diff options
| author | Yong He <yonghe@outlook.com> | 2021-10-18 12:19:45 -0700 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2021-10-18 12:19:45 -0700 |
| commit | 2f44d9e01234911dd563f0456b9d861fd8db286d (patch) | |
| tree | e2727f31654ebc93bae6a1de4b25586be6fb9d10 /tools/gfx-unit-test/mutable-shader-object.slang | |
| parent | 87e7c49fbfccd54be0d1cee61fba8f309b1f792e (diff) | |
GFX: implement mutable shader objects. (#1963)
* GFX: implement mutable shader objects.
* Revert unnecessary changes
* Revert more changes.
* Fix clang errors.
* Fix clang/gcc errors.
* Fix clang errors.
* Remove CPU test.
* Fix after merge.
* Fix after merge.
* Remove gl test
* Code review fixes.
* Fixing all vk validation errors.
* Flush test output more often.
* Fix a crash in `specializeDynamicAssociatedTypeLookup`.
* temporarily disable std-lib-serialize test to see what happens
* Fix crashes.
* Make sure cpu gfx unit tests are properly disabled on TeamCity.
* Disable cpu test.
* Fix.
* Fix cuda.
* Disable nv-ray-tracing-motion-blur
Co-authored-by: Yong He <yhe@nvidia.com>
Diffstat (limited to 'tools/gfx-unit-test/mutable-shader-object.slang')
| -rw-r--r-- | tools/gfx-unit-test/mutable-shader-object.slang | 68 |
1 files changed, 68 insertions, 0 deletions
diff --git a/tools/gfx-unit-test/mutable-shader-object.slang b/tools/gfx-unit-test/mutable-shader-object.slang new file mode 100644 index 000000000..42d520f13 --- /dev/null +++ b/tools/gfx-unit-test/mutable-shader-object.slang @@ -0,0 +1,68 @@ +// mutable-shader-object.slang + +// This is a copy of `shader-object.slang` in `shader-object` example +// for use by mutable-shader-object gfx unit test. + +// This file implements a simple compute shader that transforms +// input floating point numbers stored in a `RWStructuredBuffer`. +// Specifically, for each number x from input buffer, compute +// f(x) and store the result back in the same buffer. + +// The compute shader supports multiple transformation functions, +// such add(x, c) which returns x+c, or mul(x, c) which returns x*c. +// This functions are implemented as types that conforms to the +// `ITransformer` interface. + +// The main entry point function takes a parameter of `ITransformer` +// type, and applies the transformation to numbers in the input +// buffer. By defining the shader parameter using interfaces, +// we enable the flexiblity to generate either specialized compute +// kernels that performs specific transformation or a general +// kernel that can perform any transformations encoded by the +// parameter at run-time, without changing any shader code or +// host-application logic for setting and preparing shader parameters. + +// Defines the transformer interface, which implements a single +// `transform` operation. +interface ITransformer +{ + float transform(float x); +} + +// Represents a transform function f(x) = x + c. +struct AddTransformer : ITransformer +{ + float c; + float transform(float x) { return x + c; } +}; + +// Represents a transform function f(x) = x * c. +struct MulTransformer : ITransformer +{ + float c; + float transform(float x) { return x * c; } +}; + +// Represents a composite function f(x) = f0(f1(x)); +struct CompositeTransformer : ITransformer +{ + ITransformer func0; + ITransformer func1; + float transform(float x) + { + return func0.transform(func1.transform(x)); + } +}; + +// Main entry-point. Applies the transformation encoded by `transformer` +// to all elements in `buffer`. +[shader("compute")] +[numthreads(4,1,1)] +void computeMain( + uint3 sv_dispatchThreadID : SV_DispatchThreadID, + uniform RWStructuredBuffer<float> buffer, + uniform ITransformer transformer) +{ + var input = buffer[sv_dispatchThreadID.x]; + buffer[sv_dispatchThreadID.x] = transformer.transform(input); +} |
