Commit Graph

17 Commits

Author SHA1 Message Date
f23b2878cc Shader: Add fallback for LDG from "ube" buffer ranges. (#4027)
We have a conversion from LDG on the compute shader to a special constant buffer binding that's used to exceed hardware limits on compute, but it was only running if the byte offset could be identified. The fallback that checks all of the bindings at runtime only checks the storage buffers.

This PR adds checking ube ranges to the LoadGlobal fallback. This extends the changes in #4011 to only check ube entries which are accessed by the shader.

Fixes particles affected by the wind in The Legend of Zelda: Breath of the Wild. May fix other weird issues with compute shaders in some games.

Try a bunch of games and drivers to make sure they don't blow up loading constants willynilly from searchable buffers.
2022-12-06 23:15:44 +00:00
bbb24d8c7e Restrict shader storage buffer search when match fails (#4011)
* Restrict storage buffer search when match fails

* Shader cache version bump
2022-12-05 19:11:32 +00:00
4523a73f75 Propagate Shader phi nodes with the same source value from all blocks (#3457)
* Propagate Shader phi nodes with the same source value from all incoming blocks

* Shader cache version bump
2022-07-12 00:36:58 +02:00
142cededd4 Implement Shader Instructions SUATOM and SURED (#2090)
* Initial Implementation

* Further improvements (no support for float/64-bit types)

* Merge atomic and reduce instructions, add missing format switch

* Fix rebase issues.

* Not used.

* Whoops. Fixed.

* Partial implementation of inc/dec, cleanup and TODOs

* Remove testing path

* Address Feedback
2021-08-31 02:51:57 -03:00
3b90adcd1d Fix shaders with mixed PBK and SSY addresses on the stack (#2329)
* Fix shaders with mixed PBK and SSY addresses on the stack

* Address PR feedback and nits
2021-06-03 01:41:53 +02:00
49745cfa37 Move shader resource descriptor creation out of the backend (#2290)
* Move shader resource descriptor creation out of the backend

* Remove now unused code, and other nits

* Shader cache version bump

* Nits

* Set format for bindless image load/store

* Fix buffer write flag
2021-05-19 23:15:26 +02:00
40e276c9b5 Improve shader global memory to storage pass (#2200)
* Improve shader global memory to storage pass

* Formatting and more comments

* Shader cache version bump
2021-04-18 12:31:39 +02:00
053dcfdb05 Use multiple dest operands for shader call instructions (#1975)
* Use multiple dest operands for shader call instructions

* Passing opNode is no longer needed
2021-02-01 11:13:38 +11:00
4b7c7dab9e Support multiple destination operands on shader IR and shuffle predicates (#1964)
* Support multiple destination operands on shader IR and shuffle predicates

* Cache version change
2021-01-28 10:59:47 +11:00
49f970d5bd Implement CAL and RET shader instructions (#1618)
* Add support for CAL and RET shader instructions

* Remove unused stuff

* Fix a bug that could cause the wrong values to be passed to a function

* Avoid repopulating function id dictionary every time

* PR feedback

* Fix vertex shader A/B merge
2020-10-25 17:00:44 -03:00
e13154c83d Implement shader LEA instruction and improve bindless image load/store (#1355) 2020-07-04 01:48:44 +02:00
5795bb1528 Support separate textures and samplers (#1216)
* Support separate textures and samplers

* Add missing bindless flag, fix SNORM format on buffer textures

* Add missing separation

* Add comments about the new handles
2020-05-27 16:07:10 +02:00
gdk
6a98c643ca Add a pass to turn global memory access into storage access, and do all storage related transformations on IR 2020-01-09 02:13:00 +01:00
gdk
769c02235f Add ATOMS, LDS, POPC, RED, STS and VOTE shader instructions, start changing the way how global memory is handled 2020-01-09 02:13:00 +01:00
gdk
3ab5c23f49 Add partial support for array of samplers, and add pass to identify them from bindless texture accesses 2020-01-09 02:13:00 +01:00
gdk
278a4c317c Implement BFI, BRK, FLO, FSWZADD, PBK, SHFL and TXD shader instructions, misc. fixes 2020-01-09 02:13:00 +01:00
gdk
1876b346fe Initial work 2020-01-09 02:13:00 +01:00