Commit graph

1539 commits

Author SHA1 Message Date
kd-11 eee7bf0c97 vk: Don't load depth-stencil textures on the transfer queue 2023-07-12 03:20:08 +03:00
oltolm 0c94606fcf
Make compile with msvc, clang and gcc on Windows 2023-07-11 21:40:30 +03:00
kd-11 186fe53279 rsx: Fix the default shadow codegen path 2023-07-09 19:33:48 +03:00
kd-11 10af0e8c93 vk: Disable anisotropy if the anisotropic level is meaningless 2023-07-04 09:31:51 +03:00
kd-11 d13cf0e29a rsx: Fix shader interpreter
- It broke ages ago
2023-07-04 09:31:51 +03:00
kd-11 69eb1401a9 vk/gl: Fix debug overlay stats 2023-07-04 09:31:51 +03:00
kd-11 4decad6f6b rsx: Force compatibility checks when applying fast-path optimizations 2023-07-04 09:31:51 +03:00
kd-11 8f98e572e1 rsx: Use designated initializer for the dangerous copy_region_descriptor type
- We need to move more dangerous braced initializations to c++20
- Also adds a base address variable which will come in handy
2023-07-04 09:31:51 +03:00
kd-11 c818f8af82 vk: Fix namespace pollution in sync module 2023-07-04 09:31:51 +03:00
kd-11 8ebf24d8e7 vk: Do not clear memory if we're going to overwrite it with a copy anyway 2023-07-04 09:31:51 +03:00
kd-11 b788e05396 rsx: Implement GPU copy ellision through coordinate transform
- TODO: Some corner cases still exist where format may not be a match after a cache merge.
2023-07-04 09:31:51 +03:00
kd-11 d77a78cdf1 rsx: Rework texture coordinate handling to support clamping and a more sane scale-bias setup 2023-07-04 09:31:51 +03:00
kd-11 89c81d9f22 rsx: Switch common codegen to use the glsl scripts 2023-07-04 09:31:51 +03:00
kd-11 931a2c516f vk/rsx: Fix build errors and warnings 2023-06-30 18:38:14 +03:00
kd-11 2f9e1cf4df vk: Get rid of some spammy debug code 2023-06-30 18:38:14 +03:00
kd-11 68d1c38f3b vk: Minor refactor to make early-exit branches trivial and easier to read 2023-06-30 18:38:14 +03:00
kd-11 c4acea6c82 vk: Fix some regressions in synchronization2 behavior 2023-06-30 18:38:14 +03:00
kd-11 7ebabddfe0 vk: Fix GPU hang 2023-06-30 18:38:14 +03:00
kd-11 a047ddffc0 vk: Restructure gpu labels and add them as an event backend 2023-06-30 18:38:14 +03:00
kd-11 92992a5406 vk: Use pipeline barriers for proper event sync 2023-06-30 18:38:14 +03:00
kd-11 9ecfecb83e Constify 2023-06-27 16:41:30 +03:00
kd-11 ec4565263d Dump extensions info to the log 2023-06-27 16:41:30 +03:00
kd-11 ed1da364e5 vk: Fix a bunch of corner cases when syncing GPU and host memory 2023-06-27 16:41:30 +03:00
kd-11 942dbbc8fa vk: Add explicit sync when doing copy_image_to_buffer 2023-06-27 16:41:30 +03:00
kd-11 3e8f9cbbf2 Add flags to handle sync domain with v2 events 2023-06-27 16:41:30 +03:00
kd-11 322b5dac79 Properly enable synchronization2 feature 2023-06-27 16:41:30 +03:00
kd-11 5a85fee3f1 Silence logspam about unused parameter 2023-06-27 16:41:30 +03:00
kd-11 0a0ef50caf Try to fix build 2023-06-23 15:21:17 +03:00
kd-11 e20f1587cd Workaround for old compilers
Xcode is not yet c++20 compliant
2023-06-23 15:21:17 +03:00
kd-11 715e3856f2 vk: Update async compute (fast) to use the new sync API 2023-06-22 14:59:58 +03:00
kd-11 850166eca1 vk: Reimplement events using synchronization2 extension 2023-06-22 14:59:58 +03:00
kd-11 d5654d64d5 Nicer variable names 2023-06-16 15:29:36 +03:00
kd-11 06459eb37b vk: Latch query copy requests to reduce number of vulkan commands used 2023-06-16 15:29:36 +03:00
kd-11 49c6c2c529 vk: Do not hard-sync on first sign of fragmentation
- It is very likely that the resource would be available if we just waited a while for the GPU queue to clear.
- Instead of also discarding the current work, first check if we can get by without a hard sync.
2023-06-14 01:12:11 +03:00
kd-11 29f3eec957 vk: On pool recovery, rescan the subpool table in case a new pool really cannot be created 2023-06-14 01:12:11 +03:00
kd-11 c1d875d841 vk: Handle VK_ERROR_FRAGMENTATION when allocating descriptor pools 2023-06-14 01:12:11 +03:00
kd-11 27349c5c69 vk: Cleaner approach to object lifetime management 2023-06-10 02:54:16 +03:00
kd-11 872aa2b358 vk: Improved OOM handling
- Don't aggressively delete surfaces unless we're really crashing
- Release barrier resources before destroying the surface in case of fatal crash
2023-06-10 02:54:16 +03:00
kd-11 4d565eade6 rsx: Drop format information from the vertex cache and use a simpler model that is much faster 2023-06-10 01:36:10 +03:00
kd-11 97f7461aa9 rsx: Allow the vertex cache to be enabled independently of MTRSX 2023-06-10 01:36:10 +03:00
kd-11 7e2eaf43ca rsx: Capture stats about vertex cache performance 2023-06-10 01:36:10 +03:00
kd-11 6f834e99d2 vk: Disable conditional rendering on MacOS 2023-06-06 22:05:54 +03:00
kd-11 86c7b31b6d Force disable conditional render on macOS 2023-06-06 22:05:54 +03:00
kd-11 ac0c60b740 vk: Rename gc_wrapper_t to gc_callback_t which is more relevant 2023-06-01 19:25:50 +03:00
kd-11 f2f95a20a3 vk: Minor improvement to descriptor and garbage collector interaction 2023-06-01 19:25:50 +03:00
kd-11 12f213ffad rsx: Trim the number of in-flight invalidated resources (temp cache)
- This drastically improves memory allocation behavior.
  Holding too many invalidated resources can lead to a cascading overallocation error as old resources hold refs to even older resources and nothing gets deleted.
2023-06-01 19:25:50 +03:00
kd-11 6a2ea62f82 vk: Minor cleanup 2023-05-30 16:50:08 +03:00
kd-11 6cfd817b5a vk: Fix descriptor-related crashes 2023-05-30 16:50:08 +03:00
kd-11 c99ef4b09f vk: Refactor descriptor handling 2023-05-30 16:50:08 +03:00
kd-11 10171c19c3 vk: Unify descriptor allocation
- Pool management should be a backend implementation detail.
2023-05-30 16:50:08 +03:00