fix bs 2, start tile num calc

Author: Amir-19

flashinfer/cute_dsl/blockscaled_gemm.py

@@ -2927,6 +2927,8 @@ def __call__(
             ),
         ) if c_mc_ptr is not None else None
         #TODO(asamani): urgent fix this is just for dev
+        # this should be calculated based on how many total tiles we need to work
+        # on
         barrier_flag_tensor = cute.make_tensor(
             barrier_flag_ptr,
             layout=cute.make_ordered_layout((404,), order=(0,)),