Fixes an issue where TestHarnessRunner was managing to reserve the space
below stack again, resulting in stack growth breaking. Would typically
only show up when using the vixl simulator under gdb for some reason.
This is likely the last bandage on this code before it gets completely
rewritten to be more readable.
ldswpal doesn't overwrite the source register and only reads the bits
required for the sized operation.
Not sure exactly why we were doing a copy here.
Removing it means improving Skyrim's hottest code block, as seen in #3472
This reduces pointless constant usage. For now, it's no net change to
instcountci, but it should make it easier to get wins later. Hopefully.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Doesn't quite match the libc code directly because it uses `[gs:eax]`
with both having the sign bit set and we can't deal with that with ASM
tests. So match the behaviour in a different way.
This is what we'll actually ship (I hope), so that's the config we want to
track long-term. It's also a lot more managable resulting asm.
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
Reimagining of #3355 without any json generators or new concepts.
Fixes some mislabeling of system calls. Some getting inlined when they
shouldn't be, a lot not getting inlined when they can be.
This really cleans up the syscall implementation, all syscalls that can
be passthrough implementations require a very small two line
declaration.
Additionally cleans up a bit of implementation cruft where some
passthrough syscalls were using the glibc syscall handler, and some were
using the glibc implementation. We have had multiple issues in the past
where the glibc implementation does something subtly different than the
raw syscall and breaks things. Now all passthrough handlers do a system
call directly, removing at least one indirection and some ambiguity.
This makes it significantly easier to add new passthrough syscalls as
well. Only need to do a version check and add the three lines per
syscall. Which there are new syscalls incoming that we will want to add.
Tangible improvements:
- Syscalls are lower overhead than ever.
- When I'm adding more syscalls I have less chance of mucking it up.
Doing very little work with a bunch of instructions.
Hottest block in the Windows version of Psychonauts, it's just doing a
matrix swizzle but in the worst possible way.
this gets rid of the awkward non-flag SBB case, which streamlines SBB. while
getting better codegen for the demon opcode (-:
Signed-off-by: Alyssa Rosenzweig <alyssa@rosenzweig.io>
