RPCSX/rpcsx
1
0
Fork 0
mirror of https://github.com/RPCSX/rpcsx.git synced 2026-03-22 05:06:01 +01:00
Code Issues Actions Packages Projects Releases Wiki Activity
11123 commits 5 branches 1 tag 150 MiB
Commit graph

3013 commits

Author SHA1 Message Date
Nekotekina 95aeebe4b5 sys_rsx: add cpu_flag::wait 2020-10-30 18:00:25 +03:00
Nekotekina 605d57c541 sys_event: cleanup (replace vm::temporary_unlock) 
Also made minor changes in sys_rsx.cpp.
Removed unused exception std headers.
 2020-10-30 17:49:07 +03:00
Nekotekina f972fa26a4 Derive RSX Replay thread from cpu_thread 
Its id is set to 0, so fix some id_type() usages.
 2020-10-30 17:36:11 +03:00
Nekotekina 13de773486 Remove some vm::reservation_lock instances 2020-10-27 17:56:19 +03:00
kd-11 b32eecb5a7
rsx: Driver compatibility improvements (#9131) 
* rsx: Refactor vertex clip emit to avoid using f64 unnecessarily

- Fixes driver crash on intel

* vk: Add NVIDIA driver version check

- Warn if user has outdated drivers with known problems
 2020-10-27 13:22:15 +03:00
kd-11 18ca3ed449 rsx: Block-level reservation access 2020-10-25 20:21:04 +03:00
Nekotekina 306593a0c5 Revert "atomic.cpp: fixup for WaitOnAddress path" 
This reverts commit 3b8bce1bed.
 2020-10-21 09:54:22 +03:00
Nekotekina 3b8bce1bed atomic.cpp: fixup for WaitOnAddress path 
Also fix wait quantum.
 2020-10-21 08:18:27 +03:00
kd-11 a90801e2aa vk: Add VK_FORMAT_D32_SFLOAT to format conversion table 
- This format is required to emulate RSX_FORMAT_CLASS_D16_FLOAT
 2020-10-18 19:30:40 +03:00
Nekotekina 492ed27495 RSX: fix rsx::nv406e::semaphore_release partially 
Properly release reservation (non-TSX path).
At least update and notify reservation (TSX).
 2020-10-15 20:58:59 +03:00
Nekotekina dcff8c2637 Fix remaining vm::reservation_lock usages (for now) 
Optimization can be restored later.
 2020-10-13 12:04:59 +03:00
Nekotekina 346a1d4433 vm: rewrite reservation bits 
Implement classic unique/shared locking concept.
Implement vm::reservation_light_op.
 2020-10-10 13:58:48 +03:00
kd-11 bca3a3f4ed
vk: Open CB before doing frame cleanup so that callbacks work (#9041) 2020-10-10 10:02:55 +01:00
kd-11 d5f7e7b179 vk: Reimplement GPU query management 2020-10-06 12:02:53 +03:00
Eladash f4ca6f02a1 PPU: Implement support for 128-byte reservations coherency 2020-09-28 22:34:42 +03:00
kd-11 04ff7913b4 rsx/codegen: Workaround for borked hardware 
- Bitwise or does not evaluate correctly for some hardware.
  Substitute with subtraction instead.
 2020-09-28 22:18:36 +03:00
kd-11 9baef8c705 rsx: Emit simpler fragment program code 
- Optimize clamp16
- Use bfe instead of shift-and
 2020-09-27 18:56:04 +03:00
kd-11 a14a358b73 rsx: Optimize vertex decoder to generate simpler code 
- Significantly improves compilation speed by simplifying most of the code and doing something similar to LICM.
  * Actual decoding is now vectorized and performed in one step rather than in a loop.
  * Switches inside loops are removed and replaced with simple comparison. Generates much nicer (and smaller) GCN bytecode.
 2020-09-27 18:56:04 +03:00
kd-11 259844f4f3 vk: Disable spirv optimizer 
- I've not found it to be very useful and it just breaks good code right now.
  TODO: Re-enable when things improve.
 2020-09-27 18:56:04 +03:00
kd-11 9ea478008c vk: Properly initialize float64 support for SPIRV 2020-09-24 01:30:21 +03:00
Morgan Creekmore 4fe2951509 Fixed formatting 2020-09-23 20:27:30 +03:00
Morgan Creekmore ff8a94714f Add additional NV4097 methods to gcm_printing 2020-09-23 20:27:30 +03:00
Morgan Creekmore b45d6fee2d Fix methods_name indentation 2020-09-23 20:27:30 +03:00
Morgan Creekmore f44e696edf Added missing NV methods for gcm_printing 2020-09-23 20:27:30 +03:00
kd-11 a50ea09053 rsx: Properly pass format_class information during RTV/DSV resource barrier 
- Also takes the opportunity to remove repeating code in a minor refactor.
 2020-09-22 12:19:54 +03:00
kd-11 d012abd924 vk: Improve image transfer and scaling 
- Handle typeless src and dst with aliased typeless format
- Optimize typeless transfers by only dealing with affected texels.
  * Eliminates redundant dst->typeless transfer of full image (very expensive)
  * Eliminates full src->typeless transfer of full image and replaces with only affected region
  * Requires significantly smaller output buffers, saving on VRAM cost
 2020-09-22 12:19:54 +03:00
kd-11 7ed82c0791 rsx: Always force typeless copy if memory is crossing aspect boundary 2020-09-22 12:19:54 +03:00
kd-11 9db97278f3 rsx: Lower error message to warning 
- Mismatched texture handling is a TODO that will be handled with texturing rewrites
 2020-09-19 01:55:59 +03:00
kd-11 d3898fda57 rsx: Release misconfigured texture memory before attempting reupload 2020-09-19 01:55:59 +03:00
kd-11 7900780cea vk: Fix nul section crash due to unexpected format (B8) 2020-09-16 20:14:44 +03:00
kd-11 92d65ff3c2 rsx: Add support for mixed data types when sampling shadow coordinates 2020-09-15 17:37:52 +03:00
Megamouse a2da187615 HLE: localize most - if not all - exposed strings 2020-09-14 18:24:18 +02:00
Megamouse d0ffbbfc4d Qt/overlays: use Argument list for translatable strings 
This is somewhat crippled for now. It only takes a single argument in the callback
 2020-09-14 18:24:18 +02:00
Megamouse 460a933267 Qt/overlays: Localize most rsx overlays 2020-09-14 18:24:18 +02:00
kd-11 da6760ed98 vk: Simplify shadow comparison operations for non-integer formats 
- Just use hardware PCF, it makes everyone's life easier.
 2020-09-09 22:11:12 +03:00
kd-11 6380e67af9 rsx: Fix depth clipping 
- Fix special case where n=f making (f-n) = 0
- Dynamically update depth range by setting dirty bits
- Fix depth bounds when n=f and bounds test is disabled
 2020-09-08 15:33:08 +03:00
kd-11 dc465df3bc rsx: Enable support for extended range in depth buffer 
- Software clipping emulation is used here as OpenGL does not have explicit clip control.
- Hardware clip control for vulkan to be enabled after this.
 2020-09-08 15:33:08 +03:00
kd-11 2e88924cb9 rsx/gl: Refactoring and cleanup 
- Fix incorrect memory requirement calculation for D32FS8X24_PACK64 data type on GL
- Removes a lot of spaghetti code in GL backend from years of accumulation
- Retires several now-useless methods from RSX util toolbox
 2020-09-08 13:53:06 +03:00
kd-11 6d2cb94e3e gl/vk: Support swizzled data for RCB/RDB 2020-09-07 22:31:57 +03:00
kd-11 85e5b077f7 gl: Overhaul upload and download routines for textures to go through shared image_to_buffer and buffer_to_image routines. 
- This automatically adds support for depth float textures as well
 2020-09-07 18:25:54 +03:00
kd-11 85dd1b4ea9 vk: Fix fconvert job issues 
- Fix compilation bug caused by typo
- Invert to/from for consistent declarations
- Fix dst_swap when From == 2
 2020-09-07 18:25:54 +03:00
kd-11 220e86bbd1 gl: Accelerate D24X8_UINT operations 
- Adds compute decoding for D24X8_UINT on both download and upload routines
- Adds support for D24X8_UINT operations for typeless copy
 2020-09-07 18:25:54 +03:00
Bevan Weiss baf96b3eb6 RSX: Update manual string creation -> std::string() 
Replace manual string creation with call to std::string() constructor passing in char*
This appears to drastically reduce the cache impact here
 2020-09-05 10:38:32 +03:00
kd-11 3c43d8fe05 rsx: Fix execution barrier insertion 
- In case of element re-arrangement, the barrier should obey the current insertion pointer
 2020-09-04 09:34:13 +03:00
Eladash 73d23eb6e6
SPU: Implement Accurate DMA (#8822) 2020-09-02 23:58:29 +02:00
kd-11 a917f55ef8
vk/sdk: Sync with vulkan SDK v148 (#8814) 
- Sync with vulkan SDK 148
- glslang library was split into several smaller libraries
- HLSL is no longer needed
 2020-09-01 00:57:38 +03:00
kd-11 af9e217fa4 vk: Improve D16F handling 
- Adds upload and download routines. Mostly untested, which is why the error message exists
 2020-08-30 09:26:37 +03:00
kd-11 e9cdb248a0 glsl: Properly implement shadow filtering when running emulated shadow compare 
- Previous code was completely borked
 2020-08-29 02:03:09 +01:00
kd-11 e8274d5a59 vk: Fix depth format mismatch detection in copy_image 2020-08-29 02:03:09 +01:00
kd-11 d000d648b0 vk: Fix some minor spec violation 
- Stencil clear pass does not consume an image, do not bind one.
- Add push_barrier to allow push-pop semantics for texture barrier insert.
 2020-08-27 12:52:28 +03:00