Tony Wasserka
Tony Wasserka
I've dropped the first two commits in all my local trees and tested with the Fedora approach (i.e. 4230 + external toolchain file) for a while. This is a perfectly...
I'm -0 on the latest iteration of this PR. fextl was explicitly added to deal with FEXLoader's esoteric allocator requirements (i.e. keeping allocations within the guest-accessible 32-bit address space). This...
Closing for now - let's pick this up again once `Externals` is slimmed down more generally and/or `std::format` is more feature-complete.
> perhaps emitting all branches as indirect if the estimated code size is beyond some threshold Why can't we check the Immediate branch offset itself to use indirect branches only...
> Yeah for backwards branches, bloating every branch with a nop for a case that basically never happens outside of this single game seems like a bad idea and definitely...
Very interesting, actually this instruction should already be handled by expanding it to ARM encoding: https://github.com/mikage-emu/mikage-dev/blob/main/source/arm/thumb.cpp#L211-L219 Any chance you could identify where the instruction is thrown from? gdb's `catch throw`...
> I had a look and I get the following backtrace, which seems to point to this function: You might need a Debug build for proper backtraces. Actually I realized...
This now breaks Celeste startup for me (likely among some other issues, like sporadic failure of Steam startup). a378f9313bc9ff4480b74c97e86fd3389490237f works fine (launching the game in
How difficult would it be to add a central registry for blocks pending compilation that other threads can wait on? While rebasing #4479 on this, I realized the huge number...
I found a way to reduce the lock contention in 4479 even further, so my previous comment isn't critical as originally observed. Once the two PRs are merged, a central...