* Refactor memory size constants and calculations
Updated memory size handling in video decoder functions.
* Refactor video decoder frame size calculations
Updated frame size computation to use worst-case dimensions and adjusted alignment values.
* Refactor alignment logic to use Common::AlignUp
[Render.Vulkan] <Error> (shadPS4:GpuComm) vk_platform.cpp:57 DebugUtilsCallback: SYNC-HAZARD-READ-AFTER-WRITE: vkCmdBeginRendering():
READ_AFTER_WRITE hazard detected. vkCmdBeginRendering reads pRenderingInfo.pColorAttachments[0] (VkImageView 0x5f1a0000005f1a[ImageView 1920x1080x1 0x244980000:0x2aa9800 1:1 0:0 (Color)], loadOp VK_ATTACHMENT_LOAD_OP_LOAD), which was previously written during an image layout transition initiated by vkCmdPipelineBarrier2[gfx:0x213a04c50:DrawIndexAuto].
The current synchronization allows
VK_ACCESS_2_SHADER_READ_BIT accesses at VK_PIPELINE_STAGE_2_VERTEX_SHADER_BIT|VK_PIPELINE_STAGE_2_TESSELLATION_CONTROL_SHADER_BIT|VK_PIPELINE_STAGE_2_TESSELLATION_EVALUATION_SHADER_BIT|VK_PIPELINE_STAGE_2_GEOMETRY_SHADER_BIT|VK_PIPELINE_STAGE_2_FRAGMENT_SHADER_BIT|VK_PIPELINE_STAGE_2_COMPUTE_SHADER_BIT|VK_PIPELINE_STAGE_2_CLUSTER_CULLING_SHADER_BIT_HUAWEI,
VK_ACCESS_2_COLOR_ATTACHMENT_WRITE_BIT accesses at VK_PIPELINE_STAGE_2_COLOR_ATTACHMENT_OUTPUT_BIT,
but to prevent this hazard, it must allow VK_ACCESS_2_COLOR_ATTACHMENT_READ_BIT accesses at VK_PIPELINE_STAGE_2_COLOR_ATTACHMENT_OUTPUT_BIT.
* Restore old log filter behavior
Not sure exactly why this hadn't happened yet.
* Suggested change
* Update documentation to include changes
* Remove mention of debug log class
Debug is both a valid class and level, kinda confusing to use it as an example in my opinion.
* Error instead of assert
* Missing include
The case-insensitive fallback search() in GetHostPath is only
invoked for patch_path and host_path, so mods whose file or folder
capitalization does not exactly match the guest path are silently
bypassed even when the files are present. Mirror the existing
search(patch_path) pass for mods_path, placed first to preserve
mod > patch > base precedence.
Co-authored-by: Matías Buzzo <matias@mbuzzo.com>
The case-insensitive fallback search() in GetHostPath is only
invoked for patch_path and host_path, so mods whose file or folder
capitalization does not exactly match the guest path are silently
bypassed even when the files are present. Mirror the existing
search(patch_path) pass for mods_path, placed first to preserve
mod > patch > base precedence.
* Change buffer check to use num_records instead
* Fix the address fixup check
Since base_address == 0 is possible, this should specifically check against 1 (which we reserve for null buffers).
Kinda shocked TLOU/Uncharted worked at all without setting this.
* renderer: update font handling and add new font support
- Added support for multiple NotoSans fonts including Arabic, Thai, and CJK.
- Refactored font loading logic to use FontStack for better management.
- Removed deprecated NotoSansJP font and updated font paths in CMake.
* optimize atlas size
---------
Co-authored-by: w1naenator <valdis.bogdans@hotmail.com>
- Added support for multiple NotoSans fonts including Arabic, Thai, and CJK.
- Refactored font loading logic to use FontStack for better management.
- Removed deprecated NotoSansJP font and updated font paths in CMake.
Co-authored-by: w1naenator <valdis.bogdans@hotmail.com>
* renderer_vulkan: Microoptimizations from profiling
* buffer_cache: Only use streaming path on cpu modified memory
If the pages arent cpu modified, then a gpu buffer with valid data exists that can be used instead
* clang format
* fix max num_layers
* Early memory regions setup
Needed to enable flexible allocations before mapping the eboot
* Improved accuracy of module mappings
To keep our current logic intact, reserve the full memory space, then perform fixed mappings for trampoline and segments. Segment mappings use type flexible, so they properly count toward flexible usage, while trampoline mappings use type code to bypass this.
* Asserts for failed offset finding
Always check before you leap.
* Hex logging for segment sizes
* Only map game modules to flex mem.
* Revert "Asserts for failed offset finding"
This reverts commit 38083f8cbb.
* Revert "Early memory regions setup"
This reverts commit da1552d4ac.
* Initialize memory regions in MemoryManager constructor
Much simpler approach to enabling the necessary flexible mappings on startup.
* replace psn_signed_in settings with shad_net_enabled and added shadnet_server setting
* improved user manager to update new fields and add new variables
* clang
* better message :D
* prototype settings look
* put categories on top instead of on the side, plays nicer with input system
* added profiles + saving and loading framework
* add helper functions
* fix autofocusing
* add code for using embedded image files
* put settings in different files
* clang
* remove sdl_image submodule, unify functions when possible
* Always display selected profile in settings content areas
* Fix focusing, logs, highlight profile text
* persist highlight on focused buttons, add all categories placeholders
* All settings added
* navigate tabs with l1/r1
* Replace sdl mixer library with minimp3 header
* clang
* set spec before putting in audiostream
* respect main audio output device setting
* fixup
* replace file with submodule
* cleanup
* capitalize functions like the others
* move buffer to heap
* use vector for pcm buffer instead
* imguitest
* button tests
* fix gamepad nav
* placeholder hardcoded eboot path
* set focus correctly, move to own files
* get installed game information
* dynamically adjust rows
* use slider for ui scale
* launch big picture with CLI arg
* Use emulator settings for UI scale and window size
* center scrolling on focused item
* fix item focus on scrolling when navigating with keyboard or pad
* minor fixups and comments
* fix performance degradation
* adjust fonts to show TM symbol and higher overscale
* reuse and clang
* add exists check before iterator
* flatten navigation (gamepad navigation crosses child window container)
* cleanup and comments
* simplify update checker a bit
---------
Co-authored-by: georgemoralis <giorgosmrls@gmail.com>
* resource_tracking_pass: Adjust buffer type if host doesn't support float buffer atomic
* resource_tracking_pass: Implement data append/consume as buffer atomics in IR level
This was previously done in spirv backend, the implementation was exactly the same as the buffer atomics, so unify them
* ir: Bump instruction flag to 8 bytes
* frontend: Pass pc to buffer flags for better debugging when sharp tracking fails
* clang format
---------
Co-authored-by: georgemoralis <giorgosmrls@gmail.com>
* fixed Unsynchronized free_threads.size()` Read in ThreadState::Free
* fixed data race in posix_pthread_key_create , allocated and seqno are atomic now
* fixed playtime thread
* argv array size fix
* destroy mutex on object deletion with pthread_mutex_destroy
* fixed object_pool
* used atomics for thread safety
* equeue: Save filter/ident before std:move since we access them out of the locked loop
* fixed kqueues memory leak
* clean storage objects effectively
* fixed memory leak
* fix some races
* fixed race condition
* Implement mods folder
* Add logging when modified files exist
* get mod folder entries in MntPoints::IterateDirectory
* support eboot from mods folder
[Render.Vulkan] <error> (shadPS4:Main) vk_platform.cpp:58 DebugUtilsCallback: VUID-vkDestroyImage-image-01000: vkDestroyImage(): can't be called on VkImage 0x850000000085[Frame image #2] that is currently in use by VkCommandBuffer 0x560ea08752c0[CommandPool: Command Buffer 2].
The Vulkan spec states: All submitted commands that refer to image, either directly or via a VkImageView, must have completed execution (https://docs.vulkan.org/spec/latest/chapters/resources.html#VUID-vkDestroyImage-image-01000)
Finish needed because of:
[Render.Vulkan] <error> (shadPS4:Main) vk_platform.cpp:58 DebugUtilsCallback: VUID-vkEndCommandBuffer-commandBuffer-00059: vkEndCommandBuffer(): Cannot be called for VkCommandBuffer 0x55c62bfbb580[CommandPool: Command Buffer 3] when it is not in a recording state, vkBeginCommandBuffer() must first be called.
The Vulkan spec states: commandBuffer must be in the recording state (https://docs.vulkan.org/spec/latest/chapters/cmdbuffers.html#VUID-vkEndCommandBuffer-commandBuffer-00059)
