Update movmatrix.cu

Update LIT test.

Author: jiejanezhang

clang/test/dpct/asm/movmatrix.cu

@@ -1,6 +1,6 @@
 // UNSUPPORTED: cuda-8.0, cuda-9.0, cuda-9.1, cuda-9.2, cuda-10.0, cuda-10.1, cuda-10.2
 // UNSUPPORTED: v8.0, v9.0, v9.1, v9.2, v10.0, v10.1, v10.2
-// RUN: dpct --format-range=none -out-root %T/mov %s --cuda-include-path="%cuda-path/include" -- -std=c++14 -x cuda --cuda-host-only
+// RUN: dpct --format-range=none -out-root %T/movmatrix %s --cuda-include-path="%cuda-path/include" -- -std=c++14 -x cuda --cuda-host-only
 // RUN: FileCheck %s --match-full-lines --input-file %T/movmatrix/movmatrix.dp.cpp
 // RUN: %if build_lit %{icpx -c -fsycl %T/movmatrix/movmatrix.dp.cpp -o %T/movmatrix/movmatrix.dp.o %}
 
@@ -18,10 +18,9 @@ using bf16_2 = __nv_bfloat162;
 // Only .m8n8.b16
 // 
 
-__device__ inline void movmatrix(bf16_2 &dst, const bf16_2 &src) {
-
-  // CHECK: dpct::experimental::matrix::movmatrix(*(uint32_t *)(&dst), (*(uint32_t *)(&src)));;
+__global__ void movmatrix(bf16_2 &dst, const bf16_2 &src) {
 
+  // CHECK: dpct::experimental::matrix::movmatrix(*(uint32_t *)(&dst), (*(uint32_t *)(&src)));
   asm volatile("movmatrix.sync.aligned.m8n8.trans.b16 %0, %1;\n"
                : "+r"(*(uint32_t *)(&dst))
                : "r"(*(uint32_t *)(&src)));