Cuda implementation for copying data around.

src/cuda_backend.rs

@@ -381,4 +381,42 @@ impl CudaStorage {
         let device = dev.clone();
         Ok(Self { slice, device })
     }
+
+    pub(crate) fn copy_strided_src(
+        &self,
+        dst: &mut Self,
+        src_shape: &Shape,
+        src_stride: &[usize],
+        dst_offset: usize,
+    ) -> Result<()> {
+        let dims = src_shape.dims();
+        let el_count = src_shape.elem_count();
+        let cfg = LaunchConfig::for_num_elems(el_count as u32);
+        let dev = &self.device;
+        let ds = dev.htod_copy([dims, src_stride].concat())?;
+        match (&self.slice, &mut dst.slice) {
+            (CudaStorageSlice::F32(src), CudaStorageSlice::F32(dst)) => {
+                let func = dev.get_or_load_func("ucopy_f32", kernels::UNARY)?;
+                let mut dst = dst.slice_mut(dst_offset..);
+                // SAFETY: Set later by running the kernel.
+                let params = (el_count, dims.len(), &ds, src, &mut dst);
+                // SAFETY: ffi.
+                unsafe { func.launch(cfg, params) }?
+            }
+            (CudaStorageSlice::F64(src), CudaStorageSlice::F64(dst)) => {
+                let func = dev.get_or_load_func("ucopy_64", kernels::UNARY)?;
+                let mut dst = dst.slice_mut(dst_offset..);
+                // SAFETY: Set later by running the kernel.
+                let params = (el_count, dims.len(), &ds, src, &mut dst);
+                // SAFETY: ffi.
+                unsafe { func.launch(cfg, params) }?;
+            }
+            _ => {
+                return Err(CudaError::InternalError(
+                    "dtype mismatch in copy_strided op",
+                ))
+            }
+        }
+        Ok(())
+    }
 }