renderer: use complementary equations for \clip and \iclip

Basic symmetry considerations require that
  clip_func(src, clip) = iclip_func(src, 255 - clip),
but we use unrelated expressions for them:
  clip_func(src, clip) = (src * clip + 255) / 256 !=
  iclip_func(src, iclip) = max(src - iclip, 0) = max(src + clip - 255, 0).
Version with multiplication is more correct (albeit slower) as it gives
results closer to the infinite resolution reference (per compare -s 8)
in case of half-transparency. So I've picked better function
(clip variant) and derived another from it.

1 files changed, 4 insertions, 6 deletions

diff --git a/libass/ass_bitmap.c b/libass/ass_bitmap.c
index d3186b4..5e2ca77 100644
--- a/libass/ass_bitmap.c
+++ b/libass/ass_bitmap.c
@@ -395,16 +395,14 @@ void ass_add_bitmaps_c(uint8_t *dst, intptr_t dst_stride,
     }
 }
 
-void ass_sub_bitmaps_c(uint8_t *dst, intptr_t dst_stride,
-                       uint8_t *src, intptr_t src_stride,
-                       intptr_t width, intptr_t height)
+void ass_imul_bitmaps_c(uint8_t *dst, intptr_t dst_stride,
+                        uint8_t *src, intptr_t src_stride,
+                        intptr_t width, intptr_t height)
 {
-    short out;
     uint8_t* end = dst + dst_stride * height;
     while (dst < end) {
         for (unsigned j = 0; j < width; ++j) {
-            out = dst[j] - src[j];
-            dst[j] = FFMAX(out, 0);
+            dst[j] = (dst[j] * (255 - src[j]) + 255) >> 8;
         }
         dst += dst_stride;
         src += src_stride;