ood-ports-lbsd/graphics/ffmpeg/patches/patch-libavcodec_aacenc_utils_h

$OpenBSD: patch-libavcodec_aacenc_utils_h,v 1.11 2016/10/17 07:51:00 ajacoutot Exp $

aacenc: copy PRNG from the decoder

AAC encoder: simplify and speed up find_min_book

AAC encoder: Extensive improvements

aacenc_utils: add 'inline' flag to find_form_factor, silence warning

aacenc: add support for changing options based on a profile

aacenc_utils: fit find_form_factor() below 80 chars per line

aacenc: partially revert previous commits to set options via a profile

avcodec/aac_tablegen: get rid of hardcoded tables entirely

AAC encoder: improve SF range utilization

aacenc: switch to using the RNG from libavutil

avcodec/aacenc_is: replace pow(x, 0.75) by x/sqrtf(sqrtf(x))

lavc/aacenc_utils: replace sqrtf(Q*sqrtf(Q)) by precomputed value

aacenc: avoid double in quantize_bands.

aacenc_utils: Use temporary variable.

lavc/aacenc_utils: replace powf(x,y) by expf(logf(x), y)

aacenc: use the decoder's lcg PRNG

--- libavcodec/aacenc_utils.h.orig	Sat Aug 27 22:51:19 2016
+++ libavcodec/aacenc_utils.h	Sat Oct 15 18:51:24 2016
@@ -28,9 +28,10 @@
 #ifndef AVCODEC_AACENC_UTILS_H
 #define AVCODEC_AACENC_UTILS_H
 
+#include "libavutil/internal.h"
 #include "aac.h"
-#include "aac_tablegen_decl.h"
 #include "aacenctab.h"
+#include "aactab.h"
 
 #define ROUND_STANDARD 0.4054f
 #define ROUND_TO_ZERO 0.1054f
@@ -45,6 +46,11 @@ static inline void abs_pow34_v(float *out, const float
     }
 }
 
+static inline float pos_pow34(float a)
+{
+    return sqrtf(a * sqrtf(a));
+}
+
 /**
  * Quantize one coefficient.
  * @return absolute value of the quantized coefficient
@@ -61,13 +67,13 @@ static inline void quantize_bands(int *out, const floa
                                   const float rounding)
 {
     int i;
-    double qc;
     for (i = 0; i < size; i++) {
-        qc = scaled[i] * Q34;
-        out[i] = (int)FFMIN(qc + rounding, (double)maxval);
+        float qc = scaled[i] * Q34;
+        int tmp = (int)FFMIN(qc + rounding, (float)maxval);
         if (is_signed && in[i] < 0.0f) {
-            out[i] = -out[i];
+            tmp = -tmp;
         }
+        out[i] = tmp;
     }
 }
 
@@ -85,20 +91,68 @@ static inline float find_max_val(int group_len, int sw
 
 static inline int find_min_book(float maxval, int sf)
 {
-    float Q = ff_aac_pow2sf_tab[POW_SF2_ZERO - sf + SCALE_ONE_POS - SCALE_DIV_512];
-    float Q34 = sqrtf(Q * sqrtf(Q));
+    float Q34 = ff_aac_pow34sf_tab[POW_SF2_ZERO - sf + SCALE_ONE_POS - SCALE_DIV_512];
     int qmaxval, cb;
     qmaxval = maxval * Q34 + C_QUANT;
-    if      (qmaxval ==  0) cb = 0;
-    else if (qmaxval ==  1) cb = 1;
-    else if (qmaxval ==  2) cb = 3;
-    else if (qmaxval <=  4) cb = 5;
-    else if (qmaxval <=  7) cb = 7;
-    else if (qmaxval <= 12) cb = 9;
-    else                    cb = 11;
+    if (qmaxval >= (FF_ARRAY_ELEMS(aac_maxval_cb)))
+        cb = 11;
+    else
+        cb = aac_maxval_cb[qmaxval];
     return cb;
 }
 
+static inline float find_form_factor(int group_len, int swb_size, float thresh,
+                                     const float *scaled, float nzslope) {
+    const float iswb_size = 1.0f / swb_size;
+    const float iswb_sizem1 = 1.0f / (swb_size - 1);
+    const float ethresh = thresh;
+    float form = 0.0f, weight = 0.0f;
+    int w2, i;
+    for (w2 = 0; w2 < group_len; w2++) {
+        float e = 0.0f, e2 = 0.0f, var = 0.0f, maxval = 0.0f;
+        float nzl = 0;
+        for (i = 0; i < swb_size; i++) {
+            float s = fabsf(scaled[w2*128+i]);
+            maxval = FFMAX(maxval, s);
+            e += s;
+            e2 += s *= s;
+            /* We really don't want a hard non-zero-line count, since
+             * even below-threshold lines do add up towards band spectral power.
+             * So, fall steeply towards zero, but smoothly
+             */
+            if (s >= ethresh) {
+                nzl += 1.0f;
+            } else {
+                if (nzslope == 2.f)
+                    nzl += (s / ethresh) * (s / ethresh);
+                else
+                    nzl += ff_fast_powf(s / ethresh, nzslope);
+            }
+        }
+        if (e2 > thresh) {
+            float frm;
+            e *= iswb_size;
+
+            /** compute variance */
+            for (i = 0; i < swb_size; i++) {
+                float d = fabsf(scaled[w2*128+i]) - e;
+                var += d*d;
+            }
+            var = sqrtf(var * iswb_sizem1);
+
+            e2 *= iswb_size;
+            frm = e / FFMIN(e+4*var,maxval);
+            form += e2 * sqrtf(frm) / FFMAX(0.5f,nzl);
+            weight += e2;
+        }
+    }
+    if (weight > 0) {
+        return form / weight;
+    } else {
+        return 1.0f;
+    }
+}
+
 /** Return the minimum scalefactor where the quantized coef does not clip. */
 static inline uint8_t coef2minsf(float coef)
 {
@@ -128,6 +182,89 @@ static inline int quant_array_idx(const float val, con
     return index;
 }
 
+/**
+ * approximates exp10f(-3.0f*(0.5f + 0.5f * cosf(FFMIN(b,15.5f) / 15.5f)))
+ */
+static av_always_inline float bval2bmax(float b)
+{
+    return 0.001f + 0.0035f * (b*b*b) / (15.5f*15.5f*15.5f);
+}
+
+/*
+ * Compute a nextband map to be used with SF delta constraint utilities.
+ * The nextband array should contain 128 elements, and positions that don't
+ * map to valid, nonzero bands of the form w*16+g (with w being the initial
+ * window of the window group, only) are left indetermined.
+ */
+static inline void ff_init_nextband_map(const SingleChannelElement *sce, uint8_t *nextband)
+{
+    unsigned char prevband = 0;
+    int w, g;
+    /** Just a safe default */
+    for (g = 0; g < 128; g++)
+        nextband[g] = g;
+
+    /** Now really navigate the nonzero band chain */
+    for (w = 0; w < sce->ics.num_windows; w += sce->ics.group_len[w]) {
+        for (g = 0; g < sce->ics.num_swb; g++) {
+            if (!sce->zeroes[w*16+g] && sce->band_type[w*16+g] < RESERVED_BT)
+                prevband = nextband[prevband] = w*16+g;
+        }
+    }
+    nextband[prevband] = prevband; /* terminate */
+}
+
+/*
+ * Updates nextband to reflect a removed band (equivalent to
+ * calling ff_init_nextband_map after marking a band as zero)
+ */
+static inline void ff_nextband_remove(uint8_t *nextband, int prevband, int band)
+{
+    nextband[prevband] = nextband[band];
+}
+
+/*
+ * Checks whether the specified band could be removed without inducing
+ * scalefactor delta that violates SF delta encoding constraints.
+ * prev_sf has to be the scalefactor of the previous nonzero, nonspecial
+ * band, in encoding order, or negative if there was no such band.
+ */
+static inline int ff_sfdelta_can_remove_band(const SingleChannelElement *sce,
+    const uint8_t *nextband, int prev_sf, int band)
+{
+    return prev_sf >= 0
+        && sce->sf_idx[nextband[band]] >= (prev_sf - SCALE_MAX_DIFF)
+        && sce->sf_idx[nextband[band]] <= (prev_sf + SCALE_MAX_DIFF);
+}
+
+/*
+ * Checks whether the specified band's scalefactor could be replaced
+ * with another one without violating SF delta encoding constraints.
+ * prev_sf has to be the scalefactor of the previous nonzero, nonsepcial
+ * band, in encoding order, or negative if there was no such band.
+ */
+static inline int ff_sfdelta_can_replace(const SingleChannelElement *sce,
+    const uint8_t *nextband, int prev_sf, int new_sf, int band)
+{
+    return new_sf >= (prev_sf - SCALE_MAX_DIFF)
+        && new_sf <= (prev_sf + SCALE_MAX_DIFF)
+        && sce->sf_idx[nextband[band]] >= (new_sf - SCALE_MAX_DIFF)
+        && sce->sf_idx[nextband[band]] <= (new_sf + SCALE_MAX_DIFF);
+}
+
+/**
+ * linear congruential pseudorandom number generator
+ *
+ * @param   previous_val    pointer to the current state of the generator
+ *
+ * @return  Returns a 32-bit pseudorandom integer
+ */
+static av_always_inline int lcg_random(unsigned previous_val)
+{
+    union { unsigned u; int s; } v = { previous_val * 1664525u + 1013904223 };
+    return v.s;
+}
+
 #define ERROR_IF(cond, ...) \
     if (cond) { \
         av_log(avctx, AV_LOG_ERROR, __VA_ARGS__); \
@@ -138,6 +275,5 @@ static inline int quant_array_idx(const float val, con
     if (cond) { \
         av_log(avctx, AV_LOG_WARNING, __VA_ARGS__); \
     }
-
 
 #endif /* AVCODEC_AACENC_UTILS_H */