1 |
2 |
davidklun |
---------------------------------------------------------------------
|
2 |
|
|
---- ----
|
3 |
|
|
---- FPU ----
|
4 |
|
|
---- Floating Point Unit (Double precision) ----
|
5 |
|
|
---- ----
|
6 |
|
|
---- Author: David Lundgren ----
|
7 |
|
|
---- davidklun@gmail.com ----
|
8 |
|
|
---- ----
|
9 |
|
|
---------------------------------------------------------------------
|
10 |
|
|
---- ----
|
11 |
|
|
---- Copyright (C) 2009 David Lundgren ----
|
12 |
|
|
---- davidklun@gmail.com ----
|
13 |
|
|
---- ----
|
14 |
|
|
---- This source file may be used and distributed without ----
|
15 |
|
|
---- restriction provided that this copyright statement is not ----
|
16 |
|
|
---- removed from the file and that any derivative work contains ----
|
17 |
|
|
---- the original copyright notice and the associated disclaimer.----
|
18 |
|
|
---- ----
|
19 |
|
|
---- THIS SOFTWARE IS PROVIDED ``AS IS'' AND WITHOUT ANY ----
|
20 |
|
|
---- EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED ----
|
21 |
|
|
---- TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS ----
|
22 |
|
|
---- FOR A PARTICULAR PURPOSE. IN NO EVENT SHALL THE AUTHOR ----
|
23 |
|
|
---- OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, ----
|
24 |
|
|
---- INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES ----
|
25 |
|
|
---- (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE ----
|
26 |
|
|
---- GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR ----
|
27 |
|
|
---- BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF ----
|
28 |
|
|
---- LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT ----
|
29 |
|
|
---- (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT ----
|
30 |
|
|
---- OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE ----
|
31 |
|
|
---- POSSIBILITY OF SUCH DAMAGE. ----
|
32 |
|
|
---- ----
|
33 |
|
|
---------------------------------------------------------------------
|
34 |
|
|
|
35 |
|
|
LIBRARY ieee;
|
36 |
|
|
USE ieee.std_logic_1164.all;
|
37 |
|
|
USE ieee.std_logic_arith.all;
|
38 |
|
|
use ieee.std_logic_unsigned.all;
|
39 |
|
|
use ieee.std_logic_misc.all;
|
40 |
|
|
library work;
|
41 |
|
|
use work.fpupack.all;
|
42 |
|
|
|
43 |
|
|
ENTITY fpu_mul IS
|
44 |
|
|
|
45 |
|
|
PORT(
|
46 |
|
|
clk : IN std_logic;
|
47 |
|
|
rst : IN std_logic;
|
48 |
|
|
enable : IN std_logic;
|
49 |
|
|
opa : IN std_logic_vector (63 DOWNTO 0);
|
50 |
|
|
opb : IN std_logic_vector (63 DOWNTO 0);
|
51 |
|
|
sign : OUT std_logic;
|
52 |
|
|
product_7 : OUT std_logic_vector (55 DOWNTO 0);
|
53 |
|
|
exponent_5 : OUT std_logic_vector (11 DOWNTO 0)
|
54 |
|
|
);
|
55 |
|
|
|
56 |
|
|
END fpu_mul;
|
57 |
|
|
|
58 |
|
|
|
59 |
|
|
architecture rtl of fpu_mul is
|
60 |
|
|
|
61 |
|
|
signal product_shift : std_logic_vector(5 downto 0);
|
62 |
|
|
signal product_shift_2 : std_logic_vector(5 downto 0);
|
63 |
|
|
signal mantissa_a : std_logic_vector(51 downto 0);
|
64 |
|
|
signal mantissa_b : std_logic_vector(51 downto 0);
|
65 |
|
|
signal exponent_a : std_logic_vector(11 downto 0);
|
66 |
|
|
signal exponent_b : std_logic_vector(11 downto 0);
|
67 |
|
|
signal a_is_norm : std_logic;
|
68 |
|
|
signal b_is_norm : std_logic;
|
69 |
|
|
signal a_is_zero : std_logic;
|
70 |
|
|
signal b_is_zero : std_logic;
|
71 |
|
|
signal in_zero : std_logic;
|
72 |
|
|
signal exponent_terms : std_logic_vector(11 downto 0);
|
73 |
|
|
signal exponent_gt_expoffset : std_logic;
|
74 |
|
|
signal exponent_under : std_logic_vector(11 downto 0);
|
75 |
|
|
signal exponent_1 : std_logic_vector(11 downto 0);
|
76 |
|
|
signal exponent : std_logic_vector(11 downto 0);
|
77 |
|
|
signal exponent_2 : std_logic_vector(11 downto 0);
|
78 |
|
|
signal exponent_gt_prodshift : std_logic;
|
79 |
|
|
signal exponent_3 : std_logic_vector(11 downto 0);
|
80 |
|
|
signal exponent_4 : std_logic_vector(11 downto 0);
|
81 |
|
|
signal exponent_et_zero : std_logic;
|
82 |
|
|
signal mul_a : std_logic_vector(52 downto 0);
|
83 |
|
|
signal mul_b : std_logic_vector(52 downto 0);
|
84 |
|
|
signal product_a : std_logic_vector(40 downto 0);
|
85 |
|
|
signal product_b : std_logic_vector(40 downto 0);
|
86 |
|
|
signal product_c : std_logic_vector(40 downto 0);
|
87 |
|
|
signal product_d : std_logic_vector(25 downto 0);
|
88 |
|
|
signal product_e : std_logic_vector(33 downto 0);
|
89 |
|
|
signal product_f : std_logic_vector(33 downto 0);
|
90 |
|
|
signal product_g : std_logic_vector(35 downto 0);
|
91 |
|
|
signal product_h : std_logic_vector(28 downto 0);
|
92 |
|
|
signal product_i : std_logic_vector(28 downto 0);
|
93 |
|
|
signal product_j : std_logic_vector(30 downto 0);
|
94 |
|
|
signal sum_0 : std_logic_vector(41 downto 0);
|
95 |
|
|
signal sum_1 : std_logic_vector(35 downto 0);
|
96 |
|
|
signal sum_2 : std_logic_vector(41 downto 0);
|
97 |
|
|
signal sum_3 : std_logic_vector(35 downto 0);
|
98 |
|
|
signal sum_4 : std_logic_vector(36 downto 0);
|
99 |
|
|
signal sum_5 : std_logic_vector(27 downto 0);
|
100 |
|
|
signal sum_6 : std_logic_vector(29 downto 0);
|
101 |
|
|
signal sum_7 : std_logic_vector(36 downto 0);
|
102 |
|
|
signal sum_8 : std_logic_vector(30 downto 0);
|
103 |
|
|
signal product : std_logic_vector(105 downto 0);
|
104 |
|
|
signal product_1 : std_logic_vector(105 downto 0);
|
105 |
|
|
signal product_2 : std_logic_vector(105 downto 0);
|
106 |
|
|
signal product_3 : std_logic_vector(105 downto 0);
|
107 |
|
|
signal product_4 : std_logic_vector(105 downto 0);
|
108 |
|
|
signal product_5 : std_logic_vector(105 downto 0);
|
109 |
|
|
signal product_6 : std_logic_vector(105 downto 0);
|
110 |
|
|
signal product_lsb : std_logic;
|
111 |
|
|
|
112 |
|
|
begin
|
113 |
|
|
product_7 <= '0' & product_6(105 downto 52) & product_lsb;
|
114 |
|
|
exponent <= "000000000000";
|
115 |
|
|
process
|
116 |
|
|
begin
|
117 |
|
|
wait until clk'event and clk = '1';
|
118 |
|
|
if (rst = '1') then
|
119 |
|
|
sign <= '0';
|
120 |
|
|
mantissa_a <= (others =>'0');
|
121 |
|
|
mantissa_b <= (others =>'0');
|
122 |
|
|
exponent_a <= (others =>'0');
|
123 |
|
|
exponent_b <= (others =>'0');
|
124 |
|
|
a_is_norm <= '0';
|
125 |
|
|
b_is_norm <= '0';
|
126 |
|
|
a_is_zero <= '0';
|
127 |
|
|
b_is_zero <= '0';
|
128 |
|
|
in_zero <= '0';
|
129 |
|
|
exponent_terms <= (others =>'0');
|
130 |
|
|
exponent_gt_expoffset <= '0';
|
131 |
|
|
exponent_under <= (others =>'0');
|
132 |
|
|
exponent_1 <= (others =>'0');
|
133 |
|
|
exponent_2 <= (others =>'0');
|
134 |
|
|
exponent_gt_prodshift <= '0';
|
135 |
|
|
exponent_3 <= (others =>'0');
|
136 |
|
|
exponent_4 <= (others =>'0');
|
137 |
|
|
exponent_et_zero <= '0';
|
138 |
|
|
mul_a <= (others =>'0');
|
139 |
|
|
mul_b <= (others =>'0');
|
140 |
|
|
product_a <= (others =>'0');
|
141 |
|
|
product_b <= (others =>'0');
|
142 |
|
|
product_c <= (others =>'0');
|
143 |
|
|
product_d <= (others =>'0');
|
144 |
|
|
product_e <= (others =>'0');
|
145 |
|
|
product_f <= (others =>'0');
|
146 |
|
|
product_g <= (others =>'0');
|
147 |
|
|
product_h <= (others =>'0');
|
148 |
|
|
product_i <= (others =>'0');
|
149 |
|
|
product_j <= (others =>'0');
|
150 |
|
|
sum_0 <= (others =>'0');
|
151 |
|
|
sum_1 <= (others =>'0');
|
152 |
|
|
sum_2 <= (others =>'0');
|
153 |
|
|
sum_3 <= (others =>'0');
|
154 |
|
|
sum_4 <= (others =>'0');
|
155 |
|
|
sum_5 <= (others =>'0');
|
156 |
|
|
sum_6 <= (others =>'0');
|
157 |
|
|
sum_7 <= (others =>'0');
|
158 |
|
|
sum_8 <= (others =>'0');
|
159 |
|
|
product <= (others =>'0');
|
160 |
|
|
product_1 <= (others =>'0');
|
161 |
|
|
product_2 <= (others =>'0');
|
162 |
|
|
product_3 <= (others =>'0');
|
163 |
|
|
product_4 <= (others =>'0');
|
164 |
|
|
product_5 <= (others =>'0');
|
165 |
|
|
product_6 <= (others =>'0');
|
166 |
|
|
product_lsb <= '0';
|
167 |
|
|
exponent_5 <= (others =>'0');
|
168 |
|
|
product_shift <= (others =>'0');
|
169 |
|
|
product_shift_2 <= (others =>'0');
|
170 |
|
|
elsif (enable = '1') then
|
171 |
|
|
sign <= opa(63) xor opb(63);
|
172 |
|
|
exponent_a <= '0' & opa(62 downto 52);
|
173 |
|
|
exponent_b <= '0' & opb(62 downto 52);
|
174 |
|
|
mantissa_a <= opa(51 downto 0);
|
175 |
|
|
mantissa_b <= opb(51 downto 0);
|
176 |
|
|
a_is_norm <= or_reduce(exponent_a);
|
177 |
|
|
b_is_norm <= or_reduce(exponent_b);
|
178 |
|
|
a_is_zero <= not or_reduce(opa(62 downto 0));
|
179 |
|
|
b_is_zero <= not or_reduce(opb(62 downto 0));
|
180 |
|
|
in_zero <= a_is_zero or b_is_zero;
|
181 |
|
|
exponent_terms <= exponent_a + exponent_b + ( "0000000000" & not a_is_norm) +
|
182 |
|
|
("0000000000" & not b_is_norm);
|
183 |
|
|
if (exponent_terms > "001111111101") then
|
184 |
|
|
exponent_gt_expoffset <= '1';
|
185 |
|
|
else
|
186 |
|
|
exponent_gt_expoffset <= '0';
|
187 |
|
|
end if;
|
188 |
|
|
exponent_under <= "001111111110" - exponent_terms;
|
189 |
|
|
exponent_1 <= exponent_terms - "001111111110";
|
190 |
|
|
if (exponent_gt_expoffset = '1') then
|
191 |
|
|
exponent_2 <= exponent_1;
|
192 |
|
|
else
|
193 |
|
|
exponent_2 <= exponent;
|
194 |
|
|
end if;
|
195 |
|
|
if (exponent_2 > product_shift_2) then
|
196 |
|
|
exponent_gt_prodshift <= '1';
|
197 |
|
|
else
|
198 |
|
|
exponent_gt_prodshift <= '0';
|
199 |
|
|
end if;
|
200 |
|
|
exponent_3 <= exponent_2 - product_shift_2;
|
201 |
|
|
if (exponent_gt_prodshift = '1') then
|
202 |
|
|
exponent_4 <= exponent_3;
|
203 |
|
|
else
|
204 |
|
|
exponent_4 <= exponent;
|
205 |
|
|
end if;
|
206 |
|
|
if (exponent_4 = "000000000000") then
|
207 |
|
|
exponent_et_zero <= '1';
|
208 |
|
|
else
|
209 |
|
|
exponent_et_zero <= '0';
|
210 |
|
|
end if;
|
211 |
|
|
mul_a <= a_is_norm & mantissa_a;
|
212 |
|
|
mul_b <= b_is_norm & mantissa_b;
|
213 |
|
|
product_a <= mul_a(23 downto 0) * mul_b(16 downto 0);
|
214 |
|
|
product_b <= mul_a(23 downto 0) * mul_b(33 downto 17);
|
215 |
|
|
product_c <= mul_a(23 downto 0) * mul_b(50 downto 34);
|
216 |
|
|
product_d <= mul_a(23 downto 0) * mul_b(52 downto 51);
|
217 |
|
|
product_e <= mul_a(40 downto 24) * mul_b(16 downto 0);
|
218 |
|
|
product_f <= mul_a(40 downto 24) * mul_b(33 downto 17);
|
219 |
|
|
product_g <= mul_a(40 downto 24) * mul_b(52 downto 34);
|
220 |
|
|
product_h <= mul_a(52 downto 41) * mul_b(16 downto 0);
|
221 |
|
|
product_i <= mul_a(52 downto 41) * mul_b(33 downto 17);
|
222 |
|
|
product_j <= mul_a(52 downto 41) * mul_b(52 downto 34);
|
223 |
|
|
sum_0 <= product_a(40 downto 17) + ( '0' & product_b);
|
224 |
|
|
sum_1 <= ('0' & sum_0(41 downto 7)) + product_e;
|
225 |
|
|
sum_2 <= sum_1(35 downto 10) + ('0' & product_c);
|
226 |
|
|
sum_3 <= ( '0' & sum_2(41 downto 7)) + product_h;
|
227 |
|
|
sum_4 <= ( '0' & sum_3) + product_f;
|
228 |
|
|
sum_5 <= ('0' & sum_4(36 downto 10)) + product_d;
|
229 |
|
|
sum_6 <= sum_5(27 downto 7) + ('0' & product_i);
|
230 |
|
|
sum_7 <= sum_6 + ('0' & product_g);
|
231 |
|
|
sum_8 <= sum_7(36 downto 17) + product_j;
|
232 |
|
|
product <= sum_8 & sum_7(16 downto 0) & sum_5(6 downto 0) & sum_4(9 downto 0) & sum_2(6 downto 0) &
|
233 |
|
|
sum_1(9 downto 0) & sum_0(6 downto 0) & product_a(16 downto 0);
|
234 |
|
|
product_1 <= shr(product, exponent_under);
|
235 |
|
|
if (exponent_gt_prodshift = '1') then
|
236 |
|
|
product_5 <= product_3;
|
237 |
|
|
else
|
238 |
|
|
product_5 <= product_4;
|
239 |
|
|
end if;
|
240 |
|
|
if (exponent_gt_expoffset = '1') then
|
241 |
|
|
product_2 <= product;
|
242 |
|
|
else
|
243 |
|
|
product_2 <= product_1;
|
244 |
|
|
end if;
|
245 |
|
|
product_3 <= shl(product_2, product_shift_2);
|
246 |
|
|
product_4 <= shl(product_2, exponent_2);
|
247 |
|
|
if (exponent_gt_prodshift = '1') then
|
248 |
|
|
product_5 <= product_3;
|
249 |
|
|
else
|
250 |
|
|
product_5 <= product_4;
|
251 |
|
|
end if;
|
252 |
|
|
if (exponent_et_zero = '1') then
|
253 |
|
|
product_6 <= shr(product_5, conv_std_logic_vector('1', 106));
|
254 |
|
|
else
|
255 |
|
|
product_6 <= product_5;
|
256 |
|
|
end if;
|
257 |
|
|
product_lsb <= or_reduce(product_6(51 downto 0));
|
258 |
|
|
if (in_zero = '1') then
|
259 |
|
|
exponent_5 <= "000000000000";
|
260 |
|
|
else
|
261 |
|
|
exponent_5 <= exponent_4;
|
262 |
|
|
end if;
|
263 |
|
|
product_shift <= count_zeros_mul(product(105 downto 0));
|
264 |
|
|
product_shift_2 <= product_shift;
|
265 |
|
|
end if;
|
266 |
|
|
end process;
|
267 |
|
|
end rtl;
|