Subversion Repositories ion

--------------------------------------------------------------------------------

-- mips_cache_stub.vhdl -- 1-word cache module

--

-- This module has the same interface and logic as a real cache but the cache

-- memory is just 1 word for each of code and data.

--

-- It interfaces the CPU to the following:

--

--  1.- Internal 32-bit-wide BRAM for read only

--  2.- Internal 32-bit I/O bus

--  3.- External 16-bit wide SRAM

--

-- The SRAM memory interface signals are meant to connect directly to FPGA pins

-- and all outputs are registered (tco should be minimal).

-- SRAM data inputs are NOT registered, though. They go through a couple muxes

-- before reaching the first register so watch out for tsetup.

-- The SRAM is assumed to be fast enough to read or write in a clock cycle.

--

-- Obviously this module provides no performance gain; on the contrary, by

-- coupling the CPU to slow external memory (16 bit bus) it actually slows it

-- down. The purpose of this module is just to test the SRAM interface and the

-- cache logic and timing.

--

--------------------------------------------------------------------------------

-- External FPGA signals

--

-- This module has signals meant to connect directly to FPGA pins: the SRAM

-- interface. They are either direct register outputs or at most with an 

-- intervening 2-mux, in order to minimize the Tco (clock-to-output).

--

-- The Tco of these signals has to be accounted for in the real SRAM interface.

-- For example, under Quartus-2 and with a Cyclone-2 grade -7 device, the

-- worst Tco for the SRAM data pins is below 5 ns, enough to use a 10ns SRAM

-- with a 20 ns clock cycle.

-- Anyway, you need to take care of this yourself.

--

--------------------------------------------------------------------------------

-- Interface to CPU

--

-- 1.- All signals coming from the CPU are registered.

-- 2.- All CPU inputs come directly from a register, or at most have a 2-mux in

--     between.

-- 

-- This means this block will not degrade the timing performance of the system, 

-- as long as its logic is shallower than the current bottleneck (the ALU).

--

--------------------------------------------------------------------------------

-- KNOWN TROUBLE:

-- 

-- Apart from the very rough looks of the code, there's a few known problems:

--

-- 1.- Write cycles too long

--      In order to guarantee setup and hold times for WE controlled write 

--      cycles, two extra clock cycles are inserted for each SRAM write access.

--      This is the most reliable way and the easiest but probably not the best.

--      Until I come up with something better, write cycles to SRAM are going

--      to be very slow.

-- 

-- 2.- Access to unmapped areas will crash the CPU

--      A couple states are missing in the state machine for handling accesses 

--      to unmapped areas. I haven't yet decided how to handle that (return 

--      zero, trigger trap, mirror another mapped area...).

--

-- 3.- Code refills from SRAM is unimplemented yet

--      To be done for sheer lack of time.

--

-- 4.- Does not work as a real 1-word cache yet

--      That functionality is still missing, all accesses 'miss'. It should be

--      implemented, as a way to test the real cache logic on a small scale.

--

--------------------------------------------------------------------------------

library ieee;

use ieee.std_logic_1164.all;

use ieee.std_logic_arith.all;

use ieee.std_logic_unsigned.all;

use work.mips_pkg.all;

entity mips_cache_stub is

    generic (

        BRAM_ADDR_SIZE : integer := 10;

        SRAM_ADDR_SIZE : integer := 17

    );

    port(

        clk             : in std_logic;

        reset           : in std_logic;

        -- Interface to CPU core

        data_rd_addr    : in std_logic_vector(31 downto 0);

        data_rd         : out std_logic_vector(31 downto 0);

        data_rd_vma     : in std_logic;

        code_rd_addr    : in std_logic_vector(31 downto 2);

        code_rd         : out std_logic_vector(31 downto 0);

        code_rd_vma     : in std_logic;

        data_wr_addr    : in std_logic_vector(31 downto 2);

        byte_we         : in std_logic_vector(3 downto 0);

        data_wr         : in std_logic_vector(31 downto 0);

        mem_wait        : out std_logic;

        cache_enable    : in std_logic;

        -- interface to FPGA i/o devices

        io_rd_data      : in std_logic_vector(31 downto 0);

        io_rd_addr      : out std_logic_vector(31 downto 2);

        io_wr_addr      : out std_logic_vector(31 downto 2);

        io_wr_data      : out std_logic_vector(31 downto 0);

        io_rd_vma       : out std_logic;

        io_byte_we      : out std_logic_vector(3 downto 0);

        -- interface to synchronous 32-bit-wide FPGA BRAM (possibly used as ROM)

        bram_rd_data    : in std_logic_vector(31 downto 0);

        bram_wr_data    : out std_logic_vector(31 downto 0);

        bram_rd_addr    : out std_logic_vector(BRAM_ADDR_SIZE+1 downto 2);

        bram_wr_addr    : out std_logic_vector(BRAM_ADDR_SIZE+1 downto 2);

        bram_byte_we    : out std_logic_vector(3 downto 0);

        bram_data_rd_vma: out std_logic;

        -- interface to asynchronous 16-bit-wide EXTERNAL SRAM

        sram_address    : out std_logic_vector(SRAM_ADDR_SIZE downto 1);

        sram_databus    : inout std_logic_vector(15 downto 0);

        sram_byte_we_n  : out std_logic_vector(1 downto 0);

        sram_oe_n       : out std_logic

    );

end entity mips_cache_stub;

architecture stub of mips_cache_stub is

-- Wait state counter -- we're supporting static memory from 10 to >100 ns

subtype t_wait_state_counter is std_logic_vector(2 downto 0);

-- state machines: definition of states -----------------------------

type t_code_cache_state is (

    code_normal,                -- 

    code_wait_for_dcache,       -- wait for D-cache to stop using the buses

    code_refill_bram_0,         -- pc in bram_rd_addr

    code_refill_bram_1,         -- op in bram_rd

    code_refill_bram_2,         -- op in code_rd 

    code_refill_sram_0,         -- FIXME code refill from SRAM unimplemented

    code_refill_sram_1,

    code_refill_sram_2,

    code_bug                    -- caught an error in the state machine

   );

-- I-cache state machine state register & next state

signal cps, cns :           t_code_cache_state;

-- Wait state counter, formally part of the state machine register

signal code_wait_ctr :      t_wait_state_counter;

type t_data_cache_state is (

    data_normal,

    data_refill_sram_0,         -- rd addr in SRAM addr bus (low hword)

    data_refill_sram_1,         -- rd addr in SRAM addr bus (high hword)

    data_refill_bram_0,         -- rd addr in bram_rd_addr

    data_refill_bram_1,         -- rd data in bram_rd_data

    data_read_io_0,             -- rd addr on io_rd_addr, io_vma active

    data_read_io_1,             -- rd data on io_rd_data

    data_write_io_0,            -- wr addr & data in io_wr_*, io_byte_we active

    data_writethrough_sram_0a,  -- wr addr & data in SRAM buses (low hword)

    data_writethrough_sram_0b,  -- WE asserted

    data_writethrough_sram_0c,  -- WE deasserted

    data_writethrough_sram_1a,  -- wr addr & data in SRAM buses (high hword)

    data_writethrough_sram_1b,  -- WE asserted

    data_writethrough_sram_1c,  -- WE deasserted

    data_ignore_write,          -- hook for raising error flag FIXME untested

    data_ignore_read,           -- hook for raising error flag FIXME untested

    data_bug                    -- caught an error in the state machine

   );

-- D-cache state machine state register & next state

signal dps, dns :           t_data_cache_state;

-- Wait state counter, formally part of the state machine register

signal dws_ctr, dws :       t_wait_state_counter;

signal load_dws_ctr :       std_logic;

signal dws_wait_done :      std_logic;

-- CPU interface registers ------------------------------------------

signal data_rd_addr_reg :   t_pc;

signal data_wr_addr_reg :   t_pc;

signal code_rd_addr_reg :   t_pc;

signal data_wr_reg :        std_logic_vector(31 downto 0);

signal byte_we_reg :        std_logic_vector(3 downto 0);

-- SRAM interface ---------------------------------------------------

-- Stores first (high) HW read from SRAM

signal sram_rd_data_reg :   std_logic_vector(31 downto 16);

-- Data read from SRAM, valid in refill_1

signal sram_rd_data :       t_word;

-- I-cache -- most of this is unimplemented -------------------------

subtype t_code_tag is std_logic_vector(23 downto 2);

signal code_cache_tag :     t_code_tag;

signal code_cache_tag_store : t_code_tag;

signal code_cache_store :   t_word;

-- code word read from cache

signal code_cache_rd :      t_word;

-- raised whel code_cache_rd is not valid due to a cache miss

signal code_miss :          std_logic;

-- '1' when the I-cache state machine stalls the pipeline (mem_wait)

signal code_wait :          std_logic;

-- D-cache -- most of this is unimplemented -------------------------

subtype t_data_tag is std_logic_vector(23 downto 2);

signal data_cache_tag :     t_data_tag;

signal data_cache_tag_store : t_data_tag;

signal data_cache_store :   t_word;

-- active when there's a write waiting to be done

signal write_pending :      std_logic;

-- active when there's a read waiting to be done

signal read_pending :       std_logic;

-- data word read from cache

signal data_cache_rd :      t_word;

-- '1' when data_cache_rd is not valid due to a cache miss

signal data_miss :          std_logic;

-- '1' when the D-cache state machine stalls the pipeline (mem_wait)

signal data_wait :          std_logic;

-- Address decoding -------------------------------------------------

-- Address slices used to decode

signal code_rd_addr_mask :  t_addr_decode;

signal data_rd_addr_mask :  t_addr_decode;

signal data_wr_addr_mask :  t_addr_decode;

-- Memory map area being accessed for each of the 3 buses:

signal code_rd_attr :       t_range_attr;

signal data_rd_attr :       t_range_attr;

signal data_wr_attr :       t_range_attr;

begin

--------------------------------------------------------------------------------

-- Cache control state machines 

cache_state_machine_regs:

process(clk)

begin

   if clk'event and clk='1' then

        if reset='1' then

            cps <= code_normal;

            dps <= data_normal;

        else

            cps <= cns;

            dps <= dns;

        end if;

    end if;

end process cache_state_machine_regs;

-- (The code state machine occasionally 'waits' for the D-cache)

code_state_machine_transitions:

process(cps, dps, code_rd_vma, code_miss, code_rd_attr,

        write_pending, read_pending)

begin

    case cps is

    when code_normal =>

        -- FIXME wrong logic, these signals are not active in the same cycle

        if code_rd_vma='1' and code_miss='1' and

           read_pending='0' and write_pending='0' then

            cns <= code_refill_bram_0; -- FIXME check memory area, SRAM!

        else

            cns <= cps;

        end if;

    when code_refill_bram_0 =>

        cns <= code_refill_bram_1;

    when code_refill_bram_1 =>

        cns <= code_refill_bram_2;

    when code_refill_bram_2 =>

        if dps/=data_normal and read_pending='0' and write_pending='0' then

            cns <= code_wait_for_dcache;

        else

            cns <= code_normal;

        end if;

    when code_wait_for_dcache =>

        -- if D-cache is busy, wait for it to become idle

        if dps/=data_normal then

            cns <= cps;

        elsif code_miss='1' then

            cns <= code_refill_bram_1; -- FIXME check memory area

        else

            cns <= code_normal;

        end if;

    when code_bug =>

        -- Something weird happened, we have 1 cycle to do something like raise

        -- an error flag, etc. After 1 cycle, back to normal.

        cns <= code_normal;

    when others =>

        -- We should never arrive here. If we do we handle it in state code_bug.

        cns <= code_bug;

    end case;

end process code_state_machine_transitions;

-- This state machine does not overlap IO/BRAM/SRAM accesses for simplicity.

data_state_machine_transitions:

process(dps, write_pending, read_pending,

        data_rd_attr, data_wr_attr, dws_wait_done)

begin

    case dps is

    when data_normal =>

        if write_pending='1' then

            case data_wr_attr.mem_type is

            when MT_BRAM        => dns <= data_ignore_write;

            when MT_SRAM_16B    => dns <= data_writethrough_sram_0a;

            when MT_IO_SYNC     => dns <= data_write_io_0;

            -- FIXME ignore write to undecoded area (clear pending flag)                        

            when others         => dns <= dps;

            end case;

        elsif read_pending='1' then

            case data_rd_attr.mem_type is

            when MT_BRAM        => dns <= data_refill_bram_0;

            when MT_SRAM_16B    => dns <= data_refill_sram_0;

            when MT_IO_SYNC     => dns <= data_read_io_0;

            -- FIXME ignore read from undecoded area (clear pending flag) 

            when others         => dns <= data_ignore_read;

            end case;

        else

            dns <= dps;

        end if;

    when data_write_io_0 =>

        dns <= data_normal;

    when data_read_io_0 =>

        dns <= data_read_io_1;

    when data_read_io_1 =>

        dns <= data_normal;

    when data_refill_sram_0 =>

        if dws_wait_done='1' then

            dns <= data_refill_sram_1;

        else

            dns <= dps;

        end if;

    when data_refill_sram_1 =>

        if dws_wait_done='1' then

            dns <= data_normal;

        else

            dns <= dps;

        end if;

    when data_refill_bram_0 =>

        dns <= data_refill_bram_1;

    when data_refill_bram_1 =>

        dns <= data_normal;

    when data_writethrough_sram_0a =>

        dns <= data_writethrough_sram_0b;

    when data_writethrough_sram_0b =>

        if dws_wait_done='1' then

            dns <= data_writethrough_sram_0c;

        else

            dns <= dps;

        end if;

    when data_writethrough_sram_0c =>

        dns <= data_writethrough_sram_1a;

    when data_writethrough_sram_1a =>

        dns <= data_writethrough_sram_1b;

    when data_writethrough_sram_1b =>

        if dws_wait_done='1' then

            dns <= data_writethrough_sram_1c;

        else

            dns <= dps;

        end if;

    when data_writethrough_sram_1c =>

        dns <= data_normal;

    when data_ignore_write =>

        dns <= data_normal;

    when data_ignore_read =>

        dns <= data_normal;

    when data_bug =>

        -- Something weird happened, we have 1 cycle to do something like raise

        -- an error flag, etc. After 1 cycle, back to normal.    

        dns <= data_normal;

    when others =>

        -- Should never arrive here. If we do, we handle it in state data_bug.

        dns <= data_bug;

    end case;

end process data_state_machine_transitions;

load_dws_ctr <= '1' when

    (dns=data_refill_sram_0 and dps/=data_refill_sram_0) or

    (dns=data_refill_sram_1 and dps/=data_refill_sram_1) or

    (dns=data_writethrough_sram_0a) or

    (dns=data_writethrough_sram_1a)

    else '0';

with dns select dws <=

    data_rd_attr.wait_states    when data_refill_sram_0,

    data_wr_attr.wait_states    when data_writethrough_sram_0a,

    data_wr_attr.wait_states    when data_writethrough_sram_1a,

    data_wr_attr.wait_states    when others;

data_wait_state_counter:

process(clk)

begin

    if clk'event and clk='1' then

        if reset='1' then

            dws_ctr <= (others => '0');

        else

            if load_dws_ctr='1' then

                dws_ctr <= dws;

            elsif dws_wait_done='0' then

                dws_ctr <= dws_ctr - 1;

            end if;

        end if;

    end if;

end process data_wait_state_counter;

dws_wait_done <= '1' when dws_ctr="000" else '0';

--------------------------------------------------------------------------------

-- CPU interface registers and address decoding --------------------------------

-- Everything coming and going to the CPU is registered, so that the CPU has

-- some timing marging.

cpu_data_interface_registers:

process(clk)

begin

    if clk'event and clk='1' then

        if reset='1' then

            write_pending <= '0';

            read_pending <= '0';

            byte_we_reg <= "0000";

        else

            -- Raise 'read_pending' at the 1st cycle of a read, clear it when

            -- the read (and/or refill) operation has been done.

            -- data_rd_addr_reg always has the addr of any pending read

            if data_rd_vma='1' then

                read_pending <= '1';

                data_rd_addr_reg <= data_rd_addr(31 downto 2);

            elsif dps=data_refill_sram_1 or

                  dps=data_refill_bram_1 or

                  dps=data_read_io_0 or

                  dps=data_ignore_read then

                read_pending <= '0';

            end if;

            -- Raise 'write_pending' at the 1st cycle of a read, clear it when

            -- the write (writethrough actually) operation has been done.

            -- data_wr_addr_reg always has the addr of any pending write

            if byte_we/="0000" and dps=data_normal then

                byte_we_reg <= byte_we;

                data_wr_reg <= data_wr;

                data_wr_addr_reg <= data_wr_addr;

                write_pending <= '1';

            elsif dps=data_writethrough_sram_1b or

                  dps=data_write_io_0 or

                  dps=data_ignore_write then

                write_pending <= '0';

                byte_we_reg <= "0000";

            end if;

        end if;

    end if;

end process cpu_data_interface_registers;

cpu_code_interface_registers:

process(clk)

begin

    if clk'event and clk='1' then

        -- Register code fetch addresses only when they are valid; so that

        -- code_rd_addr_reg always holds the last fetch address.

        if (cps=code_normal and code_rd_vma='1') or

            cps=code_refill_bram_2 then -- FIXME explain this term

            code_rd_addr_reg <= code_rd_addr;

        end if;

    end if;

end process cpu_code_interface_registers;

-- Address decoding ------------------------------------------------------------

-- Decoding is done on the high bits of the address only, there'll be mirroring.

-- Write to areas not explicitly decoded will be silently ignored. Reads will

-- get undefined data.

code_rd_addr_mask <= code_rd_addr_reg(31 downto t_addr_decode'low);

data_rd_addr_mask <= data_rd_addr_reg(31 downto t_addr_decode'low);

data_wr_addr_mask <= data_wr_addr_reg(31 downto t_addr_decode'low);

code_rd_attr <= decode_addr(code_rd_addr_mask);

data_rd_attr <= decode_addr(data_rd_addr_mask);

data_wr_attr <= decode_addr(data_wr_addr_mask);

--------------------------------------------------------------------------------

-- BRAM interface

-- BRAMm address can come from code or data buses

-- (note both inputs to this mux are register outputs)

bram_rd_addr <=

    data_rd_addr_reg(bram_rd_addr'high downto 2) when dps=data_refill_bram_0

    else code_rd_addr_reg(bram_rd_addr'high downto 2) ;

bram_data_rd_vma <= '1' when dps=data_refill_bram_1 else '0';

--------------------------------------------------------------------------------

-- Code cache 

-- All the tag match logic is unfinished and will be simplified away in synth.

-- CPU is wired directly to cache output, no muxes

code_rd <= code_cache_rd;

-- FIXME Actual 1-word cache functionality is unimplemented yet

code_miss <= '1'; -- always miss

-- Read cache code and tag from code store

code_cache_rd <= code_cache_store;

code_cache_tag <= code_cache_tag_store;

code_cache_memory:

process(clk)

begin

    if clk'event and clk='1' then

        if reset='1' then

            -- in the real hardware the tag store can't be reset and it's up

            -- to the SW to initialize the cache.

            code_cache_tag_store <= (others => '0');

            code_cache_store <= (others => '0');

        else

            -- Refill cache if necessary

            if cps=code_refill_bram_1 then

                code_cache_tag_store <=

                    "01" & code_rd_addr_reg(t_code_tag'high-2 downto t_code_tag'low);

                code_cache_store <= bram_rd_data;

            --elsif cps=code_refill_sram_2 then

            --    code_cache_tag_store <=

            --        "01" & code_rd_addr_reg(t_code_tag'high-2 downto t_code_tag'low);

            --    code_cache_store <= sram_rd_data;

            end if;

        end if;

    end if;

end process code_cache_memory;

--------------------------------------------------------------------------------

-- Data cache

-- CPU data input mux: direct cache output OR uncached io input

with dps select data_rd <=

    io_rd_data      when data_read_io_1,

    data_cache_rd   when others;

-- All the tag match logic is unfinished and will be simplified away in synth.

-- The 'cache' is really a single register.

data_cache_rd <= data_cache_store;

data_cache_tag <= data_cache_tag_store;

data_cache_memory:

process(clk)

begin

    if clk'event and clk='1' then

        if reset='1' then

            -- in the real hardware the tag store can't be reset and it's up

            -- to the SW to initialize the cache.

            data_cache_tag_store <= (others => '0');

            data_cache_store <= (others => '0');

        else

            -- Refill data cache if necessary

            if dps=data_refill_sram_1 then

                data_cache_tag_store <=

                    "01" & data_rd_addr_reg(t_data_tag'high-2 downto t_data_tag'low);

                data_cache_store <= sram_rd_data;

            elsif dps=data_refill_bram_1 then

                data_cache_tag_store <=

                    "01" & data_rd_addr_reg(t_data_tag'high-2 downto t_data_tag'low);

                data_cache_store <= bram_rd_data;

            end if;

        end if;

    end if;

end process data_cache_memory;

--------------------------------------------------------------------------------

-- SRAM interface

-- Note this signals are meantto be connected directly to FPGA pins (and then

-- to a SRAM, of course). They are the only signals whose tco we care about.

-- FIXME should add a SRAM CE\ signal

-- SRAM address bus (except for LSB) comes from cpu code or data addr registers

with dps select sram_address(sram_address'high downto 2) <=

    data_rd_addr_reg(sram_address'high downto 2)    when data_refill_sram_0,

    data_rd_addr_reg(sram_address'high downto 2)    when data_refill_sram_1,

    data_wr_addr_reg(sram_address'high downto 2)    when others;

-- SRAM addr bus LSB depends on the D-cache state because we read/write the

-- halfwords sequentially in successive cycles.

with dps select sram_address(1) <=

    '0'     when data_writethrough_sram_0a,

    '0'     when data_writethrough_sram_0b,

    '0'     when data_writethrough_sram_0c,

    '1'     when data_writethrough_sram_1a,

    '1'     when data_writethrough_sram_1b,

    '1'     when data_writethrough_sram_1c,

    '0'     when data_refill_sram_0,

    '1'     when data_refill_sram_1,

    '0'     when others;

-- SRAM databus i(when used for output) comes from either hword of the data

-- write register.

with dps select sram_databus <=

    data_wr_reg(31 downto 16)   when data_writethrough_sram_0a,

    data_wr_reg(31 downto 16)   when data_writethrough_sram_0b,

    data_wr_reg(31 downto 16)   when data_writethrough_sram_0c,

    data_wr_reg(15 downto  0)   when data_writethrough_sram_1a,

    data_wr_reg(15 downto  0)   when data_writethrough_sram_1b,

    data_wr_reg(15 downto  0)   when data_writethrough_sram_1c,

    (others => 'Z')             when others;

-- The byte_we is split in two similarly.

with dps select sram_byte_we_n <=

    not byte_we_reg(3 downto 2) when data_writethrough_sram_0b,

    not byte_we_reg(1 downto 0) when data_writethrough_sram_1b,

    "11"                        when others;

-- SRAM OE\ is only asserted low for read cycles

with dps select sram_oe_n <=

    '0' when data_refill_sram_0,

    '0' when data_refill_sram_1,

    '1' when others;

-- When eading from the SRAM, read word comes from read hword register and 

-- SRAM bus (read register is loaded in previous cycle).

sram_rd_data <= sram_rd_data_reg & sram_databus;

sram_input_halfword_register:

process(clk)

begin

    if clk'event and clk='1' then

        if dps=data_refill_sram_0 then