ChanServ changed the topic of #etnaviv to: #etnaviv - the home of the reverse-engineered Vivante GPU driver - Logs https://oftc.irclog.whitequark.org/etnaviv
Surkow|laptop has quit [Ping timeout: 480 seconds]
Surkow|laptop has joined #etnaviv
JohnnyonFlame has quit [Read error: Connection reset by peer]
<tomeu> lynxeye: have gone through the ioctls sent by the blob, and looks like it waits for the job to finish, then submits another job with two GL.FLUSH_CACHE commands, and then waits for it to finish. after that, the FDR shows that the output buffer doesn't have any leftover data and is fully populated
<tomeu> so I guess that was it
pcercuei has joined #etnaviv
frieder has joined #etnaviv
<tomeu> lynxeye: actually, the problem was something different
<tomeu> when buffers go above a given size, the HW's weight caching mechanism kicks in, and some values that I thought would disable it was causing the HW to oops
lynxeye has joined #etnaviv
mvlad has joined #etnaviv
Leopold__ has joined #etnaviv
Leopold has quit [Ping timeout: 480 seconds]
frieder has quit [Remote host closed the connection]
lynxeye has quit [Quit: Leaving.]
<dv_> the imx8m plus has the GC7000lite , right?
<dv_> what is the max texture size the viv GPU that is on that SoC supports?
<dv_> I have code here that tries to upload 4024x3032 RGBA data, which results in visual artifacts, while 3840x2160 seems to do OK
<austriancoder> GL_MAX_TEXTURE_SIZE: 8192
<dv_> yeah I saw that, but I was wondering if this is incorrect
<dv_> but I guess the rather strange size is more likely to be the problem
<austriancoder> dv_: thats what the blob reports
<austriancoder> on that platform
<dv_> hm would not be the first time the blob reports incorrect stuff ..
<dv_> I'll try slightly different sizes
<dv_> oh, and is it a gc7000lite or gc7000ul on that SoC?
<dv_> I can't access the blob output atm :/
<dv_> ah, now i can. it is a UL
<dv_> indeed ... width 4032 instead of 4024 produces no artifacts, so a width / stride that is an integer multiple of 16 pixels is needed
<dv_> although perhaps this is a gstgl problem
<austriancoder> dv_: I think the reported size is fine. Machine model: Boundary Devices i.MX8MQ Nitrogen8M with etnaviv model: GC7000, revision: 6214
<dv_> and, do vivante GPUs use BGRA natively?
<dv_> I remember that some GPUs used BGRA natively, and emulated RGBA by implicitly swizzling the channels
mvlad has quit [Remote host closed the connection]
pcercuei has quit [Quit: dodo]