16:06 <MoeIcenowy> I am thinking about whether my previous addition of chipMinorFeatures12 in https://github.com/etnaviv/etna_viv/commit/cded6437ff357f1f585fe8f1948c90d06128ebdb for G2D_DEC400EX is proper

16:06 <MoeIcenowy> these new chipMinorFeatures words seems to be just defined by etnaviv

16:06 <MoeIcenowy> and I just copied it from some random converter (https://github.com/gizmo98/hwdb-converter)

16:06 <MoeIcenowy> maybe bool flags instead of new words should be added

16:17 <austriancoder> MoeIcenowy: in Mesa we are using a different approach - hwdb. Oh and these chip features are not copied from a random converter but galcore used them too

16:28 <austriancoder> marex: got the hw.. will have a look

16:37 <marex> austriancoder: mp2 ?

16:40 <MoeIcenowy> austriancoder: I think new galcore uses hwdb too?

16:41 <MoeIcenowy> and I think only a few chipMinorFeatures words are really backed by Vivante

16:41 <MoeIcenowy> others are random allocations (because galcore used hwdb

16:43 <MoeIcenowy> BTW according to my reading of the vivante kernel driver, do they only use PTA table id 0?

16:43 <MoeIcenowy> (on my GC620 r5552 setup, only pta table id 0 works reliably, other ids all fail

16:46 <MoeIcenowy> despite I saw some document says 0x00394 register is "MMU_REG_TABLE_ARRAY_SIZE", writing non-0-or-1 values still do not make other pta table ids work (although the id 0 still work)

16:50 <austriancoder> MoeIcenowy: older glacore used the feature regs only.

16:50 <austriancoder> MoeIcenowy: https://christian-gmeiner.info/2024-04-12-hwdb/

16:51 <MoeIcenowy> austriancoder: yes I read about it

16:51 <MoeIcenowy> but should we introduce new chipMinorFeatures?

16:51 <MoeIcenowy> or just keep the current ones here and start to introduce bool flags?

16:52 <MoeIcenowy> I think chipMinorFeatures12 isn't used by Vivante, but just assigned by gizmo98 ?

16:56 <MoeIcenowy> I say gc_feature_database.h is just copied to mesa, but I think copying it to the Linux kernel isn't acceptable, right?

16:56 <MoeIcenowy> s/say/see/

17:00 <austriancoder> MoeIcenowy: where do you need the new chipfeature/flag?

17:08 <MoeIcenowy> austriancoder: GC620 r5552 on TH1520, G2D_DEC400EX

17:10 <MoeIcenowy> https://github.com/revyos/thead-kernel/blob/b9cf70c75d2b7482195a94e754d59f8cfc9dda2c/drivers/gpu-viv/hal/kernel/arch/gc_hal_kernel_hardware.c#L3011

17:10 <MoeIcenowy> the main issue here seems to be it's related to a HW quirk of GC620 r5552 https://github.com/revyos/thead-kernel/blob/b9cf70c75d2b7482195a94e754d59f8cfc9dda2c/drivers/gpu-viv/hal/kernel/arch/gc_hal_kernel_hardware.c#L12289

17:22 <MoeIcenowy> well temporarily this could be just connected to GC620 r5552, for the quirk workaround

18:09 <MoeIcenowy> well for the page table id 0 only, I tried to set the first page table id to 16, and it still works once (and fail when trying to switch to page table 17