[QST][CuTeDSL] How to set __launch_bound__ / .maxntid in cute dsl? #2698
Description
Are there equivalents of __launch_bound__ or .maxntid in cute dsl?
For a compiled kernel of cute dsl, even if I use cuda.arch.warpgroup_reg_alloc/dealloc, the register number I get from CUfunction_attribute.CU_FUNC_ATTRIBUTE_NUM_REGS is still not fixed as expected. So I guess the inner compiler in cute dsl dosen't provide thread number hint to help derive register number.