Hi,
I construct a new funcion to dispatch the kernel func. The call relationship is like below:
Module (A, prim_func_a)
A {
if (n > a) {
tir.tvm_call_packed("A_1", tir.stack_make_array(placeholder. shape, shape, 2, 0f, 0)
}
else {
tir.tvm_call_packed("A_2", tir.stack_make_array(placeholder. shape, shape, 2, 0f, 0)
}
Module (A_1, PrimFunc A_1)
...
Module ( A_2, primFunc A_2)
...
The Codegen process looks good to me. But when I get the data back from GPU, it seems that func A did not call A_1 or A_2 correctly. Could anyone help me about it ? Thanks in advance. Best.