Niklas Haas 3743d2851d swscale/ops: switch from AVRational to AVRational64 ...

This has two immediate consequences:

1. Fixes overflow in the range tracker for some 32-bit packed formats:

 rgb24 -> v30xbe:
   [ u8 +++X] SWS_OP_READ         : 3 elem(s) packed >> 0
     min: {0 0 0 _}, max: {255 255 255 _}
   [ u8 +++X] SWS_OP_CONVERT      : u8 -> f32
     min: {0 0 0 _}, max: {255 255 255 _}
   [f32 ...X] SWS_OP_LINEAR       : matrix3+off3 [...]
     min: {64 64 64 _}, max: {940 960 960 _}
   [f32 ...X] SWS_OP_DITHER       : 16x16 matrix + {0 3 2 -1}
     min: {64.001953 64.001953 64.001953 _}, max: {940.998047 960.998047 960.998047 _}
   [f32 +++X] SWS_OP_CONVERT      : f32 -> u32
     min: {64 64 64 _}, max: {940 960 960 _}
   [u32 +++X] SWS_OP_SWIZZLE      : 2013
     min: {64 64 64 _}, max: {960 940 960 _}
   [u32 ++++] SWS_OP_CLEAR        : {_ _ _ 1}
     min: {64 64 64 1}, max: {960 940 960 1}
   [u32 +XXX] SWS_OP_PACK         : {10 10 10 2}
-    min: {268697857 _ _ _}, max: {-264581375 _ _ _}
+    min: {268697857 _ _ _}, max: {4030385921 _ _ _}
   [u32 zXXX] SWS_OP_SWAP_BYTES
-    min: {268697857 _ _ _}, max: {-264581375 _ _ _}
+    min: {268697857 _ _ _}, max: {4030385921 _ _ _}

2. Slightly increases the accuracy of intermediate values for some linear ops:

 yuv444p10be -> rgb48be:
   [u16 zzzX] SWS_OP_READ         : 3 elem(s) planar >> 0
     min: {0 0 0 _}, max: {1023 1023 1023 _}
   [u16 +++X] SWS_OP_SWAP_BYTES
     min: {0 0 0 _}, max: {1023 1023 1023 _}
   [u16 +++X] SWS_OP_CONVERT      : u16 -> f32
     min: {0 0 0 _}, max: {1023 1023 1023 _}
   [f32 ...X] SWS_OP_LINEAR       : matrix3+off3 [...]
46.813777] [0 0 0 1 0]]
-    min: {-57290.842348 -44341.337325 -71146.813777 _}, max: {124144.718860 111375.162457 137973.627845 _}
+    min: {-57290.842348 -44341.337326 -71146.813777 _}, max: {124144.718860 111375.162457 137973.627845 _}
   [f32 ...X] SWS_OP_MAX          : {0 0 0 _} <= x
     min: {0 0 0 _}, max: {124144.718860 111375.162457 137973.627845 _}
   [f32 ...X] SWS_OP_MIN          : x <= {65535 65535 65535 _}
     min: {0 0 0 _}, max: {65535 65535 65535 _}
   [f32 +++X] SWS_OP_CONVERT      : f32 -> u16
     min: {0 0 0 _}, max: {65535 65535 65535 _}
   [u16 zzzX] SWS_OP_SWAP_BYTES
     min: {0 0 0 _}, max: {65535 65535 65535 _}
   [u16 XXXX] SWS_OP_WRITE        : 3 elem(s) packed >> 0
     (X = unused, z = byteswapped, + = exact, 0 = zero)

Importantly, none of the changes affect the actual operation list, just the
range tracking metadata.

Sponsored-by: Sovereign Tech Fund
Signed-off-by: Niklas Haas <git@haasn.dev>

2026-06-25 01:20:44 +02:00

..

Makefile

…

ops.c

swscale/ops: switch from AVRational to AVRational64

2026-06-25 01:20:44 +02:00

ops.h

swscale/graph: support allocating hardware intermediate frames

2026-05-22 14:05:21 +09:00

spvasm.h

swscale/vulkan: implement filtered reads for the SPIRV backend

2026-05-30 12:10:06 +09:00

vulkan_glslang.c

…

vulkan_shaderc.c

…

vulkan.c

…