Add `requirePrelude()` intrinsic function. (#3250)

* Add `requirePrelude()` intrinsic function.

* Fix.

---------

Co-authored-by: Yong He <yhe@nvidia.com>

1 files changed, 38 insertions, 0 deletions

diff --git a/tests/cuda/require-prelude.slang b/tests/cuda/require-prelude.slang
new file mode 100644
index 000000000..a5e7fb19b
--- /dev/null
+++ b/tests/cuda/require-prelude.slang
@@ -0,0 +1,38 @@
+
+//TEST(compute):COMPARE_COMPUTE(filecheck-buffer=CHECK):-cpu -compute -output-using-type
+//TEST(compute):COMPARE_COMPUTE(filecheck-buffer=CHECK):-cuda -compute -output-using-type 
+//TEST(compute):COMPARE_COMPUTE(filecheck-buffer=CHECK):-cpu -compute -output-using-type
+
+int myIntrinsicFunc()
+{
+    __target_switch
+    {
+    case cpp:
+    case hlsl:
+    case cuda:
+        __requirePrelude(R"(#define MY_CUDA_INTRINSIC 100)");
+        __intrinsic_asm "(MY_CUDA_INTRINSIC)";
+    }
+}
+
+int myIntrinsicFunc1()
+{
+    __target_switch
+    {
+    case cpp:
+    case hlsl:
+    case cuda:
+        __requirePrelude(R"(#define MY_CUDA_INTRINSIC 100)");
+        __intrinsic_asm "(MY_CUDA_INTRINSIC + 1)";
+    }
+}
+
+//TEST_INPUT:ubuffer(data=[0 0 0 0], stride=4):out,name=outputBuffer
+RWStructuredBuffer<uint> outputBuffer;
+
+[numthreads(1, 1, 1)]
+void computeMain(uint3 dispatchThreadID : SV_DispatchThreadID)
+{
+    // CHECK: 201
+    outputBuffer[0] = myIntrinsicFunc() + myIntrinsicFunc1();
+}