Commit graph

548 commits

Author SHA1 Message Date
kd-11 0a2992839b rsx/gl/vk: Simulate z clipping with selective depth clamp
- The scale offset matrix is fine but on real hardware the z results seem to be independent of near/far clipping distances
-- If depth falls within near/far, clamp depth value to [0,1]
2018-01-19 12:03:57 +03:00
kd-11 71f69d1d48
rsx/overlays: Introduce 'native' HUD UI and implement some common dialogs (#4011) 2018-01-17 19:14:00 +03:00
Jake 7ca2c444cc rsx: Fix depth clipping 2018-01-14 20:50:55 +03:00
kd-11 ee009ec99c rsx: Robustness fixes
- Track last working state and reset to it if RSX starts to desync
-- This is especially useful when running vulkan since the renderer will easily outpace the rest of the system when merely recording draw commands
- Ignore empty sets
-- Mark empty/invalid IB sets as having 0 element counts.
2018-01-02 21:17:56 +03:00
kd-11 d496dbecad rsx: Implement depth clamping 2017-12-31 12:43:40 +03:00
kd-11 688569e5c8 gl: Remove flag test optimization as it does not work properly
- Most texture cache operations will break sampler state and the extra complexity is not worth the very small speed bump
2017-12-31 12:43:40 +03:00
kd-11 320fb5f94c rsx/fp: Fix alphakill 2017-12-22 20:08:14 +03:00
kd-11 4a0c4259f0 c++ is hard
- Remove unnecessary const definitions
2017-12-22 20:08:14 +03:00
kd-11 b1a1c0251f rsx: Implement variable point size 2017-12-18 10:45:37 +03:00
kd-11 6891323c18 rsx: framebuffer textures do not have mipmaps!
- Force mipmap count to 1 if sampling from an RTV/DSV
- TODO: Better wcb flush detection, it should be better to re-upload the texture after it has been dwnloaded if expected mipmaps are > 1
2017-12-18 10:45:37 +03:00
kd-11 7c7cd4153e rsx: Framebuffer setup fixes
- Sometimes square renders are done to surfaces with pitch=64 and re-uploaded with swizzle scanning
-- This setup avoids discarding targets if they are square and pitch == 64
2017-12-18 10:45:37 +03:00
kd-11 6dfe32c6d2 fix linux builds 2017-12-18 10:45:37 +03:00
kd-11 95966a467e rsx: Texture cache fixes
- Handle blit resources in a more consistent way
- TODO: Handle some corner cases (piyotama)
2017-12-18 10:45:37 +03:00
kd-11 ac0022483a rsx: Implement delayed swizzle remap for blit engine resources
- Fixes remap vectors for memory copied via blit engine as it has no context
2017-12-18 10:45:37 +03:00
kd-11 69eb483089 rsx: Framebuffer setup fix
- Stencil flag is meaningless if surface does not support stencil
2017-12-06 12:55:49 +03:00
kd-11 90c2324e47 rsx: Program cache fixes
- Reorganize storage hash vs ucode hash
- Scan for actual fragment program start in case leading NOPed code precedes the actual instructions
-- e.g FEAR2 Demo has over 32k of padding before actual program code that messes up hashes
2017-12-04 18:22:18 +03:00
kd-11 960f26f84e rsx:: fixes for framebuffer setup
- provide a DSV if depth and/or stencil testing is enabled
2017-12-04 18:22:18 +03:00
kd-11 cdd4fd9867 rsx/fp: Explicitly insert global functions.
- Functions such as pack/unpack ops must exist before the shared gather functions are declared
2017-12-04 18:22:18 +03:00
kd-11 3fbc960c44 gl/vk: Better handling of inter-format data copies
- RGBA8->RG16 does not require special instructions so the overlay pass is disabled for OGL
2017-12-01 21:00:50 +03:00
kd-11 bec6c1a939 gl: Implement rgba8 -> fp16 casting pass
-- Is this actually necessary? The two format are binary compatible
2017-12-01 21:00:50 +03:00
kd-11 08b829dc22 rsx: wcb scaling fixes 2017-12-01 21:00:50 +03:00
kd-11 9d27ac359b gl: Minor wcb tweaks 2017-12-01 21:00:50 +03:00
kd-11 da1e97618b rsx: Changes to surface pitch handling
- Zeta pitch is ignored by real HW for some reason
- Monitor ptch value changes as well since they may affect disabled surfaces
- TODO: Verify if MRT pitch is really taken into consideration
2017-12-01 21:00:50 +03:00
kd-11 3350e7ee55 rsx: Fixes for framebuffer setup code 2017-12-01 21:00:50 +03:00
kd-11 75d0ff177d gl: Avoid null deref 2017-12-01 21:00:50 +03:00
kd-11 3bfdcf698d rsx: More tuning for depth/color buffer selection 2017-12-01 21:00:50 +03:00
kd-11 51891039dd rsx: improve depth/color contested memory allocation heuristics
- Needs more tests to prove correct behaviour
2017-12-01 21:00:50 +03:00
kd-11 07fe5f87ae gl: Fix face winding
- Face winding is always calculated based off the top left corner
2017-12-01 21:00:50 +03:00
kd-11 de5a4fe083 rsx: Reimplement depth <-> RGBA reinterpretation code
- Implements proper channel order for fp24-ARGB8 conversion
- Takes swizzle remap into account when reconstructing source bytes
2017-12-01 21:00:50 +03:00
kd-11 5cf2d774f3 fp32 precision on GPUs is embarassing
- Division seems to suffer from drift easily on GPUs due to limited precision
2017-12-01 21:00:50 +03:00
kd-11 ccc0383f75 vulkan: Implement overlay shader passes
- Implements vk::overlay_pass and vk::depth_convert_pass
- Also added a sanity check in RSX core for depth replace shaders
2017-12-01 21:00:50 +03:00
kd-11 680ca1d12a rsx: Zcull refactoring and vulkan implementation 2017-12-01 21:00:50 +03:00
kd-11 30269e5bad rsx/gl: Use strict method of gathering the depth buffer bits
- Performance difference on modern GPUs is negligible
2017-12-01 21:00:50 +03:00
kd-11 07f61db24a rsx: Try to take the context for the surface creation into account 2017-12-01 21:00:50 +03:00
kd-11 63f261a66d rsx: Improve framebuffer check heuristics for contested memory buffers 2017-12-01 21:00:50 +03:00
kd-11 33f3a3e014 rsx: Major fixes
- Handle aliased depth + color target by disabling depth writes. This looks to be the correct way
- Add support for generic passes that cannot be done using general imaging operations. Lays the framework for tons of features and effects
- Implement RGBA->D24D8 casting. Sometimes games will split depth texture into RGBA8 then use the new RGBA8 as a depth texture directly
-- This happens alot in ps3 games and I'm not sure why. Its likely the ps3 did not sample fp values with linear filtering so this is a workaround
-- Only implemented for openGL at the moment
-- Requires a workaround for an AMD driver bug
2017-12-01 21:00:50 +03:00
kd-11 8646f51fa3 rsx: Add more validation to framebuffer setups. Game devs sometimes do crazy things 2017-12-01 21:00:50 +03:00
Zion Nimchuk 3a9ae2df9e silence warnings in RSX stuff 2017-11-30 18:07:19 +03:00
kd-11 be6b5922dd rsx: research native texel byte order on cpu readback (WCB) [WIP] 2017-11-20 15:18:57 +03:00
kd-11 b2a7eee1ec rsx: Bump shader cache ver and fix blit engine crash
- Disables blit operations if the target will have a size of 0 in any dimension
- Bumps shader cache ver to 1.1
2017-11-09 14:39:50 +03:00
kd-11 4e9160104a rsx/vk/gl: Cleanup and refector glsl::getFunctionImpl
- Both backends now generate very similar code
2017-11-08 13:15:34 +03:00
kd-11 242611aa46 gl/qt: Catch segfaults in wglDeleteContext with SEH (AMD windows driver)
- In rare cases the driver derefs a nullptr and dies, taking the emulator with it
- From testing, it seems the vram is indeed freed when this happens so its "safe" to continue
2017-11-08 13:15:34 +03:00
kd-11 3730b9d1da rsx: More fixes
- Support for raster offsets in surface descriptors (looks to be unused)
- Do not tag disabled render targets when using MRT (pitch = 64)
- Add missing notify_surface_changed() call for openGL
2017-11-08 13:15:34 +03:00
kd-11 0d35363ea0 rsx: Minor optimization - dont update sampler state if texture state registers have not changed 2017-11-08 13:15:34 +03:00
kd-11 75504b3f5e rsx: Rework context handling and stop leaking the GL ccontext
- GL contexts are external handles that require manual lifecycle management
2017-11-08 13:15:34 +03:00
kd-11 300a36d3d6 rsx: Fixes for cubemap reconstruction
- Do not abort generation if sides are missing, replace with blank surfaces instead
- Make cubemaps scale with res scaling
2017-11-08 13:15:34 +03:00
kd-11 60c7a508a7 rsx: Refactor create_subresource_view(deferred_subresource&) and implement a subresource cache
- This limits the number of times an image is copied and improves performance
2017-11-08 13:15:34 +03:00
kd-11 1fa18757fc rsx: Implement render-to-cubemap; Also simplify unnormalized samplers [WIP, DELETE SHADER CACHE, VERY SLOW]
- Enables real-time cubemap reflections
- TODO: Vulkan is broke; rsx is very slow with this feature
2017-11-08 13:15:34 +03:00
kd-11 fbb7186e66 rsx/gl: Addendum - Fix fragment shader to consume texture scale parameters 2017-11-08 13:15:34 +03:00
kd-11 0961a43997 rsx: Implement 1D<->2D image type casts 2017-11-08 13:15:34 +03:00