00:49 <clever> lets say that the core-local irq 7 and 9 are both pending and enabled, and i unmask irq's, which one does the standard say the cpu should service first?

00:50 <jrtc27> 7

00:51 <jrtc27> privileged spec assigns a static order on the interrupt identities defined by the spec

00:51 <clever> is it more complex then just closer to 0 being higher priority?

00:51 <jrtc27> yes

00:51 <jrtc27> external, then software, then timer

00:51 <jrtc27> machine before supervisor

00:51 <clever> ah

00:52 <jrtc27> see the mip/mie section in the priv spec

00:52 <jrtc27> towards the end

00:52 <clever> so i cant just find the first set bit, and go with that

00:55 <clever> jrtc27: ah yep, i see, it says that the priority is fixed for simplicity

00:56 <smaeul> see also section 5.1 of the AIA spec

01:01 <clever> it also looks like 7 and 9 is a bit mixed up

01:01 <clever> 7 is machine timer, while 9 is supervisor external

01:01 <clever> i think i want 7 and 11 instead

01:14 <sorear> it's unclear how much license implementations have to have nonzero latency for interrupts, make states where an interrupt could be taken but hasn't yet visible, etc

01:22 <clever> jrtc27: https://github.com/cleverca22/mini-rv32ima/blob/nix/mini-rv32ima/mini-rv32ima.h#L154-L174 i think that will do it

01:25 <clever> software interrupts not implemented currently

01:28 <clever> i'm not even sure supervisor mode is implemented

01:41 <sorear> > Is "incomplete" in that it didn't implement the tons of the spec that Linux doesn't (and you shouldn't) use.

01:58 <clever> jrtc27: oh, i notice a minor problem in terms of how irq's are handled, the emulator only checks the irq state after executing 1024 opcodes

09:43 <drmpeg> Ubuntu 24.04 works very nicely on the VF2.