v4.2 release. (#2587)

* Fix default cluster callback values to 1 to avoid profiler failure when these values are not set in command line.

* v4.2 release.

test/unit/cute/ampere/cooperative_gemm.cu

@@ -609,6 +609,8 @@ TEST(SM89_CuTe_Ada, CooperativeGemm_e5m2e5m2f32_MMA) {
   test_cooperative_gemm_col_major_layout<thread_block_size, MaxVecBits, TA, TB, TC>(shape_mnk, tiled_mma);
 }
 
+#if (__CUDACC_VER_MAJOR__ > 12) || (__CUDACC_VER_MAJOR__ == 12 && __CUDACC_VER_MINOR__ >= 8)
+
 TEST(SM89_CuTe_Ada, CooperativeGemm_e4m3e4m3f16_MMA) {
   using TA = cutlass::float_e4m3_t;
   using TB = cutlass::float_e4m3_t;
@@ -680,3 +682,5 @@ TEST(SM89_CuTe_Ada, CooperativeGemm_e5m2e5m2f16_MMA) {
 
   test_cooperative_gemm_col_major_layout<thread_block_size, MaxVecBits, TA, TB, TC>(shape_mnk, tiled_mma);
 }
+
+#endif

test/unit/cute/turing/movm.cu

@@ -44,7 +44,7 @@ __global__ void
 movm_test_device(uint16_t* g_in, uint16_t* g_out)
 {
   int tid = threadIdx.x;
-  
+
   // load input gmem -> register
   uint32_t reg = reinterpret_cast<uint32_t*>(g_in)[tid];
 
@@ -128,7 +128,7 @@ TEST(SM75_CuTe_Turing, Movm)
   //
   // CuTe MOVM
   //
-  
+
   {
   thrust::device_vector<uint16_t> d_out(count);