<tomeu>
lynxeye: well, after two weeks of combing through all the communication between the kernel and the HW, I finally figured out what was going on:
<tomeu>
the configuration of the TP units that userspace was passing to the kernel was wrong, but I didn't realize because of a bug in the struct layout in the tooling
<tomeu>
so it seemed that mesa pwas generating the same configuration as the blob, but it wasn't
<tomeu>
(one bit was 0 instead of 1)
<austriancoder>
\o/
<tomeu>
a bit of a bitter sweet moment, as I'm happy I'm past it, but man, what a stupid waste of time...
<austriancoder>
wasting time happens from time to time when doing RE
<tomeu>
kudos to my son, who kept me awake in the middle of the night with his cough, and thinking about all this I thought of comparing the binaries instead of going through the tooling
pcercuei has joined #etnaviv
<tomeu>
it is comforting though that only once the issue was in the kernel