* Fix flops calculation and tensor b stride calculation in the example 36 * Fix datatype * Update gather_scatter_fusion.cu