Untitled

{"evidence_level":"手算、功能执行","aos_soa_model":{"n":4096,"line_bytes":64,"cache_bytes":32768,"aos_loads":4096,"aos_misses":1024,"soa_loads":4096,"soa_misses":256,"interpretation":"reading one float from a 16-byte struct uses one line per 4 points; contiguous float array uses one line per 16 values"},"blocking_model":{"n":192,"block":32,"elem_bytes":4,"plain_b_row_bytes":768,"blocked_a_tile_bytes":4096,"blocked_b_tile_bytes":4096,"blocked_c_tile_bytes":4096,"three_tile_working_set_bytes":12288,"scope":"working-set arithmetic only; not a cache-cycle simulator"}}