Le 2014-08-08 15:42, jim@synthworks.com a écrit : > Has anyone tested the performance of bit_vector? Excellent question. and then these numbers should be contrasted with - other types (std_logic_vector, integer) - different simulators and vendors (many forbid you this !) - different languages I did a testbench some years ago for integer-intensive computations, using Wikipedia's SHA pseudocode. It's the archetypic example of integer-bound computations, with a profile that looks like usual case for SoC/Computers. http://ygdes.com/GHDL/int_bool/ it tests these types : integer, integer with range, std_ulogic_vector, bit_vector but I only remember that I got a speedup of 400 by adding bool/shift to integers in GHDL. I think that std_logic_vector was more or less as fast as bit_vector, but this can not be generalised to other simulators (VHPI implementation is certainly not portable). GHDL does not "optimise" or "pack" bit ectors, AFAIK. The script and all the files are available, so i could just re-run it. Is anybody willing to test it with other simulators ? (if their NDA allows..........) YG -- This message has been scanned for viruses and dangerous content by MailScanner, and is believed to be clean.Received on Fri Aug 8 07:05:43 2014
This archive was generated by hypermail 2.1.8 : Fri Aug 08 2014 - 07:06:01 PDT