Docs for the hardware/perf counters?

[bridgman]

There's no such thing as a GPU utilization % per se. Perf counters are much lower level (e.g., number of quads passing through some interface or cache hit rates in some cache). Fences are actually a better metric since they give you an idea of how much work is scheduled to execute rather than a snapshot of what various blocks are doing.

Right. Adjusting power management based on the work that is going to be done always seemed a lot smarter than adjusting it based on what you just finished doing

[curaga]

Ah, I was afraid it'd be that. To test any kernel modifications I'd need to reboot my production box, not exactly making things easy. Can't really test radeon changes in a Qemu VM.

[curaga]

Well I now have things properly reading via /dev/mem Now to the more scary kernel part. I must say, the minute+ compile-test times are not exactly inviting with the reboots and such.

[bridgman]

[curaga]

That would apply if I were getting paid to hack the kernel