mirrors/gstreamer
mirrors/gstreamer
1
1
Fork 0
mirror of https://gitlab.freedesktop.org/gstreamer/gstreamer.git synced 2024-11-24 10:41:04 +00:00
Code Issues 1 Projects Releases Wiki Activity
gstreamer/gst/compositor/compositororc.orc

366 lines
7.6 KiB
Text
Raw Normal View History

compositor: Add a new compositor based on the new GstVideoAggregator base class It is a replacement for videomixer with a similare API Co-Authored by: Thibault Saunier <tsaunier@gnome.org> https://bugzilla.gnome.org/show_bug.cgi?id=731919
2014-05-22 17:46:02 +00:00
.function compositor_orc_splat_u32
.dest 4 d1 guint32
.param 4 p1 guint32
copyl d1, p1
.function compositor_orc_memcpy_u32
.dest 4 d1 guint32
.source 4 s1 guint32
copyl d1, s1
.function compositor_orc_blend_u8
.flags 2d
.dest 1 d1 guint8
.source 1 s1 guint8
.param 2 p1
.temp 2 t1
.temp 2 t2
compositor: Revert most of previous patch. The calculation doesn't produce the same results. Keep just the change to divide alpha by 255 instead of 256, for slightly better accuracy
2015-03-15 16:53:33 +00:00
.const 1 c1 8
compositor: Add a new compositor based on the new GstVideoAggregator base class It is a replacement for videomixer with a similare API Co-Authored by: Thibault Saunier <tsaunier@gnome.org> https://bugzilla.gnome.org/show_bug.cgi?id=731919
2014-05-22 17:46:02 +00:00
convubw t1, d1
convubw t2, s1
subw t2, t2, t1
mullw t2, t2, p1
shlw t1, t1, c1
addw t2, t1, t2
shruw t2, t2, c1
convsuswb d1, t2
.function compositor_orc_blend_argb
.flags 2d
.dest 4 d guint8
.source 4 s guint8
.param 2 alpha
.temp 4 t
.temp 2 tw
.temp 1 tb
.temp 4 a
.temp 8 d_wide
.temp 8 s_wide
.temp 8 a_wide
.const 4 a_alpha 0x000000ff
loadl t, s
convlw tw, t
convwb tb, tw
splatbl a, tb
x4 convubw a_wide, a
x4 mullw a_wide, a_wide, alpha
compositor: Fix blending functions Correctly calculate alpha in a few places by dividing by 255, not 256. Fix the argb and bgra blending functions to avoid an off-by-one error in the calculations, so painting with alpha = 0xff doesn't ever bleed through from behind
2015-03-15 13:22:14 +00:00
x4 div255w a_wide, a_wide
compositor: Add a new compositor based on the new GstVideoAggregator base class It is a replacement for videomixer with a similare API Co-Authored by: Thibault Saunier <tsaunier@gnome.org> https://bugzilla.gnome.org/show_bug.cgi?id=731919
2014-05-22 17:46:02 +00:00
x4 convubw s_wide, t
compositor: Revert most of previous patch. The calculation doesn't produce the same results. Keep just the change to divide alpha by 255 instead of 256, for slightly better accuracy
2015-03-15 16:53:33 +00:00
loadl t, d
x4 convubw d_wide, t
x4 subw s_wide, s_wide, d_wide
compositor: Add a new compositor based on the new GstVideoAggregator base class It is a replacement for videomixer with a similare API Co-Authored by: Thibault Saunier <tsaunier@gnome.org> https://bugzilla.gnome.org/show_bug.cgi?id=731919
2014-05-22 17:46:02 +00:00
x4 mullw s_wide, s_wide, a_wide
compositor: Fix blending functions Correctly calculate alpha in a few places by dividing by 255, not 256. Fix the argb and bgra blending functions to avoid an off-by-one error in the calculations, so painting with alpha = 0xff doesn't ever bleed through from behind
2015-03-15 13:22:14 +00:00
compositor: Add a new compositor based on the new GstVideoAggregator base class It is a replacement for videomixer with a similare API Co-Authored by: Thibault Saunier <tsaunier@gnome.org> https://bugzilla.gnome.org/show_bug.cgi?id=731919
2014-05-22 17:46:02 +00:00
x4 div255w s_wide, s_wide
x4 addw d_wide, d_wide, s_wide
x4 convwb t, d_wide
orl t, t, a_alpha
storel d, t
.function compositor_orc_blend_bgra
.flags 2d
.dest 4 d guint8
.source 4 s guint8
.param 2 alpha
.temp 4 t
.temp 4 t2
.temp 2 tw
.temp 1 tb
.temp 4 a
.temp 8 d_wide
.temp 8 s_wide
.temp 8 a_wide
.const 4 a_alpha 0xff000000
loadl t, s
shrul t2, t, 24
convlw tw, t2
convwb tb, tw
splatbl a, tb
x4 convubw a_wide, a
x4 mullw a_wide, a_wide, alpha
compositor: Fix blending functions Correctly calculate alpha in a few places by dividing by 255, not 256. Fix the argb and bgra blending functions to avoid an off-by-one error in the calculations, so painting with alpha = 0xff doesn't ever bleed through from behind
2015-03-15 13:22:14 +00:00
x4 div255w a_wide, a_wide
compositor: Add a new compositor based on the new GstVideoAggregator base class It is a replacement for videomixer with a similare API Co-Authored by: Thibault Saunier <tsaunier@gnome.org> https://bugzilla.gnome.org/show_bug.cgi?id=731919
2014-05-22 17:46:02 +00:00
x4 convubw s_wide, t
compositor: Revert most of previous patch. The calculation doesn't produce the same results. Keep just the change to divide alpha by 255 instead of 256, for slightly better accuracy
2015-03-15 16:53:33 +00:00
loadl t, d
x4 convubw d_wide, t
x4 subw s_wide, s_wide, d_wide
compositor: Add a new compositor based on the new GstVideoAggregator base class It is a replacement for videomixer with a similare API Co-Authored by: Thibault Saunier <tsaunier@gnome.org> https://bugzilla.gnome.org/show_bug.cgi?id=731919
2014-05-22 17:46:02 +00:00
x4 mullw s_wide, s_wide, a_wide
x4 div255w s_wide, s_wide
compositor: Fix blending functions Correctly calculate alpha in a few places by dividing by 255, not 256. Fix the argb and bgra blending functions to avoid an off-by-one error in the calculations, so painting with alpha = 0xff doesn't ever bleed through from behind
2015-03-15 13:22:14 +00:00
compositor: Add a new compositor based on the new GstVideoAggregator base class It is a replacement for videomixer with a similare API Co-Authored by: Thibault Saunier <tsaunier@gnome.org> https://bugzilla.gnome.org/show_bug.cgi?id=731919
2014-05-22 17:46:02 +00:00
x4 addw d_wide, d_wide, s_wide
x4 convwb t, d_wide
orl t, t, a_alpha
storel d, t
.function compositor_orc_overlay_argb
.flags 2d
.dest 4 d guint8
.source 4 s guint8
.param 2 alpha
.temp 4 t
.temp 2 tw
.temp 1 tb
.temp 8 alpha_s
.temp 8 alpha_s_inv
.temp 8 alpha_d
.temp 4 a
.temp 8 d_wide
.temp 8 s_wide
.const 4 xfs 0xffffffff
.const 4 a_alpha 0x000000ff
.const 4 a_alpha_inv 0xffffff00
compositor: Fix blending functions Correctly calculate alpha in a few places by dividing by 255, not 256. Fix the argb and bgra blending functions to avoid an off-by-one error in the calculations, so painting with alpha = 0xff doesn't ever bleed through from behind
2015-03-15 13:22:14 +00:00
# calc source alpha as alpha_s = alpha_s * alpha / 255
compositor: Add a new compositor based on the new GstVideoAggregator base class It is a replacement for videomixer with a similare API Co-Authored by: Thibault Saunier <tsaunier@gnome.org> https://bugzilla.gnome.org/show_bug.cgi?id=731919
2014-05-22 17:46:02 +00:00
loadl t, s
convlw tw, t
convwb tb, tw
splatbl a, tb
x4 convubw alpha_s, a
x4 mullw alpha_s, alpha_s, alpha
compositor: Fix blending functions Correctly calculate alpha in a few places by dividing by 255, not 256. Fix the argb and bgra blending functions to avoid an off-by-one error in the calculations, so painting with alpha = 0xff doesn't ever bleed through from behind
2015-03-15 13:22:14 +00:00
x4 div255w alpha_s, alpha_s
compositor: Add a new compositor based on the new GstVideoAggregator base class It is a replacement for videomixer with a similare API Co-Authored by: Thibault Saunier <tsaunier@gnome.org> https://bugzilla.gnome.org/show_bug.cgi?id=731919
2014-05-22 17:46:02 +00:00
x4 convubw s_wide, t
x4 mullw s_wide, s_wide, alpha_s
# calc destination alpha as alpha_d = (255-alpha_s) * alpha_d / 255
loadpl a, xfs
x4 convubw alpha_s_inv, a
x4 subw alpha_s_inv, alpha_s_inv, alpha_s
loadl t, d
convlw tw, t
convwb tb, tw
splatbl a, tb
x4 convubw alpha_d, a
x4 mullw alpha_d, alpha_d, alpha_s_inv
x4 div255w alpha_d, alpha_d
x4 convubw d_wide, t
x4 mullw d_wide, d_wide, alpha_d
# calc final pixel as pix_d = pix_s*alpha_s + pix_d*alpha_d*(255-alpha_s)/255
x4 addw d_wide, d_wide, s_wide
# calc the final destination alpha_d = alpha_s + alpha_d * (255-alpha_s)/255
x4 addw alpha_d, alpha_d, alpha_s
# now normalize the pix_d by the final alpha to make it associative
x4 divluw, d_wide, d_wide, alpha_d
# pack the new alpha into the correct spot
x4 convwb t, d_wide
andl t, t, a_alpha_inv
x4 convwb a, alpha_d
andl a, a, a_alpha
orl t, t, a
storel d, t
compositor: Add support for crossfade blending Crossfading is a bit more complex than just having two pads with the right keyframes as the blending is not exactly the same. The difference is in the way we compute the alpha channel, in the case of crossfading, we have to compute an additive operation between the destination and the source (factored by the alpha property of both the input pad alpha property and the crossfading ratio) basically so that the crossfade result of 2 opaque frames is also fully opaque at any time in the crossfading process, avoid bleeding through the layer blending. Some rationnal can be found in https://phabricator.freedesktop.org/T7773. https://bugzilla.gnome.org/show_bug.cgi?id=784827
2017-07-06 18:26:21 +00:00
.function compositor_orc_overlay_argb_addition
.flags 2d
.dest 4 d guint8
.source 4 s guint8
.param 2 alpha
.temp 4 t
.temp 2 tw
.temp 1 tb
.temp 8 alpha_s
.temp 8 alpha_s_inv
.temp 8 alpha_factor
.temp 8 alpha_d
.temp 4 a
.temp 8 d_wide
.temp 8 s_wide
.const 4 xfs 0xffffffff
.const 4 a_alpha 0x000000ff
.const 4 a_alpha_inv 0xffffff00
# calc source alpha as alpha_s = alpha_s * alpha / 255
loadl t, s
convlw tw, t
convwb tb, tw
splatbl a, tb
x4 convubw alpha_s, a
x4 mullw alpha_s, alpha_s, alpha
x4 div255w alpha_s, alpha_s
x4 convubw s_wide, t
x4 mullw s_wide, s_wide, alpha_s
# calc destination alpha as alpha_factor = (255-alpha_s) * alpha_factor / factor
loadpl a, xfs
x4 convubw alpha_s_inv, a
x4 subw alpha_s_inv, alpha_s_inv, alpha_s
loadl t, d
convlw tw, t
convwb tb, tw
splatbl a, tb
x4 convubw alpha_factor, a
x4 mullw alpha_factor, alpha_factor, alpha_s_inv
x4 div255w alpha_factor, alpha_factor
x4 convubw d_wide, t
x4 mullw d_wide, d_wide, alpha_factor
# calc final pixel as pix_d = pix_s*alpha_s + pix_d*alpha_factor*(255-alpha_s)/255
x4 addw d_wide, d_wide, s_wide
# calc the alpha factor alpha_factor = alpha_s + alpha_factor * (255-alpha_s)/255
x4 addw alpha_factor, alpha_factor, alpha_s
# now normalize the pix_d by the final alpha to make it associative
x4 divluw, d_wide, d_wide, alpha_factor
# calc the final global alpha_d = alpha_d + (alpha_s * (alpha / 255))
loadl t, d
convlw tw, t
convwb tb, tw
splatbl a, tb
x4 convubw alpha_d, a
x4 addw alpha_d, alpha_d, alpha_s
# pack the new alpha into the correct spot
x4 convwb t, d_wide
andl t, t, a_alpha_inv
x4 convwb a, alpha_d
andl a, a, a_alpha
orl t, t, a
storel d, t
compositor: Add a new compositor based on the new GstVideoAggregator base class It is a replacement for videomixer with a similare API Co-Authored by: Thibault Saunier <tsaunier@gnome.org> https://bugzilla.gnome.org/show_bug.cgi?id=731919
2014-05-22 17:46:02 +00:00
.function compositor_orc_overlay_bgra
.flags 2d
.dest 4 d guint8
.source 4 s guint8
.param 2 alpha
.temp 4 t
.temp 4 t2
.temp 2 tw
.temp 1 tb
.temp 8 alpha_s
.temp 8 alpha_s_inv
.temp 8 alpha_d
.temp 4 a
.temp 8 d_wide
.temp 8 s_wide
.const 4 xfs 0xffffffff
.const 4 a_alpha 0xff000000
.const 4 a_alpha_inv 0x00ffffff
compositor: Fix blending functions Correctly calculate alpha in a few places by dividing by 255, not 256. Fix the argb and bgra blending functions to avoid an off-by-one error in the calculations, so painting with alpha = 0xff doesn't ever bleed through from behind
2015-03-15 13:22:14 +00:00
# calc source alpha as alpha_s = alpha_s * alpha / 255
compositor: Add a new compositor based on the new GstVideoAggregator base class It is a replacement for videomixer with a similare API Co-Authored by: Thibault Saunier <tsaunier@gnome.org> https://bugzilla.gnome.org/show_bug.cgi?id=731919
2014-05-22 17:46:02 +00:00
loadl t, s
shrul t2, t, 24
convlw tw, t2
convwb tb, tw
splatbl a, tb
x4 convubw alpha_s, a
x4 mullw alpha_s, alpha_s, alpha
compositor: Fix blending functions Correctly calculate alpha in a few places by dividing by 255, not 256. Fix the argb and bgra blending functions to avoid an off-by-one error in the calculations, so painting with alpha = 0xff doesn't ever bleed through from behind
2015-03-15 13:22:14 +00:00
x4 div255w alpha_s, alpha_s
compositor: Add a new compositor based on the new GstVideoAggregator base class It is a replacement for videomixer with a similare API Co-Authored by: Thibault Saunier <tsaunier@gnome.org> https://bugzilla.gnome.org/show_bug.cgi?id=731919
2014-05-22 17:46:02 +00:00
x4 convubw s_wide, t
x4 mullw s_wide, s_wide, alpha_s
# calc destination alpha as alpha_d = (255-alpha_s) * alpha_d / 255
loadpl a, xfs
x4 convubw alpha_s_inv, a
x4 subw alpha_s_inv, alpha_s_inv, alpha_s
loadl t, d
shrul t2, t, 24
convlw tw, t2
convwb tb, tw
splatbl a, tb
x4 convubw alpha_d, a
x4 mullw alpha_d, alpha_d, alpha_s_inv
x4 div255w alpha_d, alpha_d
x4 convubw d_wide, t
x4 mullw d_wide, d_wide, alpha_d
# calc final pixel as pix_d = pix_s*alpha_s + pix_d*alpha_d*(255-alpha_s)/255
x4 addw d_wide, d_wide, s_wide
# calc the final destination alpha_d = alpha_s + alpha_d * (255-alpha_s)/255
x4 addw alpha_d, alpha_d, alpha_s
# now normalize the pix_d by the final alpha to make it associative
x4 divluw, d_wide, d_wide, alpha_d
# pack the new alpha into the correct spot
x4 convwb t, d_wide
andl t, t, a_alpha_inv
x4 convwb a, alpha_d
andl a, a, a_alpha
orl t, t, a
storel d, t
compositor: Add support for crossfade blending Crossfading is a bit more complex than just having two pads with the right keyframes as the blending is not exactly the same. The difference is in the way we compute the alpha channel, in the case of crossfading, we have to compute an additive operation between the destination and the source (factored by the alpha property of both the input pad alpha property and the crossfading ratio) basically so that the crossfade result of 2 opaque frames is also fully opaque at any time in the crossfading process, avoid bleeding through the layer blending. Some rationnal can be found in https://phabricator.freedesktop.org/T7773. https://bugzilla.gnome.org/show_bug.cgi?id=784827
2017-07-06 18:26:21 +00:00
.function compositor_orc_overlay_bgra_addition
.flags 2d
.dest 4 d guint8
.source 4 s guint8
.param 2 alpha
.temp 4 t
.temp 4 t2
.temp 2 tw
.temp 1 tb
.temp 8 alpha_s
.temp 8 alpha_s_inv
.temp 8 alpha_factor
.temp 8 alpha_d
.temp 4 a
.temp 8 d_wide
.temp 8 s_wide
.const 4 xfs 0xffffffff
.const 4 a_alpha 0xff000000
.const 4 a_alpha_inv 0x00ffffff
# calc source alpha as alpha_s = alpha_s * alpha / 255
loadl t, s
shrul t2, t, 24
convlw tw, t2
convwb tb, tw
splatbl a, tb
x4 convubw alpha_s, a
x4 mullw alpha_s, alpha_s, alpha
x4 div255w alpha_s, alpha_s
x4 convubw s_wide, t
x4 mullw s_wide, s_wide, alpha_s
# calc destination alpha as alpha_factor = (255-alpha_s) * alpha_factor / 255
loadpl a, xfs
x4 convubw alpha_s_inv, a
x4 subw alpha_s_inv, alpha_s_inv, alpha_s
loadl t, d
shrul t2, t, 24
convlw tw, t2
convwb tb, tw
splatbl a, tb
x4 convubw alpha_factor, a
x4 mullw alpha_factor, alpha_factor, alpha_s_inv
x4 div255w alpha_factor, alpha_factor
x4 convubw d_wide, t
x4 mullw d_wide, d_wide, alpha_factor
# calc final pixel as pix_d = pix_s*alpha_s + pix_d*alpha_factor*(255-alpha_s)/255
x4 addw d_wide, d_wide, s_wide
# calc the final destination alpha_factor = alpha_s + alpha_factor * (255-alpha_s)/255
x4 addw alpha_factor, alpha_factor, alpha_s
# now normalize the pix_d by the final alpha to make it associative
x4 divluw, d_wide, d_wide, alpha_factor
# calc the final global alpha_d = alpha_d + (alpha_s * (alpha / 255))
loadl t, d
shrul t2, t, 24
convlw tw, t2
convwb tb, tw
splatbl a, tb
x4 convubw alpha_d, a
x4 addw alpha_d, alpha_d, alpha_s
# pack the new alpha into the correct spot
x4 convwb t, d_wide
andl t, t, a_alpha_inv
x4 convwb a, alpha_d
andl a, a, a_alpha
orl t, t, a
storel d, t
Reference in a new issue Copy permalink