Load generic kernels from a single native device image

cuda-core/Cargo.toml

@@ -17,7 +17,7 @@ cuda-bindings = { version = "=0.4.0", path = "../cuda-bindings" }
 cuda-core-derive = { version = "=0.4.0", path = "../cuda-core-derive" }
 anyhow = "1"
 half = { version = "2" }
-oxide-artifacts = { version = "0.2.1", features = ["object-read"] }
+oxide-artifacts = { version = "0.2.1", path = "../cuda-oxide/crates/oxide-artifacts", features = ["object-read"] }
 
 [dev-dependencies]
 trybuild = "1"

cuda-core/src/simt/embedded.rs

@@ -46,8 +46,19 @@ impl EmbeddedModule {
     pub fn load(&self, ctx: &Arc<CudaContext>) -> Result<Arc<CudaModule>, EmbeddedModuleError> {
         let image =
             loadable_payload(&self.bundle).expect("EmbeddedModule always has a loadable payload");
-        ctx.load_module_from_image(image)
-            .map_err(EmbeddedModuleError::Driver)
+        ctx.load_module_from_image_with_launch_facts(
+            image,
+            self.bundle
+                .entries
+                .iter()
+                .filter_map(|entry| {
+                    entry
+                        .launch_facts
+                        .map(|facts| (entry.symbol.clone(), facts))
+                })
+                .collect(),
+        )
+        .map_err(EmbeddedModuleError::Driver)
     }
 }