indent

[ffmpeg] / libavcodec / dsputil.c

diff --git a/libavcodec/dsputil.c b/libavcodec/dsputil.c
index cfe3666b5c25fa1d5687a95974af35abd06a2d43..a9661eb2b5ef65a486a581d6c57811c90f66d8b3 100644 (file)
--- a/libavcodec/dsputil.c
+++ b/libavcodec/dsputil.c
@@ -29,7 +29,6 @@
 
 #include "avcodec.h"
 #include "dsputil.h"
-#include "mpegvideo.h"
 #include "simple_idct.h"
 #include "faandct.h"
 #include "faanidct.h"
@@ -152,6 +151,8 @@ static const uint8_t simple_mmx_permutation[64]={
         0x32, 0x3A, 0x36, 0x3B, 0x33, 0x3E, 0x37, 0x3F,
 };
 
+static const uint8_t idct_sse2_row_perm[8] = {0, 4, 1, 5, 2, 6, 3, 7};
+
 void ff_init_scantable(uint8_t *permutation, ScanTable *st, const uint8_t *src_scantable){
     int i;
     int end;
@@ -3697,7 +3698,7 @@ static int rd8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int
     DECLARE_ALIGNED_8 (uint64_t, aligned_bak[stride]);
     DCTELEM * const temp= (DCTELEM*)aligned_temp;
     uint8_t * const bak= (uint8_t*)aligned_bak;
-    int i, last, run, bits, level, distoration, start_i;
+    int i, last, run, bits, level, distortion, start_i;
     const int esc_length= s->ac_esc_length;
     uint8_t * length;
     uint8_t * last_length;
@@ -3764,9 +3765,9 @@ static int rd8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int
 
     s->dsp.idct_add(bak, stride, temp);
 
-    distoration= s->dsp.sse[1](NULL, bak, src1, stride, 8);
+    distortion= s->dsp.sse[1](NULL, bak, src1, stride, 8);
 
-    return distoration + ((bits*s->qscale*s->qscale*109 + 64)>>7);
+    return distortion + ((bits*s->qscale*s->qscale*109 + 64)>>7);
 }
 
 static int bit8x8_c(/*MpegEncContext*/ void *c, uint8_t *src1, uint8_t *src2, int stride, int h){
@@ -3929,20 +3930,74 @@ void ff_vector_fmul_add_add_c(float *dst, const float *src0, const float *src1,
         dst[i*step] = src0[i] * src1[i] + src2[i] + src3;
 }
 
-void ff_float_to_int16_c(int16_t *dst, const float *src, int len){
+void ff_vector_fmul_window_c(float *dst, const float *src0, const float *src1, const float *win, float add_bias, int len){
+    int i,j;
+    dst += len;
+    win += len;
+    src0+= len;
+    for(i=-len, j=len-1; i<0; i++, j--) {
+        float s0 = src0[i];
+        float s1 = src1[j];
+        float wi = win[i];
+        float wj = win[j];
+        dst[i] = s0*wj - s1*wi + add_bias;
+        dst[j] = s0*wi + s1*wj + add_bias;
+    }
+}
+
+static av_always_inline int float_to_int16_one(const float *src){
+    int_fast32_t tmp = *(const int32_t*)src;
+    if(tmp & 0xf0000){
+        tmp = (0x43c0ffff - tmp)>>31;
+        // is this faster on some gcc/cpu combinations?
+//      if(tmp > 0x43c0ffff) tmp = 0xFFFF;
+//      else                 tmp = 0;
+    }
+    return tmp - 0x8000;
+}
+
+void ff_float_to_int16_c(int16_t *dst, const float *src, long len){
     int i;
-    for(i=0; i<len; i++) {
-        int_fast32_t tmp = ((const int32_t*)src)[i];
-        if(tmp & 0xf0000){
-            tmp = (0x43c0ffff - tmp)>>31;
-            // is this faster on some gcc/cpu combinations?
-//          if(tmp > 0x43c0ffff) tmp = 0xFFFF;
-//          else                 tmp = 0;
+    for(i=0; i<len; i++)
+        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){
+    int i,j,c;
+    if(channels==2){
+        for(i=0; i<len; i++){
+            dst[2*i]   = float_to_int16_one(src[0]+i);
+            dst[2*i+1] = float_to_int16_one(src[1]+i);
         }
-        dst[i] = tmp - 0x8000;
+    }else{
+        for(c=0; c<channels; c++)
+            for(i=0, j=c; i<len; i++, j+=channels)
+                dst[j] = float_to_int16_one(src[c]+i);
     }
 }
 
+static void add_int16_c(int16_t * v1, int16_t * v2, int order)
+{
+    while (order--)
+       *v1++ += *v2++;
+}
+
+static void sub_int16_c(int16_t * v1, int16_t * v2, int order)
+{
+    while (order--)
+        *v1++ -= *v2++;
+}
+
+static int32_t scalarproduct_int16_c(int16_t * v1, int16_t * v2, int order, int shift)
+{
+    int res = 0;
+
+    while (order--)
+        res += (*v1++ * *v2++) >> shift;
+
+    return res;
+}
+
 #define W0 2048
 #define W1 2841 /* 2048*sqrt (2)*cos (1*pi/16) */
 #define W2 2676 /* 2048*sqrt (2)*cos (2*pi/16) */
@@ -4427,7 +4482,12 @@ void dsputil_init(DSPContext* c, AVCodecContext *avctx)
     c->vector_fmul = vector_fmul_c;
     c->vector_fmul_reverse = vector_fmul_reverse_c;
     c->vector_fmul_add_add = ff_vector_fmul_add_add_c;
+    c->vector_fmul_window = ff_vector_fmul_window_c;
     c->float_to_int16 = ff_float_to_int16_c;
+    c->float_to_int16_interleave = ff_float_to_int16_interleave_c;
+    c->add_int16 = add_int16_c;
+    c->sub_int16 = sub_int16_c;
+    c->scalarproduct_int16 = scalarproduct_int16_c;
 
     c->shrink[0]= ff_img_copy_plane;
     c->shrink[1]= ff_shrink22;
@@ -4477,6 +4537,10 @@ void dsputil_init(DSPContext* c, AVCodecContext *avctx)
         for(i=0; i<64; i++)
             c->idct_permutation[i]= (i&0x24) | ((i&3)<<3) | ((i>>3)&3);
         break;
+    case FF_SSE2_IDCT_PERM:
+        for(i=0; i<64; i++)
+            c->idct_permutation[i]= (i&0x38) | idct_sse2_row_perm[i&7];
+        break;
     default:
         av_log(avctx, AV_LOG_ERROR, "Internal error, IDCT permutation not set\n");
     }