ecolab-nus / lisa Goto Github PK

Hi, I'm wondering where my_graph_loop.dot is？With CGRA-ME, I can get graph_loop.dot by make in the benchmark's folder. Is it correct to just rename it?
By the way, what's the meaning of param b, in "$CGRA_MAPPER -m 2 --II 20 --inef --arch_name cgra_me_4_4 -g ./benchmarks/microbench/conv2/my_graph_loop.dot b --xml ./arch/simple/target_arch/arch-homo-orth_4_4.xml --cgra_x 6 --cgra_y 6"? I think "$CGRA_MAPPER --help" does not eplain it.

First, let me talk about my experimental example. I run cholesky, its II is 1, and the cgra structure is arch-homo-orth_4_4.xml. I have the following doubts:

1. I found that some operators in the result are placed in the same PE, including the load operator and sub operator,and even there is no edge between these two operators. Why can this be achieved?
2. I found that the final mapping position of two operators contained in some edges is not adjacent, but across a grid or diagonal. How is this routed? After all, I am using arch-homo-orth_4_4.xml, which should not support crossing a grid or diagonal.

Is arch-homo-orth_4_4_leftmostmemory.xml the 4×4 CGRA with less memory connectivity in the paper? I find that arch-homo-orth_4_4_leftmostmemory.xml and arch-homo-orth_4_4.xml are identical. And I run the cholesky with arch-homo-orth_4_4_leftmostmemory.xml, it turns out that the load node is not placed on the left-most.

I have finished the task 'python3 run_exper.py gnn_lisa 0 0 13', but I cannot find the result file or folder. Would you please tell me how to solve this problem? Thank you.