1. rsx: Rework section synchronization using the new memory mirrors
2. rsx: Tweaks
- Simplify peeking into the current rsx::thread instance.
Use a simple rsx::get_current_renderer instead of asking fxm for the same
- Fix global rsx super memory shm block management
3. rsx: Improve memory validation. test_framebuffer() and
tag_framebuffer() are simplified due to mirror support
4. rsx: Only write back confirmed memory range to avoid overapproximation errors in blit engine
5. rsx: Explicitly mark clobbered flushable sections as dirty to have them
removed
6. rsx: Cumulative fixes
- Reimplement rsx::buffered_section management routines
- blit engine subsections are not hit-tested against confirmed/committed memory range
Not all applications are 'honest' about region bounds, making the real cpu range useless for blit ops
- Reimplements the AMD workaround using an identity buffer to avoid the performance hit of doing multiple glDrawArrays for every single compiled set
- Reimplements first/count allocation using a scratch buffer to reduce allocation overhead when large number of draw calls is used
- Improve dirty state tracking affecting program state
- vk: Refactor out transform constants upload into a separate channel to avoid if possible
transform data uploads are quite expensive
- Introduces a gpu program analyser step to examine shader contents before attempting compilation or cache search
- Avoids detecting shader as being different because of unused textures having state changes
- Adds better program size detection for vertex programs
- Improved vertex program decompiler
- Properly support CAL type instructions
- Support jumping over instructions marked with a termination marker with BRA/CAL class opcodes
- Fix SRC checks and abort
- Fix CC register initialization
- NOTE: Even unused SRC registers have to be valid (usually referencing in.POS)
- Readback does not work at all with float textures on AMD openGL
Driver throws a bogus OUT_OF_MEMORY error regardless of amount of VRAM and system RAM available
- vk: Clear dirty textures before copying 'old contents' in case the old data does not fill the new region
- rsx: Properly decode border color - seems to be in BGRA format
- vk: better approximation of border color to better choose between the presets
- vk: Individually clear color images outside render pass and without scissor
- vk: Fix renderpass selection for clear overlay pass
- vk: Include scissor region when emulating clear mask
NOTES:
- vk: Completely avoid using vkClearXXXXimage - its 'broken' on nvidia drivers
Spec is vague about the function so its not an actual bug
ClearAttachment is clearly defined as bypassing bound state which works correctly
- TODO: Implement memory sampling to simulate loading precleared memory if cell used memset to preinitialize the framebuffer
Autoclear depth to 1|255 and color to 0 is hacky!
- gl/vk: Fix subresource copy/blit
- gl/vk: Fix default_component_map reading
- vk: Reimplement cell readback path and improve software channel decoder
- Properly name the subresource layout field - its in blocks not bytes!
- Implement d24s8 upload from memory correctly
- Do not ignore DEPTH_FLOAT textures - they are depth textures and abide by the depth compare rules
- NOTE: Redirection of 16-bit textures is not implemented yet
Primary:
- Fix SET_SURFACE_CLEAR channel mask - it has been wrong for all these
years! Layout is RGBA not ARGB/BGRA like other registers
Other Fixes:
- vk: Implement subchannel clears using overla pass
- vk: Simplify and clean up state management
- gl: Fix nullptr deref in case of failed subresource copy
- vk/gl: Ignore float buffer clears as hardware seems to do
- Ignore unlocked blit sections [TODO]
- Do not attempt blit on hw if bytesize is unsupported
- gl: Implement typeless memory transfers
Uses pbo to handle type-agnostic memory transfer
- Mainly affected are colormasks and read swizzles
NOTES:
- Writes to G write to the second and fourth component (YW)
- Writes to B write to first and third component (XZ)
- This means the actual format layout is BGBG (RGBA) making RG mapping actually GR
- Clear does not seem to have any intended effect on this format (TLOU)
Fixes the following issues on Tales of Vesperia which requires SRM.
- Blacked out scene after the sleeping dog now renders correctly
- Ghosting effect. The ghosting was most noticeable as a delay between the character rendering and the cell shading around the character. This appears to be gone with this change.
- GL queries share the target binding (not asynchronous!)
- Discard active queries by closing them, leave closed queries alone (nothing to be done for discard op)
- Reimplements render target views used for sampling
- Optimizes access using an encoded control token
- Adds proper encoding for 24-bit textures (DRGB8 -> ORGB/OBGR)
- Adds proper encoding for ABGR textures (ABGR8 -> ARGB8)
- Silence some compiler warnings as well
- TODO: Real texture views for OGL current method is a hack
- Separate TXB from TXL: They are completely different!
- Properly perform TMU emulation in the fragment shader. Implemens SRGB conversion and alphakill at the moment
- Properly perform ROP emulation in the fragment shader. Implements FRAMEBUFFER_SRGB. While support on the chip looks to be incomplete (and wierd), it does work
- Document some more bits in SHADER_CONTROL register
gl/vk: Bump shader cache version
gl/vk: Disable anisotropic override when strict mode enabled as it is proven to alter some games negatively
gl: Clamp buffer view range to not exceed the backing buffer size. Also add assert for the same condition
- Track asynchronous operations in RSX core
- Add read barriers to force pending writes to finish.
Fixes zcull delay flicker in all UE3 titles without forcing hard stall
- Increase zcull latency as all writes should be synchronized now
- ZCULL unit emulation rewritten
- ZCULL reports are now deferred avoiding pipeline stalls
- Minor optimizations; replaced std::mutex with shared_mutex where contention is rare
- Silence unnecessary error message
- Small improvement to out of memory handling for vulkan and slightly bump vertex buffer heap
According to the NV_fragment_program spec, its not feasible to have 16-bit depth wries
NOTE: NV_fragement_program precedes NV_fragment_program2 which is very
close to what RSX consumes. It is hardware from that era afterall
- Forces Bitcast of texture data if input format cannot possibly be the
same as the existing texture format
- rsx: Other minor improvements to texture cache :-
- remove obsolete blit engine incompatibility warning. The texture will be re-uploaded if it is indeed incompatible
- Implement warn_once and err_once to avoid spamming the log with systemic errors
- Track mispredicted flushes
- Reswizzle bitcasted texture data to native layout
TODO: Also needs reshuffle according to input remap vector
- A new step is added between decompilation and pipeline object creation allowing for properties to be updated based on shader contents
- Allos masking off attachment writes that are unmodified in the shader
- gl: Do not call makeCurrent every flip - it is already called in set_current()
- gl: Improve ring buffer behaviour; use sliding window to view buffers larger than maximum viewable hardware range
NV hardware can only view 128M at a time
- gl/vk: Bump transform constant heap size When lots of draw calls are issued, the heap is exhaused very fast (8k per draw)
- gl: Remove CLIENT_STORAGE_BIT from ring buffers. Performance is marginally better without this flag (at least on windows)
- Identify depth textures reaching the gpu via shader_read upload path
- Use correct timestamp counter for opengl
- inline draw_state::test_property because msvc doesnt do it for us
- Fix for texture barriers
- vulkan: Rework texture cache handling of depth surfaces
- Support for scaled depth blit using overlay pass
- Support proper readback of D24S8 in both D32F_S8 and D24U_S8 variants
- Optimize the depth conversion routines with SSE
- vulkan: Replace slow single element copy with std::memcpy
- Check heap status before attempting blit operations
- Bump guard size on upload buffer as well
- Implement flush-always behaviour to partially fix readback from a currently bound fbo
- Without this, only the first read is correct, as more draws are added the results become 'wrong'
- Fixes WCB and cpublit behviour
- Synchronize blit_dst surfaces to avoid data loss when gpu texture scaling is used
- Its still faster in such cases to disable gpu texture scaling but some types cannot be disabled without force cpu blit (e.g framebuffer transfers)
- Memory management tuning
- rsx: on-demand texture cache rescanning for unprotected sections
- rsx: Only framebuffer resources are upscaled
- Do not resize regular blit engine resources
- Lazy initialize readback buffer when using opengl
-- These measures should help minimize vram usage
- For some surfaces, dimensions are passed via the log2 bits rather than surface pitch
-- This is similar to the setup for nv406e and probably means the surfaces are padded and swizzled
- Do not assume texture2D when creating new textures
- Flag invalid texture cache if readonly texture is trampled by fbo memory.
Avoids binding a stale handle to the pipeline and is rare enough that it should not hurt performance
- Do not add unused subroutines in shaders unless necessary
-- makes shaders easier to read and disassembled spir-v has less clutter
- glsl: Replace switch block with lookup table
- vulkan: Do not assume an aux frame context must exist in a well defined state as set in init_buffers() since the request might be external (via overlays path)
- gl: Do not bother waiting for idle before servicing external flip requests
- gl: Queue overlay cleanup requests to ensure only glthread attempts touching the context
- overlays: Do not compute size metrics for invalid/unsupported glyphs
- opengl driver optimization for nvidia. On nvidia glTextureBufferRange performance is horrendous
-- Initialize texture buffer to whole buffer at startup and use absolute offsets to read data instead
-- Over 2x performance in some cases (Resogun, TNT racers)
- gl/vk: Do not flip non-existent display buffers. Fixes spec violation at boot in TNT racers demo
- whitespace fixes for sys_rsx
- Implement low bit decode override flags for 2-component textures
- Properly implement alot of texture remaps according to the autotest results
rsx: Do not unnecessarily shuffle WZYX->RGBA unless we have proof
- From looking at format swizzles, this is incorrect
- 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]
- 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.
- 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
- 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
- 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
- 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
- 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
- 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
- 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
- Use AA mode to predict surface compression. Compression mode is useless without AA activated
- Rewrites most image subresource fetch routines to use the new heuristic
- Fix rsx:🧵:find_tile. FEED000(X) can be substituted for (X) in the code
-- Fixes alot of failures when looking for tiled regions
rsx: Fix antialiased unnormalized coords
- scaling factors are inverse to allow proper coordinates to be computed in fs
- Remove generic throws from the rsx pipeline. Stops the rsx thread from silently dying leaving the emulator in a hung state
- Hackplement add_signed and reverse_subtract_signed blend modes
- Reimplement fragment program fetch and rewrite texture upload mechanism
-- All of these steps should only be done at most once per draw call
-- Eliminates continously checking the surface store for overlapping addresses as well
addenda - critical fixes
- gl: Bind TIU before starting texture operations as they will affect the currently bound texture
- vk: Reuse sampler objects if possible
- rsx: Support for depth resampling for depth textures obtained via blit engine
vk/rsx: Minor fixes
- Fix accidental imageview dereference when using WCB if texture memory occupies FB memory
- Invalidate dirty framebuffers (strict mode only)
- Normalize line endings because VS is dumb
to flush_all is up to date. Should prevent recursive exceptions
Partially revert Jarves' fix to invalidate cache on tile unbind. This will
need alot more work. Fixes hangs
- Identify minimize/restore events as separate from regular resize and do not react to them
- Enable message queue consumption after loading the shaders cache. Also hides the frame in this step
-- This fixes the 'start fullscreen' bug when running vulkan
- This allows for better handling of deferred flushes.
-- There's still no guarantee that cache contents will have changed between the set acquisition and following flush operation
-- Hopefully this is rare enough that it doesnt cause serious issues for now
- vk: Always reopen primary command buffers. They should only be closed in flush_command_queue
- If uploading a texture and there are collisions with protected buffers, do not rebuild the cache
- Perform writes via flush before reprotecting pages that were not trampled
- Only flush no pages once
- Refactor invalidate memory functions into one function
- Add cached object rebuilding functionality to avoid throwing away useful memory on an invalidate
- Added debug monitoring of texture unit VRAM usage
- Emulate primitive restart in software whenever we get the chance
- Ensure PRIMITIVE_RESTART is never active when LIST topologies are active
- Reimplement TRIANGLE_FAN, POLYGON and QUAD expansion
- This communication is important in communicating window events. Helps properly synchronize swapchain management on vulkan and stops nvidia crashing
- Do not block the message queue lest the driver detect window as not responding
rsx: Conditional lock hack removed
vulkan - Fixes
- Remove unused texture class
- Fix native pitch calculation (WCB)
rsx: Catch hanging begin/end pairs when flushing deferred draw calls
vulkan: Register DXT compressed formats
vulkan: Register depth formats
gl: Workaround for 'texture stitching' when gathering flip surface
- TODO: Add a proper flip hack option
rsx: Fix texture memory size calculation
- DXT textures dont have real pitch. Since pitch is used to calculate memory size, make sure it always evaluates to rsx_size
rsx: Fix cpu copy detection
rsx: Validate blit dst surface and dont make assumptions about region blit order
- Also relax restrictions on memory owned by the blit engine if strict rendering is not enabled
rsx: Fix depth texture detection
rsx: Do not manually offset into dst. The overlapped range check does so automatically
rsx: Minor optimizations
rsx: Minor fixes
- Fix to detect incompatible formats when using GPU texture scaling and show message
- Better 'is_depth_texture' algorithm to eliminate false positives
- Limits buffer size to min 720 in the Y axis (1024 section causes conflicts in some cases - TODO)
rsx: Fixups to allow large textures for blit operation
- Also includes checks for both leaking sections and blit regions for vulkan
hotfix for hanging when using WCB
addendum - unlock both ro and no blocks before attempting to copy memory blocks
gl: Fixups for ARB_explicit_uniform_location
- Forces glsl v 430 to make use of the extension
rsx/vk: Rework texture cache to minimize recursive access violations
- Also modifies the vulkan commandbuffer begin/end/submit mechanism
gl: Fix cached_texture_section::is_flushable to take memory protection into account
rsx: Fix blit dst offset calculation
gl/vk/rsx: Refactoring; unify texture cache code
gl: Fixups
- Removes rsx::gl::texture class and leave gl::texture intact
- Simplify texture create and upload mechanisms
- Re-enable texture uploads with the new texture cache mechanism
rsx: texture cache - check if bit region fits into dst texture before attempting to copy
gl/vk: Cleanup
- Set initial texture layout to DST_OPTIMAL since it has no data in it anyway at the start
- Move structs outside of classes to avoid clutter
gl: Fix multidraw [WIP]
rsx: Ignore vertex base when data source is generated using arithmetic
vk: Check pending flag before doing fence poke
vk/gl: Fix for inlined array and immediate draws
rsx: Collapse joined draws when batching
rsx: Blit engine improvements
- Always handle blits to and from framebuffers through the GPU
- Handle depth surfaces properly when using GL
- Check for format mismatches when blitting to the surface store [WIP]
- Make each frame context own its own memory
- Fix GPU blit
- Fix image layout transitions in flip
vk: Improve frame-local memory usage tracking to prevent overwrites
- Also slightly bumps VRAM requirements for stream buffers to help with running out of storage
- Fixes flickering and missing graphics in some cases. Flickering is still there and needs more work
vk: Up vertex attribute heap size and increase the guard size on it
vulkan: Reorganize memory management
vulkan: blit cleanup
vulkan: blit engine improvements
- Override existing image mapping when conflicts detected
- Allow blitting of depth/stencil surfaces
Fix rsx offscreen-render-to-display-buffer-blit surface reads
- Also, properly scale display output height if reading from compressed tile
gl: Fix broken dst height computation
- The extra padding is only there to force power-of-2 sizes and isnt used
gl: Ignore compression scaling if output is rendered to in a renderpass
rsx/gl/vk: Cleanup for GPU texture scaling. Initial impl [WIP]
- TODO: Refactor more shared code into RSX/common
- QUAD_STRIP evaluates to TRIANGLE_STRIP in memory. The memory layout is identical.
- The only difference between the two modes would be the primitive_ID but that doesnt matter on RSX
- Its worth noting that results will be different between the two modes if input vertices are non-coplanar for every set of N verts
rsx/vk/shaders_cache: Move vp control mask to dynamic state
rsx/vk/gl: adds a shader cache for GL. Also Separates pipeline storage for each backend
rsx: Add more texture state variables to the cache
- Updates vulkan to use GPU vertex processing
- Rewrites vulkan to buffer entire frames and present when first available to avoid stalls
- Move more state into dynamic descriptors to reduce progam cache misses; Fix render pass conflicts before texture access
- Discards incomplete cb at destruction to avoid refs to destroyed objects
- Move set_viewport to the uninterruptible block before drawing in case cb is switched before we're ready
- Manage frame contexts separately for easier async frame management
- Avoid wasteful create-destroy cycles when sampling rtts
- Significant gains from greatly reduced CPU work
- Also reorders command submission in end() to improve throughput
- Refactors most of the vertex buffer handling
- All vertex processing is moved GPU side
* Fix always-true conditions in sceNp module
* gl_render_targets: useless check on unsigned variable, possible bug
* fixed UB in crypto utility functions
* copy-paste error in vk::init_default_resources
* pass strings by const ref
* Dont copy vectors. Make sure copies are not needed because functions are used in a multi-threaded context.
* Rework vertex attribute binding for vulkan. Allows always providing a buffer view to the pipeline even if the game has the attribute disabled as long as it is consumed by the vertex shader.
- Improvements to framebuffer usage; Avoid creating new resources every frame
- Handle null fragment program properly
- Collect vertex upload statistics
- vk: Pre-initialize 'unused' varying registers in the vertex shader in case it gets matched with a fs that consumes it
-- Fixes a crash about fog_c not being declared
gl/dx12/vk: Handle null fragment program
- cleanup - use yield semantic instead of sleep(0) as yield is more cross-platform
-- sleep(0) is a windows specific scheduler hint
- vk: Do not select first available format when choosing a swapchain format
- gl/vk: Ignore rendering zero sized framebuffers/scissors
- fp: Re-enable range clamp on fp16 registers; fix fx12 clamping [-2, 2]
* Update VKFragmentProgram.cpp
added missing exponent parameter
* fixed misplaced exponent in VKFragmentProgram.cpp
parameter that belonged to pow() was being passed to exp() instead, causing the shader compilation to fail
* fix for opengl fog_mode exponential2
same fix as the vulkan version
* directx fog_mode exponential2 fix
misplaced parameter
* directx fog_mode exponential2_abs fix
* vulkan fog_mode exponential2_abs fix
* opengl fog_mode exponential2 fix
Improve scaling and separate sampler state from texture state
gl: Unify all texture cache objects under one structure separate by use case
gl: Texture cache fixes
- Acquire lock when finding matching textures
- Account for swizzled surfaces when deciding whether to cpu memcpy
- Handle swizzled images on the GPU
* gl: Only bind attrib textures on thread startup
* gl: Persistent mapped buffers
* gl: Fix emulated primitives in an inlined array
* gl: Do not re-update program information every draw call
* gl/vk: s1 type is signed normalized not unsigned normalized
* gl/rsx: Allow disabling of persistent buffers for debugging
gl: Large heap size is more practical
gl: Fix a bug with legacy opengl buffers
* gl/rsx: Allow emulation of unsupported attribute formats
* gl: Fix typos and remove dprints
gl: cleanup debug prints
* ui: Move the GL legacy buffer toggle to the left pane
* vk/gl: Fix cmp type, its range is [-1,1] not [0,1] SNORM_INT
* gl/rsx: Implement platform-agnostic text overlays
gl: Restore performance metrics using new text out helper
gl/rsx: Refactor text generation class
* vk: Enable text overlay
gl/vk: Polish overlay counters implementation
gl: Better resource shutdown for text writer
* gl: Optimization, do not rebind TIUs every frame. Speedup
* gl: Optimizations and improvements to vertex upload code
* gl/vk: Texture format swizzles
vk: Texture format fix
vk: Fix YX format swizzles
* rsx: Decode vertex texture index
* vk: Zero-initialize some more structs
* gl: Clean up fragment program generation code
* vk: Enable alpha kill
* vk: Fix surface clear; redirect output for surface_type:b
* vk: Tie renderpass to program object to avoid incompatible passes
`unveil<>` renamed to `fmt_unveil<>`, now packs args to u64 imitating va_args
`bijective...` removed, `cfg::enum_entry` now uses formatting system
`fmt_class_string<>` added, providing type-specific "%s" handler function
Added `fmt::append`, removed `fmt::narrow` (too obscure)
Utilities/cfmt.h: C-style format template function (WIP)
Minor formatting fixes and cleanup
gl: stream uniform data using stream buffer
gl: vertex streaming improvements and bugfixes
gl: add basic timing info check for profiling
gl: ebo streaming fixes and enhancements
* gl/vk/dx12: Fix vertex shader code generation for buggy games
dx12: revert vsh attribute changes
* vk: dynamically flush command buffers if we exceed available resources
* dx12: do not prepare flip texture if it has not been initialized
* Optimizations
1) Some headers simplified for better compilation time
2) Some templates simplified for smaller executable size
3) Eliminate std::future to fix compilation for mingw64
4) PKG installation can be cancelled now
5) cellGame fixes
6) XAudio2 fix for mingw64
7) PPUInterpreter bug fixed (Clang)
* any_pod<> implemented
Aliases: any16, any32, any64
rsx::make_command fixed
Moves GL, minidx12, OpenAL, stblib to 3rdparty
Fixes AppVeyor and CMakeLists (travis-ci.yml doesn't need any changes)
Points directories in the VS solution files to the new ones on the
externals directory
Includes stuff with the < > syntax instead of " "
* get_placed_texture_storage_size returns more accurate result (fix crash in Outrun)
* Factors lot of code and use integer type more carrefully
* Treat warning as error in TextureUtils.cpp
Support vertex instancing in vertex shader using VertexID
Relax OpenGL requirements by removing 4.5 features
Use EXT version of TexBufferRange; Implement buffer copy using TexBuffer
Apply travis workaround by danilaml
Fix vertex upload in in case of inlined array
Decode textures using TextureUtils
Shut compiler up about missing return value (unsure)
Revert "Shut compiler up about missing return value (unsure)"
This reverts commit db43ba448776e6c0d6c90909b4037b6f698a545c.
Change some functions in rsx::gl::texture to static.
Replace large switch block in gl::textures::mandates_expansion with simple if tests
Also fixed newlines, CellSailEvent, fixed CellMusic changes, fixed
console_write, fixed L10n changes and removed the login dialog for now,
until a more proper implementation.
