[SYCL] Enable native FP add/sub tests for CUDA (#169)

Signed-off-by: John Pennycook <john.pennycook@intel.com>

Author: Pennycook

SYCL/AtomicRef/add-native.cpp

@@ -1,11 +1,12 @@
 // RUN: %clangxx -fsycl -fsycl-unnamed-lambda -DSYCL_USE_NATIVE_FP_ATOMICS \
-// RUN: -fsycl-targets=%sycl_triple %s -o %t.out
+// RUN: -fsycl-targets=%sycl_triple %s -o %t.out \
+// RUN: -Xsycl-target-backend=nvptx64-nvidia-cuda-sycldevice --cuda-gpu-arch=sm_60
 // RUN: %HOST_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // TODO: Remove items from UNSUPPORTED once corresponding backends support
 // "native" implementation
-// UNSUPPORTED: cpu, cuda
+// UNSUPPORTED: cpu
 
 #include <CL/sycl.hpp>
 #include <algorithm>

SYCL/AtomicRef/sub-native.cpp

@@ -1,11 +1,12 @@
 // RUN: %clangxx -fsycl -fsycl-unnamed-lambda -DSYCL_USE_NATIVE_FP_ATOMICS \
-// RUN: -fsycl-targets=%sycl_triple %s -o %t.out
+// RUN: -fsycl-targets=%sycl_triple %s -o %t.out \
+// RUN: -Xsycl-target-backend=nvptx64-nvidia-cuda-sycldevice --cuda-gpu-arch=sm_60
 // RUN: %HOST_RUN_PLACEHOLDER %t.out
 // RUN: %GPU_RUN_PLACEHOLDER %t.out
 // RUN: %CPU_RUN_PLACEHOLDER %t.out
 // TODO: Remove items from UNSUPPORTED once corresponding backends support
 // "native" implementation
-// UNSUPPORTED: cpu, cuda
+// UNSUPPORTED: cpu
 
 #include <CL/sycl.hpp>
 #include <algorithm>