METIS is awesome and very useful for discrete math community. Now I am building a software dealing with dynamic weights of computation ML graph. Our operations and tensor are highly optimized and have shared codes in hardware, hence summation of the cost is dynamic. Also we have topical order constrain.
We have develop a solver. But to utilize the Multi-level framework, one of useful things I am trying to understand is how "idx_t Match_SHEM(ctrl_t ctrl, graph_t graph)" and "refine_KwayPartition" is coded. Is there any explanation here ? @karypis
METIS is awesome and very useful for discrete math community. Now I am building a software dealing with dynamic weights of computation ML graph. Our operations and tensor are highly optimized and have shared codes in hardware, hence summation of the cost is dynamic. Also we have topical order constrain.
We have develop a solver. But to utilize the Multi-level framework, one of useful things I am trying to understand is how "idx_t Match_SHEM(ctrl_t ctrl, graph_t graph)" and "refine_KwayPartition" is coded. Is there any explanation here ? @karypis