1 |
2 |
dmitryr |
// ========== Copyright Header Begin ==========================================
|
2 |
|
|
//
|
3 |
|
|
// OpenSPARC T1 Processor File: sparc_ffu_dp.v
|
4 |
|
|
// Copyright (c) 2006 Sun Microsystems, Inc. All Rights Reserved.
|
5 |
|
|
// DO NOT ALTER OR REMOVE COPYRIGHT NOTICES.
|
6 |
|
|
//
|
7 |
|
|
// The above named program is free software; you can redistribute it and/or
|
8 |
|
|
// modify it under the terms of the GNU General Public
|
9 |
|
|
// License version 2 as published by the Free Software Foundation.
|
10 |
|
|
//
|
11 |
|
|
// The above named program is distributed in the hope that it will be
|
12 |
|
|
// useful, but WITHOUT ANY WARRANTY; without even the implied warranty of
|
13 |
|
|
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
|
14 |
|
|
// General Public License for more details.
|
15 |
|
|
//
|
16 |
|
|
// You should have received a copy of the GNU General Public
|
17 |
|
|
// License along with this work; if not, write to the Free Software
|
18 |
|
|
// Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301, USA.
|
19 |
|
|
//
|
20 |
|
|
// ========== Copyright Header End ============================================
|
21 |
|
|
///////////////////////////////////////////////////////////////////////
|
22 |
|
|
/*
|
23 |
|
|
// Module Name: sparc_ffu_dp
|
24 |
|
|
// Description: This is the ffu datapath. It stores the 2 128 bit operands
|
25 |
|
|
// and the result (puts result in the 1st source to save space).
|
26 |
|
|
*/
|
27 |
|
|
|
28 |
|
|
`include "iop.h"
|
29 |
|
|
|
30 |
|
|
module sparc_ffu_dp (/*AUTOARG*/
|
31 |
|
|
// Outputs
|
32 |
|
|
so, dp_frf_data, ffu_lsu_data, dp_vis_rs1_data, dp_vis_rs2_data,
|
33 |
|
|
dp_ctl_rs2_sign, dp_ctl_fsr_fcc, dp_ctl_fsr_rnd, dp_ctl_fsr_tem,
|
34 |
|
|
dp_ctl_fsr_aexc, dp_ctl_fsr_cexc, dp_ctl_ld_fcc,
|
35 |
|
|
dp_ctl_gsr_mask_e, dp_ctl_gsr_scale_e, dp_ctl_synd_out_low,
|
36 |
|
|
dp_ctl_synd_out_high,
|
37 |
|
|
// Inputs
|
38 |
|
|
rclk, se, si, ctl_dp_rst_l, frf_dp_data, cpx_fpu_data,
|
39 |
|
|
lsu_ffu_ld_data, vis_dp_rd_data, ctl_dp_wsr_data_w2, ctl_dp_sign,
|
40 |
|
|
ctl_dp_exc_w2, ctl_dp_fcc_w2, ctl_dp_ftt_w2, ctl_dp_noshift64_frf,
|
41 |
|
|
ctl_dp_shift_frf_right, ctl_dp_shift_frf_left,
|
42 |
|
|
ctl_dp_zero_low32_frf, ctl_dp_output_sel_rs1,
|
43 |
|
|
ctl_dp_output_sel_rs2, ctl_dp_output_sel_frf,
|
44 |
|
|
ctl_dp_output_sel_fsr, ctl_dp_noflip_lsu, ctl_dp_flip_lsu,
|
45 |
|
|
ctl_dp_noflip_fpu, ctl_dp_flip_fpu, ctl_dp_rs2_frf_read,
|
46 |
|
|
ctl_dp_rs2_sel_vis, ctl_dp_rs2_sel_fpu_lsu, ctl_dp_rs2_keep_data,
|
47 |
|
|
ctl_dp_rd_ecc, ctl_dp_fp_thr, ctl_dp_fsr_sel_old,
|
48 |
|
|
ctl_dp_fsr_sel_ld, ctl_dp_fsr_sel_fpu, ctl_dp_gsr_wsr_w2,
|
49 |
|
|
ctl_dp_thr_e, ctl_dp_new_rs1, ctl_dp_ecc_sel_frf
|
50 |
|
|
) ;
|
51 |
|
|
input rclk;
|
52 |
|
|
input se;
|
53 |
|
|
input si;
|
54 |
|
|
input ctl_dp_rst_l;
|
55 |
|
|
input [77:0] frf_dp_data;
|
56 |
|
|
input [63:0] cpx_fpu_data;
|
57 |
|
|
input [63:0] lsu_ffu_ld_data;
|
58 |
|
|
input [63:0] vis_dp_rd_data;
|
59 |
|
|
|
60 |
|
|
input [36:0] ctl_dp_wsr_data_w2;
|
61 |
|
|
|
62 |
|
|
input [1:0] ctl_dp_sign; // sign after abs or neg
|
63 |
|
|
input [9:0] ctl_dp_exc_w2;
|
64 |
|
|
input [7:0] ctl_dp_fcc_w2;
|
65 |
|
|
input [2:0] ctl_dp_ftt_w2;
|
66 |
|
|
|
67 |
|
|
// mux selects
|
68 |
|
|
input ctl_dp_noshift64_frf; // choose output from FRF
|
69 |
|
|
input ctl_dp_shift_frf_right;
|
70 |
|
|
input ctl_dp_shift_frf_left;
|
71 |
|
|
|
72 |
|
|
input ctl_dp_zero_low32_frf;
|
73 |
|
|
|
74 |
|
|
input ctl_dp_output_sel_rs1; // choose output to lsu
|
75 |
|
|
input ctl_dp_output_sel_rs2;
|
76 |
|
|
input ctl_dp_output_sel_frf;
|
77 |
|
|
input ctl_dp_output_sel_fsr;
|
78 |
|
|
|
79 |
|
|
input ctl_dp_noflip_lsu;// inputs from lsu and fpu
|
80 |
|
|
input ctl_dp_flip_lsu;
|
81 |
|
|
input ctl_dp_noflip_fpu;
|
82 |
|
|
input ctl_dp_flip_fpu;
|
83 |
|
|
|
84 |
|
|
input ctl_dp_rs2_frf_read; // choose r2
|
85 |
|
|
input ctl_dp_rs2_sel_vis;
|
86 |
|
|
input ctl_dp_rs2_sel_fpu_lsu;
|
87 |
|
|
input ctl_dp_rs2_keep_data;
|
88 |
|
|
input ctl_dp_rd_ecc;
|
89 |
|
|
|
90 |
|
|
input [3:0] ctl_dp_fp_thr;
|
91 |
|
|
|
92 |
|
|
input [3:0] ctl_dp_fsr_sel_old, // choose what to update FSR with
|
93 |
|
|
ctl_dp_fsr_sel_ld,
|
94 |
|
|
ctl_dp_fsr_sel_fpu;
|
95 |
|
|
|
96 |
|
|
input [3:0] ctl_dp_gsr_wsr_w2;
|
97 |
|
|
input [3:0] ctl_dp_thr_e;
|
98 |
|
|
|
99 |
|
|
|
100 |
|
|
// rs1 selects
|
101 |
|
|
input ctl_dp_new_rs1;
|
102 |
|
|
|
103 |
|
|
// 2:1 mux selects
|
104 |
|
|
input ctl_dp_ecc_sel_frf;
|
105 |
|
|
|
106 |
|
|
// outputs
|
107 |
|
|
output so;
|
108 |
|
|
output [63:0] dp_frf_data;
|
109 |
|
|
output [63:0] ffu_lsu_data;
|
110 |
|
|
output [63:0] dp_vis_rs1_data;
|
111 |
|
|
output [63:0] dp_vis_rs2_data;
|
112 |
|
|
output [1:0] dp_ctl_rs2_sign; // sign for rs2
|
113 |
|
|
output [7:0] dp_ctl_fsr_fcc;
|
114 |
|
|
output [1:0] dp_ctl_fsr_rnd;
|
115 |
|
|
output [4:0] dp_ctl_fsr_tem;
|
116 |
|
|
output [4:0] dp_ctl_fsr_aexc;
|
117 |
|
|
output [4:0] dp_ctl_fsr_cexc;
|
118 |
|
|
|
119 |
|
|
output [7:0] dp_ctl_ld_fcc;
|
120 |
|
|
|
121 |
|
|
output [31:0] dp_ctl_gsr_mask_e;
|
122 |
|
|
output [4:0] dp_ctl_gsr_scale_e;
|
123 |
|
|
|
124 |
|
|
|
125 |
|
|
output [6:0] dp_ctl_synd_out_low; // signals for ecc errors
|
126 |
|
|
output [6:0] dp_ctl_synd_out_high;
|
127 |
|
|
|
128 |
|
|
wire clk;
|
129 |
|
|
wire reset;
|
130 |
|
|
// local signals
|
131 |
|
|
wire [63:0] fpu_ffu_data;
|
132 |
|
|
wire [63:0] lsu_ffu_ld_data_d1;
|
133 |
|
|
wire [63:0] rs2_rd_data; // stores both the rs2 and rd data
|
134 |
|
|
wire [63:0] rs2_rd_data_next;
|
135 |
|
|
wire [63:0] write_data; // needed since block loads are pipelined
|
136 |
|
|
wire [63:0] rs2_data_changed;
|
137 |
|
|
wire [63:0] local_rd_data;
|
138 |
|
|
wire [63:0] rs1_data;
|
139 |
|
|
wire [63:0] rs1_data_next;
|
140 |
|
|
wire [63:0] shifted_frf_data;
|
141 |
|
|
wire [63:0] new_frf_data;
|
142 |
|
|
wire [63:0] lsu_fpu_data;
|
143 |
|
|
wire [63:0] frf_data_in;
|
144 |
|
|
wire [6:0] synd_in_low; // input ecc for lower word
|
145 |
|
|
wire [6:0] synd_in_h; // input ecc for upper word
|
146 |
|
|
wire [63:0] corr_data_next;
|
147 |
|
|
wire [63:0] corr_data;
|
148 |
|
|
wire [63:0] ecc_data_in;
|
149 |
|
|
|
150 |
|
|
wire [27:0] current_fsr,
|
151 |
|
|
t0_fsr,
|
152 |
|
|
t1_fsr,
|
153 |
|
|
t2_fsr,
|
154 |
|
|
t3_fsr;
|
155 |
|
|
wire [27:0] t0_fsr_nxt,
|
156 |
|
|
t1_fsr_nxt,
|
157 |
|
|
t2_fsr_nxt,
|
158 |
|
|
t3_fsr_nxt;
|
159 |
|
|
wire [27:0] t0_ldfsr_data,
|
160 |
|
|
t0_fpufsr_data;
|
161 |
|
|
wire [27:0] t1_ldfsr_data,
|
162 |
|
|
t1_fpufsr_data;
|
163 |
|
|
wire [27:0] t2_ldfsr_data,
|
164 |
|
|
t2_fpufsr_data;
|
165 |
|
|
wire [27:0] t3_ldfsr_data,
|
166 |
|
|
t3_fpufsr_data;
|
167 |
|
|
|
168 |
|
|
wire [36:0] gsr_e;
|
169 |
|
|
wire [36:0] t0_gsr;
|
170 |
|
|
wire [36:0] t0_gsr_nxt;
|
171 |
|
|
wire [36:0] t1_gsr;
|
172 |
|
|
wire [36:0] t1_gsr_nxt;
|
173 |
|
|
wire [36:0] t2_gsr;
|
174 |
|
|
wire [36:0] t2_gsr_nxt;
|
175 |
|
|
wire [36:0] t3_gsr;
|
176 |
|
|
wire [36:0] t3_gsr_nxt;
|
177 |
|
|
|
178 |
|
|
assign reset = ~ctl_dp_rst_l;
|
179 |
|
|
assign clk= rclk;
|
180 |
|
|
|
181 |
|
|
dff_s #(64) cpx_reg(.din(cpx_fpu_data[63:0]),
|
182 |
|
|
.q (fpu_ffu_data[63:0]),
|
183 |
|
|
.clk (clk), .se(se), .si(), .so());
|
184 |
|
|
|
185 |
|
|
// flop for lsu data. the data is flopped in ffu, but the vld is flopped in the lsu.
|
186 |
|
|
// This is for timing reasons on the valid bit and Sanjay didn't want to redo the
|
187 |
|
|
// lsu dp for the data portion
|
188 |
|
|
dff_s #(64) lsu_data_dff(.din(lsu_ffu_ld_data[63:0]), .clk(clk), .q(lsu_ffu_ld_data_d1[63:0]),
|
189 |
|
|
.se(se), .si(), .so());
|
190 |
|
|
assign dp_ctl_ld_fcc[7:0] = {lsu_ffu_ld_data_d1[37:32], lsu_ffu_ld_data_d1[11:10]};
|
191 |
|
|
|
192 |
|
|
///////////////////////////////////////////////
|
193 |
|
|
// Input from FRF (shift as needed for singles)
|
194 |
|
|
// The data needs to be shifted around because these are 64 bit reads but
|
195 |
|
|
// the required data might be in either the upper or lower 32 bits for
|
196 |
|
|
// singles. If it is a double then the data is left alone.
|
197 |
|
|
// If it is a single move and the source and target have the same alignment
|
198 |
|
|
// then no change happens. If it is a single move and the source and target
|
199 |
|
|
// have different alignments the operands get moved into place for the write.
|
200 |
|
|
// If it is data that will be sent to the lsu the data is moved into the lower
|
201 |
|
|
// 32 bits. If the data will be sent to the fpu the data is moved to the upper
|
202 |
|
|
// 32 bits (if not there already)
|
203 |
|
|
///////////////////////////////////////////////
|
204 |
|
|
assign frf_data_in[63:32] = frf_dp_data[70:39];
|
205 |
|
|
assign frf_data_in[31:0] = frf_dp_data[31:0];
|
206 |
|
|
mux3ds #(64) frf_input_mux(.dout(shifted_frf_data[63:0]),
|
207 |
|
|
.in0(frf_data_in[63:0]),
|
208 |
|
|
.in1({32'b0, frf_data_in[63:32]}),
|
209 |
|
|
.in2({frf_data_in[31:0], 32'b0}),
|
210 |
|
|
.sel0(ctl_dp_noshift64_frf),
|
211 |
|
|
.sel1(ctl_dp_shift_frf_right),
|
212 |
|
|
.sel2(ctl_dp_shift_frf_left));
|
213 |
|
|
assign new_frf_data[63:32] = shifted_frf_data[63:32];
|
214 |
|
|
assign new_frf_data[31:0] = shifted_frf_data[31:0] & {32{~ctl_dp_zero_low32_frf}};
|
215 |
|
|
|
216 |
|
|
mux4ds #(64) lsu_fpu_input_mux(.dout(lsu_fpu_data[63:0]),
|
217 |
|
|
.in0(lsu_ffu_ld_data_d1[63:0]),
|
218 |
|
|
.in1({lsu_ffu_ld_data_d1[31:0], 32'b0}),
|
219 |
|
|
.in2(fpu_ffu_data[63:0]),
|
220 |
|
|
.in3({32'b0, fpu_ffu_data[63:32]}),
|
221 |
|
|
.sel0(ctl_dp_noflip_lsu),
|
222 |
|
|
.sel1(ctl_dp_flip_lsu),
|
223 |
|
|
.sel2(ctl_dp_noflip_fpu),
|
224 |
|
|
.sel3(ctl_dp_flip_fpu));
|
225 |
|
|
|
226 |
|
|
// Data to FRF
|
227 |
|
|
dp_buffer #(64) frf_out_buf(.in(write_data[63:0]), .dout (dp_frf_data[63:0]));
|
228 |
|
|
|
229 |
|
|
|
230 |
|
|
|
231 |
|
|
// Data to LSU
|
232 |
|
|
// Mux for lsu data between two sets of data and the direct
|
233 |
|
|
// frf output for stores
|
234 |
|
|
mux4ds #(64) output_mux(.dout (ffu_lsu_data[63:0]),
|
235 |
|
|
.in0 (rs2_rd_data[63:0]),
|
236 |
|
|
.in1 (rs1_data[63:0]),
|
237 |
|
|
.in2 (shifted_frf_data[63:0]),
|
238 |
|
|
.in3 ({26'b0, current_fsr[27:20], 2'b0, current_fsr[19:15], 6'b0, current_fsr[14:12], 2'b0, current_fsr[11:0]}),
|
239 |
|
|
.sel0 (ctl_dp_output_sel_rs2),
|
240 |
|
|
.sel1 (ctl_dp_output_sel_rs1),
|
241 |
|
|
.sel2 (ctl_dp_output_sel_frf),
|
242 |
|
|
.sel3 (ctl_dp_output_sel_fsr));
|
243 |
|
|
|
244 |
|
|
// RS2 can take value from frf (with modification to sign), from lsu
|
245 |
|
|
// or keep value
|
246 |
|
|
// The modification to the sign bits allows for FABS and FNEG
|
247 |
|
|
assign dp_ctl_rs2_sign[1:0] = {new_frf_data[63], new_frf_data[31]};
|
248 |
|
|
|
249 |
|
|
assign rs2_data_changed[63:0] = {ctl_dp_sign[1], new_frf_data[62:32],
|
250 |
|
|
ctl_dp_sign[0], new_frf_data[30:0]};
|
251 |
|
|
|
252 |
|
|
dp_mux2es #(64) local_rd_mux(.dout(local_rd_data[63:0]),
|
253 |
|
|
.in0(rs2_data_changed[63:0]),
|
254 |
|
|
.in1(corr_data[63:0]),
|
255 |
|
|
.sel(ctl_dp_rd_ecc));
|
256 |
|
|
|
257 |
|
|
mux4ds #(64) rs2_rd_mux(.dout (rs2_rd_data_next[63:0]),
|
258 |
|
|
.in0 (local_rd_data[63:0]),
|
259 |
|
|
.in1 (vis_dp_rd_data[63:0]),
|
260 |
|
|
.in2 (lsu_fpu_data[63:0]),
|
261 |
|
|
.in3 (rs2_rd_data[63:0]),
|
262 |
|
|
.sel0 (ctl_dp_rs2_frf_read),
|
263 |
|
|
.sel1 (ctl_dp_rs2_sel_vis),
|
264 |
|
|
.sel2 (ctl_dp_rs2_sel_fpu_lsu),
|
265 |
|
|
.sel3 (ctl_dp_rs2_keep_data));
|
266 |
|
|
|
267 |
|
|
dff_s #(64) rs2_rd_dff(.din (rs2_rd_data_next[63:0]),
|
268 |
|
|
.q (rs2_rd_data[63:0]),
|
269 |
|
|
.clk (clk), .se(se), .si(), .so());
|
270 |
|
|
assign dp_vis_rs2_data[63:0] = rs2_rd_data[63:0];
|
271 |
|
|
dff_s #(64) write_data_dff(.din(rs2_rd_data[63:0]),
|
272 |
|
|
.q(write_data[63:0]),
|
273 |
|
|
.clk(clk), .se(se), .si(), .so());
|
274 |
|
|
|
275 |
|
|
////////////////////////////////////////////////////////
|
276 |
|
|
// RS1
|
277 |
|
|
////////////////////////////////////////////////////////
|
278 |
|
|
// RS1 next either takes value from frf or keeps value
|
279 |
|
|
dp_mux2es #(64) rs1_mux(.dout (rs1_data_next[63:0]),
|
280 |
|
|
.in0 (rs1_data[63:0]),
|
281 |
|
|
.in1 (new_frf_data[63:0]),
|
282 |
|
|
.sel (ctl_dp_new_rs1));
|
283 |
|
|
|
284 |
|
|
dff_s #(64) rs1_dff(.din (rs1_data_next[63:0]),
|
285 |
|
|
.q (rs1_data[63:0]),
|
286 |
|
|
.clk (clk), .se(se), .si(), .so());
|
287 |
|
|
assign dp_vis_rs1_data[63:0] = rs1_data[63:0];
|
288 |
|
|
|
289 |
|
|
|
290 |
|
|
/////////////////////////////////////////////////////////
|
291 |
|
|
// FSR
|
292 |
|
|
/////////////////////////////////////////////////////////
|
293 |
|
|
// FSR takes data from load
|
294 |
|
|
// fsr is set by ldfsr, ldxfsr, or any fpu operation
|
295 |
|
|
assign t0_ldfsr_data[27:0] = {ctl_dp_fcc_w2[7:2], // fcc3,2,1
|
296 |
|
|
lsu_ffu_ld_data_d1[31:30], // RND mode
|
297 |
|
|
//2'b0, // rsvd
|
298 |
|
|
lsu_ffu_ld_data_d1[27:23], // TEM
|
299 |
|
|
//6'b0, // NS, rsvd, ver
|
300 |
|
|
t0_fsr[14:12], // ftt
|
301 |
|
|
//2'b0, // qne, rsvd
|
302 |
|
|
lsu_ffu_ld_data_d1[11:0]}; // fcc0, aexc, cexc
|
303 |
|
|
|
304 |
|
|
assign t0_fpufsr_data[27:0] = {ctl_dp_fcc_w2[7:2],
|
305 |
|
|
t0_fsr[21:20], // rnd
|
306 |
|
|
t0_fsr[19:15], // TEM
|
307 |
|
|
ctl_dp_ftt_w2[2:0], // ftt
|
308 |
|
|
ctl_dp_fcc_w2[1:0],
|
309 |
|
|
ctl_dp_exc_w2[9:0]};
|
310 |
|
|
|
311 |
|
|
assign t1_ldfsr_data[27:0] = {ctl_dp_fcc_w2[7:2], // fcc3,2,1
|
312 |
|
|
lsu_ffu_ld_data_d1[31:30], // RND mode
|
313 |
|
|
//2'b0, // rsvd
|
314 |
|
|
lsu_ffu_ld_data_d1[27:23], // TEM
|
315 |
|
|
//6'b0, // NS, rsvd, ver
|
316 |
|
|
t1_fsr[14:12], // ftt
|
317 |
|
|
//2'b0, // qne, rsvd
|
318 |
|
|
lsu_ffu_ld_data_d1[11:0]}; // fcc0, aexc, cexc
|
319 |
|
|
|
320 |
|
|
assign t1_fpufsr_data[27:0] = {ctl_dp_fcc_w2[7:2],
|
321 |
|
|
t1_fsr[21:20], // rnd
|
322 |
|
|
t1_fsr[19:15], // TEM
|
323 |
|
|
ctl_dp_ftt_w2[2:0], // ftt
|
324 |
|
|
ctl_dp_fcc_w2[1:0],
|
325 |
|
|
ctl_dp_exc_w2[9:0]};
|
326 |
|
|
|
327 |
|
|
assign t2_ldfsr_data[27:0] = {ctl_dp_fcc_w2[7:2], // fcc3,2,1
|
328 |
|
|
lsu_ffu_ld_data_d1[31:30], // RND mode
|
329 |
|
|
//2'b0, // rsvd
|
330 |
|
|
lsu_ffu_ld_data_d1[27:23], // TEM
|
331 |
|
|
//6'b0, // NS, rsvd, ver
|
332 |
|
|
t2_fsr[14:12], // ftt
|
333 |
|
|
//2'b0, // qne, rsvd
|
334 |
|
|
lsu_ffu_ld_data_d1[11:0]}; // fcc0, aexc, cexc
|
335 |
|
|
|
336 |
|
|
assign t2_fpufsr_data[27:0] = {ctl_dp_fcc_w2[7:2],
|
337 |
|
|
t2_fsr[21:20], // rnd
|
338 |
|
|
t2_fsr[19:15], // TEM
|
339 |
|
|
ctl_dp_ftt_w2[2:0], // ftt
|
340 |
|
|
ctl_dp_fcc_w2[1:0],
|
341 |
|
|
ctl_dp_exc_w2[9:0]};
|
342 |
|
|
|
343 |
|
|
assign t3_ldfsr_data[27:0] = {ctl_dp_fcc_w2[7:2], // fcc3,2,1
|
344 |
|
|
lsu_ffu_ld_data_d1[31:30], // RND mode
|
345 |
|
|
//2'b0, // rsvd
|
346 |
|
|
lsu_ffu_ld_data_d1[27:23], // TEM
|
347 |
|
|
//6'b0, // NS, rsvd, ver
|
348 |
|
|
t3_fsr[14:12], // ftt
|
349 |
|
|
//2'b0, // qne, rsvd
|
350 |
|
|
lsu_ffu_ld_data_d1[11:0]}; // fcc0, aexc, cexc
|
351 |
|
|
|
352 |
|
|
assign t3_fpufsr_data[27:0] = {ctl_dp_fcc_w2[7:2],
|
353 |
|
|
t3_fsr[21:20], // rnd
|
354 |
|
|
t3_fsr[19:15], // TEM
|
355 |
|
|
ctl_dp_ftt_w2[2:0], // ftt
|
356 |
|
|
ctl_dp_fcc_w2[1:0],
|
357 |
|
|
ctl_dp_exc_w2[9:0]};
|
358 |
|
|
|
359 |
|
|
`ifdef FPGA_SYN_1THREAD
|
360 |
|
|
|
361 |
|
|
mux3ds #28 fsr0_mux(.dout (t0_fsr_nxt[27:0]),
|
362 |
|
|
.in0 (t0_fsr[27:0]),
|
363 |
|
|
.in1 (t0_ldfsr_data[27:0]),
|
364 |
|
|
.in2 (t0_fpufsr_data[27:0]),
|
365 |
|
|
.sel0 (ctl_dp_fsr_sel_old[0]),
|
366 |
|
|
.sel1 (ctl_dp_fsr_sel_ld[0]),
|
367 |
|
|
.sel2 (ctl_dp_fsr_sel_fpu[0]));
|
368 |
|
|
// FSR registers
|
369 |
|
|
// need only 28 flops for FSR since rest are always 0
|
370 |
|
|
dffr_s #28 fsr0_reg(.din (t0_fsr_nxt[27:0]),
|
371 |
|
|
.q (t0_fsr[27:0]),
|
372 |
|
|
.rst(reset),
|
373 |
|
|
.clk (clk), .se(se), .si(), .so());
|
374 |
|
|
assign current_fsr[27:0] = t0_fsr[27:0];
|
375 |
|
|
|
376 |
|
|
`else
|
377 |
|
|
|
378 |
|
|
mux3ds #28 fsr0_mux(.dout (t0_fsr_nxt[27:0]),
|
379 |
|
|
.in0 (t0_fsr[27:0]),
|
380 |
|
|
.in1 (t0_ldfsr_data[27:0]),
|
381 |
|
|
.in2 (t0_fpufsr_data[27:0]),
|
382 |
|
|
.sel0 (ctl_dp_fsr_sel_old[0]),
|
383 |
|
|
.sel1 (ctl_dp_fsr_sel_ld[0]),
|
384 |
|
|
.sel2 (ctl_dp_fsr_sel_fpu[0]));
|
385 |
|
|
mux3ds #28 fsr1_mux(.dout (t1_fsr_nxt[27:0]),
|
386 |
|
|
.in0 (t1_fsr[27:0]),
|
387 |
|
|
.in1 (t1_ldfsr_data[27:0]),
|
388 |
|
|
.in2 (t1_fpufsr_data[27:0]),
|
389 |
|
|
.sel0 (ctl_dp_fsr_sel_old[1]),
|
390 |
|
|
.sel1 (ctl_dp_fsr_sel_ld[1]),
|
391 |
|
|
.sel2 (ctl_dp_fsr_sel_fpu[1]));
|
392 |
|
|
mux3ds #28 fsr2_mux(.dout (t2_fsr_nxt[27:0]),
|
393 |
|
|
.in0 (t2_fsr[27:0]),
|
394 |
|
|
.in1 (t2_ldfsr_data[27:0]),
|
395 |
|
|
.in2 (t2_fpufsr_data[27:0]),
|
396 |
|
|
.sel0 (ctl_dp_fsr_sel_old[2]),
|
397 |
|
|
.sel1 (ctl_dp_fsr_sel_ld[2]),
|
398 |
|
|
.sel2 (ctl_dp_fsr_sel_fpu[2]));
|
399 |
|
|
mux3ds #28 fsr3_mux(.dout (t3_fsr_nxt[27:0]),
|
400 |
|
|
.in0 (t3_fsr[27:0]),
|
401 |
|
|
.in1 (t3_ldfsr_data[27:0]),
|
402 |
|
|
.in2 (t3_fpufsr_data[27:0]),
|
403 |
|
|
.sel0 (ctl_dp_fsr_sel_old[3]),
|
404 |
|
|
.sel1 (ctl_dp_fsr_sel_ld[3]),
|
405 |
|
|
.sel2 (ctl_dp_fsr_sel_fpu[3]));
|
406 |
|
|
|
407 |
|
|
// FSR registers
|
408 |
|
|
// need only 28 flops for FSR since rest are always 0
|
409 |
|
|
dffr_s #28 fsr0_reg(.din (t0_fsr_nxt[27:0]),
|
410 |
|
|
.q (t0_fsr[27:0]),
|
411 |
|
|
.rst(reset),
|
412 |
|
|
.clk (clk), .se(se), .si(), .so());
|
413 |
|
|
dffr_s #28 fsr1_reg(.din (t1_fsr_nxt[27:0]),
|
414 |
|
|
.q (t1_fsr[27:0]),
|
415 |
|
|
.rst(reset),
|
416 |
|
|
.clk (clk), .se(se), .si(), .so());
|
417 |
|
|
dffr_s #28 fsr2_reg(.din (t2_fsr_nxt[27:0]),
|
418 |
|
|
.q (t2_fsr[27:0]),
|
419 |
|
|
.rst(reset),
|
420 |
|
|
.clk (clk), .se(se), .si(), .so());
|
421 |
|
|
dffr_s #28 fsr3_reg(.din (t3_fsr_nxt[27:0]),
|
422 |
|
|
.q (t3_fsr[27:0]),
|
423 |
|
|
.rst(reset),
|
424 |
|
|
.clk (clk), .se(se), .si(), .so());
|
425 |
|
|
|
426 |
|
|
// Current FSR
|
427 |
|
|
mux4ds #28 curr_fsr_mux(.dout (current_fsr[27:0]),
|
428 |
|
|
.in0 (t0_fsr[27:0]),
|
429 |
|
|
.in1 (t1_fsr[27:0]),
|
430 |
|
|
.in2 (t2_fsr[27:0]),
|
431 |
|
|
.in3 (t3_fsr[27:0]),
|
432 |
|
|
.sel0 (ctl_dp_fp_thr[0]),
|
433 |
|
|
.sel1 (ctl_dp_fp_thr[1]),
|
434 |
|
|
.sel2 (ctl_dp_fp_thr[2]),
|
435 |
|
|
.sel3 (ctl_dp_fp_thr[3]));
|
436 |
|
|
`endif // !`ifdef FPGA_SYN_1THREAD
|
437 |
|
|
|
438 |
|
|
assign dp_ctl_fsr_fcc = {current_fsr[27:22], current_fsr[11:10]};
|
439 |
|
|
assign dp_ctl_fsr_rnd = current_fsr[21:20];
|
440 |
|
|
assign dp_ctl_fsr_tem = current_fsr[19:15];
|
441 |
|
|
assign dp_ctl_fsr_aexc = current_fsr[9:5];
|
442 |
|
|
assign dp_ctl_fsr_cexc = current_fsr[4:0];
|
443 |
|
|
|
444 |
|
|
////////////////////////////////////////////////////////////
|
445 |
|
|
// ECC generation and correction
|
446 |
|
|
////////////////////////////////////////////////////////////
|
447 |
|
|
dp_mux2es #(64) ecc_mux(.dout(ecc_data_in[63:0]),
|
448 |
|
|
.in0(rs2_rd_data[63:0]),
|
449 |
|
|
.in1({frf_dp_data[70:39], frf_dp_data[31:0]}),
|
450 |
|
|
.sel(ctl_dp_ecc_sel_frf));
|
451 |
|
|
|
452 |
|
|
assign synd_in_low[6:0] = {7{ctl_dp_ecc_sel_frf}} & frf_dp_data[38:32];
|
453 |
|
|
assign synd_in_h[6:0] = {7{ctl_dp_ecc_sel_frf}} & frf_dp_data[77:71];
|
454 |
|
|
|
455 |
|
|
zzecc_sctag_ecc39 ecccor_low(.din(ecc_data_in[31:0]),
|
456 |
|
|
.parity(synd_in_low[6:0]),
|
457 |
|
|
.dout(corr_data_next[31:0]),
|
458 |
|
|
.pflag(dp_ctl_synd_out_low[6]),
|
459 |
|
|
.cflag(dp_ctl_synd_out_low[5:0]));
|
460 |
|
|
|
461 |
|
|
zzecc_sctag_ecc39 ecccor_high(.din(ecc_data_in[63:32]),
|
462 |
|
|
.parity(synd_in_h[6:0]),
|
463 |
|
|
.dout(corr_data_next[63:32]),
|
464 |
|
|
.pflag(dp_ctl_synd_out_high[6]),
|
465 |
|
|
.cflag(dp_ctl_synd_out_high[5:0]));
|
466 |
|
|
|
467 |
|
|
|
468 |
|
|
dff_s #(64) ecc_corr_data(.din(corr_data_next[63:0]), .q(corr_data[63:0]),
|
469 |
|
|
.clk(clk), .se(se), .si(), .so());
|
470 |
|
|
|
471 |
|
|
|
472 |
|
|
////////////////////////////////////////////////
|
473 |
|
|
// GSR Storage
|
474 |
|
|
////////////////////////////////////////////////
|
475 |
|
|
// GSR registers
|
476 |
|
|
// need only 37 flops for GSR since rest are always 0
|
477 |
|
|
// and the align and rnd fields are in the ctl block
|
478 |
|
|
`ifdef FPGA_SYN_1THREAD
|
479 |
|
|
dffr_s #37 gsr0_reg(.din (t0_gsr_nxt[36:0]),
|
480 |
|
|
.q (t0_gsr[36:0]),
|
481 |
|
|
.rst(reset),
|
482 |
|
|
.clk (clk), .se(se), .si(), .so());
|
483 |
|
|
assign t0_gsr_nxt[36:0] = t0_gsr[36:0];
|
484 |
|
|
assign gsr_e[36:0] = t0_gsr[36:0];
|
485 |
|
|
|
486 |
|
|
`else
|
487 |
|
|
|
488 |
|
|
dffr_s #37 gsr0_reg(.din (t0_gsr_nxt[36:0]),
|
489 |
|
|
.q (t0_gsr[36:0]),
|
490 |
|
|
.rst(reset),
|
491 |
|
|
.clk (clk), .se(se), .si(), .so());
|
492 |
|
|
dffr_s #37 gsr1_reg(.din (t1_gsr_nxt[36:0]),
|
493 |
|
|
.q (t1_gsr[36:0]),
|
494 |
|
|
.rst(reset),
|
495 |
|
|
.clk (clk), .se(se), .si(), .so());
|
496 |
|
|
dffr_s #37 gsr2_reg(.din (t2_gsr_nxt[36:0]),
|
497 |
|
|
.q (t2_gsr[36:0]),
|
498 |
|
|
.rst(reset),
|
499 |
|
|
.clk (clk), .se(se), .si(), .so());
|
500 |
|
|
dffr_s #37 gsr3_reg(.din (t3_gsr_nxt[36:0]),
|
501 |
|
|
.q (t3_gsr[36:0]),
|
502 |
|
|
.rst(reset),
|
503 |
|
|
.clk (clk), .se(se), .si(), .so());
|
504 |
|
|
|
505 |
|
|
dp_mux2es #(37) gsr0_mux(.dout(t0_gsr_nxt[36:0]),
|
506 |
|
|
.in0(t0_gsr[36:0]),
|
507 |
|
|
.in1(ctl_dp_wsr_data_w2[36:0]),
|
508 |
|
|
.sel(ctl_dp_gsr_wsr_w2[0]));
|
509 |
|
|
dp_mux2es #(37) gsr1_mux(.dout(t1_gsr_nxt[36:0]),
|
510 |
|
|
.in0(t1_gsr[36:0]),
|
511 |
|
|
.in1(ctl_dp_wsr_data_w2[36:0]),
|
512 |
|
|
.sel(ctl_dp_gsr_wsr_w2[1]));
|
513 |
|
|
dp_mux2es #(37) gsr2_mux(.dout(t2_gsr_nxt[36:0]),
|
514 |
|
|
.in0(t2_gsr[36:0]),
|
515 |
|
|
.in1(ctl_dp_wsr_data_w2[36:0]),
|
516 |
|
|
.sel(ctl_dp_gsr_wsr_w2[2]));
|
517 |
|
|
dp_mux2es #(37) gsr3_mux(.dout(t3_gsr_nxt[36:0]),
|
518 |
|
|
.in0(t3_gsr[36:0]),
|
519 |
|
|
.in1(ctl_dp_wsr_data_w2[36:0]),
|
520 |
|
|
.sel(ctl_dp_gsr_wsr_w2[3]));
|
521 |
|
|
|
522 |
|
|
|
523 |
|
|
// GSR_E
|
524 |
|
|
mux4ds #37 curr_gsr_mux(.dout (gsr_e[36:0]),
|
525 |
|
|
.in0 (t0_gsr[36:0]),
|
526 |
|
|
.in1 (t1_gsr[36:0]),
|
527 |
|
|
.in2 (t2_gsr[36:0]),
|
528 |
|
|
.in3 (t3_gsr[36:0]),
|
529 |
|
|
.sel0 (ctl_dp_thr_e[0]),
|
530 |
|
|
.sel1 (ctl_dp_thr_e[1]),
|
531 |
|
|
.sel2 (ctl_dp_thr_e[2]),
|
532 |
|
|
.sel3 (ctl_dp_thr_e[3]));
|
533 |
|
|
`endif // !`ifdef FPGA_SYN_1THREAD
|
534 |
|
|
|
535 |
|
|
assign dp_ctl_gsr_scale_e[4:0] = gsr_e[4:0];
|
536 |
|
|
assign dp_ctl_gsr_mask_e[31:0] = gsr_e[36:5];
|
537 |
|
|
|
538 |
|
|
|
539 |
|
|
|
540 |
|
|
|
541 |
|
|
endmodule // sparc_ffu_dp
|