OpenCores
URL https://opencores.org/ocsvn/ft816float/ft816float/trunk

Subversion Repositories ft816float

[/] [ft816float/] [trunk/] [rtl/] [verilog2/] [DFPDivide96.sv] - Blame information for rev 88

Go to most recent revision | Details | Compare with Previous | View Log

Line No. Rev Author Line
1 75 robfinch
`timescale 1ns / 1ps
2
// ============================================================================
3
//        __
4
//   \\__/ o\    (C) 2006-2022  Robert Finch, Waterloo
5
//    \  __ /    All rights reserved.
6
//     \/_//     robfinch@finitron.ca
7
//       ||
8
//
9
//      DFPDivide96.sv
10
//    - decimal floating point divider
11
//    - parameterized width
12
//
13
//
14
// BSD 3-Clause License
15
// Redistribution and use in source and binary forms, with or without
16
// modification, are permitted provided that the following conditions are met:
17
//
18
// 1. Redistributions of source code must retain the above copyright notice, this
19
//    list of conditions and the following disclaimer.
20
//
21
// 2. Redistributions in binary form must reproduce the above copyright notice,
22
//    this list of conditions and the following disclaimer in the documentation
23
//    and/or other materials provided with the distribution.
24
//
25
// 3. Neither the name of the copyright holder nor the names of its
26
//    contributors may be used to endorse or promote products derived from
27
//    this software without specific prior written permission.
28
//
29
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
30
// AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
31
// IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
32
// DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
33
// FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
34
// DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
35
// SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
36
// CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
37
// OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
38
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
39
//
40
//      Floating Point Divider
41
//
42
//Properties:
43
//+-inf * +-inf = -+inf    (this is handled by exOver)
44
//+-inf * 0     = QNaN
45
//+-0 / +-0      = QNaN
46
// ============================================================================
47
 
48
import DFPPkg::*;
49
 
50
`define QINFDIV         4'd2
51
`define QZEROZERO       4'd3
52
 
53
module DFPDivide96(rst, clk, ce, ld, op, a, b, o, done, sign_exe, overflow, underflow);
54
parameter N=25;
55
// FADD is a constant that makes the divider width a multiple of four and includes eight extra bits.
56
input rst;
57
input clk;
58
input ce;
59
input ld;
60
input op;
61
input  DFP96 a, b;
62
output DFP96UD o;
63
output reg done;
64
output sign_exe;
65
output overflow;
66
output underflow;
67
 
68
// registered outputs
69
reg sign_exe=0;
70
reg inf=0;
71
reg     overflow=0;
72
reg     underflow=0;
73
 
74
reg so, sxo;
75
reg [11:0] xo;
76
reg [(N+1)*4*2-1:0] mo;
77
 
78
DFP96U au, bu;
79
DFPUnpack96 u01 (a, au);
80
DFPUnpack96 u02 (b, bu);
81
 
82
// constants
83
wire [11:0] infXp = 12'hBFF;    // infinite / NaN - all ones
84
wire [11:0] bias = 12'h5FF;
85
// The following is the value for an exponent of zero, with the offset
86
// eg. 8'h7f for eight bit exponent, 11'h7ff for eleven bit exponent, etc.
87
// The following is a template for a quiet nan. (MSB=1)
88
wire [N*4-1:0] qNaN  = {4'h1,{(N-1)*4{1'b0}}};
89
 
90
// variables
91
wire [(N+2)*4*2-1:0] divo;
92
 
93
// Operands
94
reg sa, sb;                     // sign bit
95
reg [N*4-1:0] siga, sigb;
96
reg az, bz;
97
reg aInf, bInf;
98
reg aNan,bNan;
99
wire done1;
100
wire signed [7:0] lzcnt;
101
 
102
// -----------------------------------------------------------
103
// Clock #1
104
// - decode the input operands
105
// - derive basic information
106
// - calculate fraction
107
// -----------------------------------------------------------
108
reg ld1;
109
always @(posedge clk)
110
        if (ce) sa <= au.sign;
111
always @(posedge clk)
112
        if (ce) sb <= bu.sign;
113
always @(posedge clk)
114
        if (ce) siga <= au.sig;
115
always @(posedge clk)
116
        if (ce) sigb <= bu.sig;
117
always @(posedge clk)
118
        if (ce) az <= au.exp==12'd0 && au.sig==100'd0;
119
always @(posedge clk)
120
        if (ce) bz <= bu.exp==12'd0 && bu.sig==100'd0;
121
always @(posedge clk)
122
        if (ce) aInf <= au.infinity;
123
always @(posedge clk)
124
        if (ce) bInf <= bu.infinity;
125
always @(posedge clk)
126
        if (ce) aNan <= au.nan;
127
always @(posedge clk)
128
        if (ce) bNan <= bu.nan;
129
ft_delay #(.WID(1), .DEP(1)) udly1 (.clk(clk), .ce(ce), .i(ld), .o(ld1));
130
 
131
// -----------------------------------------------------------
132
// Clock #2 to N
133
// - calculate fraction
134
// -----------------------------------------------------------
135
wire done3a,done3;
136
// Perform divide
137 82 robfinch
dfdiv #(N+2) u2 (.clk(clk), .ld(ld1), .a({siga,8'b0}), .b({sigb,8'b0}), .q(divo), .r(), .done(done1), .lzcnt(lzcnt));
138 80 robfinch
//wire [7:0] lzcnt_bin = lzcnt[3:0] + (lzcnt[7:4] * 10);
139
wire [(N+2)*4*2-1:0] divo1 = divo[(N+2)*4*2-1:0] << ({lzcnt-1,2'b0});//WAS FPWID=128?+44
140 75 robfinch
ft_delay #(.WID(1), .DEP(3)) u3 (.clk(clk), .ce(ce), .i(done1), .o(done3a));
141
assign done3 = done1&done3a;
142
 
143
// -----------------------------------------------------------
144
// Clock #N+1
145
// - calculate exponent
146
// - calculate fraction
147
// - determine when a NaN is output
148
// -----------------------------------------------------------
149
// Compute the exponent.
150
// - correct the exponent for denormalized operands
151
// - adjust the difference by the bias (add 127)
152
// - also factor in the different decimal position for division
153
reg [13:0] ex1; // sum of exponents
154
reg qNaNOut;
155
 
156
always @(posedge clk)
157 85 robfinch
  if (ce) ex1 <= au.exp - bu.exp + bias - ((lzcnt >= N) ? lzcnt-N-1 : 0);
158 75 robfinch
 
159
always @(posedge clk)
160
  if (ce) qNaNOut <= (az&bz)|(aInf&bInf);
161
 
162
wire over = 1'b0;
163
wire under = &ex1[13:12];
164
reg [3:0] st;
165
 
166
// -----------------------------------------------------------
167
// Clock #N+3
168
// -----------------------------------------------------------
169
always_ff @(posedge clk)
170
// Simulation likes to see these values reset to zero on reset. Otherwise the
171
// values propagate in sim as X's.
172
if (rst) begin
173
        xo <= 1'd0;
174
        mo <= 1'd0;
175
        so <= 1'd0;
176
        sign_exe <= 1'd0;
177
        overflow <= 1'd0;
178
        underflow <= 1'd0;
179
        done <= 1'b1;
180
end
181
else if (ce) begin
182
  done <= 1'b0;
183
        if (done3&done1) begin
184
          done <= 1'b1;
185
 
186
                casez({qNaNOut|aNan|bNan,bInf,bz,over,under})
187
                5'b1????:               xo <= infXp;    // NaN exponent value
188
                5'b01???:               xo <= 1'd0;             // divide by inf
189
                5'b001??:               xo <= infXp;    // divide by zero
190
                5'b0001?:               xo <= infXp;    // overflow
191
                5'b00001:               xo <= 1'd0;             // underflow
192
                default:                xo <= ex1;      // normal or underflow: passthru neg. exp. for normalization
193
                endcase
194
 
195
                casez({aNan,bNan,qNaNOut,bInf,bz,over,aInf&bInf,az&bz})
196
                8'b1???????:  begin mo <= {4'h1,au[N*4-1:0],{(N+1)*4-1{1'b0}}}; st[3] <= 1'b1; end
197
                8'b01??????:  begin mo <= {4'h1,bu[N*4-1:0],{(N+1)*4-1{1'b0}}}; st[3] <= 1'b1; end
198
                8'b001?????:    begin mo <= {4'h1,qNaN[N*4-1:0]|{aInf,1'b0}|{az,bz},{(N+1)*4-1{1'b0}}}; st[3] <= 1'b1; end
199
                8'b0001????:    begin mo <= {(N+1)*4*2-1{1'd0}};        st[3] <= 1'b0; end      // div by inf
200
                8'b00001???:    begin mo <= {(N+1)*4*2-1{1'd0}};        st[3] <= 1'b0; end      // div by zero
201
                8'b000001??:    begin mo <= {(N+1)*4*2-1{1'd0}};        st[3] <= 1'b0; end      // Inf exponent
202
                8'b0000001?:    begin mo <= {4'h1,qNaN|`QINFDIV,{(N+1)*4-1{1'b0}}};     st[3] <= 1'b1; end      // infinity / infinity
203
                8'b00000001:    begin mo <= {4'h1,qNaN|`QZEROZERO,{(N+1)*4-1{1'b0}}};   st[3] <= 1'b1; end      // zero / zero
204
                default:                begin mo <= divo1[(N+2)*4*2-1:8];       st[3] <= 1'b0; end      // plain div
205
                endcase
206
 
207
                sign_exe        <= sa & sb;
208
                overflow        <= over;
209
                underflow       <= under;
210
 
211
                o.nan <= aNan|bNan|qNaNOut;
212
                o.snan <= aNan|bNan|qNaNOut;
213
                o.qnan <= 1'b0;
214
                o.infinity <= over|aInf;
215
                o.sign <= sa ^ sb;
216
                o.exp <= xo;
217
                o.sig <= mo;
218
        end
219
end
220
 
221
endmodule
222
 
223
module DFPDivide96nr(rst, clk, ce, ld, op, a, b, o, rm, done, sign_exe, inf, overflow, underflow);
224
parameter N=25;
225
input rst;
226
input clk;
227
input ce;
228
input ld;
229
input op;
230
input  DFP96 a, b;
231
output DFP96 o;
232
input [2:0] rm;
233
output sign_exe;
234
output done;
235
output inf;
236
output overflow;
237
output underflow;
238
 
239
DFP96UD o1;
240
wire sign_exe1, inf1, overflow1, underflow1;
241
DFP96UN fpn0;
242
wire done1, done1a;
243
 
244
DFPDivide96    #(.N(N)) u1 (rst, clk, ce, ld, op, a, b, o1, done1, sign_exe1, overflow1, underflow1);
245
DFPNormalize96 #(.N(N)) u2(.clk(clk), .ce(ce), .under_i(underflow1), .i(o1), .o(fpn0) );
246
DFPRound96     #(.N(N)) u3(.clk(clk), .ce(ce), .rm(rm), .i(fpn0), .o(o) );
247
delay2      #(1)   u4(.clk(clk), .ce(ce), .i(sign_exe1), .o(sign_exe));
248
delay2      #(1)   u5(.clk(clk), .ce(ce), .i(inf1), .o(inf));
249
delay2      #(1)   u6(.clk(clk), .ce(ce), .i(overflow1), .o(overflow));
250
delay2      #(1)   u7(.clk(clk), .ce(ce), .i(underflow1), .o(underflow));
251 81 robfinch
ft_delay        #(.WID(1),.DEP(14))   u8(.clk(clk), .ce(ce), .i(done1), .o(done1a));
252 75 robfinch
assign done = done1&done1a;
253
 
254
endmodule
255
 

powered by: WebSVN 2.1.0

© copyright 1999-2024 OpenCores.org, equivalent to Oliscience, all rights reserved. OpenCores®, registered trademark.