RepoMirrors
/
mpv
mirror of https://github.com/mpv-player/mpv
1
0
Fork 0
Code Issues Releases Wiki Activity
mpv/test/ref/draw_bmp.txt

220 lines
17 KiB
Plaintext
Raw Normal View History

draw_bmp: add integer blending for 8 bit formats Whatever it's worth. Instead of doing a pretty stupid conversion to float, just blend it directly. This works for most RGB formats that are 8 bits per component or below (the latter because we expand packed fringe RGB formats for simplicity). For higher bit depth RGB this would need extra code.
2020-05-12 21:10:14 +00:00
0bgr = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
0rgb = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
abgr = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrap, a=unknown, ca=unknown, ca_f=unknown
argb = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrap, a=unknown, ca=unknown, ca_f=unknown
draw_bmp: rewrite draw_bmp.c is the software blender for subtitles and OSD. It's used by encoding mode (burning subtitles), and some VOs, like vo_drm, vo_x11, vo_xv, and possibly more. This changes the algorithm from upsampling the video to 4:4:4 and then blending to downsampling the OSD and then blending directly to video. This has far-reaching consequences for its internals, and results in an effective rewrite. Since I wanted to avoid un-premultiplying, all blending is done with premultiplied alpha. That's actually the sane thing to do. The old code just didn't do it, because it's very weird in YUV fixed point. Essentially, you'd have to compensate for the chroma centering constant by subtracting src_alpha/255*128. This seemed so hairy (especially with correct rounding and high bit depths involved) that I went for using float. I think it turned out mostly OK, although it's more complex and less maintainable than before. reinit() is certainly a bit too long. While it should be possible to optimize the RGB path more (for example by blending directly instead of doing the stupid float conversion), this is probably slower. vo_xv users probably lose in this, because it takes the slowest path (due to subsampling requirements and using YUV). Why this rewrite? Nobody knows. I simply forgot the reason. But you'll have it anyway. Whether or not this would have required a full rewrite, at least it supports target alpha now (you can for example hard sub transparent PNGs, if you ever wanted to use mpv for this). Remove the check in vf_sub. The new draw_bmp.c is not as reliant on libswscale anymore (mostly uses repack.c now), and osd.c shows an error message on missing support instead now. Formats with chroma subsampling of 4 are not supported, because FFmpeg doesn't provide pixfmt definitions for alpha variants. We could provide those ourselves (relatively trivial), but why bother.
2020-05-09 16:01:07 +00:00
ayuv64 = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuva444pf, a=unknown, ca=unknown, ca_f=unknown
ayuv64be = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuva444pf, a=unknown, ca=unknown, ca_f=unknown
bayer_bggr16= no
bayer_bggr16be= no
bayer_bggr8 = no
bayer_gbrg16= no
bayer_gbrg16be= no
bayer_gbrg8 = no
bayer_grbg16= no
bayer_grbg16be= no
bayer_grbg8 = no
bayer_rggb16= no
bayer_rggb16be= no
bayer_rggb8 = no
draw_bmp: add integer blending for 8 bit formats Whatever it's worth. Instead of doing a pretty stupid conversion to float, just blend it directly. This works for most RGB formats that are 8 bits per component or below (the latter because we expand packed fringe RGB formats for simplicity). For higher bit depth RGB this would need extra code.
2020-05-12 21:10:14 +00:00
bgr0 = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
bgr24 = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
draw_bmp: rewrite draw_bmp.c is the software blender for subtitles and OSD. It's used by encoding mode (burning subtitles), and some VOs, like vo_drm, vo_x11, vo_xv, and possibly more. This changes the algorithm from upsampling the video to 4:4:4 and then blending to downsampling the OSD and then blending directly to video. This has far-reaching consequences for its internals, and results in an effective rewrite. Since I wanted to avoid un-premultiplying, all blending is done with premultiplied alpha. That's actually the sane thing to do. The old code just didn't do it, because it's very weird in YUV fixed point. Essentially, you'd have to compensate for the chroma centering constant by subtracting src_alpha/255*128. This seemed so hairy (especially with correct rounding and high bit depths involved) that I went for using float. I think it turned out mostly OK, although it's more complex and less maintainable than before. reinit() is certainly a bit too long. While it should be possible to optimize the RGB path more (for example by blending directly instead of doing the stupid float conversion), this is probably slower. vo_xv users probably lose in this, because it takes the slowest path (due to subsampling requirements and using YUV). Why this rewrite? Nobody knows. I simply forgot the reason. But you'll have it anyway. Whether or not this would have required a full rewrite, at least it supports target alpha now (you can for example hard sub transparent PNGs, if you ever wanted to use mpv for this). Remove the check in vf_sub. The new draw_bmp.c is not as reliant on libswscale anymore (mostly uses repack.c now), and osd.c shows an error message on missing support instead now. Formats with chroma subsampling of 4 are not supported, because FFmpeg doesn't provide pixfmt definitions for alpha variants. We could provide those ourselves (relatively trivial), but why bother.
2020-05-09 16:01:07 +00:00
bgr4 = no
draw_bmp: add integer blending for 8 bit formats Whatever it's worth. Instead of doing a pretty stupid conversion to float, just blend it directly. This works for most RGB formats that are 8 bits per component or below (the latter because we expand packed fringe RGB formats for simplicity). For higher bit depth RGB this would need extra code.
2020-05-12 21:10:14 +00:00
bgr444 = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
bgr444be = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
draw_bmp: don't make strange decisions on broken iknput csp params This checked params->color.space for being RGB. If the colorspace is unset, this did dumb things because even if the imgfmt was a RGB one, the colorspace was not set to RGB. This actually also happened to the tests. (Short-cutting RGB like this is actually wrong, since RGB could still have strange gamma or primaries, which would warrant a full conversion. So you'd need to check for these other parameters as well. To be fixed later.)
2020-05-12 20:43:29 +00:00
bgr48 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
bgr48be = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
draw_bmp: add integer blending for 8 bit formats Whatever it's worth. Instead of doing a pretty stupid conversion to float, just blend it directly. This works for most RGB formats that are 8 bits per component or below (the latter because we expand packed fringe RGB formats for simplicity). For higher bit depth RGB this would need extra code.
2020-05-12 21:10:14 +00:00
bgr4_byte = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
bgr555 = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
bgr555be = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
bgr565 = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
bgr565be = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
bgr8 = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
bgra = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrap, a=unknown, ca=unknown, ca_f=unknown
draw_bmp: don't make strange decisions on broken iknput csp params This checked params->color.space for being RGB. If the colorspace is unset, this did dumb things because even if the imgfmt was a RGB one, the colorspace was not set to RGB. This actually also happened to the tests. (Short-cutting RGB like this is actually wrong, since RGB could still have strange gamma or primaries, which would warrant a full conversion. So you'd need to check for these other parameters as well. To be fixed later.)
2020-05-12 20:43:29 +00:00
bgra64 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrapf32, a=unknown, ca=unknown, ca_f=unknown
bgra64be = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrapf32, a=unknown, ca=unknown, ca_f=unknown
draw_bmp: rewrite draw_bmp.c is the software blender for subtitles and OSD. It's used by encoding mode (burning subtitles), and some VOs, like vo_drm, vo_x11, vo_xv, and possibly more. This changes the algorithm from upsampling the video to 4:4:4 and then blending to downsampling the OSD and then blending directly to video. This has far-reaching consequences for its internals, and results in an effective rewrite. Since I wanted to avoid un-premultiplying, all blending is done with premultiplied alpha. That's actually the sane thing to do. The old code just didn't do it, because it's very weird in YUV fixed point. Essentially, you'd have to compensate for the chroma centering constant by subtracting src_alpha/255*128. This seemed so hairy (especially with correct rounding and high bit depths involved) that I went for using float. I think it turned out mostly OK, although it's more complex and less maintainable than before. reinit() is certainly a bit too long. While it should be possible to optimize the RGB path more (for example by blending directly instead of doing the stupid float conversion), this is probably slower. vo_xv users probably lose in this, because it takes the slowest path (due to subsampling requirements and using YUV). Why this rewrite? Nobody knows. I simply forgot the reason. But you'll have it anyway. Whether or not this would have required a full rewrite, at least it supports target alpha now (you can for example hard sub transparent PNGs, if you ever wanted to use mpv for this). Remove the check in vf_sub. The new draw_bmp.c is not as reliant on libswscale anymore (mostly uses repack.c now), and osd.c shows an error message on missing support instead now. Formats with chroma subsampling of 4 are not supported, because FFmpeg doesn't provide pixfmt definitions for alpha variants. We could provide those ourselves (relatively trivial), but why bother.
2020-05-09 16:01:07 +00:00
cuda = no
d3d11 = no
d3d11va_vld = no
drm_prime = no
dxva2_vld = no
draw_bmp: add integer blending for 8 bit formats Whatever it's worth. Instead of doing a pretty stupid conversion to float, just blend it directly. This works for most RGB formats that are 8 bits per component or below (the latter because we expand packed fringe RGB formats for simplicity). For higher bit depth RGB this would need extra code.
2020-05-12 21:10:14 +00:00
gbrap = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrap, a=unknown, ca=unknown, ca_f=unknown
draw_bmp: don't make strange decisions on broken iknput csp params This checked params->color.space for being RGB. If the colorspace is unset, this did dumb things because even if the imgfmt was a RGB one, the colorspace was not set to RGB. This actually also happened to the tests. (Short-cutting RGB like this is actually wrong, since RGB could still have strange gamma or primaries, which would warrant a full conversion. So you'd need to check for these other parameters as well. To be fixed later.)
2020-05-12 20:43:29 +00:00
gbrap10 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrapf32, a=unknown, ca=unknown, ca_f=unknown
gbrap10be = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrapf32, a=unknown, ca=unknown, ca_f=unknown
gbrap12 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrapf32, a=unknown, ca=unknown, ca_f=unknown
gbrap12be = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrapf32, a=unknown, ca=unknown, ca_f=unknown
gbrap16 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrapf32, a=unknown, ca=unknown, ca_f=unknown
gbrap16be = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrapf32, a=unknown, ca=unknown, ca_f=unknown
gbrapf32 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrapf32, a=unknown, ca=unknown, ca_f=unknown
gbrapf32be = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrapf32, a=unknown, ca=unknown, ca_f=unknown
draw_bmp: add integer blending for 8 bit formats Whatever it's worth. Instead of doing a pretty stupid conversion to float, just blend it directly. This works for most RGB formats that are 8 bits per component or below (the latter because we expand packed fringe RGB formats for simplicity). For higher bit depth RGB this would need extra code.
2020-05-12 21:10:14 +00:00
gbrp = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
draw_bmp: don't make strange decisions on broken iknput csp params This checked params->color.space for being RGB. If the colorspace is unset, this did dumb things because even if the imgfmt was a RGB one, the colorspace was not set to RGB. This actually also happened to the tests. (Short-cutting RGB like this is actually wrong, since RGB could still have strange gamma or primaries, which would warrant a full conversion. So you'd need to check for these other parameters as well. To be fixed later.)
2020-05-12 20:43:29 +00:00
gbrp1 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
gbrp10 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
gbrp10be = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
gbrp12 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
gbrp12be = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
gbrp14 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
gbrp14be = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
gbrp16 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
gbrp16be = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
gbrp2 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
gbrp3 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
gbrp4 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
gbrp5 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
gbrp6 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
gbrp9 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
gbrp9be = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
gbrpf32 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
gbrpf32be = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
draw_bmp: rewrite draw_bmp.c is the software blender for subtitles and OSD. It's used by encoding mode (burning subtitles), and some VOs, like vo_drm, vo_x11, vo_xv, and possibly more. This changes the algorithm from upsampling the video to 4:4:4 and then blending to downsampling the OSD and then blending directly to video. This has far-reaching consequences for its internals, and results in an effective rewrite. Since I wanted to avoid un-premultiplying, all blending is done with premultiplied alpha. That's actually the sane thing to do. The old code just didn't do it, because it's very weird in YUV fixed point. Essentially, you'd have to compensate for the chroma centering constant by subtracting src_alpha/255*128. This seemed so hairy (especially with correct rounding and high bit depths involved) that I went for using float. I think it turned out mostly OK, although it's more complex and less maintainable than before. reinit() is certainly a bit too long. While it should be possible to optimize the RGB path more (for example by blending directly instead of doing the stupid float conversion), this is probably slower. vo_xv users probably lose in this, because it takes the slowest path (due to subsampling requirements and using YUV). Why this rewrite? Nobody knows. I simply forgot the reason. But you'll have it anyway. Whether or not this would have required a full rewrite, at least it supports target alpha now (you can for example hard sub transparent PNGs, if you ever wanted to use mpv for this). Remove the check in vf_sub. The new draw_bmp.c is not as reliant on libswscale anymore (mostly uses repack.c now), and osd.c shows an error message on missing support instead now. Formats with chroma subsampling of 4 are not supported, because FFmpeg doesn't provide pixfmt definitions for alpha variants. We could provide those ourselves (relatively trivial), but why bother.
2020-05-09 16:01:07 +00:00
gray = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayf32, a=unknown, ca=unknown, ca_f=unknown
gray10 = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayf32, a=unknown, ca=unknown, ca_f=unknown
gray10be = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayf32, a=unknown, ca=unknown, ca_f=unknown
gray12 = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayf32, a=unknown, ca=unknown, ca_f=unknown
gray12be = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayf32, a=unknown, ca=unknown, ca_f=unknown
gray14 = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayf32, a=unknown, ca=unknown, ca_f=unknown
gray14be = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayf32, a=unknown, ca=unknown, ca_f=unknown
gray16 = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayf32, a=unknown, ca=unknown, ca_f=unknown
gray16be = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayf32, a=unknown, ca=unknown, ca_f=unknown
gray9 = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayf32, a=unknown, ca=unknown, ca_f=unknown
gray9be = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayf32, a=unknown, ca=unknown, ca_f=unknown
grayaf32 = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayaf32, a=unknown, ca=unknown, ca_f=unknown
grayf32 = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayf32, a=unknown, ca=unknown, ca_f=unknown
grayf32be = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayf32, a=unknown, ca=unknown, ca_f=unknown
mediacodec = no
mmal = no
monob = align=8:1 ov=yap8 , ov_f=grayaf32, v_f=grayf32, a=unknown, ca=unknown, ca_f=unknown
monow = align=8:1 ov=yap8 , ov_f=grayaf32, v_f=grayf32, a=unknown, ca=unknown, ca_f=unknown
nv12 = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
nv16 = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
nv20 = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
nv20be = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
nv21 = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
nv24 = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuv444pf, a=unknown, ca=unknown, ca_f=unknown
nv42 = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuv444pf, a=unknown, ca=unknown, ca_f=unknown
opencl = no
p010 = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
p010be = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
p016 = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
p016be = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
pal8 = no
qsv = no
draw_bmp: add integer blending for 8 bit formats Whatever it's worth. Instead of doing a pretty stupid conversion to float, just blend it directly. This works for most RGB formats that are 8 bits per component or below (the latter because we expand packed fringe RGB formats for simplicity). For higher bit depth RGB this would need extra code.
2020-05-12 21:10:14 +00:00
rgb0 = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
rgb24 = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
draw_bmp: don't make strange decisions on broken iknput csp params This checked params->color.space for being RGB. If the colorspace is unset, this did dumb things because even if the imgfmt was a RGB one, the colorspace was not set to RGB. This actually also happened to the tests. (Short-cutting RGB like this is actually wrong, since RGB could still have strange gamma or primaries, which would warrant a full conversion. So you'd need to check for these other parameters as well. To be fixed later.)
2020-05-12 20:43:29 +00:00
rgb30 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
draw_bmp: rewrite draw_bmp.c is the software blender for subtitles and OSD. It's used by encoding mode (burning subtitles), and some VOs, like vo_drm, vo_x11, vo_xv, and possibly more. This changes the algorithm from upsampling the video to 4:4:4 and then blending to downsampling the OSD and then blending directly to video. This has far-reaching consequences for its internals, and results in an effective rewrite. Since I wanted to avoid un-premultiplying, all blending is done with premultiplied alpha. That's actually the sane thing to do. The old code just didn't do it, because it's very weird in YUV fixed point. Essentially, you'd have to compensate for the chroma centering constant by subtracting src_alpha/255*128. This seemed so hairy (especially with correct rounding and high bit depths involved) that I went for using float. I think it turned out mostly OK, although it's more complex and less maintainable than before. reinit() is certainly a bit too long. While it should be possible to optimize the RGB path more (for example by blending directly instead of doing the stupid float conversion), this is probably slower. vo_xv users probably lose in this, because it takes the slowest path (due to subsampling requirements and using YUV). Why this rewrite? Nobody knows. I simply forgot the reason. But you'll have it anyway. Whether or not this would have required a full rewrite, at least it supports target alpha now (you can for example hard sub transparent PNGs, if you ever wanted to use mpv for this). Remove the check in vf_sub. The new draw_bmp.c is not as reliant on libswscale anymore (mostly uses repack.c now), and osd.c shows an error message on missing support instead now. Formats with chroma subsampling of 4 are not supported, because FFmpeg doesn't provide pixfmt definitions for alpha variants. We could provide those ourselves (relatively trivial), but why bother.
2020-05-09 16:01:07 +00:00
rgb4 = no
draw_bmp: add integer blending for 8 bit formats Whatever it's worth. Instead of doing a pretty stupid conversion to float, just blend it directly. This works for most RGB formats that are 8 bits per component or below (the latter because we expand packed fringe RGB formats for simplicity). For higher bit depth RGB this would need extra code.
2020-05-12 21:10:14 +00:00
rgb444 = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
rgb444be = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
draw_bmp: don't make strange decisions on broken iknput csp params This checked params->color.space for being RGB. If the colorspace is unset, this did dumb things because even if the imgfmt was a RGB one, the colorspace was not set to RGB. This actually also happened to the tests. (Short-cutting RGB like this is actually wrong, since RGB could still have strange gamma or primaries, which would warrant a full conversion. So you'd need to check for these other parameters as well. To be fixed later.)
2020-05-12 20:43:29 +00:00
rgb48 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
rgb48be = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
draw_bmp: add integer blending for 8 bit formats Whatever it's worth. Instead of doing a pretty stupid conversion to float, just blend it directly. This works for most RGB formats that are 8 bits per component or below (the latter because we expand packed fringe RGB formats for simplicity). For higher bit depth RGB this would need extra code.
2020-05-12 21:10:14 +00:00
rgb4_byte = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
rgb555 = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
rgb555be = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
rgb565 = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
rgb565be = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
rgb8 = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrp, a=unknown, ca=unknown, ca_f=unknown
rgba = align=1:1 ov=unknown, ov_f=gbrap, v_f=gbrap, a=unknown, ca=unknown, ca_f=unknown
draw_bmp: don't make strange decisions on broken iknput csp params This checked params->color.space for being RGB. If the colorspace is unset, this did dumb things because even if the imgfmt was a RGB one, the colorspace was not set to RGB. This actually also happened to the tests. (Short-cutting RGB like this is actually wrong, since RGB could still have strange gamma or primaries, which would warrant a full conversion. So you'd need to check for these other parameters as well. To be fixed later.)
2020-05-12 20:43:29 +00:00
rgba64 = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrapf32, a=unknown, ca=unknown, ca_f=unknown
rgba64be = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrapf32, a=unknown, ca=unknown, ca_f=unknown
draw_bmp: rewrite draw_bmp.c is the software blender for subtitles and OSD. It's used by encoding mode (burning subtitles), and some VOs, like vo_drm, vo_x11, vo_xv, and possibly more. This changes the algorithm from upsampling the video to 4:4:4 and then blending to downsampling the OSD and then blending directly to video. This has far-reaching consequences for its internals, and results in an effective rewrite. Since I wanted to avoid un-premultiplying, all blending is done with premultiplied alpha. That's actually the sane thing to do. The old code just didn't do it, because it's very weird in YUV fixed point. Essentially, you'd have to compensate for the chroma centering constant by subtracting src_alpha/255*128. This seemed so hairy (especially with correct rounding and high bit depths involved) that I went for using float. I think it turned out mostly OK, although it's more complex and less maintainable than before. reinit() is certainly a bit too long. While it should be possible to optimize the RGB path more (for example by blending directly instead of doing the stupid float conversion), this is probably slower. vo_xv users probably lose in this, because it takes the slowest path (due to subsampling requirements and using YUV). Why this rewrite? Nobody knows. I simply forgot the reason. But you'll have it anyway. Whether or not this would have required a full rewrite, at least it supports target alpha now (you can for example hard sub transparent PNGs, if you ever wanted to use mpv for this). Remove the check in vf_sub. The new draw_bmp.c is not as reliant on libswscale anymore (mostly uses repack.c now), and osd.c shows an error message on missing support instead now. Formats with chroma subsampling of 4 are not supported, because FFmpeg doesn't provide pixfmt definitions for alpha variants. We could provide those ourselves (relatively trivial), but why bother.
2020-05-09 16:01:07 +00:00
uyvy422 = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
uyyvyy411 = no
vaapi = no
vaapi_idct = no
vaapi_moco = no
vdpau = no
vdpau_output= no
videotoolbox= no
vulkan = no
test: update to new ffmpeg pixfmts Mainly, X2RGB10BE is added. Add our own unpack test for this format. Also, swscale seems to have added support for GBRPF conversion.
2020-06-17 14:20:37 +00:00
x2rgb10be = align=1:1 ov=unknown, ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
draw_bmp: rewrite draw_bmp.c is the software blender for subtitles and OSD. It's used by encoding mode (burning subtitles), and some VOs, like vo_drm, vo_x11, vo_xv, and possibly more. This changes the algorithm from upsampling the video to 4:4:4 and then blending to downsampling the OSD and then blending directly to video. This has far-reaching consequences for its internals, and results in an effective rewrite. Since I wanted to avoid un-premultiplying, all blending is done with premultiplied alpha. That's actually the sane thing to do. The old code just didn't do it, because it's very weird in YUV fixed point. Essentially, you'd have to compensate for the chroma centering constant by subtracting src_alpha/255*128. This seemed so hairy (especially with correct rounding and high bit depths involved) that I went for using float. I think it turned out mostly OK, although it's more complex and less maintainable than before. reinit() is certainly a bit too long. While it should be possible to optimize the RGB path more (for example by blending directly instead of doing the stupid float conversion), this is probably slower. vo_xv users probably lose in this, because it takes the slowest path (due to subsampling requirements and using YUV). Why this rewrite? Nobody knows. I simply forgot the reason. But you'll have it anyway. Whether or not this would have required a full rewrite, at least it supports target alpha now (you can for example hard sub transparent PNGs, if you ever wanted to use mpv for this). Remove the check in vf_sub. The new draw_bmp.c is not as reliant on libswscale anymore (mostly uses repack.c now), and osd.c shows an error message on missing support instead now. Formats with chroma subsampling of 4 are not supported, because FFmpeg doesn't provide pixfmt definitions for alpha variants. We could provide those ourselves (relatively trivial), but why bother.
2020-05-09 16:01:07 +00:00
xvmc = no
xyz12 = align=1:1 ov=gbrap , ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
xyz12be = align=1:1 ov=gbrap , ov_f=gbrapf32, v_f=gbrpf32, a=unknown, ca=unknown, ca_f=unknown
y1 = no
y210 = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
y210be = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
ya16 = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayaf32, a=unknown, ca=unknown, ca_f=unknown
ya16be = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayaf32, a=unknown, ca=unknown, ca_f=unknown
ya8 = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayaf32, a=unknown, ca=unknown, ca_f=unknown
yap16 = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayaf32, a=unknown, ca=unknown, ca_f=unknown
yap8 = align=1:1 ov=yap8 , ov_f=grayaf32, v_f=grayaf32, a=unknown, ca=unknown, ca_f=unknown
yuv410p = no
yuv410pf = no
yuv411p = no
yuv411pf = no
yuv420p = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
yuv420p10 = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
yuv420p10be = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
yuv420p12 = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
yuv420p12be = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
yuv420p14 = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
yuv420p14be = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
yuv420p16 = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
yuv420p16be = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
yuv420p9 = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
yuv420p9be = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
yuv420pf = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuv420pf, a=gray, ca=gray, ca_f=grayf32
yuv422p = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
yuv422p10 = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
yuv422p10be = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
yuv422p12 = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
yuv422p12be = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
yuv422p14 = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
yuv422p14be = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
yuv422p16 = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
yuv422p16be = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
yuv422p9 = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
yuv422p9be = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
yuv422pf = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
yuv440p = no
yuv440p10 = no
yuv440p10be = no
yuv440p12 = no
yuv440p12be = no
yuv440pf = no
yuv444p = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuv444pf, a=unknown, ca=unknown, ca_f=unknown
yuv444p10 = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuv444pf, a=unknown, ca=unknown, ca_f=unknown
yuv444p10be = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuv444pf, a=unknown, ca=unknown, ca_f=unknown
yuv444p12 = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuv444pf, a=unknown, ca=unknown, ca_f=unknown
yuv444p12be = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuv444pf, a=unknown, ca=unknown, ca_f=unknown
yuv444p14 = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuv444pf, a=unknown, ca=unknown, ca_f=unknown
yuv444p14be = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuv444pf, a=unknown, ca=unknown, ca_f=unknown
yuv444p16 = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuv444pf, a=unknown, ca=unknown, ca_f=unknown
yuv444p16be = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuv444pf, a=unknown, ca=unknown, ca_f=unknown
yuv444p9 = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuv444pf, a=unknown, ca=unknown, ca_f=unknown
yuv444p9be = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuv444pf, a=unknown, ca=unknown, ca_f=unknown
yuv444pf = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuv444pf, a=unknown, ca=unknown, ca_f=unknown
yuva410pf = no
yuva411pf = no
yuva420p = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuva420pf, a=gray, ca=gray, ca_f=grayf32
yuva420p10 = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuva420pf, a=gray, ca=gray, ca_f=grayf32
yuva420p10be= align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuva420pf, a=gray, ca=gray, ca_f=grayf32
yuva420p16 = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuva420pf, a=gray, ca=gray, ca_f=grayf32
yuva420p16be= align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuva420pf, a=gray, ca=gray, ca_f=grayf32
yuva420p9 = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuva420pf, a=gray, ca=gray, ca_f=grayf32
yuva420p9be = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuva420pf, a=gray, ca=gray, ca_f=grayf32
yuva420pf = align=2:2 ov=yuva420p, ov_f=yuva420pf, v_f=yuva420pf, a=gray, ca=gray, ca_f=grayf32
yuva422p = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuva422pf, a=gray, ca=gray, ca_f=grayf32
yuva422p10 = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuva422pf, a=gray, ca=gray, ca_f=grayf32
yuva422p10be= align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuva422pf, a=gray, ca=gray, ca_f=grayf32
yuva422p12 = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuva422pf, a=gray, ca=gray, ca_f=grayf32
yuva422p12be= align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuva422pf, a=gray, ca=gray, ca_f=grayf32
yuva422p16 = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuva422pf, a=gray, ca=gray, ca_f=grayf32
yuva422p16be= align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuva422pf, a=gray, ca=gray, ca_f=grayf32
yuva422p9 = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuva422pf, a=gray, ca=gray, ca_f=grayf32
yuva422p9be = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuva422pf, a=gray, ca=gray, ca_f=grayf32
yuva422pf = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuva422pf, a=gray, ca=gray, ca_f=grayf32
yuva440pf = no
yuva444p = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuva444pf, a=unknown, ca=unknown, ca_f=unknown
yuva444p10 = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuva444pf, a=unknown, ca=unknown, ca_f=unknown
yuva444p10be= align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuva444pf, a=unknown, ca=unknown, ca_f=unknown
yuva444p12 = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuva444pf, a=unknown, ca=unknown, ca_f=unknown
yuva444p12be= align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuva444pf, a=unknown, ca=unknown, ca_f=unknown
yuva444p16 = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuva444pf, a=unknown, ca=unknown, ca_f=unknown
yuva444p16be= align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuva444pf, a=unknown, ca=unknown, ca_f=unknown
yuva444p9 = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuva444pf, a=unknown, ca=unknown, ca_f=unknown
yuva444p9be = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuva444pf, a=unknown, ca=unknown, ca_f=unknown
yuva444pf = align=1:1 ov=yuva444p, ov_f=yuva444pf, v_f=yuva444pf, a=unknown, ca=unknown, ca_f=unknown
yuvj411p = no
yuvj422p = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
yuvj440p = no
yuyv422 = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32
yvyu422 = align=2:1 ov=yuva422p, ov_f=yuva422pf, v_f=yuv422pf, a=gray, ca=gray, ca_f=grayf32