| 11:45 |
amberglint |
hi asciilifeform |
| 11:45 |
amberglint |
I have a question about http://logs.bitdash.io/pest/2023-08-21#1029969 |
| 11:45 |
amberglint |
have you considered cutting the mmu emulation to the bone by hardwiring the address translation 1-to-1? |
| 11:45 |
amberglint |
it's not like you need multiple address spaces inside the emulator anyway as you can simply run it in multiple host processes |
| 11:45 |
amberglint |
afaik Linux can be coaxed to run without a "proper" mmu |
| 11:54 |
amberglint |
here's a recent tutorial, for RISC-V rather than MIPS, but I think it shouldn't be too different: https://github.com/vowstar/k210-linux-nommu |
| |
~ 1 hours 54 minutes ~ |
| 13:48 |
asciilifeform |
http://logs.nosuchlabs.com/log/asciilifeform/2023-08-22#1115181 << considered mmu-less linux, yes. but not had time to experiment further. |
| 13:48 |
dulapbot |
Logged on 2023-08-22 11:45:13 amberglint: have you considered cutting the mmu emulation to the bone by hardwiring the address translation 1-to-1? |
| 13:49 |
asciilifeform |
the problem is, you still need good chunk of addr translation logic for the memory-mapped i/o. |
| 13:50 |
asciilifeform |
subj. |
| 13:51 |
asciilifeform |
but yes would be massively lighter if mmuless. |