<ayaka>
it doesn't, it would harm, I could start up
<LiquidAcid>
ayaka, mali is not upstream ;)
<LiquidAcid>
why would someone backport something that mostly works with a very specific kernel version (vendor kernel) anyway
<ayaka>
I though somebody would like to modification those source to make it work with upstream kernel
<ayaka>
and release with the upstream update
<LiquidAcid>
i guess that "somebody" exists, but just liking something is usually not enough :)
<ayaka>
either it is not public nor well known
<LiquidAcid>
or it simply doesn't exist
<ayaka>
it could be
<ayaka>
basically, I want to solve the output speed of decoder
<LiquidAcid>
not sure what that has to do with mali?
<ayaka>
someone complaining that the output of decoder is NV12, which making it necessary to convert to RGB before output
<LiquidAcid>
which decoder are we talking about?
<ayaka>
LiquidAcid, the MFC v5
<javier__>
ayaka: I think the problem is more that you are using videoconvert for color space conversion
<LiquidAcid>
ayaka, but the MFC has nothing to do with mali
<ayaka>
that is right
<ayaka>
here comes two solutions
<javier__>
in theory the Exynos general scaler (gscaler) should be able to convert he NV12/21 format to the RGB supported by the display controller
<LiquidAcid>
ayaka, and v5 afaik can _only_ output nv12
<ayaka>
one using the GLES to draw the output which supports nv12
<LiquidAcid>
iirc that is a hardware limitation
<javier__>
ayaka: so you could either chain the s5p-mfc with exynos-gsc v4l2 m2m or use the exynos-drm gsc support
<ayaka>
the other may for the other platform but not exynos, directly render to video controller support nv12
<ayaka>
javier__, do you mean the fimc? I don't know which is gsc
<LiquidAcid>
ayaka, fimc is only available on exynos4
<LiquidAcid>
ayaka, gscaler is the counterpart on exynos5
<ayaka>
yes I know, you see I use the MFC v5, so it is exynos4
<javier__>
ayaka: right, I thought you were using exynos5, not exynos4
<LiquidAcid>
ayaka, while both share an infrastructure in DRM (IPP), IPP is more or less broken, so using it in its current state is not recommended
<javier__>
ayaka: forgot the mapping between Exynos SoC and MFC IP block versions :)
<ayaka>
I hope the fimc would works fine, last time I test it with gstreamer, I meet a Colorspaces in encoder
<LiquidAcid>
ayaka, you can also use the fimc via the v4l2 subsystem, probably ndufresne can tell you more about this
<javier__>
ayaka: so *I think* you could use s5p-mfc -> exynos4-is -> exynos-drm
<javier__>
but yes, ndufresne has an exynos4 and I know he was using s5p-mfc and the fimc
<ayaka>
javier__, btw, does exynos release the jpeg encoder/decoder example code?
<LiquidAcid>
and until marek fixes up IPP or creates something else, the only viable option is to use v4l2
<javier__>
ayaka: I don't know tbh, probably the state of s5p-jpeg in mainline isn't that great...
<ayaka>
I looked back the mail list, a samsung guy said he would released a example in a hour but it never
<javier__>
ayaka: do you have a link?
<LiquidAcid>
ayaka, have you search the tizen repos=
<LiquidAcid>
?
<LiquidAcid>
ayaka, libcamsrcjpegenc-hw-e3250 could be related
<ayaka>
LiquidAcid, I see
<javier__>
ayaka: probably worth answering to the thread you mentioned and ask for the example. If you give me the email address, I can also ping him internally on Monday and let you know
<ayaka>
javier__, I fogot it, finding
<LiquidAcid>
javier__, which division are you again?