mirror of
git://projects.qi-hardware.com/nn-usb-fpga.git
synced 2025-04-21 12:27:27 +03:00
Adding plasma example
This commit is contained in:
139
plasma/logic/pipeline.vhd
Normal file
139
plasma/logic/pipeline.vhd
Normal file
@@ -0,0 +1,139 @@
|
||||
---------------------------------------------------------------------
|
||||
-- TITLE: Pipeline
|
||||
-- AUTHOR: Steve Rhoads (rhoadss@yahoo.com)
|
||||
-- DATE CREATED: 6/24/02
|
||||
-- FILENAME: pipeline.vhd
|
||||
-- PROJECT: Plasma CPU core
|
||||
-- COPYRIGHT: Software placed into the public domain by the author.
|
||||
-- Software 'as is' without warranty. Author liable for nothing.
|
||||
-- DESCRIPTION:
|
||||
-- Controls the three stage pipeline by delaying the signals:
|
||||
-- a_bus, b_bus, alu/shift/mult_func, c_source, and rs_index.
|
||||
---------------------------------------------------------------------
|
||||
library ieee;
|
||||
use ieee.std_logic_1164.all;
|
||||
use work.mlite_pack.all;
|
||||
|
||||
--Note: sigD <= sig after rising_edge(clk)
|
||||
entity pipeline is
|
||||
port(clk : in std_logic;
|
||||
reset : in std_logic;
|
||||
a_bus : in std_logic_vector(31 downto 0);
|
||||
a_busD : out std_logic_vector(31 downto 0);
|
||||
b_bus : in std_logic_vector(31 downto 0);
|
||||
b_busD : out std_logic_vector(31 downto 0);
|
||||
alu_func : in alu_function_type;
|
||||
alu_funcD : out alu_function_type;
|
||||
shift_func : in shift_function_type;
|
||||
shift_funcD : out shift_function_type;
|
||||
mult_func : in mult_function_type;
|
||||
mult_funcD : out mult_function_type;
|
||||
reg_dest : in std_logic_vector(31 downto 0);
|
||||
reg_destD : out std_logic_vector(31 downto 0);
|
||||
rd_index : in std_logic_vector(5 downto 0);
|
||||
rd_indexD : out std_logic_vector(5 downto 0);
|
||||
|
||||
rs_index : in std_logic_vector(5 downto 0);
|
||||
rt_index : in std_logic_vector(5 downto 0);
|
||||
pc_source : in pc_source_type;
|
||||
mem_source : in mem_source_type;
|
||||
a_source : in a_source_type;
|
||||
b_source : in b_source_type;
|
||||
c_source : in c_source_type;
|
||||
c_bus : in std_logic_vector(31 downto 0);
|
||||
pause_any : in std_logic;
|
||||
pause_pipeline : out std_logic);
|
||||
end; --entity pipeline
|
||||
|
||||
architecture logic of pipeline is
|
||||
signal rd_index_reg : std_logic_vector(5 downto 0);
|
||||
signal reg_dest_reg : std_logic_vector(31 downto 0);
|
||||
signal reg_dest_delay : std_logic_vector(31 downto 0);
|
||||
signal c_source_reg : c_source_type;
|
||||
signal pause_enable_reg : std_logic;
|
||||
begin
|
||||
|
||||
--When operating in three stage pipeline mode, the following signals
|
||||
--are delayed by one clock cycle: a_bus, b_bus, alu/shift/mult_func,
|
||||
--c_source, and rd_index.
|
||||
pipeline3: process(clk, reset, a_bus, b_bus, alu_func, shift_func, mult_func,
|
||||
rd_index, rd_index_reg, pause_any, pause_enable_reg,
|
||||
rs_index, rt_index,
|
||||
pc_source, mem_source, a_source, b_source, c_source, c_source_reg,
|
||||
reg_dest, reg_dest_reg, reg_dest_delay, c_bus)
|
||||
variable pause_mult_clock : std_logic;
|
||||
variable freeze_pipeline : std_logic;
|
||||
begin
|
||||
if (pc_source /= FROM_INC4 and pc_source /= FROM_OPCODE25_0) or
|
||||
mem_source /= MEM_FETCH or
|
||||
(mult_func = MULT_READ_LO or mult_func = MULT_READ_HI) then
|
||||
pause_mult_clock := '1';
|
||||
else
|
||||
pause_mult_clock := '0';
|
||||
end if;
|
||||
|
||||
freeze_pipeline := not (pause_mult_clock and pause_enable_reg) and pause_any;
|
||||
pause_pipeline <= pause_mult_clock and pause_enable_reg;
|
||||
rd_indexD <= rd_index_reg;
|
||||
|
||||
-- The value written back into the register bank, signal reg_dest is tricky.
|
||||
-- If reg_dest comes from the ALU via the signal c_bus, it is already delayed
|
||||
-- into stage #3, because a_busD and b_busD are delayed. If reg_dest comes from
|
||||
-- c_memory, pc_current, or pc_plus4 then reg_dest hasn't yet been delayed into
|
||||
-- stage #3.
|
||||
-- Instead of delaying c_memory, pc_current, and pc_plus4, these signals
|
||||
-- are multiplexed into reg_dest which is then delayed. The decision to use
|
||||
-- the already delayed c_bus or the delayed value of reg_dest (reg_dest_reg) is
|
||||
-- based on a delayed value of c_source (c_source_reg).
|
||||
|
||||
if c_source_reg = C_FROM_ALU then
|
||||
reg_dest_delay <= c_bus; --delayed by 1 clock cycle via a_busD & b_busD
|
||||
else
|
||||
reg_dest_delay <= reg_dest_reg; --need to delay 1 clock cycle from reg_dest
|
||||
end if;
|
||||
reg_destD <= reg_dest_delay;
|
||||
|
||||
if reset = '1' then
|
||||
a_busD <= ZERO;
|
||||
b_busD <= ZERO;
|
||||
alu_funcD <= ALU_NOTHING;
|
||||
shift_funcD <= SHIFT_NOTHING;
|
||||
mult_funcD <= MULT_NOTHING;
|
||||
reg_dest_reg <= ZERO;
|
||||
c_source_reg <= "000";
|
||||
rd_index_reg <= "000000";
|
||||
pause_enable_reg <= '0';
|
||||
elsif rising_edge(clk) then
|
||||
if freeze_pipeline = '0' then
|
||||
if (rs_index = "000000" or rs_index /= rd_index_reg) or
|
||||
(a_source /= A_FROM_REG_SOURCE or pause_enable_reg = '0') then
|
||||
a_busD <= a_bus;
|
||||
else
|
||||
a_busD <= reg_dest_delay; --rs from previous operation (bypass stage)
|
||||
end if;
|
||||
|
||||
if (rt_index = "000000" or rt_index /= rd_index_reg) or
|
||||
(b_source /= B_FROM_REG_TARGET or pause_enable_reg = '0') then
|
||||
b_busD <= b_bus;
|
||||
else
|
||||
b_busD <= reg_dest_delay; --rt from previous operation
|
||||
end if;
|
||||
|
||||
alu_funcD <= alu_func;
|
||||
shift_funcD <= shift_func;
|
||||
mult_funcD <= mult_func;
|
||||
reg_dest_reg <= reg_dest;
|
||||
c_source_reg <= c_source;
|
||||
rd_index_reg <= rd_index;
|
||||
end if;
|
||||
|
||||
if pause_enable_reg = '0' and pause_any = '0' then
|
||||
pause_enable_reg <= '1'; --enable pause_pipeline
|
||||
elsif pause_mult_clock = '1' then
|
||||
pause_enable_reg <= '0'; --disable pause_pipeline
|
||||
end if;
|
||||
end if;
|
||||
|
||||
end process; --pipeline3
|
||||
|
||||
end; --logic
|
||||
Reference in New Issue
Block a user