avcodec/ffv1: Use dual run coder for fltmap

This improves compression by 0.1% overall and 44% for the changed table
I tried several other things but so far this is the best
compromise between complexity and compression

This can also be extended to 32 and 64bit floats

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

2 files changed, 32 insertions, 8 deletions

diff --git a/libavcodec/ffv1dec_template.c b/libavcodec/ffv1dec_template.c
index 2769775c36..01e61e1ae8 100644
--- a/libavcodec/ffv1dec_template.c
+++ b/libavcodec/ffv1dec_template.c
@@ -159,12 +159,25 @@ static int RENAME(decode_rgb_frame)(FFV1Context *f, FFV1SliceContext *sc,
         for (int p= 0; p<3 + transparency; p++) {
             int j = 0;
             int lu = 0;
-            uint8_t state[2] = {128, 128};
+            uint8_t state[2][32];
+            memset(state, 128, sizeof(state));
+
             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;
+                int run = get_symbol_inline(&sc->c, state[lu], 0);
+                if (run > 65536U - i)
+                    return AVERROR_INVALIDDATA;
+                if (lu) {
+                    lu ^= !run;
+                    while (run--) {
+                        sc->fltmap[p][j++] = i ^ ((i&0x8000) ? 0 : 0x7FFF);
+                        i++;
+                    }
+                } else {
+                    i += run;
+                    if (i != 65536)
+                        sc->fltmap[p][j++] = i ^ ((i&0x8000) ? 0 : 0x7FFF);
+                    lu ^= !run;
+                }
             }
         }
     }
diff --git a/libavcodec/ffv1enc_template.c b/libavcodec/ffv1enc_template.c
index 4c7cd2e647..af3354497a 100644
--- a/libavcodec/ffv1enc_template.c
+++ b/libavcodec/ffv1enc_template.c
@@ -177,15 +177,26 @@ static int RENAME(encode_rgb_frame)(FFV1Context *f, FFV1SliceContext *sc,
         for (int p= 0; p<3 + transparency; p++) {
             int j = 0;
             int lu = 0;
-            uint8_t state[2] = {128, 128};
+            uint8_t state[2][32];
+            int run = 0;
+            memset(state, 128, sizeof(state));
             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;
+
+                if (lu == u) {
+                    run ++;
+                } else {
+                    put_symbol_inline(&sc->c, state[lu], run, 0, NULL, NULL);
+                    if (run == 0)
+                        lu = u;
+                    run = 0;
+                }
             }
+            if (run)
+                put_symbol(&sc->c, state[lu], run, 0);
         }
     }