search

ptitSeb / box64

Box64 - Linux Userspace x86_64 Emulator with a twist, targeted at ARM64 Linux devices
https://box86.org
MIT License
3.9k stars 287 forks source link

[LONGAARCH_DYNAREC] Added 00...05 and 28...2D opcodes #1297

Closed factfinding closed 8 months ago

factfinding commented 8 months ago

[LONGAARCH] More renamed arch to LA64 [LONGAARCH_DYNAREC] Added 00...05 and 28...2D opcodes

ksco commented 8 months ago

I think we should try to utilize the lbt extension from the first day.

ksco commented 8 months ago

.. so we don't have to deal with eflags and x87 emulation, it's slow (well, compared to the lbt extension) and painful.

ptitSeb commented 8 months ago

this extension is present on all cpu right?

ksco commented 8 months ago

this extension is present on all cpu right?

Yes, I think so.

factfinding commented 8 months ago

Sorry, I am not an internal employee of Loongson Corporation, so I can only use the Loongarch instruction set that Loongson has already made public to implement dynamic binary. The LBT instruction is not publicly available, so I don't know how to use it. Loongarch and riscv are very similar, so I just imitated riscv to implement dynamic binary.

ksco commented 8 months ago

Me neither ;) but there are some reverse-engineered documents we can refer: https://github.com/jiegec/la-inst/blob/master/LBT.md, https://github.com/loongson-community/loongarch-opcodes/blob/develop/lbt.txt. I'll look into it and see if it's enough for us to use LBT.

ptitSeb commented 8 months ago

As I see it, it's fine to use RiSCV implementation as a 1st draft for the Loongarch dynarec. LBT extension is clearly a plus, but it can come after, as a replacement. And probably not all cases are covert by LBT anyway.