Making it easier to work with shaders https://git.yummers.dev/slang.git/atom?h=master 2022-05-18T17:57:37+00:00 2022-05-18T17:57:37+00:00 Alexey Panteleev alpanteleev@nvidia.com 2022-05-18T17:57:37+00:00 urn:sha1:69cb6e8f300d77e74bd2c7dfe15d12e10b38f512 See https://github.com/shader-slang/slang/issues/2213 2018-11-29T15:48:38+00:00 Tim Foley tfoleyNV@users.noreply.github.com 2018-11-29T15:48:38+00:00 urn:sha1:e5cc4660c634a0dd35a9813e03192d380f253332 The `Type` infrastructure uses a class hierarchy, but blindly `dynamic_cast`ing to a desired case doesn't always give the expected result, because a `Type` could represent a `typedef` (a `NamedExpressionType`) that itself resolves to, e.g, a vector type (a `VectorExpressionType`). In that case a `dynamic_cast<VectorExpressionType*>(someType)` would fail, even though the type logically represents a vector. The `Type::As<T>()` method is designed to handle this case, by "looking through" simple `typedef`s to get at the real definition of a type. The fix in this case is to use `Type::As<T>()` at various points in the reflection code (`reflection.cpp`) instead of `dynamic_cast`. This problem surfaced with a `StructuredBuffer<float2>` not reflecting correctly, because the element type (`float2`) is actually a `typedef` (for `vector<float,2>`), so I've included a test case that stresses that case. Getting the right output in the test required tweaking the `slang-reflection-test` tool to produce additional output for resource types (currently narrowed down to only affect structured buffers to avoid large diffs in expected test outputs).