From 5eb0f2a425b6e1e0f821b52972b7af75de3480e2 Mon Sep 17 00:00:00 2001
From: Loren Merritt <lorenm@u.washington.edu>
Date: Wed, 16 Jul 2008 00:50:12 +0000
Subject: [PATCH] float_to_int16_interleave: change src to an array of pointers
 instead of assuming it's contiguous. this has no immediate effect, but will
 allow it to be used in more codecs.

Originally committed as revision 14252 to svn://svn.ffmpeg.org/ffmpeg/trunk
---
 libavcodec/dsputil.c          | 10 +++++-----
 libavcodec/dsputil.h          |  2 +-
 libavcodec/i386/dsputil_mmx.c | 20 ++++++++++----------
 libavcodec/vorbis_dec.c       |  6 +++++-
 4 files changed, 21 insertions(+), 17 deletions(-)

diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c
index 212e2415da..a9661eb2b5 100644
--- a/libavcodec/dsputil.c
+++ b/libavcodec/dsputil.c
@@ -3962,17 +3962,17 @@ void ff_float_to_int16_c(int16_t *dst, const float *src, long len){
         dst[i] = float_to_int16_one(src+i);
 }
 
-void ff_float_to_int16_interleave_c(int16_t *dst, const float *src, long len, int channels){
+void ff_float_to_int16_interleave_c(int16_t *dst, const float **src, long len, int channels){
     int i,j,c;
     if(channels==2){
         for(i=0; i<len; i++){
-            dst[2*i]   = float_to_int16_one(src+i);
-            dst[2*i+1] = float_to_int16_one(src+i+len);
+            dst[2*i]   = float_to_int16_one(src[0]+i);
+            dst[2*i+1] = float_to_int16_one(src[1]+i);
         }
     }else{
-        for(c=0; c<channels; c++, src+=len)
+        for(c=0; c<channels; c++)
             for(i=0, j=c; i<len; i++, j+=channels)
-                dst[j] = float_to_int16_one(src+i);
+                dst[j] = float_to_int16_one(src[c]+i);
     }
 }
 
diff --git a/libavcodec/dsputil.h b/libavcodec/dsputil.h
index e44eaa2c48..7a47b87b44 100644
--- a/libavcodec/dsputil.h
+++ b/libavcodec/dsputil.h
@@ -372,7 +372,7 @@ typedef struct DSPContext {
     /* C version: convert floats from the range [384.0,386.0] to ints in [-32768,32767]
      * simd versions: convert floats from [-32768.0,32767.0] without rescaling and arrays are 16byte aligned */
     void (*float_to_int16)(int16_t *dst, const float *src, long len);
-    void (*float_to_int16_interleave)(int16_t *dst, const float *src, long len, int channels);
+    void (*float_to_int16_interleave)(int16_t *dst, const float **src, long len, int channels);
 
     /* (I)DCT */
     void (*fdct)(DCTELEM *block/* align 16*/);
diff --git a/libavcodec/i386/dsputil_mmx.c b/libavcodec/i386/dsputil_mmx.c
index c507023936..90865e80c7 100644
--- a/libavcodec/i386/dsputil_mmx.c
+++ b/libavcodec/i386/dsputil_mmx.c
@@ -2156,32 +2156,32 @@ static void float_to_int16_sse2(int16_t *dst, const float *src, long len){
 
 #define FLOAT_TO_INT16_INTERLEAVE(cpu, body) \
 /* gcc pessimizes register allocation if this is in the same function as float_to_int16_interleave_sse2*/\
-static av_noinline void float_to_int16_interleave2_##cpu(int16_t *dst, const float *src, long len, int channels){\
-    DECLARE_ALIGNED_16(int16_t, tmp[len*channels]);\
+static av_noinline void float_to_int16_interleave2_##cpu(int16_t *dst, const float **src, long len, int channels){\
+    DECLARE_ALIGNED_16(int16_t, tmp[len]);\
     int i,j,c;\
-    float_to_int16_##cpu(tmp, src, len*channels);\
     for(c=0; c<channels; c++){\
-        int16_t *ptmp = tmp+c*len;\
+        float_to_int16_##cpu(tmp, src[c], len);\
         for(i=0, j=c; i<len; i++, j+=channels)\
-            dst[j] = ptmp[i];\
+            dst[j] = tmp[i];\
     }\
 }\
 \
-static void float_to_int16_interleave_##cpu(int16_t *dst, const float *src, long len, int channels){\
+static void float_to_int16_interleave_##cpu(int16_t *dst, const float **src, long len, int channels){\
     if(channels==1)\
-        float_to_int16_##cpu(dst, src, len);\
+        float_to_int16_##cpu(dst, src[0], len);\
     else if(channels>2)\
         float_to_int16_interleave2_##cpu(dst, src, len, channels);\
     else{\
-        float *src1;\
+        const float *src0 = src[0];\
+        const float *src1 = src[1];\
         asm volatile(\
             "shl $2, %0 \n"\
             "add %0, %1 \n"\
             "add %0, %2 \n"\
-            "lea (%2,%0), %3 \n"\
+            "add %0, %3 \n"\
             "neg %0 \n"\
             body\
-            :"+r"(len), "+r"(dst), "+r"(src), "=r"(src1)\
+            :"+r"(len), "+r"(dst), "+r"(src0), "+r"(src1)\
         );\
     }\
 }
diff --git a/libavcodec/vorbis_dec.c b/libavcodec/vorbis_dec.c
index a338141b92..0f0c17b617 100644
--- a/libavcodec/vorbis_dec.c
+++ b/libavcodec/vorbis_dec.c
@@ -1553,6 +1553,8 @@ static int vorbis_decode_frame(AVCodecContext *avccontext,
 {
     vorbis_context *vc = avccontext->priv_data ;
     GetBitContext *gb = &(vc->gb);
+    const float *channel_ptrs[vc->audio_channels];
+    int i;
 
     int_fast16_t len;
 
@@ -1579,7 +1581,9 @@ static int vorbis_decode_frame(AVCodecContext *avccontext,
 
     AV_DEBUG("parsed %d bytes %d bits, returned %d samples (*ch*bits) \n", get_bits_count(gb)/8, get_bits_count(gb)%8, len);
 
-    vc->dsp.float_to_int16_interleave(data, vc->channel_residues, len, vc->audio_channels);
+    for(i=0; i<vc->audio_channels; i++)
+        channel_ptrs[i] = vc->channel_residues+i*len;
+    vc->dsp.float_to_int16_interleave(data, channel_ptrs, len, vc->audio_channels);
     *data_size=len*2*vc->audio_channels;
 
     return buf_size ;