmacro
+
+;%1 == instruction
+;%2 == 1 if float, 0 if int
+;%3 == 0 if 3-operand (xmm, xmm, xmm), 1 if 4-operand (xmm, xmm, xmm, imm)
+%macro AVX_INSTR 3
+    %macro %1 2-8 fnord, fnord, fnord, %1, %2, %3
+        %ifidn %3, fnord
+            RUN_AVX_INSTR %6, %7, %8, 2, %1, %2
+        %elifidn %4, fnord
+            RUN_AVX_INSTR %6, %7, %8, 3, %1, %2, %3
+        %elifidn %5, fnord
+            RUN_AVX_INSTR %6, %7, %8, 4, %1, %2, %3, %4
+        %else
+            RUN_AVX_INSTR %6, %7, %8, 5, %1, %2, %3, %4, %5
+        %endif
+    %endmacro
+%endmacro
+
+AVX_INSTR addpd, 1, 0
+AVX_INSTR addps, 1, 0
+AVX_INSTR addsd, 1, 0
+AVX_INSTR addss, 1, 0
+AVX_INSTR addsubpd, 1, 0
+AVX_INSTR addsubps, 1, 0
+AVX_INSTR andpd, 1, 0
+AVX_INSTR andps, 1, 0
+AVX_INSTR andnpd, 1, 0
+AVX_INSTR andnps, 1, 0
+AVX_INSTR blendpd, 1, 0
+AVX_INSTR blendps, 1, 0
+AVX_INSTR blendvpd, 1, 0
+AVX_INSTR blendvps, 1, 0
+AVX_INSTR cmppd, 1, 0
+AVX_INSTR cmpps, 1, 0
+AVX_INSTR cmpsd, 1, 0
+AVX_INSTR cmpss, 1, 0
+AVX_INSTR divpd, 1, 0
+AVX_INSTR divps, 1, 0
+AVX_INSTR divsd, 1, 0
+AVX_INSTR divss, 1, 0
+AVX_INSTR dppd, 1, 0
+AVX_INSTR dpps, 1, 0
+AVX_INSTR haddpd, 1, 0
+AVX_INSTR haddps, 1, 0
+AVX_INSTR hsubpd, 1, 0
+AVX_INSTR hsubps, 1, 0
+AVX_INSTR maxpd, 1, 0
+AVX_INSTR maxps, 1, 0
+AVX_INSTR maxsd, 1, 0
+AVX_INSTR maxss, 1, 0
+AVX_INSTR minpd, 1, 0
+AVX_INSTR minps, 1, 0
+AVX_INSTR minsd, 1, 0
+AVX_INSTR minss, 1, 0
+AVX_INSTR mpsadbw, 0, 1
+AVX_INSTR mulpd, 1, 0
+AVX_INSTR mulps, 1, 0
+AVX_INSTR mulsd, 1, 0
+AVX_INSTR mulss, 1, 0
+AVX_INSTR orpd, 1, 0
+AVX_INSTR orps, 1, 0
+AVX_INSTR packsswb, 0, 0
+AVX_INSTR packssdw, 0, 0
+AVX_INSTR packuswb, 0, 0
+AVX_INSTR packusdw, 0, 0
+AVX_INSTR paddb, 0, 0
+AVX_INSTR paddw, 0, 0
+AVX_INSTR paddd, 0, 0
+AVX_INSTR paddq, 0, 0
+AVX_INSTR paddsb, 0, 0
+AVX_INSTR paddsw, 0, 0
+AVX_INSTR paddusb, 0, 0
+AVX_INSTR paddusw, 0, 0
+AVX_INSTR palignr, 0, 1
+AVX_INSTR pand, 0, 0
+AVX_INSTR pandn, 0, 0
+AVX_INSTR pavgb, 0, 0
+AVX_INSTR pavgw, 0, 0
+AVX_INSTR pblendvb, 0, 0
+AVX_INSTR pblendw, 0, 1
+AVX_INSTR pcmpestri, 0, 0
+AVX_INSTR pcmpestrm, 0, 0
+AVX_INSTR pcmpistri, 0, 0
+AVX_INSTR pcmpistrm, 0, 0
+AVX_INSTR pcmpeqb, 0, 0
+AVX_INSTR pcmpeqw, 0, 0
+AVX_INSTR pcmpeqd, 0, 0
+AVX_INSTR pcmpeqq, 0, 0
+AVX_INSTR pcmpgtb, 0, 0
+AVX_INSTR pcmpgtw, 0, 0
+AVX_INSTR pcmpgtd, 0, 0
+AVX_INSTR pcmpgtq, 0, 0
+AVX_INSTR phaddw, 0, 0
+AVX_INSTR phaddd, 0, 0
+AVX_INSTR phaddsw, 0, 0
+AVX_INSTR phsubw, 0, 0
+AVX_INSTR phsubd, 0, 0
+AVX_INSTR phsubsw, 0, 0
+AVX_INSTR pmaddwd, 0, 0
+AVX_INSTR pmaddubsw, 0, 0
+AVX_INSTR pmaxsb, 0, 0
+AVX_INSTR pmaxsw, 0, 0
+AVX_INSTR pmaxsd, 0, 0
+AVX_INSTR pmaxub, 0, 0
+AVX_INSTR pmaxuw, 0, 0
+AVX_INSTR pmaxud, 0, 0
+AVX_INSTR pminsb, 0, 0
+AVX_INSTR pminsw, 0, 0
+AVX_INSTR pminsd, 0, 0
+AVX_INSTR pminub, 0, 0
+AVX_INSTR pminuw, 0, 0
+AVX_INSTR pminud, 0, 0
+AVX_INSTR pmulhuw, 0, 0
+AVX_INSTR pmulhrsw, 0, 0
+AVX_INSTR pmulhw, 0, 0
+AVX_INSTR pmullw, 0, 0
+AVX_INSTR pmulld, 0, 0
+AVX_INSTR pmuludq, 0, 0
+AVX_INSTR pmuldq, 0, 0
+AVX_INSTR por, 0, 0
+AVX_INSTR psadbw, 0, 0
+AVX_INSTR pshufb, 0, 0
+AVX_INSTR psignb, 0, 0
+AVX_INSTR psignw, 0, 0
+AVX_INSTR psignd, 0, 0
+AVX_INSTR psllw, 0, 0
+AVX_INSTR pslld, 0, 0
+AVX_INSTR psllq, 0, 0
+AVX_INSTR pslldq, 0, 0
+AVX_INSTR psraw, 0, 0
+AVX_INSTR psrad, 0, 0
+AVX_INSTR psrlw, 0, 0
+AVX_INSTR psrld, 0, 0
+AVX_INSTR psrlq, 0, 0
+AVX_INSTR psrldq, 0, 0
+AVX_INSTR psubb, 0, 0
+AVX_INSTR psubw, 0, 0
+AVX_INSTR psubd, 0, 0
+AVX_INSTR psubq, 0, 0
+AVX_INSTR psubsb, 0, 0
+AVX_INSTR psubsw, 0, 0
+AVX_INSTR psubusb, 0, 0
+AVX_INSTR psubusw, 0, 0
+AVX_INSTR punpckhbw, 0, 0
+AVX_INSTR punpckhwd, 0, 0
+AVX_INSTR punpckhdq, 0, 0
+AVX_INSTR punpckhqdq, 0, 0
+AVX_INSTR punpcklbw, 0, 0
+AVX_INSTR punpcklwd, 0, 0
+AVX_INSTR punpckldq, 0, 0
+AVX_INSTR punpcklqdq, 0, 0
+AVX_INSTR pxor, 0, 0
+AVX_INSTR shufps, 0, 1
+AVX_INSTR subpd, 1, 0
+AVX_INSTR subps, 1, 0
+AVX_INSTR subsd, 1, 0
+AVX_INSTR subss, 1, 0
+AVX_INSTR unpckhpd, 1, 0
+AVX_INSTR unpckhps, 1, 0
+AVX_INSTR unpcklpd, 1, 0
+AVX_INSTR unpcklps, 1, 0
+AVX_INSTR xorpd, 1, 0
+AVX_INSTR xorps, 1, 0
+
+; 3DNow instructions, for sharing code between AVX, SSE and 3DN
+AVX_INSTR pfadd, 1, 0
+AVX_INSTR pfsub, 1, 0
+AVX_INSTR pfmul, 1, 0