09:07 pH5: tomeu: I think the bit_len==8 encoding for zero runs might work like this on imx8mp: https://gitlab.freedesktop.org/pH5/mesa/-/commit/7ff4f5a5bd3a58f4ce83744d200130d92f815980
09:07 pH5: so that's bit_len==0 encoding for run_length_table indices 0 and 1, and bit_len==8 for 2 to 5, allowing table length 6.
09:08 pH5: I have not made test cases that produce a longer table yet.