summaryrefslogtreecommitdiffstats
path: root/tests
diff options
context:
space:
mode:
authorYong He <yonghe@outlook.com>2023-03-17 15:57:22 -0700
committerGitHub <noreply@github.com>2023-03-17 15:57:22 -0700
commit7f11f883d0781952f002b3aa3222a3aa0040f18a (patch)
tree08eaf10fef39211fbc3f124679bfe8a35775a5a7 /tests
parent4b55bf6d75bdeed087728505a1c9b43d3a99af8d (diff)
Add support for emitting cuda kernel and host functions. (#2712)
* Add support for emitting cuda kernel and host functions. * Update test. * Fix cuda preamble emit. --------- Co-authored-by: Yong He <yhe@nvidia.com>
Diffstat (limited to 'tests')
-rw-r--r--tests/autodiff/cuda-kernel-export.slang14
1 files changed, 13 insertions, 1 deletions
diff --git a/tests/autodiff/cuda-kernel-export.slang b/tests/autodiff/cuda-kernel-export.slang
index 0db4d8cea..54442498b 100644
--- a/tests/autodiff/cuda-kernel-export.slang
+++ b/tests/autodiff/cuda-kernel-export.slang
@@ -1,4 +1,4 @@
-//DISABLED_TEST:SIMPLE: -target cuda -line-directive-mode none
+//DISABLE_TEST:SIMPLE: -target cuda -line-directive-mode none
// Verify that we can output a cuda device function with [CudaDeviceExport].
// Disabled until we have FileCheck.
@@ -27,3 +27,15 @@ void diffF(inout DifferentialPair<MixedType> m, float dout)
{
__bwd_diff(f)(m, dout);
}
+
+[CudaKernel]
+void myKernel(float* inValues, float* outValues)
+{
+ outValues[0] = sin(inValues[0]);
+}
+
+[CudaHost]
+public __extern_cpp void runCompute(float *inValues, float *outValues, uint3 dispathcSize)
+{
+ __dispatch_kernel(myKernel, uint3(128, 1, 1), dispathcSize)(inValues, outValues);
+} \ No newline at end of file