avcodec/ffv1: flip half of float16 and Compactify floats

float16 (and more so float32) have many odd values
half the values are negative, many are larger than "1.0"
and many values are very close to 0.

Storing the 16bits as is, looses compression because of the mixture
of dense and sparse regions and also many completely unused ones.

This simply remaps the 65536 values so no unused values remain
This improves compression by about 1.5% for the ACES_OT_VWG_SampleFrames testset
(this testset contains all kind of funny values including many images
 with negative rgb values)

The space needed for the map is insignificant compared to the
compression gained

This patch also flips half the float range as it can be done
using the same table.

Sponsored-by: Sovereign Tech Fund
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

libavcodec/ffv1.h

@@ -105,6 +105,7 @@ typedef struct FFV1SliceContext {
             uint64_t (*rc_stat2[MAX_QUANT_TABLES])[32][2];
         };
     };
+    uint16_t   fltmap[4][65536];
 } FFV1SliceContext;
 
 typedef struct FFV1Context {

libavcodec/ffv1dec_template.c

@@ -155,6 +155,20 @@ static int RENAME(decode_rgb_frame)(FFV1Context *f, FFV1SliceContext *sc,
 
     memset(RENAME(sc->sample_buffer), 0, 8 * (w + 6) * sizeof(*RENAME(sc->sample_buffer)));
 
+    if (f->flt) {
+        for (int p= 0; p<3 + transparency; p++) {
+            int j = 0;
+            int lu = 0;
+            uint8_t state[2] = {128, 128};
+            for (int i= 0; i<65536; i++) {
+                int u = get_rac(&sc->c, state + lu);
+                sc->fltmap[p][j] = i ^ ((i&0x8000) ? 0 : 0x7FFF);
+                j+= u;
+                lu = u;
+            }
+        }
+    }
+
     for (y = 0; y < h; y++) {
         for (p = 0; p < 3 + transparency; p++) {
             int ret;
@@ -185,6 +199,13 @@ static int RENAME(decode_rgb_frame)(FFV1Context *f, FFV1SliceContext *sc,
                 b += g;
                 r += g;
             }
+            if (f->flt) {
+                r = sc->fltmap[0][r & 0xFFFF];
+                g = sc->fltmap[1][g & 0xFFFF];
+                b = sc->fltmap[2][b & 0xFFFF];
+                if (transparency)
+                    a = sc->fltmap[3][a & 0xFFFF];
+            }
 
             if (lbd)
                 *((uint32_t*)(src[0] + x*4 + stride[0]*y)) = b + ((unsigned)g<<8) + ((unsigned)r<<16) + ((unsigned)a<<24);

libavcodec/ffv1enc_template.c

@@ -148,6 +148,47 @@ static int RENAME(encode_rgb_frame)(FFV1Context *f, FFV1SliceContext *sc,
     memset(RENAME(sc->sample_buffer), 0, ring_size * MAX_PLANES *
            (w + 6) * sizeof(*RENAME(sc->sample_buffer)));
 
+    if (f->flt) {
+        memset(sc->fltmap, 0, sizeof(sc->fltmap));
+
+        for (y = 0; y < h; y++) {
+            for (x = 0; x < w; x++) {
+                int b, g, r, av_uninit(a);
+
+                if (sizeof(TYPE) == 4 || transparency) {
+                    g = *((const uint16_t *)(src[0] + x*2 + stride[0]*y));
+                    b = *((const uint16_t *)(src[1] + x*2 + stride[1]*y));
+                    r = *((const uint16_t *)(src[2] + x*2 + stride[2]*y));
+                    if (transparency)
+                        a = *((const uint16_t *)(src[3] + x*2 + stride[3]*y));
+                } else {
+                    b = *((const uint16_t *)(src[0] + x*2 + stride[0]*y));
+                    g = *((const uint16_t *)(src[1] + x*2 + stride[1]*y));
+                    r = *((const uint16_t *)(src[2] + x*2 + stride[2]*y));
+                }
+
+                sc->fltmap[0][r] = 1;
+                sc->fltmap[1][g] = 1;
+                sc->fltmap[2][b] = 1;
+                if (transparency)
+                    sc->fltmap[3][a] = 1;
+            }
+        }
+        for (int p= 0; p<3 + transparency; p++) {
+            int j = 0;
+            int lu = 0;
+            uint8_t state[2] = {128, 128};
+            for (int i= 0; i<65536; i++) {
+                int ri = i ^ ((i&0x8000) ? 0 : 0x7FFF);
+                int u = sc->fltmap[p][ri];
+                sc->fltmap[p][ri] = j;
+                j+= u;
+                put_rac(&sc->c, state + lu, u);
+                lu = u;
+            }
+        }
+    }
+
     for (y = 0; y < h; y++) {
         for (i = 0; i < ring_size; i++)
             for (p = 0; p < MAX_PLANES; p++)
@@ -180,6 +221,14 @@ static int RENAME(encode_rgb_frame)(FFV1Context *f, FFV1SliceContext *sc,
                 r = *((const uint16_t *)(src[2] + x*2 + stride[2]*y));
             }
 
+            if (f->flt) {
+                r = sc->fltmap[0][r];
+                g = sc->fltmap[1][g];
+                b = sc->fltmap[2][b];
+                if (transparency)
+                    a = sc->fltmap[3][a];
+            }
+
             if (sc->slice_coding_mode != 1) {
                 b -= g;
                 r -= g;