Design of MAC unit (dsp processors) using VHDL -

my project design of 32bit mac(multiply , accumlate) unit using reversible logic. project , have designed 32bit mulitplier , 64 bit adder using reversible logic. now, in next step want design 64 bit accumlator takes value adder , stores , adds previous value present in it. not getting idea how design accumlator. please in completion of project.

a basic vhdl accumulator can implemented in few lines of code. how you decide implement it, , additional features necessary going depend on specific requirements.

for example:

• are inputs signed or unsigned?
• what type of inputs?
• does accumulator saturate, or roll over?

here sample unsigned accumulator give idea of need implement (based on source):

library ieee; use ieee.std_logic_1164.all; use ieee.numeric_std.all;  entity accumulator   port (     din:  in  std_logic_vector(3 downto 0);     clk:  in  std_logic;     rst:  in  std_logic;     dout: out std_logic_vector(3 downto 0)   ); end entity accumulator;  architecture behave of accumulator    signal acc_value : std_logic_vector(3 downto 0);  begin    process(clk)   begin     if rising_edge(clk)       if rst='1'         acc_value <= (others => '0'); -- reset accumulated value 0       else         acc_value <= std_logic_vector( unsigned(acc_value) + unsigned(din) );       end if;     end if;   end process;    -- assign output   dout <= acc_value;  end behave; 

to describe design in words: every clock cycle on rising edge, data input din interpreted unsigned value, , added accumulated value acc_value. if rst input asserted, instead of accumulating din input, accumulated value cleared 0. value of accumulator presented on output of block, dout.

based on interfacing with, might want consider following changes/modifications:

• perhaps din should signed or unsigned types instead of std_logic_vector. recommend this, depends on how representing values in other places of design.
• dout signed or unsigned value instead of std_logic_vector - depends requirements.
• in case, acc_value, accumulated value register, rollover if values accumulated high. maybe want generate error condition when happens, or perform check ensure saturate @ maximum value of acc_value instead.
• acc_value need not same width din -- twice wide (or whatever requirements are). wider is, more can accumulate before rollover condition occurs.