Hi, I have a question about tvm computation, can we generated the target code without vector instructions, even I write the code like this:
n = 1
A = te.placeholder((n,), name="A")
B = te.placeholder((n,), name="B")
C = te.compute(A.shape, lambda i: A[i] + B[i], name="C")
it still generated vector instructions in the final executable program, my new hardware environment doesn’t support vector instructions, and I want to verify this new hardware environment.
I need a simple function to be used to get a tvm module, then I can use c++ loading this module to get a executable program, and this executable program can’t have vector instruction.
Right now, with my function, llvm will generate the object file (myfunction.o) that contains vector instruction, could anybody give me some advice ? Thanks!