100% browser CPU usage even though guest OS "idle" about jslm32

[ubercomp]

Yeah... This one is a bummer. Unfortunately, as far as I know (and I might very well be wrong here), the architecture doesn't give me a way to know that the guest is idle. If I were to get serious about this, i.e. build an emulator to run real applications, the 100% CPU issue would of course have to be solved, which means I would probably have to go for Intel or ARM. The reason I did LatticeMico32 was just because I wanted to know if I could and LM32 was the easiest target I could find. Also, I wanted a "dynamic recompilator" instead of just an interpreter.

[ysangkok]

According to Wikipedia, the HLT instruction requires ring 0 access.

As I see it, that leaves us with two options:

• Find the place in the Linux kernel where the HLT instruction would be called and make the emulator wait e.g. 10 ms when the instruction pointer reaches that address.
• Modify the instruction set and use the instruction in Linux. The requires patching Linux, so I guess it's a bad idea. Though it's a bit less hackish in some aspects, especially if it could be done in a backward-compatible way.

[ysangkok]

Closing this since it is unactionable (we can't know whether the guest is idle).