Merge remote-tracking branch 'qatar/master'

* qatar/master:
  avplay: use libavresample for sample format conversion and channel mixing
  Fix compilation with YASM/NASM without AVX support.
  WMAL: do not output last frame again if nothing was decoded in current packet
  WMAL: do not start decoding if frame does not end in current packet
  adpcm-thp: fix invalid array indexing
  ppc: add const where needed in scalarproduct_int16_altivec()
  ppc: remove shift parameter from scalarproduct_int16_altivec()
  ppc: dsputil: do unaligned block accesses correctly
  dvenc: do not call dsputil functions with stride not a multiple of 16
  APIchanges: fill in some dates and commit hashes

Conflicts:
	doc/APIchanges
	ffplay.c
	libavcodec/adpcm.c

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

7 files changed, 131 insertions, 139 deletions

diff --git a/doc/APIchanges b/doc/APIchanges
index 0867fa4fe0..57187692ff 100644
--- a/doc/APIchanges
+++ b/doc/APIchanges
@@ -2,14 +2,14 @@ Never assume the API of libav* to be stable unless at least 1 month has passed
 since the last major version increase.
 
 The last version increases were:
-libavcodec:  2012-01-27
-libavdevice: 2011-04-18
-libavfilter: 2011-04-18
-libavformat: 2012-01-27
-libavresample: 2012-xx-xx
-libpostproc: 2011-04-18
-libswscale:  2011-06-20
-libavutil:   2011-04-18
+libavcodec:    2012-01-27
+libavdevice:   2011-04-18
+libavfilter:   2011-04-18
+libavformat:   2012-01-27
+libavresample: 2012-04-24
+libpostproc:   2011-04-18
+libswscale:    2011-06-20
+libavutil:     2011-04-18
 
 
 API changes, most recent first:
@@ -23,13 +23,13 @@ API changes, most recent first:
 2012-03-26 - a67d9cf - lavfi 2.66.100
   Add avfilter_fill_frame_from_{audio_,}buffer_ref() functions.
 
-2012-04-25 - xxxxxxx - lavu 51.29.0 - cpu.h
+2012-04-25 - 3527a73 - lavu 51.29.0 - cpu.h
   Add av_parse_cpu_flags()
 
-2012-xx-xx - xxxxxxx - lavr 0.0.0
+2012-04-24 - c8af852 - lavr 0.0.0
   Add libavresample audio conversion library
 
-2012-xx-xx - xxxxxxx - lavu 51.28.0 - audio_fifo.h
+2012-04-20 - 0c0d1bc - lavu 51.28.0 - audio_fifo.h
   Add audio FIFO functions:
     av_audio_fifo_free()
     av_audio_fifo_alloc()
@@ -41,10 +41,10 @@ API changes, most recent first:
     av_audio_fifo_size()
     av_audio_fifo_space()
 
-2012-xx-xx - xxxxxxx - lavfi 2.16.0 - avfiltergraph.h
-  Add avfilter_graph_parse2()
+2012-04-14 - lavfi 2.16.0 - avfiltergraph.h
+  d7bcc71 Add avfilter_graph_parse2().
 
-2012-xx-xx - xxxxxxx - lavu 51.27.0 - samplefmt.h
+2012-04-08 - 4d693b0 - lavu 51.27.0 - samplefmt.h
   Add av_get_packed_sample_fmt() and av_get_planar_sample_fmt()
 
 2012-03-21 - b75c67d - lavu 51.43.100
diff --git a/ffplay.c b/ffplay.c
index 4489b34cca..4cc1a82d98 100644
--- a/ffplay.c
+++ b/ffplay.c
@@ -40,7 +40,6 @@
 #include "libavformat/avformat.h"
 #include "libavdevice/avdevice.h"
 #include "libswscale/swscale.h"
-#include "libavcodec/audioconvert.h"
 #include "libavutil/opt.h"
 #include "libavcodec/avfft.h"
 #include "libswresample/swresample.h"
@@ -2130,7 +2129,8 @@ static int audio_decode_frame(VideoState *is, double *pts_ptr)
             /* if no pts, then compute it */
             pts = is->audio_clock;
             *pts_ptr = pts;
-            is->audio_clock += (double)data_size / (dec->channels * dec->sample_rate * av_get_bytes_per_sample(dec->sample_fmt));
+            is->audio_clock += (double)data_size /
+                (dec->channels * dec->sample_rate * av_get_bytes_per_sample(dec->sample_fmt));
 #ifdef DEBUG
             {
                 static double last_clock;
@@ -2373,9 +2373,9 @@ static void stream_component_close(VideoState *is, int stream_index)
         SDL_CloseAudio();
 
         packet_queue_end(&is->audioq);
+        av_free_packet(&is->audio_pkt);
         if (is->swr_ctx)
             swr_free(&is->swr_ctx);
-        av_free_packet(&is->audio_pkt);
         av_freep(&is->audio_buf1);
         is->audio_buf = NULL;
         av_freep(&is->frame);
diff --git a/libavcodec/adpcm.c b/libavcodec/adpcm.c
index c1f6ee0a5b..b86168acb8 100644
--- a/libavcodec/adpcm.c
+++ b/libavcodec/adpcm.c
@@ -1212,12 +1212,14 @@ static int adpcm_decode_frame(AVCodecContext *avctx, void *data,
         int prev[2][2];
         int ch;
 
-        for (i = 0; i < 32; i++)
-            table[0][i] = sign_extend(bytestream2_get_be16u(&gb), 16);
+        for (i = 0; i < 2; i++)
+            for (n = 0; n < 16; n++)
+                table[i][n] = sign_extend(bytestream2_get_be16u(&gb), 16);
 
         /* Initialize the previous sample.  */
-        for (i = 0; i < 4; i++)
-            prev[i>>1][i&1] = sign_extend(bytestream2_get_be16u(&gb), 16);
+        for (i = 0; i < 2; i++)
+            for (n = 0; n < 2; n++)
+                prev[i][n] = sign_extend(bytestream2_get_be16u(&gb), 16);
 
         for (ch = 0; ch <= st; ch++) {
             samples = (short *)c->frame.data[0] + ch;
diff --git a/libavcodec/dv.c b/libavcodec/dv.c
index 5bb84793b5..cd2cd4ba8f 100644
--- a/libavcodec/dv.c
+++ b/libavcodec/dv.c
@@ -668,7 +668,7 @@ static int dv_encode_video_segment(AVCodecContext *avctx, void *arg)
     int mb_x, mb_y, c_offset, linesize, y_stride;
     uint8_t*  y_ptr;
     uint8_t*  dif;
-    LOCAL_ALIGNED_8(uint8_t, scratch, [64]);
+    LOCAL_ALIGNED_8(uint8_t, scratch, [128]);
     EncBlockInfo  enc_blks[5*DV_MAX_BPM];
     PutBitContext pbs[5*DV_MAX_BPM];
     PutBitContext* pb;
@@ -723,10 +723,10 @@ static int dv_encode_video_segment(AVCodecContext *avctx, void *arg)
                     b[0] = c_ptr[0]; b[1] = c_ptr[1]; b[2] = c_ptr[2]; b[3] = c_ptr[3];
                     b[4] =     d[0]; b[5] =     d[1]; b[6] =     d[2]; b[7] =     d[3];
                     c_ptr += linesize;
-                    b += 8;
+                    b += 16;
                 }
                 c_ptr = scratch;
-                linesize = 8;
+                linesize = 16;
             }
 
             vs_bit_size += dv_init_enc_block(    enc_blk++, c_ptr           , linesize, s, 1);
diff --git a/libavcodec/ppc/dsputil_altivec.c b/libavcodec/ppc/dsputil_altivec.c
index c4c9dbaf9f..de8b85891b 100644
--- a/libavcodec/ppc/dsputil_altivec.c
+++ b/libavcodec/ppc/dsputil_altivec.c
@@ -34,7 +34,9 @@ static int sad16_x2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size
     int i;
     int s;
     const vector unsigned char zero = (const vector unsigned char)vec_splat_u8(0);
-    vector unsigned char *tv;
+    vector unsigned char perm1 = vec_lvsl(0, pix2);
+    vector unsigned char perm2 = vec_add(perm1, vec_splat_u8(1));
+    vector unsigned char pix2l, pix2r;
     vector unsigned char pix1v, pix2v, pix2iv, avgv, t5;
     vector unsigned int sad;
     vector signed int sumdiffs;
@@ -45,14 +47,11 @@ static int sad16_x2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size
         /* Read unaligned pixels into our vectors. The vectors are as follows:
            pix1v: pix1[0]-pix1[15]
            pix2v: pix2[0]-pix2[15]      pix2iv: pix2[1]-pix2[16] */
-        tv = (vector unsigned char *) pix1;
-        pix1v = vec_perm(tv[0], tv[1], vec_lvsl(0, pix1));
-
-        tv = (vector unsigned char *) &pix2[0];
-        pix2v = vec_perm(tv[0], tv[1], vec_lvsl(0, &pix2[0]));
-
-        tv = (vector unsigned char *) &pix2[1];
-        pix2iv = vec_perm(tv[0], tv[1], vec_lvsl(0, &pix2[1]));
+        pix1v  = vec_ld( 0, pix1);
+        pix2l  = vec_ld( 0, pix2);
+        pix2r  = vec_ld(16, pix2);
+        pix2v  = vec_perm(pix2l, pix2r, perm1);
+        pix2iv = vec_perm(pix2l, pix2r, perm2);
 
         /* Calculate the average vector */
         avgv = vec_avg(pix2v, pix2iv);
@@ -79,7 +78,8 @@ static int sad16_y2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size
     int i;
     int s;
     const vector unsigned char zero = (const vector unsigned char)vec_splat_u8(0);
-    vector unsigned char *tv;
+    vector unsigned char perm = vec_lvsl(0, pix2);
+    vector unsigned char pix2l, pix2r;
     vector unsigned char pix1v, pix2v, pix3v, avgv, t5;
     vector unsigned int sad;
     vector signed int sumdiffs;
@@ -95,18 +95,19 @@ static int sad16_y2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size
        Read unaligned pixels into our vectors. The vectors are as follows:
        pix2v: pix2[0]-pix2[15]
        Split the pixel vectors into shorts */
-    tv = (vector unsigned char *) &pix2[0];
-    pix2v = vec_perm(tv[0], tv[1], vec_lvsl(0, &pix2[0]));
+    pix2l = vec_ld( 0, pix2);
+    pix2r = vec_ld(15, pix2);
+    pix2v = vec_perm(pix2l, pix2r, perm);
 
     for (i = 0; i < h; i++) {
         /* Read unaligned pixels into our vectors. The vectors are as follows:
            pix1v: pix1[0]-pix1[15]
            pix3v: pix3[0]-pix3[15] */
-        tv = (vector unsigned char *) pix1;
-        pix1v = vec_perm(tv[0], tv[1], vec_lvsl(0, pix1));
+        pix1v = vec_ld(0, pix1);
 
-        tv = (vector unsigned char *) &pix3[0];
-        pix3v = vec_perm(tv[0], tv[1], vec_lvsl(0, &pix3[0]));
+        pix2l = vec_ld( 0, pix3);
+        pix2r = vec_ld(15, pix3);
+        pix3v = vec_perm(pix2l, pix2r, perm);
 
         /* Calculate the average vector */
         avgv = vec_avg(pix2v, pix3v);
@@ -137,7 +138,10 @@ static int sad16_xy2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_siz
     uint8_t *pix3 = pix2 + line_size;
     const vector unsigned char zero = (const vector unsigned char)vec_splat_u8(0);
     const vector unsigned short two = (const vector unsigned short)vec_splat_u16(2);
-    vector unsigned char *tv, avgv, t5;
+    vector unsigned char avgv, t5;
+    vector unsigned char perm1 = vec_lvsl(0, pix2);
+    vector unsigned char perm2 = vec_add(perm1, vec_splat_u8(1));
+    vector unsigned char pix2l, pix2r;
     vector unsigned char pix1v, pix2v, pix3v, pix2iv, pix3iv;
     vector unsigned short pix2lv, pix2hv, pix2ilv, pix2ihv;
     vector unsigned short pix3lv, pix3hv, pix3ilv, pix3ihv;
@@ -157,11 +161,10 @@ static int sad16_xy2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_siz
        Read unaligned pixels into our vectors. The vectors are as follows:
        pix2v: pix2[0]-pix2[15]  pix2iv: pix2[1]-pix2[16]
        Split the pixel vectors into shorts */
-    tv = (vector unsigned char *) &pix2[0];
-    pix2v = vec_perm(tv[0], tv[1], vec_lvsl(0, &pix2[0]));
-
-    tv = (vector unsigned char *) &pix2[1];
-    pix2iv = vec_perm(tv[0], tv[1], vec_lvsl(0, &pix2[1]));
+    pix2l  = vec_ld( 0, pix2);
+    pix2r  = vec_ld(16, pix2);
+    pix2v  = vec_perm(pix2l, pix2r, perm1);
+    pix2iv = vec_perm(pix2l, pix2r, perm2);
 
     pix2hv  = (vector unsigned short) vec_mergeh(zero, pix2v);
     pix2lv  = (vector unsigned short) vec_mergel(zero, pix2v);
@@ -174,14 +177,12 @@ static int sad16_xy2_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_siz
         /* Read unaligned pixels into our vectors. The vectors are as follows:
            pix1v: pix1[0]-pix1[15]
            pix3v: pix3[0]-pix3[15]      pix3iv: pix3[1]-pix3[16] */
-        tv = (vector unsigned char *) pix1;
-        pix1v = vec_perm(tv[0], tv[1], vec_lvsl(0, pix1));
-
-        tv = (vector unsigned char *) &pix3[0];
-        pix3v = vec_perm(tv[0], tv[1], vec_lvsl(0, &pix3[0]));
+        pix1v = vec_ld(0, pix1);
 
-        tv = (vector unsigned char *) &pix3[1];
-        pix3iv = vec_perm(tv[0], tv[1], vec_lvsl(0, &pix3[1]));
+        pix2l  = vec_ld( 0, pix3);
+        pix2r  = vec_ld(16, pix3);
+        pix3v  = vec_perm(pix2l, pix2r, perm1);
+        pix3iv = vec_perm(pix2l, pix2r, perm2);
 
         /* Note that AltiVec does have vec_avg, but this works on vector pairs
            and rounds up. We could do avg(avg(a,b),avg(c,d)), but the rounding
@@ -230,7 +231,7 @@ static int sad16_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, i
     int i;
     int s;
     const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
-    vector unsigned char perm1, perm2, pix1v_low, pix1v_high, pix2v_low, pix2v_high;
+    vector unsigned char perm = vec_lvsl(0, pix2);
     vector unsigned char t1, t2, t3,t4, t5;
     vector unsigned int sad;
     vector signed int sumdiffs;
@@ -240,14 +241,10 @@ static int sad16_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, i
 
     for (i = 0; i < h; i++) {
         /* Read potentially unaligned pixels into t1 and t2 */
-        perm1 = vec_lvsl(0, pix1);
-        pix1v_high = vec_ld( 0, pix1);
-        pix1v_low  = vec_ld(15, pix1);
-        perm2 = vec_lvsl(0, pix2);
-        pix2v_high = vec_ld( 0, pix2);
-        pix2v_low  = vec_ld(15, pix2);
-        t1 = vec_perm(pix1v_high, pix1v_low, perm1);
-        t2 = vec_perm(pix2v_high, pix2v_low, perm2);
+        vector unsigned char pix2l = vec_ld( 0, pix2);
+        vector unsigned char pix2r = vec_ld(15, pix2);
+        t1 = vec_ld(0, pix1);
+        t2 = vec_perm(pix2l, pix2r, perm);
 
         /* Calculate a sum of abs differences vector */
         t3 = vec_max(t1, t2);
@@ -274,25 +271,25 @@ static int sad8_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, in
     int i;
     int s;
     const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
-    vector unsigned char perm1, perm2, permclear, *pix1v, *pix2v;
+    const vector unsigned char permclear = (vector unsigned char){255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0};
+    vector unsigned char perm1 = vec_lvsl(0, pix1);
+    vector unsigned char perm2 = vec_lvsl(0, pix2);
     vector unsigned char t1, t2, t3,t4, t5;
     vector unsigned int sad;
     vector signed int sumdiffs;
 
     sad = (vector unsigned int)vec_splat_u32(0);
 
-    permclear = (vector unsigned char){255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0};
-
     for (i = 0; i < h; i++) {
         /* Read potentially unaligned pixels into t1 and t2
            Since we're reading 16 pixels, and actually only want 8,
            mask out the last 8 pixels. The 0s don't change the sum. */
-        perm1 = vec_lvsl(0, pix1);
-        pix1v = (vector unsigned char *) pix1;
-        perm2 = vec_lvsl(0, pix2);
-        pix2v = (vector unsigned char *) pix2;
-        t1 = vec_and(vec_perm(pix1v[0], pix1v[1], perm1), permclear);
-        t2 = vec_and(vec_perm(pix2v[0], pix2v[1], perm2), permclear);
+        vector unsigned char pix1l = vec_ld( 0, pix1);
+        vector unsigned char pix1r = vec_ld(15, pix1);
+        vector unsigned char pix2l = vec_ld( 0, pix2);
+        vector unsigned char pix2r = vec_ld(15, pix2);
+        t1 = vec_and(vec_perm(pix1l, pix1r, perm1), permclear);
+        t2 = vec_and(vec_perm(pix2l, pix2r, perm2), permclear);
 
         /* Calculate a sum of abs differences vector */
         t3 = vec_max(t1, t2);
@@ -319,7 +316,7 @@ static int pix_norm1_altivec(uint8_t *pix, int line_size)
     int i;
     int s;
     const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
-    vector unsigned char *tv;
+    vector unsigned char perm = vec_lvsl(0, pix);
     vector unsigned char pixv;
     vector unsigned int sv;
     vector signed int sum;
@@ -329,8 +326,9 @@ static int pix_norm1_altivec(uint8_t *pix, int line_size)
     s = 0;
     for (i = 0; i < 16; i++) {
         /* Read in the potentially unaligned pixels */
-        tv = (vector unsigned char *) pix;
-        pixv = vec_perm(tv[0], tv[1], vec_lvsl(0, pix));
+        vector unsigned char pixl = vec_ld( 0, pix);
+        vector unsigned char pixr = vec_ld(15, pix);
+        pixv = vec_perm(pixl, pixr, perm);
 
         /* Square the values, and add them to our sum */
         sv = vec_msum(pixv, pixv, sv);
@@ -355,26 +353,25 @@ static int sse8_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, in
     int i;
     int s;
     const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
-    vector unsigned char perm1, perm2, permclear, *pix1v, *pix2v;
+    const vector unsigned char permclear = (vector unsigned char){255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0};
+    vector unsigned char perm1 = vec_lvsl(0, pix1);
+    vector unsigned char perm2 = vec_lvsl(0, pix2);
     vector unsigned char t1, t2, t3,t4, t5;
     vector unsigned int sum;
     vector signed int sumsqr;
 
     sum = (vector unsigned int)vec_splat_u32(0);
 
-    permclear = (vector unsigned char){255,255,255,255,255,255,255,255,0,0,0,0,0,0,0,0};
-
-
     for (i = 0; i < h; i++) {
         /* Read potentially unaligned pixels into t1 and t2
            Since we're reading 16 pixels, and actually only want 8,
            mask out the last 8 pixels. The 0s don't change the sum. */
-        perm1 = vec_lvsl(0, pix1);
-        pix1v = (vector unsigned char *) pix1;
-        perm2 = vec_lvsl(0, pix2);
-        pix2v = (vector unsigned char *) pix2;
-        t1 = vec_and(vec_perm(pix1v[0], pix1v[1], perm1), permclear);
-        t2 = vec_and(vec_perm(pix2v[0], pix2v[1], perm2), permclear);
+        vector unsigned char pix1l = vec_ld( 0, pix1);
+        vector unsigned char pix1r = vec_ld(15, pix1);
+        vector unsigned char pix2l = vec_ld( 0, pix2);
+        vector unsigned char pix2r = vec_ld(15, pix2);
+        t1 = vec_and(vec_perm(pix1l, pix1r, perm1), permclear);
+        t2 = vec_and(vec_perm(pix2l, pix2r, perm2), permclear);
 
         /* Since we want to use unsigned chars, we can take advantage
            of the fact that abs(a-b)^2 = (a-b)^2. */
@@ -409,7 +406,7 @@ static int sse16_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, i
     int i;
     int s;
     const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
-    vector unsigned char perm1, perm2, *pix1v, *pix2v;
+    vector unsigned char perm = vec_lvsl(0, pix2);
     vector unsigned char t1, t2, t3,t4, t5;
     vector unsigned int sum;
     vector signed int sumsqr;
@@ -418,12 +415,10 @@ static int sse16_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, i
 
     for (i = 0; i < h; i++) {
         /* Read potentially unaligned pixels into t1 and t2 */
-        perm1 = vec_lvsl(0, pix1);
-        pix1v = (vector unsigned char *) pix1;
-        perm2 = vec_lvsl(0, pix2);
-        pix2v = (vector unsigned char *) pix2;
-        t1 = vec_perm(pix1v[0], pix1v[1], perm1);
-        t2 = vec_perm(pix2v[0], pix2v[1], perm2);
+        vector unsigned char pix2l = vec_ld( 0, pix2);
+        vector unsigned char pix2r = vec_ld(15, pix2);
+        t1 = vec_ld(0, pix1);
+        t2 = vec_perm(pix2l, pix2r, perm);
 
         /* Since we want to use unsigned chars, we can take advantage
            of the fact that abs(a-b)^2 = (a-b)^2. */
@@ -451,7 +446,7 @@ static int sse16_altivec(void *v, uint8_t *pix1, uint8_t *pix2, int line_size, i
 static int pix_sum_altivec(uint8_t * pix, int line_size)
 {
     const vector unsigned int zero = (const vector unsigned int)vec_splat_u32(0);
-    vector unsigned char perm, *pixv;
+    vector unsigned char perm = vec_lvsl(0, pix);
     vector unsigned char t1;
     vector unsigned int sad;
     vector signed int sumdiffs;
@@ -463,9 +458,9 @@ static int pix_sum_altivec(uint8_t * pix, int line_size)
 
     for (i = 0; i < 16; i++) {
         /* Read the potentially unaligned 16 pixels into t1 */
-        perm = vec_lvsl(0, pix);
-        pixv = (vector unsigned char *) pix;
-        t1 = vec_perm(pixv[0], pixv[1], perm);
+        vector unsigned char pixl = vec_ld( 0, pix);
+        vector unsigned char pixr = vec_ld(15, pix);
+        t1 = vec_perm(pixl, pixr, perm);
 
         /* Add each 4 pixel group together and put 4 results into sad */
         sad = vec_sum4s(t1, sad);
@@ -484,7 +479,8 @@ static int pix_sum_altivec(uint8_t * pix, int line_size)
 static void get_pixels_altivec(DCTELEM *restrict block, const uint8_t *pixels, int line_size)
 {
     int i;
-    vector unsigned char perm, bytes, *pixv;
+    vector unsigned char perm = vec_lvsl(0, pixels);
+    vector unsigned char bytes;
     const vector unsigned char zero = (const vector unsigned char)vec_splat_u8(0);
     vector signed short shorts;
 
@@ -492,9 +488,9 @@ static void get_pixels_altivec(DCTELEM *restrict block, const uint8_t *pixels, i
         // Read potentially unaligned pixels.
         // We're reading 16 pixels, and actually only want 8,
         // but we simply ignore the extras.
-        perm = vec_lvsl(0, pixels);
-        pixv = (vector unsigned char *) pixels;
-        bytes = vec_perm(pixv[0], pixv[1], perm);
+        vector unsigned char pixl = vec_ld( 0, pixels);
+        vector unsigned char pixr = vec_ld(15, pixels);
+        bytes = vec_perm(pixl, pixr, perm);
 
         // convert the bytes into shorts
         shorts = (vector signed short)vec_mergeh(zero, bytes);
@@ -510,7 +506,9 @@ static void diff_pixels_altivec(DCTELEM *restrict block, const uint8_t *s1,
         const uint8_t *s2, int stride)
 {
     int i;
-    vector unsigned char perm, bytes, *pixv;
+    vector unsigned char perm1 = vec_lvsl(0, s1);
+    vector unsigned char perm2 = vec_lvsl(0, s2);
+    vector unsigned char bytes, pixl, pixr;
     const vector unsigned char zero = (const vector unsigned char)vec_splat_u8(0);
     vector signed short shorts1, shorts2;
 
@@ -518,17 +516,17 @@ static void diff_pixels_altivec(DCTELEM *restrict block, const uint8_t *s1,
         // Read potentially unaligned pixels
         // We're reading 16 pixels, and actually only want 8,
         // but we simply ignore the extras.
-        perm = vec_lvsl(0, s1);
-        pixv = (vector unsigned char *) s1;
-        bytes = vec_perm(pixv[0], pixv[1], perm);
+        pixl = vec_ld( 0, s1);
+        pixr = vec_ld(15, s1);
+        bytes = vec_perm(pixl, pixr, perm1);
 
         // convert the bytes into shorts
         shorts1 = (vector signed short)vec_mergeh(zero, bytes);
 
         // Do the same for the second block of pixels
-        perm = vec_lvsl(0, s2);
-        pixv = (vector unsigned char *) s2;
-        bytes = vec_perm(pixv[0], pixv[1], perm);
+        pixl = vec_ld( 0, s2);
+        pixr = vec_ld(15, s2);
+        bytes = vec_perm(pixl, pixr, perm2);
 
         // convert the bytes into shorts
         shorts2 = (vector signed short)vec_mergeh(zero, bytes);
@@ -550,17 +548,17 @@ static void diff_pixels_altivec(DCTELEM *restrict block, const uint8_t *s1,
         // Read potentially unaligned pixels
         // We're reading 16 pixels, and actually only want 8,
         // but we simply ignore the extras.
-        perm = vec_lvsl(0, s1);
-        pixv = (vector unsigned char *) s1;
-        bytes = vec_perm(pixv[0], pixv[1], perm);
+        pixl = vec_ld( 0, s1);
+        pixr = vec_ld(15, s1);
+        bytes = vec_perm(pixl, pixr, perm1);
 
         // convert the bytes into shorts
         shorts1 = (vector signed short)vec_mergeh(zero, bytes);
 
         // Do the same for the second block of pixels
-        perm = vec_lvsl(0, s2);
-        pixv = (vector unsigned char *) s2;
-        bytes = vec_perm(pixv[0], pixv[1], perm);
+        pixl = vec_ld( 0, s2);
+        pixr = vec_ld(15, s2);
+        bytes = vec_perm(pixl, pixr, perm2);
 
         // convert the bytes into shorts
         shorts2 = (vector signed short)vec_mergeh(zero, bytes);
diff --git a/libavcodec/ppc/int_altivec.c b/libavcodec/ppc/int_altivec.c
index 30ae14fa5b..57e63da9f0 100644
--- a/libavcodec/ppc/int_altivec.c
+++ b/libavcodec/ppc/int_altivec.c
@@ -79,28 +79,20 @@ static int ssd_int8_vs_int16_altivec(const int8_t *pix1, const int16_t *pix2,
     return u.score[3];
 }
 
-static int32_t scalarproduct_int16_altivec(int16_t *v1, const int16_t *v2,
-                                           int order, const int shift)
+static int32_t scalarproduct_int16_altivec(const int16_t *v1, const int16_t *v2,
+                                           int order)
 {
     int i;
     LOAD_ZERO;
-    register vec_s16 vec1, *pv;
+    const vec_s16 *pv;
+    register vec_s16 vec1;
     register vec_s32 res = vec_splat_s32(0), t;
-    register vec_u32 shifts;
     int32_t ires;
 
-    shifts = zero_u32v;
-    if(shift & 0x10) shifts = vec_add(shifts, vec_sl(vec_splat_u32(0x08), vec_splat_u32(0x1)));
-    if(shift & 0x08) shifts = vec_add(shifts, vec_splat_u32(0x08));
-    if(shift & 0x04) shifts = vec_add(shifts, vec_splat_u32(0x04));
-    if(shift & 0x02) shifts = vec_add(shifts, vec_splat_u32(0x02));
-    if(shift & 0x01) shifts = vec_add(shifts, vec_splat_u32(0x01));
-
     for(i = 0; i < order; i += 8){
-        pv = (vec_s16*)v1;
+        pv = (const vec_s16*)v1;
         vec1 = vec_perm(pv[0], pv[1], vec_lvsl(0, v1));
         t = vec_msum(vec1, vec_ld(0, v2), zero_s32v);
-        t = vec_sr(t, shifts);
         res = vec_sums(t, res);
         v1 += 8;
         v2 += 8;
@@ -114,31 +106,31 @@ static int32_t scalarproduct_and_madd_int16_altivec(int16_t *v1, const int16_t *
 {
     LOAD_ZERO;
     vec_s16 *pv1 = (vec_s16*)v1;
-    vec_s16 *pv2 = (vec_s16*)v2;
-    vec_s16 *pv3 = (vec_s16*)v3;
     register vec_s16 muls = {mul,mul,mul,mul,mul,mul,mul,mul};
-    register vec_s16 t0, t1, i0, i1;
-    register vec_s16 i2 = pv2[0], i3 = pv3[0];
+    register vec_s16 t0, t1, i0, i1, i4;
+    register vec_s16 i2 = vec_ld(0, v2), i3 = vec_ld(0, v3);
     register vec_s32 res = zero_s32v;
     register vec_u8 align = vec_lvsl(0, v2);
     int32_t ires;
     order >>= 4;
     do {
-        t0 = vec_perm(i2, pv2[1], align);
-        i2 = pv2[2];
-        t1 = vec_perm(pv2[1], i2, align);
+        i1 = vec_ld(16, v2);
+        t0 = vec_perm(i2, i1, align);
+        i2 = vec_ld(32, v2);
+        t1 = vec_perm(i1, i2, align);
         i0 = pv1[0];
         i1 = pv1[1];
         res = vec_msum(t0, i0, res);
         res = vec_msum(t1, i1, res);
-        t0 = vec_perm(i3, pv3[1], align);
-        i3 = pv3[2];
-        t1 = vec_perm(pv3[1], i3, align);
+        i4 = vec_ld(16, v3);
+        t0 = vec_perm(i3, i4, align);
+        i3 = vec_ld(32, v3);
+        t1 = vec_perm(i4, i3, align);
         pv1[0] = vec_mladd(t0, muls, i0);
         pv1[1] = vec_mladd(t1, muls, i1);
         pv1 += 2;
-        pv2 += 2;
-        pv3 += 2;
+        v2  += 8;
+        v3  += 8;
     } while(--order);
     res = vec_splat(vec_sums(res, zero_s32v), 3);
     vec_ste(res, 0, &ires);
diff --git a/libavcodec/wmalosslessdec.c b/libavcodec/wmalosslessdec.c
index de5dca333c..f4166d338e 100644
--- a/libavcodec/wmalosslessdec.c
+++ b/libavcodec/wmalosslessdec.c
@@ -1216,8 +1216,8 @@ static int decode_packet(AVCodecContext *avctx, void *data, int *got_frame_ptr,
             save_bits(s, gb, num_bits_prev_frame, 1);
 
             /* decode the cross packet frame if it is valid */
-            if (!s->packet_loss)
-                decode_frame(s);
+            if (num_bits_prev_frame < remaining_packet_bits && !s->packet_loss)
+                    decode_frame(s);
         } else if (s->num_saved_bits - s->frame_offset) {
             av_dlog(avctx, "ignoring %x previously saved bits\n",
                     s->num_saved_bits - s->frame_offset);