1 |
2 |
danv |
-------------------------------------------------------------------------------
|
2 |
|
|
--
|
3 |
3 |
danv |
-- Copyright 2020
|
4 |
2 |
danv |
-- ASTRON (Netherlands Institute for Radio Astronomy) <http://www.astron.nl/>
|
5 |
|
|
-- P.O.Box 2, 7990 AA Dwingeloo, The Netherlands
|
6 |
3 |
danv |
--
|
7 |
|
|
-- Licensed under the Apache License, Version 2.0 (the "License");
|
8 |
|
|
-- you may not use this file except in compliance with the License.
|
9 |
|
|
-- You may obtain a copy of the License at
|
10 |
|
|
--
|
11 |
|
|
-- http://www.apache.org/licenses/LICENSE-2.0
|
12 |
|
|
--
|
13 |
|
|
-- Unless required by applicable law or agreed to in writing, software
|
14 |
|
|
-- distributed under the License is distributed on an "AS IS" BASIS,
|
15 |
|
|
-- WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
16 |
|
|
-- See the License for the specific language governing permissions and
|
17 |
|
|
-- limitations under the License.
|
18 |
2 |
danv |
--
|
19 |
|
|
-------------------------------------------------------------------------------
|
20 |
|
|
|
21 |
|
|
LIBRARY IEEE, common_pkg_lib, common_ram_lib, common_mult_lib, technology_lib;
|
22 |
|
|
USE IEEE.std_logic_1164.ALL;
|
23 |
|
|
USE technology_lib.technology_select_pkg.ALL;
|
24 |
|
|
USE common_pkg_lib.common_pkg.ALL;
|
25 |
|
|
USE common_ram_lib.common_ram_pkg.ALL;
|
26 |
|
|
|
27 |
|
|
-- Purpose:
|
28 |
|
|
-- Maintain a set of accumulators and output their values at every in_sync.
|
29 |
|
|
-- Description:
|
30 |
|
|
-- . The products of two input streams are accumulated per block. The block
|
31 |
|
|
-- size is g_nof_mux*g_nof_stat. The nof accumulators is equal to the block
|
32 |
|
|
-- size. The nof blocks that get accumulated depends on in_sync, because a
|
33 |
|
|
-- new accumulation starts every time when in_sync pulses. Also when in_sync
|
34 |
|
|
-- pulses then after some latency the accumulation values of the previous
|
35 |
|
|
-- in_sync interval become available at the out_* ports.
|
36 |
|
|
-- . If g_complex = FALSE then only the real power statistic out_re is calculated,
|
37 |
|
|
-- else also the imaginary power statistic out_im. The real power statistic
|
38 |
|
|
-- is used for auto power calulations of a complex input, by connecting the
|
39 |
|
|
-- signal to both input a and b. The imaginary is power statistic is used when
|
40 |
|
|
-- the cross power needs to be calculated between 2 different complex inputs.
|
41 |
|
|
-- Remarks:
|
42 |
|
|
-- . The required accumulator width depends the input data width and the nof of
|
43 |
|
|
-- block, i.e. the nof accumulations. E.g. for 18b*18b = 36b products and
|
44 |
|
|
-- 200000 accumulations yielding 18b bit growth so in total 36b+18b = 54b for
|
45 |
|
|
-- the accumulators.
|
46 |
|
|
-- . The nof accumulators determines the size (c_mem_acc) of the internal
|
47 |
|
|
-- accumulator memory.
|
48 |
|
|
-- . Using g_nof_mux>1 allows distinghuising different streams with a block.
|
49 |
|
|
-- The g_nof_mux does not impact the address range instead it impacts the
|
50 |
|
|
-- out_val_m strobes that can be used as wr_en to the corresponding statistics
|
51 |
|
|
-- output register in a range of g_nof_mux statistics output registers.
|
52 |
|
|
|
53 |
|
|
ENTITY st_calc IS
|
54 |
|
|
GENERIC (
|
55 |
|
|
g_technology : NATURAL := c_tech_select_default;
|
56 |
|
|
g_nof_mux : NATURAL := 1;
|
57 |
|
|
g_nof_stat : NATURAL := 512;
|
58 |
|
|
g_in_dat_w : NATURAL := 18; -- = input data width
|
59 |
|
|
g_out_dat_w : NATURAL := 54; -- = accumulator width for the input data products, so >> 2*g_in_dat_w
|
60 |
|
|
g_out_adr_w : NATURAL := 9; -- = ceil_log2(g_nof_stat)
|
61 |
|
|
g_complex : BOOLEAN := FALSE
|
62 |
|
|
);
|
63 |
|
|
PORT (
|
64 |
|
|
rst : IN STD_LOGIC;
|
65 |
|
|
clk : IN STD_LOGIC;
|
66 |
|
|
clken : IN STD_LOGIC := '1';
|
67 |
|
|
in_ar : IN STD_LOGIC_VECTOR(g_in_dat_w-1 DOWNTO 0);
|
68 |
|
|
in_ai : IN STD_LOGIC_VECTOR(g_in_dat_w-1 DOWNTO 0);
|
69 |
|
|
in_br : IN STD_LOGIC_VECTOR(g_in_dat_w-1 DOWNTO 0);
|
70 |
|
|
in_bi : IN STD_LOGIC_VECTOR(g_in_dat_w-1 DOWNTO 0);
|
71 |
|
|
in_val : IN STD_LOGIC;
|
72 |
|
|
in_sync : IN STD_LOGIC;
|
73 |
|
|
out_adr : OUT STD_LOGIC_VECTOR(g_out_adr_w-1 DOWNTO 0);
|
74 |
|
|
out_re : OUT STD_LOGIC_VECTOR(g_out_dat_w-1 DOWNTO 0);
|
75 |
|
|
out_im : OUT STD_LOGIC_VECTOR(g_out_dat_w-1 DOWNTO 0);
|
76 |
|
|
out_val : OUT STD_LOGIC; -- Use when g_nof_mux = 1, else leave OPEN
|
77 |
|
|
out_val_m : OUT STD_LOGIC_VECTOR(g_nof_mux-1 DOWNTO 0) -- Use when g_nof_mux > 1, else leave OPEN
|
78 |
|
|
);
|
79 |
|
|
END;
|
80 |
|
|
|
81 |
|
|
|
82 |
|
|
ARCHITECTURE str OF st_calc IS
|
83 |
|
|
|
84 |
|
|
CONSTANT c_mux_w : NATURAL := true_log2(g_nof_mux);
|
85 |
|
|
CONSTANT c_adr_w : NATURAL := c_mux_w+g_out_adr_w; -- = = ceil_log2(g_nof_mux*g_nof_stat)
|
86 |
|
|
|
87 |
|
|
CONSTANT c_dly_rd : NATURAL := 2;
|
88 |
|
|
CONSTANT c_dly_mul : NATURAL := 3;
|
89 |
|
|
CONSTANT c_dly_acc : NATURAL := 2;
|
90 |
|
|
CONSTANT c_dly_out : NATURAL := 0;
|
91 |
|
|
|
92 |
|
|
CONSTANT c_mult_w : NATURAL := 2*g_in_dat_w;
|
93 |
|
|
|
94 |
|
|
CONSTANT c_acc_w : NATURAL := g_out_dat_w;
|
95 |
|
|
CONSTANT c_acc_hold_load : BOOLEAN := TRUE;
|
96 |
|
|
|
97 |
|
|
CONSTANT c_rd_latency : NATURAL := 2;
|
98 |
|
|
CONSTANT c_mem_acc : t_c_mem := (c_rd_latency, c_adr_w, c_acc_w, g_nof_mux*g_nof_stat, 'X'); -- 1 M9K
|
99 |
|
|
|
100 |
|
|
|
101 |
|
|
SIGNAL mult_re : STD_LOGIC_VECTOR(c_mult_w-1 DOWNTO 0);
|
102 |
|
|
SIGNAL mult_im : STD_LOGIC_VECTOR(c_mult_w-1 DOWNTO 0);
|
103 |
|
|
|
104 |
|
|
SIGNAL reg_ar : STD_LOGIC_VECTOR(in_ar'RANGE);
|
105 |
|
|
SIGNAL reg_ai : STD_LOGIC_VECTOR(in_ai'RANGE);
|
106 |
|
|
SIGNAL reg_br : STD_LOGIC_VECTOR(in_br'RANGE);
|
107 |
|
|
SIGNAL reg_bi : STD_LOGIC_VECTOR(in_bi'RANGE);
|
108 |
|
|
SIGNAL reg_val : STD_LOGIC;
|
109 |
|
|
SIGNAL reg_sync : STD_LOGIC;
|
110 |
|
|
|
111 |
|
|
SIGNAL nxt_reg_ar : STD_LOGIC_VECTOR(in_ar'RANGE);
|
112 |
|
|
SIGNAL nxt_reg_ai : STD_LOGIC_VECTOR(in_ai'RANGE);
|
113 |
|
|
SIGNAL nxt_reg_br : STD_LOGIC_VECTOR(in_br'RANGE);
|
114 |
|
|
SIGNAL nxt_reg_bi : STD_LOGIC_VECTOR(in_bi'RANGE);
|
115 |
|
|
SIGNAL nxt_reg_val : STD_LOGIC;
|
116 |
|
|
SIGNAL nxt_reg_sync : STD_LOGIC;
|
117 |
|
|
|
118 |
|
|
SIGNAL acc_load : STD_LOGIC;
|
119 |
|
|
|
120 |
|
|
SIGNAL rd_en : STD_LOGIC;
|
121 |
|
|
SIGNAL rd_adr : STD_LOGIC_VECTOR(c_adr_w-1 DOWNTO 0);
|
122 |
|
|
SIGNAL rd_re : STD_LOGIC_VECTOR(c_acc_w-1 DOWNTO 0);
|
123 |
|
|
SIGNAL rd_im : STD_LOGIC_VECTOR(c_acc_w-1 DOWNTO 0);
|
124 |
|
|
|
125 |
|
|
SIGNAL wr_en : STD_LOGIC;
|
126 |
|
|
SIGNAL wr_adr : STD_LOGIC_VECTOR(c_adr_w-1 DOWNTO 0);
|
127 |
|
|
SIGNAL wr_re : STD_LOGIC_VECTOR(c_acc_w-1 DOWNTO 0);
|
128 |
|
|
SIGNAL wr_im : STD_LOGIC_VECTOR(c_acc_w-1 DOWNTO 0);
|
129 |
|
|
|
130 |
|
|
SIGNAL out_adr_m : STD_LOGIC_VECTOR(c_adr_w-1 DOWNTO 0);
|
131 |
|
|
|
132 |
|
|
BEGIN
|
133 |
|
|
|
134 |
|
|
regs: PROCESS(rst,clk)
|
135 |
|
|
BEGIN
|
136 |
|
|
IF rst='1' THEN
|
137 |
|
|
reg_ar <= (OTHERS => '0');
|
138 |
|
|
reg_ai <= (OTHERS => '0');
|
139 |
|
|
reg_br <= (OTHERS => '0');
|
140 |
|
|
reg_bi <= (OTHERS => '0');
|
141 |
|
|
reg_val <= '0';
|
142 |
|
|
reg_sync <= '0';
|
143 |
|
|
ELSIF rising_edge(clk) THEN
|
144 |
|
|
reg_ar <= nxt_reg_ar;
|
145 |
|
|
reg_ai <= nxt_reg_ai;
|
146 |
|
|
reg_br <= nxt_reg_br;
|
147 |
|
|
reg_bi <= nxt_reg_bi;
|
148 |
|
|
reg_val <= nxt_reg_val;
|
149 |
|
|
reg_sync <= nxt_reg_sync;
|
150 |
|
|
END IF;
|
151 |
|
|
END PROCESS;
|
152 |
|
|
|
153 |
|
|
nxt_reg_ar <= in_ar WHEN in_val='1' ELSE reg_ar;
|
154 |
|
|
nxt_reg_ai <= in_ai WHEN in_val='1' ELSE reg_ai;
|
155 |
|
|
nxt_reg_br <= in_br WHEN in_val='1' ELSE reg_br;
|
156 |
|
|
nxt_reg_bi <= in_bi WHEN in_val='1' ELSE reg_bi;
|
157 |
|
|
nxt_reg_val <= in_val;
|
158 |
|
|
nxt_reg_sync <= in_sync;
|
159 |
|
|
|
160 |
|
|
-- ctrl block: generates all ctrl signals
|
161 |
|
|
ctrl: ENTITY work.st_ctrl
|
162 |
|
|
GENERIC MAP (
|
163 |
|
|
g_nof_mux => g_nof_mux,
|
164 |
|
|
g_nof_stat => g_nof_stat,
|
165 |
|
|
g_adr_w => c_adr_w,
|
166 |
|
|
g_dly_rd => c_dly_rd,
|
167 |
|
|
g_dly_mul => c_dly_mul,
|
168 |
|
|
g_dly_acc => c_dly_acc,
|
169 |
|
|
g_dly_out => c_dly_out
|
170 |
|
|
)
|
171 |
|
|
PORT MAP (
|
172 |
|
|
rst => rst,
|
173 |
|
|
clk => clk,
|
174 |
|
|
in_sync => reg_sync,
|
175 |
|
|
in_val => reg_val,
|
176 |
|
|
rd_en => rd_en,
|
177 |
|
|
rd_adr => rd_adr,
|
178 |
|
|
rd_val => OPEN,
|
179 |
|
|
mult_val => OPEN,
|
180 |
|
|
acc_load => acc_load,
|
181 |
|
|
wr_en => wr_en,
|
182 |
|
|
wr_adr => wr_adr,
|
183 |
|
|
out_val => out_val,
|
184 |
|
|
out_val_m => out_val_m,
|
185 |
|
|
out_adr => out_adr_m
|
186 |
|
|
);
|
187 |
|
|
|
188 |
|
|
out_adr <= out_adr_m(c_adr_w-1 DOWNTO c_mux_w);
|
189 |
|
|
|
190 |
|
|
-- complex multiplier: computes a * conj(b)
|
191 |
|
|
--mul: ENTITY common_lib.common_complex_mult(str)
|
192 |
|
|
mul: ENTITY common_mult_lib.common_complex_mult
|
193 |
|
|
GENERIC MAP (
|
194 |
|
|
g_technology => g_technology,
|
195 |
|
|
g_variant => "IP",
|
196 |
|
|
g_in_a_w => in_ar'LENGTH,
|
197 |
|
|
g_in_b_w => in_br'LENGTH,
|
198 |
|
|
g_out_p_w => mult_re'LENGTH,
|
199 |
|
|
g_conjugate_b => TRUE, -- use conjugate product for cross power
|
200 |
|
|
g_pipeline_input => 1,
|
201 |
|
|
g_pipeline_product => 0,
|
202 |
|
|
g_pipeline_adder => 1,
|
203 |
|
|
g_pipeline_output => 1 -- 1+0+1+1 = 3 = c_dly_mul
|
204 |
|
|
)
|
205 |
|
|
PORT MAP (
|
206 |
|
|
clk => clk,
|
207 |
|
|
clken => clken,
|
208 |
|
|
in_ar => reg_ar,
|
209 |
|
|
in_ai => reg_ai,
|
210 |
|
|
in_br => reg_br,
|
211 |
|
|
in_bi => reg_bi,
|
212 |
|
|
out_pr => mult_re,
|
213 |
|
|
out_pi => mult_im
|
214 |
|
|
);
|
215 |
|
|
|
216 |
|
|
-- accumulator for real part
|
217 |
|
|
acc_re: ENTITY work.st_acc
|
218 |
|
|
GENERIC MAP (
|
219 |
|
|
g_dat_w => c_mult_w,
|
220 |
|
|
g_acc_w => c_acc_w,
|
221 |
|
|
g_hold_load => c_acc_hold_load,
|
222 |
|
|
g_pipeline_input => 1,
|
223 |
|
|
g_pipeline_output => c_dly_acc-1
|
224 |
|
|
)
|
225 |
|
|
PORT MAP (
|
226 |
|
|
clk => clk,
|
227 |
|
|
clken => clken,
|
228 |
|
|
in_load => acc_load,
|
229 |
|
|
in_dat => mult_re,
|
230 |
|
|
in_acc => rd_re,
|
231 |
|
|
out_acc => wr_re
|
232 |
|
|
);
|
233 |
|
|
|
234 |
|
|
-- accumulator memory for real part
|
235 |
|
|
ram_re: ENTITY common_ram_lib.common_ram_r_w
|
236 |
|
|
GENERIC MAP (
|
237 |
|
|
g_technology => g_technology,
|
238 |
|
|
g_ram => c_mem_acc,
|
239 |
|
|
g_init_file => "UNUSED"
|
240 |
|
|
)
|
241 |
|
|
PORT MAP (
|
242 |
|
|
rst => rst,
|
243 |
|
|
clk => clk,
|
244 |
|
|
clken => clken,
|
245 |
|
|
wr_en => wr_en,
|
246 |
|
|
wr_adr => wr_adr,
|
247 |
|
|
wr_dat => wr_re,
|
248 |
|
|
rd_en => rd_en,
|
249 |
|
|
rd_adr => rd_adr,
|
250 |
|
|
rd_dat => rd_re,
|
251 |
|
|
rd_val => OPEN
|
252 |
|
|
);
|
253 |
|
|
|
254 |
|
|
out_re <= rd_re; -- c_dly_out = 0
|
255 |
|
|
|
256 |
|
|
-- imaginary part is optional
|
257 |
|
|
no_im: IF g_complex=FALSE GENERATE
|
258 |
|
|
out_im <= (OTHERS => '0');
|
259 |
|
|
END GENERATE;
|
260 |
|
|
|
261 |
|
|
gen_im: IF g_complex=TRUE GENERATE
|
262 |
|
|
-- accumulator
|
263 |
|
|
acc_im: ENTITY work.st_acc
|
264 |
|
|
GENERIC MAP (
|
265 |
|
|
g_dat_w => c_mult_w,
|
266 |
|
|
g_acc_w => c_acc_w,
|
267 |
|
|
g_hold_load => c_acc_hold_load,
|
268 |
|
|
g_pipeline_input => 1,
|
269 |
|
|
g_pipeline_output => c_dly_acc-1
|
270 |
|
|
)
|
271 |
|
|
PORT MAP (
|
272 |
|
|
clk => clk,
|
273 |
|
|
clken => clken,
|
274 |
|
|
in_load => acc_load,
|
275 |
|
|
in_dat => mult_im,
|
276 |
|
|
in_acc => rd_im,
|
277 |
|
|
out_acc => wr_im
|
278 |
|
|
);
|
279 |
|
|
|
280 |
|
|
-- dual port memory
|
281 |
|
|
ram_im: ENTITY common_ram_lib.common_ram_r_w
|
282 |
|
|
GENERIC MAP (
|
283 |
|
|
g_technology => g_technology,
|
284 |
|
|
g_ram => c_mem_acc,
|
285 |
|
|
g_init_file => "UNUSED"
|
286 |
|
|
)
|
287 |
|
|
PORT MAP (
|
288 |
|
|
rst => rst,
|
289 |
|
|
clk => clk,
|
290 |
|
|
clken => clken,
|
291 |
|
|
wr_en => wr_en,
|
292 |
|
|
wr_adr => wr_adr,
|
293 |
|
|
wr_dat => wr_im,
|
294 |
|
|
rd_en => rd_en,
|
295 |
|
|
rd_adr => rd_adr,
|
296 |
|
|
rd_dat => rd_im,
|
297 |
|
|
rd_val => OPEN
|
298 |
|
|
);
|
299 |
|
|
|
300 |
|
|
out_im <= rd_im; -- c_dly_out = 0
|
301 |
|
|
END GENERATE;
|
302 |
|
|
|
303 |
|
|
END str;
|