02:58 <icecream95> +MOV.i32 r48, clause_10

04:00 <icecream95> Hmm.. bi_postra_liveness is not covering blocks that can't be reached from the final function

07:58 <icecream95> PIGLIT: {"subtest": {"Call i32_func_void" : "pass"}}

07:58 <icecream95> That's about half of the work towards function call support done, another 90% to go!

08:20 <icecream95> Getting spilling to work correctly is the next thing to do

10:31 <icecream95> Wait, there isn't any need to actually have a stack if there is no recursion, which makes things simpler to implement

18:22 <greenjustin> mps: That's sort of what I ended up doing, was copying device tree information around and then editing the MODULE_DEVICE_TABLE in panfrost_drv.c to match. It probably is possible to fry your device that way though, especially if it's not a GPU that's explicitly supported.

18:23 <greenjustin> mps: In my case, MT8183 was explicitly supported, there was just some skew between the device tree in our custom build of the kernel and upstream that I had to rectify. I have no idea what could happen if you just copy device tree info from a totally different GPU. I would expect not good things?

19:04 <mps> greenjustin_: yesterday I looked with 'dtc arch/arm/boot/dts/exynos5800-peach-pi.dtb' and I see it is already there as 'compatible = "samsung,exynos5420-mali\0arm,mali-t628";'

19:05 <mps> greenjustin_: not sure should it make sense to replace '5420` with '5800'

19:07 <mps> arch/arm/boot/dts/exynos5800.dtsi already have this '#include "exynos5420.dtsi"' where gpu is set

19:08 <macc24> mps: add a &gpu node with status = "okay";

19:08 <macc24> to dts of your device

19:08 <mps> not sure how this could (if it could) be fixed, and tested

19:08 <macc24> similar to this https://github.com/torvalds/linux/blob/master/arch/arm64/boot/dts/rockchip/rk3326-odroid-go2.dts#L283

19:11 <mps> macc24: I see this &tmu_gpu { vtmu-supply = <&ldo10_reg>

19:12 <mps> macc24: so I should add `&tmu_gpu { vtmu-supply = <&ldo10_reg>

19:12 <mps> uhm

19:12 <mps> macc24: so I should add `status = "okay";` after this line?

19:12 <macc24> mps: tmu_gpu is thermal device

19:15 <mps> macc24: hm, I'm lost

19:20 <mps> ahm, I see `status = "disabled"` in gpu node

19:20 <mps> macc24: this is worth to try, thanks

19:30 <mps> macc24: greenjustin_: aieee, panfrost is detected now on boot

19:30 <mps> or module load

19:31 <macc24> mps: i'd give downstream dts a look since the exynos chips tend to be cursed

19:31 <mps> yes

19:31 <macc24> also i wouldn't get my hopes up

19:32 <mps> hmmm, how to force xorg to use it

19:34 <mps> i.e. how to force card1 instead of card0

19:34 <macc24> this is not how it works

19:53 <mps> macc24: it works out-of-the-box on gru-kevin chromebook

22:51 <icecream95> Nice.. mold has 32-bit ARM support now. I might even start doing some Mesa development on speedy again, GNU ld was so slow