rellla changed the topic of #linux-sunxi to: Allwinner/sunxi /development discussion - did you try looking at our wiki? https://linux-sunxi.org - Don't ask to ask. Just ask and wait! - https://github.com/linux-sunxi/ - Logs at http://irclog.whitequark.org/linux-sunxi - *only registered users can talk*
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
Rafael1980 has quit [Ping timeout: 258 seconds]
Rafael1980 has joined #linux-sunxi
dev1990 has quit [Quit: Konversation terminated!]
random_yanek has joined #linux-sunxi
Rafael1980 has quit [Ping timeout: 258 seconds]
Rafael1980 has joined #linux-sunxi
_whitelogger has joined #linux-sunxi
agraf_ has joined #linux-sunxi
agraf has quit [Ping timeout: 250 seconds]
agraf_ is now known as agraf
kaspter has quit [Ping timeout: 250 seconds]
kaspter has joined #linux-sunxi
random_yanek has quit [Ping timeout: 258 seconds]
curlybracket has joined #linux-sunxi
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
curlybracket has quit [Read error: No route to host]
curlybracket has joined #linux-sunxi
Rafael1980 has quit [Quit: Konversation terminated!]
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
curlybracket has quit [Read error: Connection reset by peer]
curlybracket has joined #linux-sunxi
shfil has quit [Quit: Connection closed for inactivity]
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
jaho__ has joined #linux-sunxi
random_yanek has joined #linux-sunxi
jaho_ has joined #linux-sunxi
jaho__ has quit [Ping timeout: 276 seconds]
suprothunderbolt has joined #linux-sunxi
victhor has quit [Ping timeout: 258 seconds]
cnxsoft has joined #linux-sunxi
ndufresne has joined #linux-sunxi
dddddd has quit [Remote host closed the connection]
random_yanek has quit [Ping timeout: 276 seconds]
faisal has quit [Ping timeout: 245 seconds]
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
random_yanek has joined #linux-sunxi
TheSeven has quit [Ping timeout: 252 seconds]
[7] has joined #linux-sunxi
random_yanek has quit [Ping timeout: 255 seconds]
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
ganbold has quit [Ping timeout: 246 seconds]
kaspter has quit [Read error: Connection reset by peer]
ganbold has joined #linux-sunxi
kaspter has joined #linux-sunxi
random_yanek has joined #linux-sunxi
random_yanek has quit [Ping timeout: 245 seconds]
random_yanek has joined #linux-sunxi
jstefanop has joined #linux-sunxi
jstefanop has quit [Ping timeout: 255 seconds]
AneoX has joined #linux-sunxi
AneoX_ has quit [Ping timeout: 245 seconds]
lurchi_ has joined #linux-sunxi
lurchi__ has quit [Ping timeout: 258 seconds]
jernej has joined #linux-sunxi
f0xx has joined #linux-sunxi
Gerwin_J has joined #linux-sunxi
Putti has joined #linux-sunxi
random_yanek has quit [Ping timeout: 276 seconds]
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
airwind has joined #linux-sunxi
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
Putti has quit [Ping timeout: 245 seconds]
random_yanek has joined #linux-sunxi
cmeerw has joined #linux-sunxi
HeavyMetal has quit [Ping timeout: 264 seconds]
reinforce has joined #linux-sunxi
HeavyMetal has joined #linux-sunxi
HeavyMetal has joined #linux-sunxi
clemens3_ has quit [Ping timeout: 245 seconds]
nuuuciano__ has quit [Ping timeout: 250 seconds]
tllim has quit [Read error: Connection reset by peer]
RichardG867_ has joined #linux-sunxi
RichardG867_ has quit [Changing host]
RichardG867_ has joined #linux-sunxi
msimpson has joined #linux-sunxi
RichardG867 has quit [Ping timeout: 245 seconds]
cnxsoft has quit [Quit: cnxsoft]
clemens3_ has joined #linux-sunxi
<suprothunderbolt> I'm a bit confused how to specify clocks to a codec. I see the pll-audio-base defined in code but I'm not sure how that maps to the DTS
cnxsoft has joined #linux-sunxi
<jo0nas> (I don't understand this stuff myself, just passing on)
<suprothunderbolt> doesn't look handy. That looks like the normal device tree
<suprothunderbolt> and it's only the onboard codec. I'd love an example with any I2S codec that isn't onboard. Like a chip from TI, cirrus, AD etc
cnxsoft has quit [Ping timeout: 250 seconds]
f0xx has quit [Ping timeout: 258 seconds]
random_yanek has quit [Ping timeout: 258 seconds]
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
faisal has joined #linux-sunxi
gnufan_home has quit [Quit: Leaving.]
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
megi has joined #linux-sunxi
codekipper has joined #linux-sunxi
random_yanek has joined #linux-sunxi
airwind has quit [Ping timeout: 250 seconds]
airwind has joined #linux-sunxi
<codekipper> suprothunderbolt: I've got some examples here https://github.com/codekipper/linux-sunxi/commits/sunxi-wip-a64
<suprothunderbolt> oh nice :) I'm just looking at your multichannel i2s code
<suprothunderbolt> codekipper, does the TDM stuff work okay? I'm trying to get an 6 in 8 out codec to work
<codekipper> pine64 is i2s master for their DAC. orangepi 2 is setup for slave with the audioinjector.
<codekipper> What codec are you using?,
<suprothunderbolt> TI 3168A
<codekipper> connected to what board?
<suprothunderbolt> sopine
<suprothunderbolt> A64
<codekipper> multi-channel works on the pine for HDMI but that has 4 output connections...not sure about how you would connect to an external codec like that
<suprothunderbolt> I mean using TDM
<suprothunderbolt> I haven't looked through your patch much but I saw mention of slot width etc
<codekipper> older boards (A20) had multiple output pins
<suprothunderbolt> ahh. I've just got it connected to 1 pin in and out.
<suprothunderbolt> Does your code not support TDM?
<codekipper> I would need to look at the specs and signalling diagrams...driver may not support it but the block might
<suprothunderbolt> I'm porting from a AM3358. The datasheet for the A64 suggests it supports TDM fine
<suprothunderbolt> and it's mentioned in the BSP code
<codekipper> I've only had the need to use PCM
<suprothunderbolt> with 2 channels in / out?
alexxy has quit [Ping timeout: 246 seconds]
<suprothunderbolt> I think PCM / TDM are the same thing depending on data sheet description. :) I see there are 8 slots in / out defined in the PCM channel mapping
<suprothunderbolt> codekipper, so you'd assume your multi channel code might need modification to work with a codec over a single pin?
<codekipper> the multi-channel code is currently only used for HDMI using PCM.
<suprothunderbolt> okay :)
<codekipper> all of my codecs are pcm devices.
<codekipper> I also don't see a 3168 codec in mainline....do see a pcm3168a!
popolon has joined #linux-sunxi
<suprothunderbolt> yeah, that's what I meant sorry
<suprothunderbolt> TI PCM3168A
<codekipper> I would have to look at the signalling diagrams and then at the i2s registers to see what is possible...this is new territory for both of us
<suprothunderbolt> yeah, I was planning to port the BSP code but maybe it's easier just to implement it from scratch.
<suprothunderbolt> though I'm still confused about where the clock names come from. I have to specify the MCLK signal to the codec and would like to force it to be enabled,
faisal has quit [Ping timeout: 268 seconds]
<suprothunderbolt> but any examples I can see are using clocks = <&ccu CLK_AC_DIG>; which makes sense, it's an alias for the PLL_AUDIO, but then the clock-names section I'm unsure about
Mangy_Dog has joined #linux-sunxi
<suprothunderbolt> clock-names = "mod"; I'm not sure what this means, as I can't see in the clock code where that name is defined
<DuClare> It's not in the clock code, it's in whatever uses the clock
<suprothunderbolt> oh okay, so that's the output name of the clock?
<DuClare> Just a name. If a driver needs multiple clocks, it'll select the one it needs by name
<suprothunderbolt> clocks is the input and clock-names is what the device expects it to be called
<suprothunderbolt> okay cool
<DuClare> Grep for "mod" under drivers, you'll find lines like mixer->mod_clk = devm_clk_get(dev, "mod");
<DuClare> Or of_get_clk_by_name(node, "mod")
<suprothunderbolt> yep, so the device driver is looking for scki devm_clk_get(dev, "scki"); so I define clock-names = "scki"
<codekipper> arch/arm64/boot/dts/renesas/ulcb-kf.dtsi seems to use this codec
<DuClare> Btw the required clocks should be documented somewhere under Documentation/devicetree/bindings/
<DuClare> (grep for the driver compatible string)
<suprothunderbolt> codekipper: I've got a working board using a TI AM3358 SOC
<codekipper> do you have the schematic for it?
<suprothunderbolt> yeah.
<suprothunderbolt> DuClare: thanks, I didn't realise it was that way around. Makes sense now
<suprothunderbolt> codekipper, that renesas example is defining the TDM slot mapping in a similar way to the AM3358. The Am3358 i2s system is a lot more complex than the sunxi one though, lots of data pins and you need to set the direction the serialisers
<codekipper> I look to see what settings are required to achieve what is described in Figure 52. Audio Data Format: 24-Bit TDM Format
random_yanek has quit [Ping timeout: 250 seconds]
suprothunderbolt has quit [Ping timeout: 245 seconds]
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
selfbg has joined #linux-sunxi
f0xx has joined #linux-sunxi
Rafael1980 has joined #linux-sunxi
Gerwin_J has quit [Read error: Connection reset by peer]
libv has quit [Ping timeout: 276 seconds]
libv has joined #linux-sunxi
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
mpmc has quit [Ping timeout: 276 seconds]
random_yanek has joined #linux-sunxi
mpmc has joined #linux-sunxi
victhor has joined #linux-sunxi
Gerwin_J has joined #linux-sunxi
afaerber has quit [Remote host closed the connection]
yann has joined #linux-sunxi
dddddd has joined #linux-sunxi
afaerber has joined #linux-sunxi
airwind has quit [Ping timeout: 276 seconds]
popolon has quit [Ping timeout: 276 seconds]
lurchi_ has quit [Ping timeout: 255 seconds]
lurchi_ has joined #linux-sunxi
lurchi_ is now known as lurchi__
airwind has joined #linux-sunxi
alexxy has joined #linux-sunxi
ElBarto has quit [Read error: Connection reset by peer]
ElBarto has joined #linux-sunxi
cnxsoft has joined #linux-sunxi
paulliu has joined #linux-sunxi
\\Mr_C\\ has joined #linux-sunxi
victhor has quit [Ping timeout: 252 seconds]
FergusL has quit [Quit: Ping timeout (120 seconds)]
FergusL has joined #linux-sunxi
fkluknav has quit [Remote host closed the connection]
lurchi__ is now known as lurchi_
Gerwin_J has quit [Read error: Connection reset by peer]
fkluknav has joined #linux-sunxi
megi has quit [Ping timeout: 250 seconds]
<libv> Tsvetan: are there known issues with network on the K revision of the lime2?
<libv> it seems to have a different ethernet phy
<libv> which is getting really hot
<libv> jo0nas: this is only part of the explanation
<libv> i will throw bpi u-boot on it again, and see how that fares, i had network before
<jo0nas> correct
<jo0nas> I provide infor for your 2nd message above, not for all of them
<libv> thanks :)
<jo0nas> ...and I tried to make that very clear in my message
<jo0nas> I am aware of others experiencing issues with recent revisions of lime2 - and can provide link to that information too if interested - but notice that you seem to request _official_ info so I only provided you what came officially from Olimex
<libv> the rev. G that i got as well is just happy
<jo0nas> good for you - there is an issue specific for that board as well, but good if you don't experience that
<jo0nas> s/is/is rumored to be/
Gerwin_J has joined #linux-sunxi
<libv> i have 5 Ks here, 2 of which were tested, and are acting the same
<libv> jo0nas: this sort of info should live in the wiki
<jo0nas> yes, agreed - and when I figure out what is what in those rumors I will add it (if noone beats me to it)
<libv> i am still digging as well
<jo0nas> still waiting for more detailed reports than "doesn't boot, didn't have a screen connected but maybe same issue as others talk about here..." from those owning those boards and experiencing those various issues
<libv> # CONFIG_MICREL_PHY is not set
<libv> that could explain a thing or two
<libv> nah, still getting hot
<libv> and no network
<libv> going to the bpi config
<libv> if that makes it magically work, we have a starting point
<libv> before it was an oversight when i copied the sd card from the bpi to the lime2
<libv> to test the bpi lcd on the lime2
<libv> and whatdoyouknow
<libv> network.
<libv> with bpi-m1 uboot
<libv> still getting pretty hot
<libv> CONFIG_GMAC_TX_DELAY=3 in uboot config.
<libv> fixes this
Net147 has quit [Quit: Quit]
megi has joined #linux-sunxi
Net147 has joined #linux-sunxi
<libv> that chip gets pretty uncomfortably hot
<Mangy_Dog> this reminds me for the handheld im designing im going to need to sort out some kind of heat sink for it
<Mangy_Dog> but its going to have to be a custom job
<Mangy_Dog> nothing over the top just a small alu block and a heat pipe
<libv> this is hw that will need to run 24/7 for a long time
<libv> as this is going in the fosdem video boxes
<Mangy_Dog> no idea what that is
<Mangy_Dog> other than it has something to do with video :D
<libv> fosdem is the biggest and best open source conference on the planet
<Mangy_Dog> oohh
<libv> it streamed out 27 tracks in parallel this year
<micken> anarsoul: still trying to find out what happens, but its hard since the guy is in uk without debug cable.
<libv> and it has 54 boxes spread over a university campus in brussels
<libv> we are about to severely improve the setup
<Mangy_Dog> hmm
<libv> if the lime2 itself is stable and reliable that is
<micken> anarsoul: the most odd thing is that the card with u-boot + system works in other 1080p 11"
<libv> we are designing a hdmi input daughterboard for it
<Mangy_Dog> is there a media box type distro for these boards that work fully with netflix and amazon prime? I really need to replace my old crap slow stuttery amazon prime fire stick....
<libv> i am hacking up a cable/adapter for running our existing bpi 3.5" lcds (we have 1/8th of the total production) so we have something to develop and test against while we nail down the design of the board
<libv> Mangy_Dog: no
<libv> this will hopefully manage 720p input
<libv> which is good enough for fosdem
<libv> err, not no to your actual question
afaerber has quit [Quit: Leaving]
<libv> but no, this is not a project that will make your life better
Gerwin_J has quit [Ping timeout: 264 seconds]
<Mangy_Dog> l;ol
<libv> i do think there are linux distros tailored as media centers, some of them should work with some arm devices
<libv> but i have no idea about those
lurchi_ is now known as lurchi__
<KotCzarny> does kodi have netflix plugin?
Andy-D has quit [Ping timeout: 245 seconds]
msimpson has quit [Read error: Connection reset by peer]
markvandenborre has joined #linux-sunxi
<Tsvetan> livb: not if you use our official images :)
<Tsvetan> libv rev.K use KSZ9031 PHY I doubt uboot from other board will work
<jo0nas> libv: CONFIG_GMAC_TX_DELAY=3 was reported in Debian to fix issue on newer board but possibly be a regression for older boards: https://bugs.debian.org/927397
<libv> Tsvetan: are you aware of any differences, or where i could most easily find those differences?
<Tsvetan> this is the first time I see ARM Trust Zone documented
<jo0nas> Tsvetan: that repo says it is even with upstream
<libv> in support for u-boot and the kernel for the newer revision?
<jo0nas> Tsvetan: it is in Debian Buster :-)
<KotCzarny> libv: it has branches
<Tsvetan> libv ftp://staging.olimex.com/Allwinner_Images/A20-OLinuXino/
<jo0nas> the mainline codebase from ARM is in Buster: https://tracker.debian.org/pkg/arm-trusted-firmware
<jo0nas> older (now deprecated!) Allwinner branch is also in Debian: https://tracker.debian.org/pkg/atf-allwinner
<libv> ...
<jo0nas> Tsvetan: great to have working fork of u-boot specifically for Olimex products, but even better to have mainline u-boot which supports Olimex products!
<plaes> I have the "read olime board info from eeprom" somewhere in the end of my todo-list :(
msimpson has joined #linux-sunxi
<libv> plaes: see above
<libv> added to the wiki already
<jo0nas> plaes: did you test that patch also with rev.C boards? See https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=927397
<plaes> I don't have any rev.C
<plaes> I started with rev.G :(
<libv> i will go and test this with rev G in a minute
<libv> ah
<plaes> but yeah.. we need functionality in u-boot to read board revision and then do those "fixups"
<plaes> also, eeprom contains mac address
<libv> we should start with documenting these things.
<libv> which is the easiest and least time consuming actions to take here
* jo0nas nudged bug reporter on irc (in case he missed my bug email)
fkluknav has quit [Remote host closed the connection]
fkluknav has joined #linux-sunxi
<libv> i will try the olimex u-boot after i have verified the changes to the lcd hackcable
<plaes> yeah.. TX_DELAY=4
<jo0nas> ...if you ignore other revisions than G, then yeah
kaspter has quit [Read error: Connection reset by peer]
kaspter has joined #linux-sunxi
airwind has quit [Quit: airwind]
Gerwin_J has joined #linux-sunxi
afaerber has joined #linux-sunxi
nuuuciano__ has joined #linux-sunxi
njoseph has joined #linux-sunxi
msimpson has quit [Read error: Connection reset by peer]
curlybracket has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
curlybracket has joined #linux-sunxi
Moe_Icenowy has joined #linux-sunxi
msimpson has joined #linux-sunxi
random_yanek has quit [Ping timeout: 244 seconds]
shfil has joined #linux-sunxi
MoeIcenowy has quit [Ping timeout: 258 seconds]
Moe_Icenowy is now known as MoeIcenowy
random_yanek has joined #linux-sunxi
random_yanek has quit [Max SendQ exceeded]
selfbg has quit [Remote host closed the connection]
Gerwin_J has quit [Read error: Connection reset by peer]
MoeIcenowy has quit [Quit: ZNC 1.6.5+deb1+deb9u1 - http://znc.in]
MoeIcenowy has joined #linux-sunxi
Gerwin_J has joined #linux-sunxi
kaspter has quit [Read error: Connection reset by peer]
kaspter has joined #linux-sunxi
random_yanek has joined #linux-sunxi
jstefanop has joined #linux-sunxi
a|3x has quit [Ping timeout: 245 seconds]
nuuuciano__ has quit [Ping timeout: 246 seconds]
nuuuciano__ has joined #linux-sunxi
msimpson has quit [Read error: Connection reset by peer]
a|3x has joined #linux-sunxi
msimpson has joined #linux-sunxi
\\Mr-C\\ has joined #linux-sunxi
tllim has joined #linux-sunxi
\\Mr_C\\ has quit [Ping timeout: 258 seconds]
Gerwin_J has quit [Read error: Connection reset by peer]
f0xx has quit [Ping timeout: 276 seconds]
fkluknav has quit [Remote host closed the connection]
fkluknav has joined #linux-sunxi
tl_lim has joined #linux-sunxi
\\Mr-C\\ has quit [Quit: (Read error: Connection reset by beer)]
msimpson has quit [Read error: Connection reset by peer]
Gerwin_J has joined #linux-sunxi
reinforce has quit [Quit: Leaving.]
netlynx has joined #linux-sunxi
Putti has joined #linux-sunxi
\\Mr_C\\ has joined #linux-sunxi
nuuuciano__ has quit [Ping timeout: 258 seconds]
yann has quit [Ping timeout: 246 seconds]
reinforce has joined #linux-sunxi
_whitelogger has joined #linux-sunxi
yann has joined #linux-sunxi
nuuuciano has joined #linux-sunxi
_0x5eb_ has quit [Quit: Goodbye!]
afaerber has quit [Quit: Leaving]
clemens3_ has quit [Ping timeout: 245 seconds]
wigyori has quit [Remote host closed the connection]
Nakaori has joined #linux-sunxi
faisal has joined #linux-sunxi
sunilmohan_ has joined #linux-sunxi
RichardG867_ is now known as RichardG867
chomwitt has joined #linux-sunxi
<jo0nas> sunilmohan_: there's a backlog here: https://freenode.irclog.whitequark.org/linux-sunxi/2019-04-25
<jo0nas> libv, plaes: sunilmohan_ did the investigation and Debian bugreport about lime2 rev.C regression when fixing for rev.G boards
cnxsoft has quit [Remote host closed the connection]
cnxsoft has joined #linux-sunxi
<libv> jo0nas: yes/no
<libv> the thing still gets pretty hot
<libv> i have not tried the olimex u-boot yet
<libv> but i fear it might just be the tx delay
cnxsoft has quit [Remote host closed the connection]
jstefanop has quit [Remote host closed the connection]
jstefanop has joined #linux-sunxi
jstefanop has quit [Ping timeout: 255 seconds]
cnxsoft has joined #linux-sunxi
paulliu has quit [Remote host closed the connection]
wigyori has joined #linux-sunxi
cnxsoft has quit [Remote host closed the connection]
netlynx has quit [Quit: Ex-Chat]
<jo0nas> libv: uhm I think you read me wrong: I wasn't asking a question, but introducing you to sunilmohan_ :-)
<jo0nas> ...but too bad your board is still running too hot :-(
Gerwin_J has quit [Quit: Gerwin_J]
cnxsoft has joined #linux-sunxi
jstefanop has joined #linux-sunxi
jstefanop has quit [Ping timeout: 245 seconds]
cnxsoft has quit [Remote host closed the connection]
cnxsoft has joined #linux-sunxi
clemens3_ has joined #linux-sunxi
afaerber has joined #linux-sunxi
cnxsoft has quit [Remote host closed the connection]
cnxsoft has joined #linux-sunxi
msde has joined #linux-sunxi
NeuroScr has joined #linux-sunxi
msde has quit [Remote host closed the connection]
noblock has joined #linux-sunxi
<noblock> anarsoul: It will work with MIN3
<anarsoul> noblock: hey
<noblock> Hi,
<anarsoul> noblock: I think you can just replace blocknum calculation with your code in current code
<anarsoul> and drop your compute_steps_generic() - it's basically equivalent of current code
<noblock> OK I will try.
<anarsoul> the only difference is that your code rounds up tile width/height before doing the shift
<noblock> And the +1 you removed. Is this an issue?
<anarsoul> (just check the code - it's functional equivalent of yours)
<noblock> I can check with the tables.
<anarsoul> noblock: basically if numblocks <= limit it either picks shift_w == shift_h or depending on which dimension is larger, shift_w = shift_h + 1 or shift_h = shift_w + 1
<anarsoul> noblock: see your look up table from v1 - you'll notice the pattern :)
<noblock> Yes.
<anarsoul> noblock: and that's exactly what yuq's original code does, the only difference is in calculating numblocks. Your code rounds up shifted tile width/height, yuq's code doesn't
<noblock> The implementation I have commited, is more compact.
<anarsoul> anyway, thanks for working on this issue :)
<noblock> numblock calcution is fine, I'm certain of it.
<anarsoul> noblock: yeah, the only part that I don't understand is +1 in maskt
tlwoerner_ has joined #linux-sunxi
<anarsoul> otherwise it looks reasonable
<noblock> It may be a good idea to dump the h3 state.
<noblock> or any other mali-400.
dev1990 has joined #linux-sunxi
<anarsoul> noblock: your code works fine for me with MIN3 in place on A64
yann has quit [Ping timeout: 244 seconds]
<noblock> The code was modified for the 260 issue.
<noblock> Are we certain it covers all combination on the mali-400?
<anarsoul> I tested some common resolutions. If you have a test that checks all the combinations - I can try it on mali-400.
<noblock> My last test was to use the H5 workaround with the blob, and check if some PP errors were generated.
<anarsoul> noblock: well, I don't get any PP errors with missing MIN3, but it misrenders shadow demo
<anarsoul> (shadow is missing in this case)
<noblock> I don't know, if we have something with piglit that check for misrendering.
<anarsoul> I'm not sure if it covers all the possible resolutions
<noblock> We can do a loop.
reinforce has quit [Quit: Leaving.]
<noblock> If it can check for instance that the triangle is well processed. We can just add the loop for all rendering size.
<anarsoul> technically there's only up to 16 cases all permutations of (0-3, 0-3)
gnufan_home has joined #linux-sunxi
<noblock> Yes, but the blob has some condition like width = height, or with an offset and swaped (stepw,steph).
<noblock> Maybe some limitations at the hardware level.
<anarsoul> yeah, looks like max(abs(shift_w - shift_h)) = 1.
<noblock> For now if the common resolution are working, and some texture resolutions like 256x256, 512x512.. are OK. This will be fine.
<noblock> Yes, the algorithm set the lower values for (stepw, steph) with blocknum<=blockmax.
cmeerw has quit [Ping timeout: 264 seconds]
tlwoerner_ has quit [Quit: Leaving]
<noblock> I have an H3, I can do some tests on the mali-400 part.
Putti has quit [Ping timeout: 245 seconds]
<noblock> You are right not all (stepw, steph) values are generated.
lurchi__ has quit [Ping timeout: 258 seconds]
lurchi__ has joined #linux-sunxi
<anarsoul> noblock: as I said earlier, looks like it tries to comply with max(abs(shift_w - shift_h)) = 1
<anarsoul> so difference between shift_w and shift_h is either 0 or +-1
<noblock> Yes, you are right,
<noblock> This is what the last code does, it increments w or h, only one each time.
<anarsoul> yes, and that's what original code does
<noblock> Yes.
<anarsoul> that's why I said your code is essentially equivalent of original code, the only difference is in blocknum calculation
victhor has joined #linux-sunxi
<noblock> To check for blocknum; We have to dump all combinations on mali-400.
<noblock> The calcutions must be right all the time.
<anarsoul> noblock: I believe your is correct. It totally makes sense to round up width/height
<noblock> It must be done, minimum pixel size is 16x16.
<jernej> noblock: I just tested your v3 workaround for blob on H5 and it works better than previous
<noblock> jernej: Good.
<noblock> With the v3, I had no issue with PP errors anymore; The code must cover all cases now.
<anarsoul> noblock: yeah
<anarsoul> noblock: whatever it calculates for 16x16 is fine, we'll get shift_w = shift_h = 0 :)
cnxsoft has quit [Remote host closed the connection]
<noblock> If we look at what lima does, it generate an array of pointer of blocknum size.
cnxsoft has joined #linux-sunxi
<anarsoul> oh, OK
<noblock> So this used internaly by the gpu.
<anarsoul> noblock: btw, I suspect that GP in H5 is 2x slower than in other mali450 due to this maxblock limitation
<anarsoul> if you take a look at https://en.wikipedia.org/wiki/Mali_(GPU)#Implementations you'll see that mali400 can do 55 megatriangles/s, while mali4500 can do 142 megatriangle/s, I believe H5 is somewhere in between
<noblock> The performances are good on the H5. On my board with a heatsink, it works properly at 504MHz.
<anarsoul> noblock: it's probably sufficient for your needs :)
<anarsoul> but I'm surprised that allwinner did such a modification to GPU :)
<noblock> The H5 is faster than the S905W even at 504MHz.
<anarsoul> noblock: probably due to faster memory
<noblock> Maybe.
cnxsoft has quit [Remote host closed the connection]
<anarsoul> also I doubt that you have GP-bound workload for it
<noblock> Indeed.
<anarsoul> but it'll be interesting to benchmark it
<noblock> Checking the frequency is a first step.
cnxsoft has joined #linux-sunxi
cnxsoft has quit [Remote host closed the connection]
cnxsoft has joined #linux-sunxi
clemens3_ has quit [Ping timeout: 244 seconds]
clemens3 has joined #linux-sunxi
NeuroScr has quit [Quit: NeuroScr]
cnxsoft has quit [Remote host closed the connection]
cnxsoft has joined #linux-sunxi
Mangy_Dog has quit [Ping timeout: 245 seconds]
cnxsoft has quit [Ping timeout: 246 seconds]
cnxsoft has joined #linux-sunxi
lurchi__ is now known as lurchi_
trabiator6011 has joined #linux-sunxi
Rafael1980 has quit [Quit: Konversation terminated!]
trabiator601 has quit [Ping timeout: 276 seconds]
dev1990 has quit [Quit: Konversation terminated!]
gnufan_home has quit [Quit: Leaving.]
AneoX_ has joined #linux-sunxi
AneoX has quit [Ping timeout: 276 seconds]