The GPU Accuracy level is now divided into Performance, Balanced and Accurate.
1. Performance prioritizes speed at all costs. It's faster, but it can be unstable and may have some bugs (which is expected).
2. Balanced maintains excellent performance and is safer against bugs and shader corruption.
3. Accurate is the most precise and the most expensive in terms of hardware. Only a few games still need this level to work properly.
The Release Early Fences toggle has also been removed by @PavelBARABANOV, as it's not needed anymore.
Co-authored-by: PavelBARABANOV <pavelbarabanov94@gmail.com>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3129
Reviewed-by: Caio Oliveira <caiooliveirafarias0@gmail.com>
Reviewed-by: Maufeat <sahyno1996@gmail.com>
Co-authored-by: MaranBr <maranbr@outlook.com>
Co-committed-by: MaranBr <maranbr@outlook.com>
NOTES:
regs.window_origin.flip_y MUST flip the y coordinate of any given FragCoord, we don't emulate this, this is the root cause of the error, but I'll just revert for now since it's easier
DON'T MERGE unless it's near 0.0.4 and I (or someone else) hasn't tackled this yet properly
This reverts commit 17fe74ef11.
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3075
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Reviewed-by: CamilleLaVey <camillelavey99@gmail.com>
Co-authored-by: lizzie <lizzie@eden-emu.dev>
Co-committed-by: lizzie <lizzie@eden-emu.dev>
Fixes AMD + Windows because it forces barriers to include the fragment test and color output stages explicitly, ensuring that all render pass writes are visible before later commands. Without it, AMD’s driver sometimes skipped synchronization, causing broken rendering in Final Fantasy Tactics.
PR #3069 also fixes this regression by reverting vk_scheduler.cpp in PR #180.
This PR fixes PR #180 and may be the better solution.
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/3071
Reviewed-by: crueter <crueter@eden-emu.dev>
Reviewed-by: Lizzie <lizzie@eden-emu.dev>
Co-authored-by: John <john@eden-emu.dev>
Co-committed-by: John <john@eden-emu.dev>
When viewport_scale_offset_enabled is disabled, the fallback path
previously assumed a top-left origin for both viewport and scissor.
This caused incorrect positioning or inverted geometry when the GPU
state expected a lower-left origin.
This change:
- Adjusts viewport setup: if window_origin is lower-left, shift Y and
flip height negative to emulate lower-left in Vulkan’s top-left space.
- Updates scissor setup: recalculates Y for lower-left origin and
ensures width/height fall back to 1 if zero, avoiding invalid extents.
This aligns Vulkan’s viewport/scissor behavior with Maxwell state,
fixing rendering issues in paths without scale/offset enabled.
Co-authored-by: MaranBr <maranbr@outlook.com>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/294
Co-authored-by: Shinmegumi <shinmegumi@eden-emu.dev>
Co-committed-by: Shinmegumi <shinmegumi@eden-emu.dev>
this pull should impact ninja gaiden ragebound only! it makes it playable past stage 4-1.
it contains a workaround for missing maxwell_3d's iterated_blend functionality, which fixes several graphics all over the game.
the issue causes transparency enabled blends (mostly lighting fx) to be wrongly blended into destination, turning textures into black frames.
in stage 4-1 there are lighthing layers in the foreground, causing sprites layer to become overlapped by these opaque black frames, including entire screen in a mid boss fight, making it unplayable* (players maneuvered by turning immortal option on and swinging sword all around until defeating it).
also only in stage 4-1 the fix has a short drawback: when you buff up next attack these problematique blends will be drawn back as black frames, but only for a split second, so no big deal.
this workaround was already discovered and available in PR 302, but in an unconventional way for a game specific override, so we did forbidden it. now it uses classic game specific override solution exampled in core.cpp's System::Impl::LoadOverrides method, so now i guess it's worth to merge it and deliver this to players until we harness iterated_blend control.
additionally I've slightly reworked vk_rasterizer.cpp's RasterizerVulkan::UpdateBlending, if (state_tracker.TouchBlendEquations()) {...} session.
it was made in a way that for a single blend, it exhaustly calls 48 (6 x 8) MaxwellToVK redundant functions, and declared a lambda function inside a 8 laps loop.
reworked it so that instead of 48 calls it makes only the necessary 6 calls, and then merely safely copy the result for the other 7 times.
Co-authored-by: Allison Cunha <allisonbzk@gmail.com>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2934
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Reviewed-by: Shinmegumi <shinmegumi@eden-emu.dev>
Reviewed-by: Maufeat <sahyno1996@gmail.com>
Co-authored-by: xbzk <xbzk@eden-emu.dev>
Co-committed-by: xbzk <xbzk@eden-emu.dev>
- Removed forced dynamic state 0 logic
- Restore and update the removal of broken states on certain drivers
- Inside 'vk_rasterizer.cpp' make 'UpdateDynamicStates' only check device
capabilities directly instead of relying on user settings.
- Add a 'Force Unsupported Extensions' toggle that:
"Bypasses all driver workarounds and safety checks.
May cause crashes, graphical glitches, or instability.
Only enable for testing purposes."
Cleans up EDS logic and adds new 'Force Unsupported Extensions' toggle,
Fixes vertex explosions in 'Pokemon: Legends ZA' when EDS is set to 0.
Co-authored-by: JPikachu <jpikachu.eden@gmail.com>
Co-authored-by: crueter <crueter@eden-emu.dev>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2947
Reviewed-by: Lizzie <lizzie@eden-emu.dev>
Co-authored-by: JPikachu <jpikachu@eden-emu.dev>
Co-committed-by: JPikachu <jpikachu@eden-emu.dev>
Create 3 ring buffers which rotates between buffers each frame to avoid GPU/CPU conflicts
BindMappedUniformBuffer first tries to allocate from the ring buffer and falls back to staging pool only if allocation is too large.
Note to testers:- please test the performance since it is primarily a performance optimization and also look for visual bugs.
Co-authored-by: wildcard <wildcard@eden-emu.dev>
Co-authored-by: MaranBr <maranbr@outlook.com>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2698
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Co-authored-by: Shinmegumi <shinmegumi@eden-emu.dev>
Co-committed-by: Shinmegumi <shinmegumi@eden-emu.dev>
The barrier before the CPU-upload copy was using VK_PIPELINE_STAGE_HOST_BIT. Validation rules only allow HOST as the source stage if you’re also specifying host-side access flags; inside a command buffer the GPU isn’t executing “host work,” so pairing that stage with the usual image layout transition is technically invalid.
Switching the source stage to VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT keeps the ordering guarantee we need and satisfies the spec, while the actual host visibility is still handled correctly by the preceding vmaFlushAllocation.
Co-authored-by: Ribbit <ribbit@placeholder.com>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2681
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Reviewed-by: Shinmegumi <shinmegumi@eden-emu.dev>
Co-authored-by: Ribbit <ribbit@eden-emu.dev>
Co-committed-by: Ribbit <ribbit@eden-emu.dev>
Adding the shader read barrier keeps every render/compute/transfer write visible before the image is sampled, so it prevents the “read-before-writes-finish” hazards. Without it you can get random stale frames, flickering post process passes, partially updated HUD textures, and corrupted depth-to-color conversions especially in scenes that render into an offscreen image and immediately feed that image to a shader (reflections, bloom, dynamic resolution, depth visualizers, etc.). This fix makes those R2T chains deterministic again across all Vulkan drivers.
Co-authored-by: Ribbit <ribbit@placeholder.com>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2671
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Reviewed-by: crueter <crueter@eden-emu.dev>
Co-authored-by: Ribbit <ribbit@eden-emu.dev>
Co-committed-by: Ribbit <ribbit@eden-emu.dev>
We copy pixels into a CPU-side staging buffer and then ask the GPU to read from it. On some systems those CPU writes aren’t automatically visible to the GPU unless explicitly flushed, so the GPU can sometimes read stale data. By calling buffer.Flush() immediately after writing, we force those CPU changes to become visible to the device, ensuring the GPU sees the latest frame. However, this is an emulator, so sometimes what spec says may not work cause reasons.
Co-authored-by: Ribbit <ribbit@placeholder.com>
Reviewed-on: https://git.eden-emu.dev/eden-emu/eden/pulls/2624
Reviewed-by: Shinmegumi <shinmegumi@eden-emu.dev>
Reviewed-by: MaranBr <maranbr@eden-emu.dev>
Co-authored-by: Ribbit <ribbit@eden-emu.dev>
Co-committed-by: Ribbit <ribbit@eden-emu.dev>
CamilleLaVey
Gamer64
crueter
lizzie
Maufeat
PavelBARABANOV
MaranBr
Kleidis
John
Shinmegumi
xbzk
Caio Oliveira
JPikachu
Ribbit