Commit Graph

42 Commits

Author SHA1 Message Date
Philip Rebohle 3d213efe53
[dxbc] Use SSBOs for dynamically indexed constant buffers if needed
SSBOs are tightly bound-checked on all Nvidia GPUs, so this allows for
a more accurate workaround for games relying on OOB access behaviour.
2019-10-28 15:42:46 +01:00
Philip Rebohle 6ab074c95b
[dxbc] Only use atomic append/consume optimization in compute shaders
- For fragment shaders, this isn't safe since ballots include helper invocations
- For vertex shaders, if drivers don't support subgroup operations in those
  stages, we don't want it to affect the performance of compute shadres.
2019-07-26 14:07:39 +02:00
Philip Rebohle 8644d75722
[dxbc] Use OpDemoteToHelperInvocationEXT for discards if supported 2019-07-11 19:33:17 +02:00
Philip Rebohle 06d4e06d8a
[dxbc] Enable early discard on RADV/ACO
Still can't enable on LLVM due to GPU hangs. The way we detect it
may have to change in the future, but for now, this should do.
2019-07-08 16:30:56 +02:00
Philip Rebohle 58dcf77c08
[dxbc] Disable atomic counter optimization on old RADV versions
Breaks TressFX in Tomb Raider 2013 for some reason, whereas this works
fine on 19.1 with both LLVM 8.0 and 9.0-git.
2019-06-15 12:58:31 +02:00
Philip Rebohle d5ba36443e
[dxbc] Fix useRawSsbo option
Add some break to fix some breakage.
2019-06-11 20:23:47 +02:00
Philip Rebohle d755b18ccb
[dxbc] Repurpose useRawSsbo option
Basically, abuse the minSsboAlignment property to convince the
compiler to always/never emit raw SSBOs for raw and structured
buffers, and use the actual device limit by default.
2019-06-11 20:20:40 +02:00
Philip Rebohle 0f7c23c022
[dxbc] Always use raw storage buffers if alignment requirements are met 2019-06-11 20:20:40 +02:00
Philip Rebohle d94d89c3ef
[dxbc] Add option to use subgroup ops for atomic counter operations
This can greatly reduce the number of atomic operations when using
append/consume buffers.
2019-05-15 18:49:02 +02:00
Philip Rebohle dc52212873
[dxbc] Disable Constant Buffer Range Check on AMD
The hardware already behaves as intended, no need to waste GPÜ cycles.
2019-04-30 16:07:24 +02:00
Philip Rebohle 5819a69302
[d3d11] Add option to enable constant buffer range checks 2019-04-05 20:56:32 +02:00
Philip Rebohle 0326258829 [dxbc] Use SDiv instead of ShiftRightLogical for index calc on Nvidia
Reportedly fixes Resident Evil 2 and Devil May Cry 5 on Nvidia.
2019-03-09 19:59:56 +01:00
Philip Rebohle 38c6eeed26
[dxbc] Only emit depth clamp in fragment shader if necessary
We don't need this if the depth clip extension is supported
by the driver.
2019-02-19 14:27:21 +01:00
Philip Rebohle c0b325b483
[dxbc] Add option to enable strict sm4-compliat division
SM4 is defined to return the first source operand if the divisor is zero.
Windows drivers don't do this by default, so we shouldn't do it either.
2019-02-05 21:08:13 +01:00
Philip Rebohle 4fe5929799
[dxbc] Add options to enable/disble early discard and raw ssbo use 2019-01-08 20:58:18 +01:00
Philip Rebohle 48548eb894
[dxbc] Add useRawSsbo option
ENabled when we can replace texel buffers for raw and structured
buffers with raw SSBOs.
2018-12-14 23:45:37 +01:00
Philip Rebohle f276bcd0e7
[dxbc] Check for RADV/Nvidia drivers explicitly using the new function
This way, only RADV gets locked out of using early discard whereas
AMDVLK and -PRO are able to use it.
2018-12-12 16:27:01 +01:00
Philip Rebohle 7096937c11
[dxbc] Use ballot to determine early-discard condition in fragment shaders 2018-11-27 11:50:01 +01:00
Philip Rebohle 2424cf0289
[dxbc] Disable early discard on AMD cards
Causes GPU hangs.
2018-11-24 10:09:33 +01:00
Philip Rebohle a75c596132
[d3d11] Add config option to enable TGSM initialization 2018-11-23 16:12:02 +01:00
Philip Rebohle 3e686098cc
[dxbc] Disable subgroup early discard on Nvidia GPUs
People report performance regressions up to 5% with it enabled.
2018-11-23 12:28:48 +01:00
Philip Rebohle 0418c02ac3
[dxbc] Use clustered subgroup operations on supported hardware 2018-11-21 11:17:09 +01:00
Philip Rebohle 4f76b89941
[dxbc] Use subgroup operations for early discard
Can improve performance in some complex fragment shaders.
2018-11-21 11:17:09 +01:00
Philip Rebohle a574829bb6
[dxbc] Remove DeferKill flag
This behaviour is correct, and needs to be enabled by default.
2018-11-21 11:17:09 +01:00
Philip Rebohle bd03225c14
[dxbc] Refactor DxbcOptions
A bit mask isn't good enough going forward, and we also don't
need application-specific options for now.
2018-11-09 08:41:02 +01:00
Philip Rebohle 0ba00b3f59
[dxvk] Add extended device feature structure
This allows the client API to query and enable extended
features in the future, should it become necessary, much
like the extende feature queries.
2018-07-31 16:58:25 +02:00
Philip Rebohle f7587014c7
[dxbc] Fix compiler warning 2018-06-18 15:33:35 +02:00
Philip Rebohle a607fe107f
[dxbc] Remove workarounds for the old Nvidia shader compiler 2018-06-12 22:13:53 +02:00
Philip Rebohle af62d423b2
[dxbc] Add per-device option for deferred kills 2018-05-26 17:08:39 +02:00
Philip Rebohle 58e9280891
[dxbc] Remove Tex2D -> Tex2DArray workaround 2018-05-24 11:49:12 +02:00
Philip Rebohle a5c43b5e59 [dxbc] Enable array texture hack for Dishonored 2
Partially fixes shadow rendering issues.
2018-04-23 11:09:48 +02:00
Philip Rebohle 01147492d5 [dxbc] Implemented Texture2D -> Texture2DArray mapping as a knob
Doing this for all applications is not necessary and degrades
performance in some cases.
2018-04-23 00:46:27 +02:00
Philip Rebohle e53cb72e39 [dxbc] Make DxbcOptions a set of flags
This allows for game-specific workarounds and driver-specific
workarounds to be applied in an easier fashion.
2018-04-22 23:49:41 +02:00
Philip Rebohle 021ea64103
[dxbc] Remove Nvidia driver version check
Apparently this causes a regression in Overwatch for some.
2018-04-11 19:44:45 +02:00
Philip Rebohle 8ea17e7a25
[dxbc] Implemented Nop instruction
Required for Metal Gear Solid V.
2018-04-11 01:49:39 +02:00
Philip Rebohle f3544cc8e1
[dxbc] Fix __GL_NextGenCompiler check... 2018-04-11 01:17:07 +02:00
Philip Rebohle 3ff437ddbe
[dxbc] Re-enable Nvidia workarounds if __GL_NextGenCompiler is '0'
Fixes potential regressions when using the old SPIR-V compiler.
2018-04-11 00:26:42 +02:00
Philip Rebohle 1b6442b859
[dxbc] Disable Nvidia-specific workarounds for 396.18 and above 2018-04-11 00:20:24 +02:00
Philip Rebohle 584ee6b6f0
[dxbc] Lift ShaderStorageImageReadWithoutFormat requirement 2018-03-21 12:47:53 +01:00
Philip Rebohle 0e9b7d7ccd
[dxbc] Implemented new workaround for depth-compare ops on Nvidia 2018-03-01 10:11:15 +01:00
Philip Rebohle 8bfd12067a
[dxbc] Remove packDrefValueIntoCoordinates hack
Causes more issues than it solves when the coordinate vector
gets too big.
2018-02-28 06:44:55 +01:00
Philip Rebohle f4cd90d6fa [dxbc] Implemented vendor-specific workarounds in an attemt to fix Nvidia 2018-01-07 20:05:27 +01:00