[FFmpeg-devel] [PATCH] avcodec/ffv1enc: Add -remap_optimizer option

* [FFmpeg-devel] [PATCH] avcodec/ffv1enc: Add -remap_optimizer option
@ 2025-03-28 11:41 Michael Niedermayer
  2025-03-30 20:25 ` Michael Niedermayer
  0 siblings, 1 reply; 2+ messages in thread
From: Michael Niedermayer @ 2025-03-28 11:41 UTC (permalink / raw)
  To: FFmpeg development discussions and patches

This allows tuning how much effort (time) the encoder spends on
optimizing the remap table

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
---
 doc/encoders.texi    |  4 ++++
 libavcodec/ffv1.h    |  2 +-
 libavcodec/ffv1enc.c | 52 +++++++++++++++++++++++++++++++-------------
 3 files changed, 42 insertions(+), 16 deletions(-)

diff --git a/doc/encoders.texi b/doc/encoders.texi
index f3fcc1aa60a..f987a234bc6 100644
--- a/doc/encoders.texi
+++ b/doc/encoders.texi
@@ -554,6 +554,10 @@ use 8bit default
 @item greater8bit
 use >8bit default
 @end table
+
+@item remap_optimizer
+0 - 5, default 3, how much effort the encoder puts into optimizing the remap table.
+
 @end table
 
 @anchor{flac}
diff --git a/libavcodec/ffv1.h b/libavcodec/ffv1.h
index 09118e0b7dc..578b504c9f1 100644
--- a/libavcodec/ffv1.h
+++ b/libavcodec/ffv1.h
@@ -148,7 +148,7 @@ typedef struct FFV1Context {
     int colorspace;
     int flt;
     int remap_mode;
-
+    int remap_optimizer;
 
     int use32bit;
 
diff --git a/libavcodec/ffv1enc.c b/libavcodec/ffv1enc.c
index b156ff2dc10..a09b31218db 100644
--- a/libavcodec/ffv1enc.c
+++ b/libavcodec/ffv1enc.c
@@ -1369,33 +1369,46 @@ static void encode_float32_remap(FFV1Context *f, FFV1SliceContext *sc,
                                  const uint8_t *src[4])
 {
     int pixel_num = sc->slice_width * sc->slice_height;
-    int mul_count;
+    const int max_log2_mul_count  = ((int[]){  1,  1,  1,  9,  9,  10})[f->remap_optimizer];
+    const int log2_mul_count_step = ((int[]){  1,  1,  1,  9,  9,   1})[f->remap_optimizer];
+    const int max_log2_mul        = ((int[]){  1,  8,  8,  9, 22,  22})[f->remap_optimizer];
+    const int log2_mul_step       = ((int[]){  1,  8,  1,  1,  1,   1})[f->remap_optimizer];
+    const int bruteforce_count    = ((int[]){  0,  0,  0,  1,  1,   1})[f->remap_optimizer];
+    const int stair_mode          = ((int[]){  0,  0,  0,  1,  0,   0})[f->remap_optimizer];
 
     av_assert0 (pixel_num <= 65536);
 
     for (int p= 0; p < 1 + 2*f->chroma_planes + f->transparency; p++) {
-        float score_sum[2] = {0};
-        int mul_all[2][513];
+        int best_log2_mul_count = 0;
+        float score_sum[11] = {0};
+        int mul_all[11][1025];
 
-        for (mul_count= 1; mul_count<=512; mul_count+=511) {
-            float score_tab_all[513][23] = {0};
+        for (int log2_mul_count= 0; log2_mul_count <= max_log2_mul_count; log2_mul_count += log2_mul_count_step) {
+            float score_tab_all[1025][23] = {0};
             int64_t last_val = -1;
-            int *mul_tab = mul_all[mul_count>>9];
+            int *mul_tab = mul_all[log2_mul_count];
             int last_mul_index = -1;
-            score_sum[mul_count>>9] += log2(mul_count);
+            int mul_count = 1 << log2_mul_count;
+
+            score_sum[log2_mul_count] += log2_mul_count;
             for (int i= 0; i<pixel_num; i++) {
                 int64_t val = sc->unit[p][i].val;
                 int mul_index = (val + 1LL)*mul_count >> 32;
                 if (val != last_val) {
                     float *score_tab = score_tab_all[(last_val + 1LL)*mul_count >> 32];
                     av_assert2(last_val < val);
-                    for(int si= 0; si < FF_ARRAY_ELEMS(*score_tab_all); si++) {
+                    for(int si= 0; si <= max_log2_mul; si += log2_mul_step) {
                         int64_t delta = val - last_val;
                         int mul;
                         int64_t cost;
 
                         if (last_val < 0) {
                             mul = 1;
+                        } else if (stair_mode && mul_count == 512 && si == max_log2_mul ) {
+                            if (mul_index >= 0x378/8 && mul_index <= 23 + 0x378/8) {
+                                mul = (0x800080 >> (mul_index - 0x378/8));
+                            } else
+                                mul = 1;
                         } else {
                             mul = (0x10001LL)<<si >> 16;
                         }
@@ -1414,20 +1427,29 @@ static void encode_float32_remap(FFV1Context *f, FFV1SliceContext *sc,
             for(int i= 0; i<mul_count; i++) {
                 int best_index = 0;
                 float *score_tab = score_tab_all[i];
-                for(int si= 1; si < FF_ARRAY_ELEMS(*score_tab_all); si++) {
+                for(int si= 0; si <= max_log2_mul; si += log2_mul_step) {
                     if (score_tab[si] < score_tab[ best_index ])
                         best_index = si;
                 }
-                mul_tab[i] = -((0x10001LL)<<best_index >> 16);
-                score_sum[mul_count>>9] += score_tab[ best_index ];
+                if (stair_mode && mul_count == 512 && best_index == max_log2_mul ) {
+                    if (i >= 0x378/8 && i <= 23 + 0x378/8) {
+                        mul_tab[i] = -(0x800080 >> (i - 0x378/8));
+                    } else
+                        mul_tab[i] = -1;
+                } else
+                    mul_tab[i] = -((0x10001LL)<<best_index >> 16);
+                score_sum[log2_mul_count] += score_tab[ best_index ];
             }
             mul_tab[mul_count] = 1;
 
-            score_sum[mul_count>>9] = encode_float32_remap_segment(sc, p, mul_count, mul_all[mul_count>>9], 0, 0);
+            if (bruteforce_count)
+                score_sum[log2_mul_count] = encode_float32_remap_segment(sc, p, mul_count, mul_all[log2_mul_count], 0, 0);
+
+            if (score_sum[log2_mul_count] < score_sum[best_log2_mul_count])
+                best_log2_mul_count = log2_mul_count;
         }
 
-        mul_count = score_sum[0] <= score_sum[1] ? 1 : 512;
-        encode_float32_remap_segment(sc, p, mul_count, mul_all[mul_count>>9], 1, 1);
+        encode_float32_remap_segment(sc, p, 1<<best_log2_mul_count, mul_all[best_log2_mul_count], 1, 1);
     }
 }
 
@@ -1813,7 +1835,7 @@ static const AVOption options[] = {
             { .i64 =  1 }, INT_MIN, INT_MAX, VE, .unit = "remap_mode" },
         { "flipdualrle", "Dual RLE", 0, AV_OPT_TYPE_CONST,
             { .i64 =  2 }, INT_MIN, INT_MAX, VE, .unit = "remap_mode" },
-
+    { "remap_optimizer", "Remap Optimizer", OFFSET(remap_optimizer), AV_OPT_TYPE_INT, { .i64 = 3 }, 0, 5, VE, .unit = "remap_optimizer" },
 
     { NULL }
 };
-- 
2.48.1

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".

^ permalink raw reply	[flat|nested] 2+ messages in thread