- use std::optional instead of std::unique_ptr for the Antialias (FXAA, etc) passes to avoid the extra deref
- use a pattern for deferencing the IR pointer chasing loop as suggested on the intel optimization manual
- this also removes std::vector<> overhead by using boost::container::small_vector<> (not a silver bullet but in the case of this function reduces access times)
Signed-off-by: lizzie <lizzie@eden-emu.dev>
Co-authored-by: Caio Oliveira <caiooliveirafarias0@gmail.com>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2565
Reviewed-by: Maufeat <sahyno1996@gmail.com>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Co-authored-by: lizzie <lizzie@eden-emu.dev>
Co-committed-by: lizzie <lizzie@eden-emu.dev>
The old style `add_test` invocation is not target aware which means that
trying to run the test via `ninja test` results in:
Could not find executable dynarmic_tests
Signed-off-by: Marcin Serwin <marcin@serwin.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3192
Co-authored-by: Marcin Serwin <marcin@serwin.dev>
Co-committed-by: Marcin Serwin <marcin@serwin.dev>
Clang seems to complain that fmt doesn't have a member named join so, it fails to build.
Including fmt/ranges.h fixes that, and lets eden be buildable again.
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3188
Reviewed-by: Caio Oliveira <caiooliveirafarias0@gmail.com>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Co-authored-by: SDK Chan <sdkchan@eden-emu.dev>
Co-committed-by: SDK Chan <sdkchan@eden-emu.dev>
From my tests this decreases JIT latency twofold, may be placebo.
saving reg_alloc while having it readily available is certainly a very interesting choice... afterall saving it onto %rdi is way more cheap isn't it? :)
Please test any performance rgressions, I got +20 FPS on Rain World (unlocked) off this change alone
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3150
Reviewed-by: crueter <crueter@eden-emu.dev>
Reviewed-by: Caio Oliveira <caiooliveirafarias0@gmail.com>
Co-authored-by: lizzie <lizzie@eden-emu.dev>
Co-committed-by: lizzie <lizzie@eden-emu.dev>
Successor to that old MoltenVK PR. Does a lot of cleanups within root CMakeLists.txt, hands over MoltenVK and VulkanValidationLayers to CPMUtil, and separates out common operations into my modules.
Hopefully reduces the monstrosity that is root CMakeLists.txt. Please test:
- builds on all platforms
- VulkanValidationLayers
Signed-off-by: crueter <crueter@eden-emu.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3126
Reviewed-by: Lizzie <lizzie@eden-emu.dev>
Reviewed-by: CamilleLaVey <camillelavey99@gmail.com>
So when libc starts it has to start at an entry point located into crt0, now most OSes will do "enough" setup to allow mprotect() and mmap() to be called in static ctors (remember they're called BEFORE main)
By some stupid miracle, NetBSD doesn't; this means that using those functions on NetBSD will result in spurious results
The reason why is still unknown to me, but this is also combined with the fact that allocating a big chunk of memory for the JIT will make NetBSD refuse to mprotect()/mmap() it in low memory situations (even when space is available); so I take the same approach as with solaris
Also I now make it so fastmem handlers are NOT registered for OSes that disabled fastmem, this is because they pollute sigsegv and makes debugging stupidier
Signed-off-by: lizzie lizzie@eden-emu.dev
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3092
Reviewed-by: Caio Oliveira <caiooliveirafarias0@gmail.com>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Co-authored-by: lizzie <lizzie@eden-emu.dev>
Co-committed-by: lizzie <lizzie@eden-emu.dev>
Should result in very marginally small performance gains. Basically removes the deref of vtable for EmitX64 on lto builds, so in THEORY it should be better than having to defer w.r.t all terminal handlers.
aka. we just like, inline them in one big function and keep CPU away from deference hell.
Signed-off-by: lizzie lizzie@eden-emu.dev
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3033
Reviewed-by: Caio Oliveira <caiooliveirafarias0@gmail.com>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Co-authored-by: lizzie <lizzie@eden-emu.dev>
Co-committed-by: lizzie <lizzie@eden-emu.dev>
Eventually we shall use std::unreachable() ONCE all platforms support it :)
Signed-off-by: lizzie <lizzie@eden-emu.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2889
Reviewed-by: crueter <crueter@eden-emu.dev>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Reviewed-by: Caio Oliveira <caiooliveirafarias0@gmail.com>
Co-authored-by: lizzie <lizzie@eden-emu.dev>
Co-committed-by: lizzie <lizzie@eden-emu.dev>
Uh - the other pr that had "Fix dragonfly" got commits lost and I didn't notice... oops; cherry picked them back through :)
Signed-off-by: lizzie <lizzie@eden-emu.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2860
Reviewed-by: crueter <crueter@eden-emu.dev>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Co-authored-by: lizzie <lizzie@eden-emu.dev>
Co-committed-by: lizzie <lizzie@eden-emu.dev>
- replace instances of ASSERT() with those where UNREACHABLE() should be used instead
- debuggers exist for a reason, you can't just debug an issue in dynarmic with just printing fancy text... you need to inspect values and alldat - while yes the asserts are "useful"; there is this beautiful thing called backtraces
- this will indirectly speedup the main decoder loop because of the added UNREACHABLE()
- this also removes a bunch of macros that were redundant
- the weird trick of [&](){}() is really funky, just do what everyone has done for the past 30 years and use a `do { <thing> } while(0)` :)
I may or may not have missed one assert or messed up my regex substitutions...
Signed-off-by: lizzie <lizzie@eden-emu.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2890
Reviewed-by: Shinmegumi <shinmegumi@eden-emu.dev>
Reviewed-by: Caio Oliveira <caiooliveirafarias0@gmail.com>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Co-authored-by: lizzie <lizzie@eden-emu.dev>
Co-committed-by: lizzie <lizzie@eden-emu.dev>
- std::bit_cast already on libc++ (see also #2774)
- type trait is a 5-liner for u8/u16/u32/u64
- UNREACHABLE(); can just be converted into __builtin_unreachable() (TODO: eventually phase out to std::unreachable() once all platforms support it)\
The templatery parametrization is creating many copies of the same function, polluting i-cache and just bloating code size for minimal savings (and in fact could be negative net savings because of the bloated code size) - this is mostly in preparation to a partial removal of template<size_t XXX> in functions to remove instancing (think of the function being copied x4 times... because the compiler is forced to inline it...)
Signed-off-by: lizzie lizzie@eden-emu.dev
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2775
Reviewed-by: crueter <crueter@eden-emu.dev>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Co-authored-by: lizzie <lizzie@eden-emu.dev>
Co-committed-by: lizzie <lizzie@eden-emu.dev>
Long story short, each of these LUT generates a function that is only ran once, spam the entire code with bunch of little "specialized templates" (remember, N * M * L) and basically bloat codesize for no good reason
Additionally it tries to outsmart the compiler which it's not always a good idea - herein I replace most (except 1) of those "helper" LUTs with something the compiler can actually work satisfiably
Happy reminder of the sheer amount of functions spammed:
mcl::list thing generates a number from 0 to 63 for (fsize 64), add the 0-31 for fsize 32, 0-15 for fsize 16 and you got around... 64+32+16 = 112 functions for a single parameter
Now include in the equation rounding_mode, which is about, what, 6? so 112*6 = 672
Yeah
Signed-off-by: lizzie <lizzie@eden-emu.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2801
Reviewed-by: crueter <crueter@eden-emu.dev>
Reviewed-by: Shinmegumi <shinmegumi@eden-emu.dev>
Co-authored-by: lizzie <lizzie@eden-emu.dev>
Co-committed-by: lizzie <lizzie@eden-emu.dev>
Still had the issues with libusb, but that should get solved with the other PRs anyways
Signed-off-by: lizzie <lizzie@eden-emu.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2805
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Reviewed-by: crueter <crueter@eden-emu.dev>
Co-authored-by: lizzie <lizzie@eden-emu.dev>
Co-committed-by: lizzie <lizzie@eden-emu.dev>
updates CPMUtil to
[`8f9b39fc98`](8f9b39fc98)
Makes the tooling/docs better and enforces usage of tar.gz everywhere as
opposed to zips
Signed-off-by: crueter <crueter@eden-emu.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2790
Reviewed-by: CamilleLaVey <camillelavey99@gmail.com>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Uses tags for a bunch of deps that can use them
Also adds a bunmch of scripts to tools/cpm, notably for checking hashes
and checking for updates.
TODO for the future:
- CI target to check hashes
- Weekly CI to check for updates
Need to get that other CI runner up
additional stuff
- Ports gentoo fixes
- makes solaris work (TODO: sdl2)
- way better docs
- properly separates CPMUtil as a standalone project
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2666
Reviewed-by: Lizzie <lizzie@eden-emu.dev>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Most programs are well behaved and don't cause internal IR issues. Hence, verification can be safely skipped.
Signed-off-by: lizzie <lizzie@eden-emu.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2613
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Co-authored-by: lizzie <lizzie@eden-emu.dev>
Co-committed-by: lizzie <lizzie@eden-emu.dev>
Fixes issues building on aarch64 linux with a slightly outdated system qt; also fixes linker selection process
Signed-off-by: lizzie <lizzie@eden-emu.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2612
Reviewed-by: crueter <crueter@eden-emu.dev>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Co-authored-by: lizzie <lizzie@eden-emu.dev>
Co-committed-by: lizzie <lizzie@eden-emu.dev>
ankerl::unordered_dense::map will provide better lookup times, theoretically.
fastmem trap handler usually had mutex contention for non-linkable series of blocks (block linking failures); so just use shared_mutex
Signed-off-by: lizzie <lizzie@eden-emu.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/320
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Reviewed-by: crueter <crueter@eden-emu.dev>
Co-authored-by: lizzie <lizzie@eden-emu.dev>
Co-committed-by: lizzie <lizzie@eden-emu.dev>
