Remove unneeded add bias from 3 functions.

DSPContext.vector_fmul_window()
DCADSPContext.lfe_fir()
SynthFilterContext.synth_filter_float()
Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit 80ba1ddb)

libavcodec/aacdec.c

@@ -1721,19 +1721,19 @@ static void imdct_and_windowing(AACContext *ac, SingleChannelElement *sce)
      */
     if ((ics->window_sequence[1] == ONLY_LONG_SEQUENCE || ics->window_sequence[1] == LONG_STOP_SEQUENCE) &&
             (ics->window_sequence[0] == ONLY_LONG_SEQUENCE || ics->window_sequence[0] == LONG_START_SEQUENCE)) {
-        ac->dsp.vector_fmul_window(    out,               saved,            buf,         lwindow_prev, 0, 512);
+        ac->dsp.vector_fmul_window(    out,               saved,            buf,         lwindow_prev, 512);
     } else {
         memcpy(                        out,               saved,            448 * sizeof(float));
 
         if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
-            ac->dsp.vector_fmul_window(out + 448 + 0*128, saved + 448,      buf + 0*128, swindow_prev, 0, 64);
-            ac->dsp.vector_fmul_window(out + 448 + 1*128, buf + 0*128 + 64, buf + 1*128, swindow,      0, 64);
-            ac->dsp.vector_fmul_window(out + 448 + 2*128, buf + 1*128 + 64, buf + 2*128, swindow,      0, 64);
-            ac->dsp.vector_fmul_window(out + 448 + 3*128, buf + 2*128 + 64, buf + 3*128, swindow,      0, 64);
-            ac->dsp.vector_fmul_window(temp,              buf + 3*128 + 64, buf + 4*128, swindow,      0, 64);
+            ac->dsp.vector_fmul_window(out + 448 + 0*128, saved + 448,      buf + 0*128, swindow_prev, 64);
+            ac->dsp.vector_fmul_window(out + 448 + 1*128, buf + 0*128 + 64, buf + 1*128, swindow,      64);
+            ac->dsp.vector_fmul_window(out + 448 + 2*128, buf + 1*128 + 64, buf + 2*128, swindow,      64);
+            ac->dsp.vector_fmul_window(out + 448 + 3*128, buf + 2*128 + 64, buf + 3*128, swindow,      64);
+            ac->dsp.vector_fmul_window(temp,              buf + 3*128 + 64, buf + 4*128, swindow,      64);
             memcpy(                    out + 448 + 4*128, temp, 64 * sizeof(float));
         } else {
-            ac->dsp.vector_fmul_window(out + 448,         saved + 448,      buf,         swindow_prev, 0, 64);
+            ac->dsp.vector_fmul_window(out + 448,         saved + 448,      buf,         swindow_prev, 64);
             memcpy(                    out + 576,         buf + 64,         448 * sizeof(float));
         }
     }
@@ -1741,9 +1741,9 @@ static void imdct_and_windowing(AACContext *ac, SingleChannelElement *sce)
     // buffer update
     if (ics->window_sequence[0] == EIGHT_SHORT_SEQUENCE) {
         memcpy(                    saved,       temp + 64,         64 * sizeof(float));
-        ac->dsp.vector_fmul_window(saved + 64,  buf + 4*128 + 64, buf + 5*128, swindow, 0, 64);
-        ac->dsp.vector_fmul_window(saved + 192, buf + 5*128 + 64, buf + 6*128, swindow, 0, 64);
-        ac->dsp.vector_fmul_window(saved + 320, buf + 6*128 + 64, buf + 7*128, swindow, 0, 64);
+        ac->dsp.vector_fmul_window(saved + 64,  buf + 4*128 + 64, buf + 5*128, swindow, 64);
+        ac->dsp.vector_fmul_window(saved + 192, buf + 5*128 + 64, buf + 6*128, swindow, 64);
+        ac->dsp.vector_fmul_window(saved + 320, buf + 6*128 + 64, buf + 7*128, swindow, 64);
         memcpy(                    saved + 448, buf + 7*128 + 64,  64 * sizeof(float));
     } else if (ics->window_sequence[0] == LONG_START_SEQUENCE) {
         memcpy(                    saved,       buf + 512,        448 * sizeof(float));

libavcodec/ac3dec.c

@@ -628,13 +628,13 @@ static inline void do_imdct(AC3DecodeContext *s, int channels)
             for(i=0; i<128; i++)
                 x[i] = s->transform_coeffs[ch][2*i];
             ff_imdct_half(&s->imdct_256, s->tmp_output, x);
-            s->dsp.vector_fmul_window(s->output[ch-1], s->delay[ch-1], s->tmp_output, s->window, 0, 128);
+            s->dsp.vector_fmul_window(s->output[ch-1], s->delay[ch-1], s->tmp_output, s->window, 128);
             for(i=0; i<128; i++)
                 x[i] = s->transform_coeffs[ch][2*i+1];
             ff_imdct_half(&s->imdct_256, s->delay[ch-1], x);
         } else {
             ff_imdct_half(&s->imdct_512, s->tmp_output, s->transform_coeffs[ch]);
-            s->dsp.vector_fmul_window(s->output[ch-1], s->delay[ch-1], s->tmp_output, s->window, 0, 128);
+            s->dsp.vector_fmul_window(s->output[ch-1], s->delay[ch-1], s->tmp_output, s->window, 128);
             memcpy(s->delay[ch-1], s->tmp_output+128, 128*sizeof(float));
         }
     }

libavcodec/arm/dcadsp_init_arm.c

@@ -23,7 +23,7 @@
 #include "libavcodec/dcadsp.h"
 
 void ff_dca_lfe_fir_neon(float *out, const float *in, const float *coefs,
-                         int decifactor, float scale, float bias);
+                         int decifactor, float scale);
 
 void av_cold ff_dcadsp_init_arm(DCADSPContext *s)
 {

libavcodec/arm/dcadsp_neon.S

@@ -29,7 +29,7 @@ function ff_dca_lfe_fir_neon, export=1
         cmp             r3,  #32
         moveq           r6,  #256/32
         movne           r6,  #256/64
-NOVFP   vldr            d0,  [sp, #16]          @ scale, bias
+NOVFP   vldr            s0,  [sp, #16]          @ scale
         mov             lr,  #-16
 1:
         vmov.f32        q2,  #0.0               @ v0
@@ -51,8 +51,7 @@ NOVFP   vldr            d0,  [sp, #16]          @ scale, bias
         vadd.f32        d4,  d4,  d5
         vadd.f32        d6,  d6,  d7
         vpadd.f32       d4,  d4,  d6
-        vdup.32         d5,  d0[1]
-        vmla.f32        d5,  d4,  d0[0]
+        vmul.f32        d5,  d4,  d0[0]
         vst1.32         {d5[0]},  [r0,:32]!
         vst1.32         {d5[1]},  [r4,:32]!
         bne             1b

libavcodec/arm/dsputil_init_neon.c

@@ -140,8 +140,7 @@ void ff_vp3_h_loop_filter_neon(uint8_t *, int, int *);
 
 void ff_vector_fmul_neon(float *dst, const float *src0, const float *src1, int len);
 void ff_vector_fmul_window_neon(float *dst, const float *src0,
-                                const float *src1, const float *win,
-                                float add_bias, int len);
+                                const float *src1, const float *win, int len);
 void ff_vector_fmul_scalar_neon(float *dst, const float *src, float mul,
                                 int len);
 void ff_vector_fmul_sv_scalar_2_neon(float *dst, const float *src,

libavcodec/arm/dsputil_neon.S

@@ -777,11 +777,8 @@ function ff_vector_fmul_neon, export=1
 endfunc
 
 function ff_vector_fmul_window_neon, export=1
-VFP     vdup.32         q8,  d0[0]
-NOVFP   vld1.32         {d16[],d17[]}, [sp,:32]
         push            {r4,r5,lr}
-VFP     ldr             lr,  [sp, #12]
-NOVFP   ldr             lr,  [sp, #16]
+        ldr             lr,  [sp, #12]
         sub             r2,  r2,  #8
         sub             r5,  lr,  #2
         add             r2,  r2,  r5, lsl #2
@@ -793,14 +790,12 @@ NOVFP   ldr             lr,  [sp, #16]
         vld1.64         {d4,d5},  [r3,:128]!
         vld1.64         {d6,d7},  [r4,:128], r5
 1:      subs            lr,  lr,  #4
-        vmov            q11, q8
-        vmla.f32        d22, d0,  d4
-        vmov            q10, q8
-        vmla.f32        d23, d1,  d5
+        vmul.f32        d22, d0,  d4
         vrev64.32       q3,  q3
-        vmla.f32        d20, d0,  d7
+        vmul.f32        d23, d1,  d5
         vrev64.32       q1,  q1
-        vmla.f32        d21, d1,  d6
+        vmul.f32        d20, d0,  d7
+        vmul.f32        d21, d1,  d6
         beq             2f
         vmla.f32        d22, d3,  d7
         vld1.64         {d0,d1},  [r1,:128]!

libavcodec/arm/fft_init_arm.c

@@ -34,7 +34,7 @@ void ff_synth_filter_float_neon(FFTContext *imdct,
                                 float *synth_buf_ptr, int *synth_buf_offset,
                                 float synth_buf2[32], const float window[512],
                                 float out[32], const float in[32],
-                                float scale, float bias);
+                                float scale);
 
 av_cold void ff_fft_init_arm(FFTContext *s)
 {

libavcodec/arm/synth_filter_neon.S

@@ -42,7 +42,7 @@ VFP     vpop            {d0}
 
         ldr             r5,  [sp, #9*4]         @ window
         ldr             r2,  [sp, #10*4]        @ out
-NOVFP   vldr            d0,  [sp, #12*4]        @ scale, bias
+NOVFP   vldr            s0,  [sp, #12*4]        @ scale
         add             r8,  r9,  #12*4
 
         mov             lr,  #64*4
@@ -90,10 +90,8 @@ NOVFP   vldr            d0,  [sp, #12*4]        @ scale, bias
         sub             r11, r11, #512*4
         b               2b
 3:
-        vdup.32         q8,  d0[1]
-        vdup.32         q9,  d0[1]
-        vmla.f32        q8,  q10, d0[0]
-        vmla.f32        q9,  q1,  d0[0]
+        vmul.f32        q8,  q10, d0[0]
+        vmul.f32        q9,  q1,  d0[0]
         vst1.32         {q3},     [r3,:128]
         sub             r3,  r3,  #16*4
         vst1.32         {q2},     [r3,:128]

libavcodec/atrac1.c

@@ -141,7 +141,7 @@ static int at1_imdct_block(AT1SUCtx* su, AT1Ctx *q)
 
             /* overlap and window */
             q->dsp.vector_fmul_window(&q->bands[band_num][start_pos], prev_buf,
-                                      &su->spectrum[0][ref_pos + start_pos], ff_sine_32, 0, 16);
+                                      &su->spectrum[0][ref_pos + start_pos], ff_sine_32, 16);
 
             prev_buf = &su->spectrum[0][ref_pos+start_pos + 16];
             start_pos += block_size;

libavcodec/dca.c

@@ -896,7 +896,7 @@ static void qmf_32_subbands(DCAContext * s, int chans,
         s->synth.synth_filter_float(&s->imdct,
                               s->subband_fir_hist[chans], &s->hist_index[chans],
                               s->subband_fir_noidea[chans], prCoeff,
-                              samples_out, s->raXin, scale, 0);
+                              samples_out, s->raXin, scale);
         samples_out+= 32;
 
     }
@@ -929,7 +929,7 @@ static void lfe_interpolation_fir(DCAContext *s, int decimation_select,
     /* Interpolation */
     for (deciindex = 0; deciindex < num_deci_sample; deciindex++) {
         s->dcadsp.lfe_fir(samples_out, samples_in, prCoeff, decifactor,
-                          scale, 0);
+                          scale);
         samples_in++;
         samples_out += 2 * decifactor;
     }

libavcodec/dcadsp.c

@@ -23,7 +23,7 @@
 #include "dcadsp.h"
 
 static void dca_lfe_fir_c(float *out, const float *in, const float *coefs,
-                          int decifactor, float scale, float bias)
+                          int decifactor, float scale)
 {
     float *out2 = out + decifactor;
     const float *cf0 = coefs;
@@ -39,8 +39,8 @@ static void dca_lfe_fir_c(float *out, const float *in, const float *coefs,
             v0 += s * *cf0++;
             v1 += s * *--cf1;
         }
-        *out++  = (v0 * scale) + bias;
-        *out2++ = (v1 * scale) + bias;
+        *out++  = v0 * scale;
+        *out2++ = v1 * scale;
     }
 }
 

libavcodec/dcadsp.h

@@ -21,7 +21,7 @@
 
 typedef struct DCADSPContext {
     void (*lfe_fir)(float *out, const float *in, const float *coefs,
-                    int decifactor, float scale, float bias);
+                    int decifactor, float scale);
 } DCADSPContext;
 
 void ff_dcadsp_init(DCADSPContext *s);

libavcodec/dsputil.c

@@ -3776,7 +3776,9 @@ static void vector_fmul_add_c(float *dst, const float *src0, const float *src1,
         dst[i] = src0[i] * src1[i] + src2[i];
 }
 
-void ff_vector_fmul_window_c(float *dst, const float *src0, const float *src1, const float *win, float add_bias, int len){
+static void vector_fmul_window_c(float *dst, const float *src0,
+                                 const float *src1, const float *win, int len)
+{
     int i,j;
     dst += len;
     win += len;
@@ -3786,8 +3788,8 @@ void ff_vector_fmul_window_c(float *dst, const float *src0, const float *src1, c
         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;
+        dst[i] = s0*wj - s1*wi;
+        dst[j] = s0*wi + s1*wj;
     }
 }
 
@@ -4434,7 +4436,7 @@ av_cold 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 = vector_fmul_add_c;
-    c->vector_fmul_window = ff_vector_fmul_window_c;
+    c->vector_fmul_window = vector_fmul_window_c;
     c->int32_to_float_fmul_scalar = int32_to_float_fmul_scalar_c;
     c->vector_clipf = vector_clipf_c;
     c->float_to_int16 = ff_float_to_int16_c;

libavcodec/dsputil.h

@@ -68,9 +68,6 @@ void ff_h264_luma_dc_dequant_idct_c(DCTELEM *output, DCTELEM *input, int qmul);
 void ff_svq3_luma_dc_dequant_idct_c(DCTELEM *output, DCTELEM *input, int qp);
 void ff_svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc);
 
-void ff_vector_fmul_window_c(float *dst, const float *src0, const float *src1,
-                             const float *win, float add_bias, int len);
-
 /* encoding scans */
 extern const uint8_t ff_alternate_horizontal_scan[64];
 extern const uint8_t ff_alternate_vertical_scan[64];
@@ -393,7 +390,7 @@ typedef struct DSPContext {
     /* assume len is a multiple of 8, and src arrays are 16-byte aligned */
     void (*vector_fmul_add)(float *dst, const float *src0, const float *src1, const float *src2, int len);
     /* assume len is a multiple of 4, and arrays are 16-byte aligned */
-    void (*vector_fmul_window)(float *dst, const float *src0, const float *src1, const float *win, float add_bias, int len);
+    void (*vector_fmul_window)(float *dst, const float *src0, const float *src1, const float *win, int len);
     /* assume len is a multiple of 8, and arrays are 16-byte aligned */
     void (*int32_to_float_fmul_scalar)(float *dst, const int *src, float mul, int len);
     void (*vector_clipf)(float *dst /* align 16 */, const float *src /* align 16 */, float min, float max, int len /* align 16 */);

libavcodec/ppc/float_altivec.c

@@ -90,13 +90,9 @@ static void vector_fmul_add_altivec(float *dst, const float *src0,
     }
 }
 
-static void vector_fmul_window_altivec(float *dst, const float *src0, const float *src1, const float *win, float add_bias, int len)
+static void vector_fmul_window_altivec(float *dst, const float *src0, const float *src1, const float *win, int len)
 {
-    union {
-        vector float v;
-        float s[4];
-    } vadd;
-    vector float vadd_bias, zero, t0, t1, s0, s1, wi, wj;
+    vector float zero, t0, t1, s0, s1, wi, wj;
     const vector unsigned char reverse = vcprm(3,2,1,0);
     int i,j;
 
@@ -104,8 +100,6 @@ static void vector_fmul_window_altivec(float *dst, const float *src0, const floa
     win += len;
     src0+= len;
 
-    vadd.s[0] = add_bias;
-    vadd_bias = vec_splat(vadd.v, 0);
     zero = (vector float)vec_splat_u32(0);
 
     for(i=-len*4, j=len*4-16; i<0; i+=16, j-=16) {
@@ -117,9 +111,9 @@ static void vector_fmul_window_altivec(float *dst, const float *src0, const floa
         s1 = vec_perm(s1, s1, reverse);
         wj = vec_perm(wj, wj, reverse);
 
-        t0 = vec_madd(s0, wj, vadd_bias);
+        t0 = vec_madd(s0, wj, zero);
         t0 = vec_nmsub(s1, wi, t0);
-        t1 = vec_madd(s0, wi, vadd_bias);
+        t1 = vec_madd(s0, wi, zero);
         t1 = vec_madd(s1, wj, t1);
         t1 = vec_perm(t1, t1, reverse);
 

libavcodec/synth_filter.c

@@ -24,7 +24,7 @@
 static void synth_filter_float(FFTContext *imdct,
                            float *synth_buf_ptr, int *synth_buf_offset,
                            float synth_buf2[32], const float window[512],
-                           float out[32], const float in[32], float scale, float bias)
+                           float out[32], const float in[32], float scale)
 {
     float *synth_buf= synth_buf_ptr + *synth_buf_offset;
     int i, j;
@@ -48,8 +48,8 @@ static void synth_filter_float(FFTContext *imdct,
             c += window[i + j + 32]*( synth_buf[16 + i + j - 512]);
             d += window[i + j + 48]*( synth_buf[31 - i + j - 512]);
         }
-        out[i     ] = a*scale + bias;
-        out[i + 16] = b*scale + bias;
+        out[i     ] = a*scale;
+        out[i + 16] = b*scale;
         synth_buf2[i     ] = c;
         synth_buf2[i + 16] = d;
     }

libavcodec/synth_filter.h

@@ -28,7 +28,7 @@ typedef struct SynthFilterContext {
                                float *synth_buf_ptr, int *synth_buf_offset,
                                float synth_buf2[32], const float window[512],
                                float out[32], const float in[32],
-                               float scale, float bias);
+                               float scale);
 } SynthFilterContext;
 
 void ff_synth_filter_init(SynthFilterContext *c);

libavcodec/twinvq.c

@@ -646,7 +646,6 @@ static void imdct_and_window(TwinContext *tctx, enum FrameType ftype, int wtype,
                                      prev_buf + (bsize-wsize)/2,
                                      buf1 + bsize*j,
                                      ff_sine_windows[av_log2(wsize)],
-                                     0.0,
                                      wsize/2);
         out2 += wsize;
 

libavcodec/vorbis_dec.c

@@ -1575,13 +1575,13 @@ static int vorbis_parse_audio_packet(vorbis_context *vc)
         const float *win  = vc->win[blockflag & previous_window];
 
         if (blockflag == previous_window) {
-            vc->dsp.vector_fmul_window(ret, saved, buf, win, 0, blocksize / 4);
+            vc->dsp.vector_fmul_window(ret, saved, buf, win, blocksize / 4);
         } else if (blockflag > previous_window) {
-            vc->dsp.vector_fmul_window(ret, saved, buf, win, 0, bs0 / 4);
+            vc->dsp.vector_fmul_window(ret, saved, buf, win, bs0 / 4);
             memcpy(ret+bs0/2, buf+bs0/4, ((bs1-bs0)/4) * sizeof(float));
         } else {
             memcpy(ret, saved, ((bs1 - bs0) / 4) * sizeof(float));
-            vc->dsp.vector_fmul_window(ret + (bs1 - bs0) / 4, saved + (bs1 - bs0) / 4, buf, win, 0, bs0 / 4);
+            vc->dsp.vector_fmul_window(ret + (bs1 - bs0) / 4, saved + (bs1 - bs0) / 4, buf, win, bs0 / 4);
         }
         memcpy(saved, buf + blocksize / 4, blocksize / 4 * sizeof(float));
     }

libavcodec/wmaprodec.c

@@ -1031,7 +1031,7 @@ static void wmapro_window(WMAProDecodeCtx *s)
         winlen >>= 1;
 
         s->dsp.vector_fmul_window(start, start, start + winlen,
-                                  window, 0, winlen);
+                                  window, winlen);
 
         s->channel[c].prev_block_len = s->subframe_len;
     }

libavcodec/x86/dsputil_mmx.c

@@ -2190,10 +2190,9 @@ static void vector_fmul_add_sse(float *dst, const float *src0, const float *src1
     );
 }
 
-static void vector_fmul_window_3dnow2(float *dst, const float *src0, const float *src1,
-                                      const float *win, float add_bias, int len){
 #if HAVE_6REGS
-    if(add_bias == 0){
+static void vector_fmul_window_3dnow2(float *dst, const float *src0, const float *src1,
+                                      const float *win, int len){
         x86_reg i = -len*4;
         x86_reg j = len*4-8;
         __asm__ volatile(
@@ -2220,15 +2219,10 @@ static void vector_fmul_window_3dnow2(float *dst, const float *src0, const float
             :"+r"(i), "+r"(j)
             :"r"(dst+len), "r"(src0+len), "r"(src1), "r"(win+len)
         );
-    }else
-#endif
-        ff_vector_fmul_window_c(dst, src0, src1, win, add_bias, len);
 }
 
 static void vector_fmul_window_sse(float *dst, const float *src0, const float *src1,
-                                   const float *win, float add_bias, int len){
-#if HAVE_6REGS
-    if(add_bias == 0){
+                                   const float *win, int len){
         x86_reg i = -len*4;
         x86_reg j = len*4-16;
         __asm__ volatile(
@@ -2256,10 +2250,8 @@ static void vector_fmul_window_sse(float *dst, const float *src0, const float *s
             :"+r"(i), "+r"(j)
             :"r"(dst+len), "r"(src0+len), "r"(src1), "r"(win+len)
         );
-    }else
-#endif
-        ff_vector_fmul_window_c(dst, src0, src1, win, add_bias, len);
 }
+#endif /* HAVE_6REGS */
 
 static void int32_to_float_fmul_scalar_sse(float *dst, const int *src, float mul, int len)
 {
@@ -2882,7 +2874,9 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
         }
         if(mm_flags & AV_CPU_FLAG_3DNOWEXT){
             c->vector_fmul_reverse = vector_fmul_reverse_3dnow2;
+#if HAVE_6REGS
             c->vector_fmul_window = vector_fmul_window_3dnow2;
+#endif
             if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
                 c->float_to_int16_interleave = float_to_int16_interleave_3dn2;
             }
@@ -2899,7 +2893,9 @@ void dsputil_init_mmx(DSPContext* c, AVCodecContext *avctx)
             c->vector_fmul = vector_fmul_sse;
             c->vector_fmul_reverse = vector_fmul_reverse_sse;
             c->vector_fmul_add = vector_fmul_add_sse;
+#if HAVE_6REGS
             c->vector_fmul_window = vector_fmul_window_sse;
+#endif
             c->int32_to_float_fmul_scalar = int32_to_float_fmul_scalar_sse;
             c->vector_clipf = vector_clipf_sse;
             c->float_to_int16 = float_to_int16_sse;