Added `[AutoPyBindCUDA]` for automatic kernel binding + `[PyExport]` for exporting type information (#3209)

* Initial: add a DiffTensor impl

* Auto-binding and diff tensor implementations now work

* Refactored diff-tensor implementation + added py-export for struct types

* Cleanup

* Update slang-ir-pytorch-cpp-binding.cpp

* Updated test names

* Update autodiff-data-flow.slang.expected

* Add more versions of load/store & default generic args for DiffTensorView.

* Add diagnostic for default generic arg and more tests

* Add more `[AutoPyBind]` tests

2 files changed, 53 insertions, 3 deletions

diff --git a/tests/diagnostics/autodiff-data-flow.slang.expected b/tests/diagnostics/autodiff-data-flow.slang.expected
index 301f84985..6840bfd3c 100644
--- a/tests/diagnostics/autodiff-data-flow.slang.expected
+++ b/tests/diagnostics/autodiff-data-flow.slang.expected
@@ -3,9 +3,6 @@ standard error = {
 tests/diagnostics/autodiff-data-flow.slang(15): error 41020: derivative cannot be propagated through call to non-forward-differentiable function `nonDiff`, use 'no_diff' to clarify intention.
         val = nonDiff(x * 2.0f);
                      ^
-tests/diagnostics/autodiff-data-flow.slang(22): error 41021: a differentiable function must have at least one differentiable output.
-void g(float x)
-     ^
 tests/diagnostics/autodiff-data-flow.slang(28): error 30510: loops inside a differentiable function need to provide either '[MaxIters(n)]' or '[ForceUnroll]' attribute.
     for (int i = 0; i < 5; i++) // Not ok, we can't infer the loop iterations because the body modifies induction var.
     ^~~
diff --git a/tests/diagnostics/generic-incorrect-default-arg.slang b/tests/diagnostics/generic-incorrect-default-arg.slang
new file mode 100644
index 000000000..14192293c
--- /dev/null
+++ b/tests/diagnostics/generic-incorrect-default-arg.slang
@@ -0,0 +1,53 @@
+//TEST:SIMPLE(filecheck=CHECK): -target hlsl -profile cs_5_0 -entry computeMain -line-directive-mode none
+
+// Check that user code can declare and use a generic
+// `struct` type.
+
+//TEST_INPUT:ubuffer(data=[0 0 0 0], stride=4):out,name=outputBuffer
+RWStructuredBuffer<int> outputBuffer;
+
+interface ITest
+{
+	int doThing(int x);
+};
+
+struct Impl1 : ITest
+{
+	int doThing(int x)
+	{
+		return x * 2;
+	}
+};
+
+struct Impl2
+{
+    int doSomethingElse(int x)
+    {
+        return x * 3;
+    }
+};
+
+__generic<T : ITest = Impl2>
+// CHECK: tests/diagnostics/generic-incorrect-default-arg.slang([[@LINE-1]]): error 38029: type argument 'Impl2' does not conform to the required interface 'ITest'
+struct GenStruct
+{
+	T obj;
+};
+
+int test(GenStruct gs, int val)
+{
+	return gs.obj.doThing(val);
+}
+
+[numthreads(4, 1, 1)]
+void computeMain(uint3 dispatchThreadID : SV_DispatchThreadID)
+{
+	int tid = dispatchThreadID.x;
+
+	int outVal = 0;
+
+	GenStruct<Impl1> gs;
+	outVal += test(gs, tid);
+
+	outputBuffer[tid] = outVal;
+}
\ No newline at end of file