2 use ieee.std_logic_1164.all;
3 use ieee.std_logic_arith.all;
4 use ieee.std_logic_unsigned.all;
5 use ieee.numeric_std.all;
8 use unisim.vcomponents.all;
11 use work.lx_rocon_pkg.all;
13 -- lx_rocon_top - wires the modules with the outside world
15 -- ======================================================
16 -- MASTER CPU EXTERNAL MEMORY BUS
17 -- ======================================================
19 -- Master cpu memory bus has the following wires:
21 -- - address[15..0] The address, used to mark chip enable
22 -- - data_in[31..0] The data coming to bus
23 -- - data_out[31..0] The data coming from bus, multiplexed
24 -- - bls[3..0] Write enable for respective bytes
26 entity lx_rocon_top is
30 --clk_cpu : in std_logic;
31 clk_50m : in std_logic;
33 cs0_xc : in std_logic;
36 bls : in std_logic_vector(3 downto 0);
37 address : in std_logic_vector(15 downto 0);
38 data : inout std_logic_vector(31 downto 0);
40 irc0_a : in std_logic;
41 irc0_b : in std_logic;
42 irc0_index : in std_logic;
43 irc0_mark : in std_logic;
45 irc1_a : in std_logic;
46 irc1_b : in std_logic;
47 irc1_index : in std_logic;
48 irc1_mark : in std_logic;
50 irc2_a : in std_logic;
51 irc2_b : in std_logic;
52 irc2_index : in std_logic;
53 irc2_mark : in std_logic;
55 irc3_a : in std_logic;
56 irc3_b : in std_logic;
57 irc3_index : in std_logic;
58 irc3_mark : in std_logic;
60 irc4_a : in std_logic;
61 irc4_b : in std_logic;
62 irc4_index : in std_logic;
63 irc4_mark : in std_logic;
65 irc5_a : in std_logic;
66 irc5_b : in std_logic;
67 irc5_index : in std_logic;
68 irc5_mark : in std_logic;
70 irc6_a : in std_logic;
71 irc6_b : in std_logic;
72 irc6_index : in std_logic;
73 irc6_mark : in std_logic;
75 irc7_a : in std_logic;
76 irc7_b : in std_logic;
77 irc7_index : in std_logic;
78 irc7_mark : in std_logic;
82 s1_clk_in : in std_logic;
83 s1_miso : in std_logic;
84 s1_sync_in : in std_logic;
86 s1_clk_out : out std_logic;
87 s1_mosi : out std_logic;
88 s1_sync_out : out std_logic
92 architecture Behavioral of lx_rocon_top is
95 signal reset_s : std_logic;
96 signal init_s : std_logic;
97 -- Peripherals on the memory buses
98 -- Master to Tumbl DMEM / IMEM (Master)
99 signal tumbl_out_s : std_logic_vector(31 downto 0);
100 signal tumbl_ce_s : std_logic;
101 -- Measurement (Master)
102 signal meas_out_s : std_logic_vector(31 downto 0);
103 signal meas_ce_s : std_logic;
104 -- Master to Tumbl XMEM
105 signal master_tumbl_xmem_out_s : std_logic_vector(31 downto 0);
106 signal master_tumbl_xmem_ce_s : std_logic;
107 signal master_tumbl_xmem_lock_s : std_logic;
109 signal irc_proc_out_s : std_logic_vector(31 downto 0);
110 signal irc_proc_ce_s : std_logic;
111 signal irc_proc_next_ce_s : std_logic;
113 signal lxmaster_out_s : std_logic_vector(15 downto 0);
114 signal lxmaster_ce_s : std_logic;
115 signal lxmaster_next_ce_s : std_logic;
116 -- Signals for external bus transmission
117 signal data_i_s : std_logic_vector(31 downto 0);
118 signal data_o_s : std_logic_vector(31 downto 0);
119 -- Signals for internal transaction
120 signal last_address_s : std_logic_vector(15 downto 0);
121 signal next_last_address_s : std_logic_vector(15 downto 0);
122 signal next_address_hold_s : std_logic;
123 signal address_hold_s : std_logic;
124 signal last_rd_s : std_logic;
125 signal next_last_rd_s : std_logic;
126 signal last_bls_s : std_logic_vector(3 downto 0); -- prev bls_f_s (active 1)
127 signal next_last_bls_s : std_logic_vector(3 downto 0);
129 -- Reading logic for Master CPU:
130 -- Broadcast rd only till ta (transaction acknowledge)
131 -- is received, then latch the data till the state of
132 -- rd or address changes
134 -- Data latching is synchronous - it's purpose is to
135 -- provide stable data for CPU on the bus
136 signal cs0_xc_f_s : std_logic;
137 signal rd_f_s : std_logic; -- Filtered RD
138 signal i_rd_s : std_logic; -- Internal bus RD (active 1)
139 -- signal next_i_rd_s : std_logic;
140 signal last_i_rd_s : std_logic; -- Delayed RD bus, used for latching
141 signal next_last_i_rd_s : std_logic;
142 signal i_rd_cycle2_s : std_logic; -- Some internal subsystems provide
143 signal next_i_rd_cycle2_s : std_logic; -- data only after 2 cycles
145 signal address_f_s : std_logic_vector(15 downto 0); -- Filtered address
147 signal data_f_s : std_logic_vector(31 downto 0); -- Filterred input data
149 signal data_read_s : std_logic_vector(31 downto 0); -- Latched read data
150 signal next_data_read_s : std_logic_vector(31 downto 0);
153 signal bls_f_s : std_logic_vector(3 downto 0); -- Filtered BLS (active 1)
154 signal i_bls_s : std_logic_vector(3 downto 0); -- Internal BLS (active 1)
155 signal next_i_bls_s : std_logic_vector(3 downto 0);
157 signal data_write_s : std_logic_vector(31 downto 0); -- Data broadcasted to write
158 signal next_data_write_s : std_logic_vector(31 downto 0);
161 signal tumbl_bls_s : std_logic_vector(3 downto 0);
162 signal tumbl_address_s : std_logic_vector(14 downto 0);
163 signal tumbl_data_i_s : std_logic_vector(31 downto 0);
165 signal tumbl_xmemb_o_s : CORE2DMEMB_Type;
166 signal tumbl_xmemb_i_s : DMEMB2CORE_Type;
167 signal tumbl_xmemb_sel_s : std_logic;
169 -- signal s0 : std_logic;
170 -- signal s1 : std_logic;
171 -- signal s2 : std_logic;
174 attribute REGISTER_DUPLICATION : string;
175 attribute REGISTER_DUPLICATION of rd : signal is "NO";
176 attribute REGISTER_DUPLICATION of rd_f_s : signal is "NO";
177 attribute REGISTER_DUPLICATION of bls : signal is "NO";
178 attribute REGISTER_DUPLICATION of bls_f_s : signal is "NO";
179 attribute REGISTER_DUPLICATION of address : signal is "NO";
180 attribute REGISTER_DUPLICATION of address_f_s : signal is "NO";
181 attribute REGISTER_DUPLICATION of cs0_xc : signal is "NO";
186 memory_bus_tumbl: bus_tumbl
193 address_i => address_f_s(11 downto 0),
195 data_o => tumbl_out_s,
197 xmemb_o => tumbl_xmemb_o_s,
198 xmemb_i => tumbl_xmemb_i_s,
199 xmemb_sel_o => tumbl_xmemb_sel_s
203 memory_bus_measurement: bus_measurement
209 address_i => address_f_s(1 downto 0),
216 memory_bus_irc: bus_irc
222 address_i => tumbl_address_s(4 downto 0),
223 next_ce_i => irc_proc_next_ce_s,
224 data_i => tumbl_data_i_s,
225 data_o => irc_proc_out_s,
226 bls_i => tumbl_bls_s,
228 irc_i(0).a => irc0_a,
229 irc_i(0).b => irc0_b,
230 irc_i(0).index => irc0_index,
231 irc_i(0).mark => irc0_mark,
233 irc_i(1).a => irc1_a,
234 irc_i(1).b => irc1_b,
235 irc_i(1).index => irc1_index,
236 irc_i(1).mark => irc1_mark,
238 irc_i(2).a => irc2_a,
239 irc_i(2).b => irc2_b,
240 irc_i(2).index => irc2_index,
241 irc_i(2).mark => irc2_mark,
243 irc_i(3).a => irc3_a,
244 irc_i(3).b => irc3_b,
245 irc_i(3).index => irc3_index,
246 irc_i(3).mark => irc3_mark,
248 irc_i(4).a => irc4_a,
249 irc_i(4).b => irc4_b,
250 irc_i(4).index => irc4_index,
251 irc_i(4).mark => irc4_mark,
253 irc_i(5).a => irc5_a,
254 irc_i(5).b => irc5_b,
255 irc_i(5).index => irc5_index,
256 irc_i(5).mark => irc5_mark,
258 irc_i(6).a => irc6_a,
259 irc_i(6).b => irc6_b,
260 irc_i(6).index => irc6_index,
261 irc_i(6).mark => irc6_mark,
263 irc_i(7).a => irc7_a,
264 irc_i(7).b => irc7_b,
265 irc_i(7).index => irc7_index,
266 irc_i(7).mark => irc7_mark
270 memory_bus_lxmaster: bus_lxmaster
276 address_i => tumbl_address_s(10 downto 0),
277 next_ce_i => lxmaster_next_ce_s,
278 data_i => tumbl_data_i_s(15 downto 0),
279 data_o => lxmaster_out_s,
280 bls_i => tumbl_bls_s(1 downto 0),
282 clock_i => s1_clk_in,
284 sync_i => s1_sync_in,
286 clock_o => s1_clk_out,
288 sync_o => s1_sync_out
301 -- s1_sync_out <= s2;
317 data_i_s <= data_write_s;
320 tumbl_bls_s <= i_bls_s when (master_tumbl_xmem_lock_s = '1')
321 else tumbl_xmemb_o_s.bls when (tumbl_xmemb_sel_s = '1')
323 tumbl_address_s <= address_f_s(14 downto 0) when (master_tumbl_xmem_lock_s = '1')
324 else tumbl_xmemb_o_s.addr when (tumbl_xmemb_sel_s = '1')
325 else (others => '0');
326 tumbl_data_i_s <= data_i_s when (master_tumbl_xmem_lock_s = '1')
327 else tumbl_xmemb_o_s.data when (tumbl_xmemb_sel_s = '1')
328 else (others => '0');
330 tumbl_xmemb_i_s.int <= '0'; -- No interrupt
331 -- Enable clken only when available for Tumbl
332 tumbl_xmemb_i_s.clken <= not master_tumbl_xmem_lock_s;
337 process(cs0_xc_f_s, rd_f_s, last_rd_s, i_rd_cycle2_s, last_i_rd_s,
338 bls_f_s, last_bls_s, data_f_s, data_write_s,
339 data_o_s, data_read_s, last_address_s, address_f_s)
342 next_i_rd_cycle2_s <= '0';
343 next_address_hold_s <= '0';
345 -- Check if we have chip select
346 if cs0_xc_f_s = '1' then
351 if last_rd_s = '0' or (last_address_s /= address_f_s) then
353 next_i_rd_cycle2_s <= '1';
354 next_last_i_rd_s <= '1';
355 elsif i_rd_cycle2_s = '1' then -- FIXME it seems that some internal
356 i_rd_s <= '1'; -- peripherals demands 2 cycles to read
357 next_last_i_rd_s <= '1';
360 next_last_i_rd_s <= '0';
363 if last_i_rd_s = '1' then
364 -- Latch data we just read - they are valid in this cycle
365 next_data_read_s <= data_o_s;
367 next_data_read_s <= data_read_s;
370 -- -- Not reading, anything goes
371 -- data_read_s <= (others => 'X');
372 next_data_read_s <= data_read_s;
374 next_last_i_rd_s <= '0';
377 next_last_rd_s <= rd_f_s;
379 -- Data for write are captured only when BLS signals are stable
380 if bls_f_s /= "0000" then
381 next_data_write_s <= data_f_s;
382 next_address_hold_s <= '1';
384 next_data_write_s <= data_write_s;
387 if (bls_f_s /= "0000") or (rd_f_s = '1') then
388 next_last_address_s <= address_f_s;
390 next_last_address_s <= last_address_s;
393 next_last_rd_s <= '0';
395 next_last_i_rd_s <= '0';
397 next_i_bls_s <= "0000";
398 next_data_write_s <= data_write_s;
399 next_data_read_s <= data_read_s;
400 next_last_address_s <= last_address_s;
403 -- Data for write are captured at/before BLS signals are negated
404 -- and actual write cycle takes place exacly after BLS negation
405 if ((last_bls_s and not bls_f_s) /= "0000") or
406 ((last_bls_s /= "0000") and (cs0_xc_f_s = '0')) then
407 next_i_bls_s <= last_bls_s;
408 next_last_bls_s <= "0000";
409 next_address_hold_s <= '1';
411 next_i_bls_s <= "0000";
412 if cs0_xc_f_s = '1' then
413 next_last_bls_s <= bls_f_s;
415 next_last_bls_s <= "0000" ;
426 wait until clk_50m = '1' and clk_50m'event;
428 address_hold_s <= next_address_hold_s;
430 -- Synchronized external signals with main clock domain
431 cs0_xc_f_s <= not cs0_xc;
435 if address_hold_s = '0' then
436 address_f_s <= address;
438 address_f_s <= next_last_address_s;
441 -- Synchronoust state andvance to next period
442 last_bls_s <= next_last_bls_s;
443 last_rd_s <= next_last_rd_s;
444 i_bls_s <= next_i_bls_s;
445 -- i_rd_s <= next_i_rd_s;
446 i_rd_cycle2_s <= next_i_rd_cycle2_s;
447 last_i_rd_s <= next_last_i_rd_s;
448 data_write_s <= next_data_write_s;
449 last_address_s <= next_last_address_s;
450 data_read_s <= next_data_read_s;
452 -- ======================================================
454 -- ======================================================
456 -- Just copy these to their desired next state
457 irc_proc_ce_s <= irc_proc_next_ce_s;
458 lxmaster_ce_s <= lxmaster_next_ce_s;
462 -- Do the actual wiring here
464 process(cs0_xc_f_s, i_bls_s, address_f_s, tumbl_out_s, meas_out_s, master_tumbl_xmem_out_s)
467 -- Inactive by default
470 master_tumbl_xmem_ce_s <= '0';
471 data_o_s <= (others => '0');
473 if cs0_xc_f_s = '1' or i_bls_s /= "0000" then
475 -- Memory Map (16-bit address @ 32-bit each)
477 -- Each address is seen as 32-bit entry now
478 -- 0x0000 - 0x0FFF: Tumbl IMEM / DMEM
479 -- 0x1FFC - 0x1FFF: Measurement
480 -- 0x8000 - 0x8FFF: Tumbl BUS
482 if address_f_s < "0001000000000000" then -- Tumbl
484 data_o_s <= tumbl_out_s;
485 elsif address_f_s(15 downto 2) = "00011111111111" then -- Measurement
487 data_o_s <= meas_out_s;
488 elsif address_f_s(15) = '1' then -- Tumbl External BUS
489 master_tumbl_xmem_ce_s <= '1';
490 data_o_s <= master_tumbl_xmem_out_s;
497 -- If RD and BLS is not high, we must keep DATA at high impedance
498 -- or the FPGA collides with SDRAM (damaging each other)
500 process(cs0_xc, rd, data_read_s)
503 -- CS0 / RD / BLS are active LOW
504 if cs0_xc = '0' and rd = '0' then
505 -- Don't risk flipping (between data_o_s and latched data_read_s, it's better to wait)
506 -- Maybe check this later.
507 -- if last_i_rd_s = '1' then
514 data <= (others => 'Z');
519 -- Outputs from Tumbl (enabling and address muxing) and Master CPU
521 process(tumbl_xmemb_sel_s, tumbl_xmemb_o_s, master_tumbl_xmem_ce_s, address_f_s, i_rd_s, i_bls_s)
522 variable addr_v : std_logic_vector(14 downto 0); -- This space is visible by both (32-bit)
523 variable sel_v : std_logic;
527 irc_proc_next_ce_s <= '0';
528 lxmaster_next_ce_s <= '0';
529 master_tumbl_xmem_lock_s <= '0';
531 addr_v := (others => '0');
534 -- Check who is accessing
535 if master_tumbl_xmem_ce_s = '1' and (i_rd_s = '1' or i_bls_s /= "0000") then
536 -- Master blocks Tumbl
537 master_tumbl_xmem_lock_s <= '1';
538 addr_v := address_f_s(14 downto 0);
541 addr_v := tumbl_xmemb_o_s.addr;
546 -- IRC: 0x0800 - 0x081F (32-bit address)
547 -- LX MASTER: 0x1000 - 0x17FF (32-bit address)
548 if addr_v(14 downto 5) = "0001000000" then
549 irc_proc_next_ce_s <= '1';
550 elsif addr_v(14 downto 11) = "0010" then
551 lxmaster_next_ce_s <= '1';
557 -- Inputs to Tumbl (enabling and address muxing)
559 process(irc_proc_ce_s, irc_proc_out_s, lxmaster_ce_s, lxmaster_out_s, tumbl_xmemb_i_s)
562 tumbl_xmemb_i_s.data <= (others => 'X');
564 -- NOTE: This is input to Tumbl EXEQ - with MUL instruction for input > 18-bit,
565 -- (i.e. more DSPs in a sequence), this already has tough timing constraints
566 -- and SmartXplorer has to be used with XiSE or use Synplify.
567 if irc_proc_ce_s = '1' then
568 tumbl_xmemb_i_s.data <= irc_proc_out_s;
569 elsif lxmaster_ce_s = '1' then
570 tumbl_xmemb_i_s.data(15 downto 0) <= lxmaster_out_s;
571 tumbl_xmemb_i_s.data(31 downto 16) <= (others => '0');
574 master_tumbl_xmem_out_s <= tumbl_xmemb_i_s.data;