0
|
1 ;; Pipeline model for ST Microelectronics Loongson-2E/2F cores.
|
|
2
|
|
3 ;; Copyright (C) 2008 Free Software Foundation, Inc.
|
|
4 ;; Contributed by CodeSourcery.
|
|
5 ;;
|
|
6 ;; GCC is free software; you can redistribute it and/or modify
|
|
7 ;; it under the terms of the GNU General Public License as published by
|
|
8 ;; the Free Software Foundation; either version 3, or (at your option)
|
|
9 ;; any later version.
|
|
10 ;;
|
|
11 ;; GCC is distributed in the hope that it will be useful,
|
|
12 ;; but WITHOUT ANY WARRANTY; without even the implied warranty of
|
|
13 ;; MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
|
14 ;; GNU General Public License for more details.
|
|
15 ;;
|
|
16 ;; You should have received a copy of the GNU General Public License
|
|
17 ;; along with GCC; see the file COPYING3. If not see
|
|
18 ;; <http://www.gnu.org/licenses/>.
|
|
19
|
|
20 ;; Automaton for integer instructions.
|
|
21 (define_automaton "ls2_alu")
|
|
22
|
|
23 ;; ALU1 and ALU2.
|
|
24 ;; We need to query these units to adjust round-robin counter.
|
|
25 (define_query_cpu_unit "ls2_alu1_core,ls2_alu2_core" "ls2_alu")
|
|
26
|
|
27 ;; Pseudo units to help modeling of ALU1/2 round-robin dispatch strategy.
|
|
28 (define_cpu_unit "ls2_alu1_turn,ls2_alu2_turn" "ls2_alu")
|
|
29
|
|
30 ;; Pseudo units to enable/disable ls2_alu[12]_turn units.
|
|
31 ;; ls2_alu[12]_turn unit can be subscribed only after ls2_alu[12]_turn_enabled
|
|
32 ;; unit is subscribed.
|
|
33 (define_cpu_unit "ls2_alu1_turn_enabled,ls2_alu2_turn_enabled" "ls2_alu")
|
|
34 (presence_set "ls2_alu1_turn" "ls2_alu1_turn_enabled")
|
|
35 (presence_set "ls2_alu2_turn" "ls2_alu2_turn_enabled")
|
|
36
|
|
37 ;; Reservations for ALU1 (ALU2) instructions.
|
|
38 ;; Instruction goes to ALU1 (ALU2) and makes next ALU1/2 instruction to
|
|
39 ;; be dispatched to ALU2 (ALU1).
|
|
40 (define_reservation "ls2_alu1"
|
|
41 "(ls2_alu1_core+ls2_alu2_turn_enabled)|ls2_alu1_core")
|
|
42 (define_reservation "ls2_alu2"
|
|
43 "(ls2_alu2_core+ls2_alu1_turn_enabled)|ls2_alu2_core")
|
|
44
|
|
45 ;; Reservation for ALU1/2 instructions.
|
|
46 ;; Instruction will go to ALU1 iff ls2_alu1_turn_enabled is subscribed and
|
|
47 ;; switch the turn to ALU2 by subscribing ls2_alu2_turn_enabled.
|
|
48 ;; Or to ALU2 otherwise.
|
|
49 (define_reservation "ls2_alu"
|
|
50 "(ls2_alu1_core+ls2_alu1_turn+ls2_alu2_turn_enabled)
|
|
51 |(ls2_alu1_core+ls2_alu1_turn)
|
|
52 |(ls2_alu2_core+ls2_alu2_turn+ls2_alu1_turn_enabled)
|
|
53 |(ls2_alu2_core+ls2_alu2_turn)")
|
|
54
|
|
55 ;; Automaton for floating-point instructions.
|
|
56 (define_automaton "ls2_falu")
|
|
57
|
|
58 ;; FALU1 and FALU2.
|
|
59 ;; We need to query these units to adjust round-robin counter.
|
|
60 (define_query_cpu_unit "ls2_falu1_core,ls2_falu2_core" "ls2_falu")
|
|
61
|
|
62 ;; Pseudo units to help modeling of FALU1/2 round-robin dispatch strategy.
|
|
63 (define_cpu_unit "ls2_falu1_turn,ls2_falu2_turn" "ls2_falu")
|
|
64
|
|
65 ;; Pseudo units to enable/disable ls2_falu[12]_turn units.
|
|
66 ;; ls2_falu[12]_turn unit can be subscribed only after
|
|
67 ;; ls2_falu[12]_turn_enabled unit is subscribed.
|
|
68 (define_cpu_unit "ls2_falu1_turn_enabled,ls2_falu2_turn_enabled" "ls2_falu")
|
|
69 (presence_set "ls2_falu1_turn" "ls2_falu1_turn_enabled")
|
|
70 (presence_set "ls2_falu2_turn" "ls2_falu2_turn_enabled")
|
|
71
|
|
72 ;; Reservations for FALU1 (FALU2) instructions.
|
|
73 ;; Instruction goes to FALU1 (FALU2) and makes next FALU1/2 instruction to
|
|
74 ;; be dispatched to FALU2 (FALU1).
|
|
75 (define_reservation "ls2_falu1"
|
|
76 "(ls2_falu1_core+ls2_falu2_turn_enabled)|ls2_falu1_core")
|
|
77 (define_reservation "ls2_falu2"
|
|
78 "(ls2_falu2_core+ls2_falu1_turn_enabled)|ls2_falu2_core")
|
|
79
|
|
80 ;; Reservation for FALU1/2 instructions.
|
|
81 ;; Instruction will go to FALU1 iff ls2_falu1_turn_enabled is subscribed and
|
|
82 ;; switch the turn to FALU2 by subscribing ls2_falu2_turn_enabled.
|
|
83 ;; Or to FALU2 otherwise.
|
|
84 (define_reservation "ls2_falu"
|
|
85 "(ls2_falu1+ls2_falu1_turn+ls2_falu2_turn_enabled)
|
|
86 |(ls2_falu1+ls2_falu1_turn)
|
|
87 |(ls2_falu2+ls2_falu2_turn+ls2_falu1_turn_enabled)
|
|
88 |(ls2_falu2+ls2_falu2_turn)")
|
|
89
|
|
90 ;; The following 4 instructions each subscribe one of
|
|
91 ;; ls2_[f]alu{1,2}_turn_enabled units according to this attribute.
|
|
92 ;; These instructions are used in mips.c: sched_ls2_dfa_post_advance_cycle.
|
|
93
|
|
94 (define_attr "ls2_turn_type" "alu1,alu2,falu1,falu2,unknown"
|
|
95 (const_string "unknown"))
|
|
96
|
|
97 ;; Subscribe ls2_alu1_turn_enabled.
|
|
98 (define_insn "ls2_alu1_turn_enabled_insn"
|
|
99 [(unspec [(const_int 0)] UNSPEC_LOONGSON_ALU1_TURN_ENABLED_INSN)]
|
|
100 "TUNE_LOONGSON_2EF"
|
|
101 { gcc_unreachable (); }
|
|
102 [(set_attr "ls2_turn_type" "alu1")])
|
|
103
|
|
104 (define_insn_reservation "ls2_alu1_turn_enabled" 0
|
|
105 (eq_attr "ls2_turn_type" "alu1")
|
|
106 "ls2_alu1_turn_enabled")
|
|
107
|
|
108 ;; Subscribe ls2_alu2_turn_enabled.
|
|
109 (define_insn "ls2_alu2_turn_enabled_insn"
|
|
110 [(unspec [(const_int 0)] UNSPEC_LOONGSON_ALU2_TURN_ENABLED_INSN)]
|
|
111 "TUNE_LOONGSON_2EF"
|
|
112 { gcc_unreachable (); }
|
|
113 [(set_attr "ls2_turn_type" "alu2")])
|
|
114
|
|
115 (define_insn_reservation "ls2_alu2_turn_enabled" 0
|
|
116 (eq_attr "ls2_turn_type" "alu2")
|
|
117 "ls2_alu2_turn_enabled")
|
|
118
|
|
119 ;; Subscribe ls2_falu1_turn_enabled.
|
|
120 (define_insn "ls2_falu1_turn_enabled_insn"
|
|
121 [(unspec [(const_int 0)] UNSPEC_LOONGSON_FALU1_TURN_ENABLED_INSN)]
|
|
122 "TUNE_LOONGSON_2EF"
|
|
123 { gcc_unreachable (); }
|
|
124 [(set_attr "ls2_turn_type" "falu1")])
|
|
125
|
|
126 (define_insn_reservation "ls2_falu1_turn_enabled" 0
|
|
127 (eq_attr "ls2_turn_type" "falu1")
|
|
128 "ls2_falu1_turn_enabled")
|
|
129
|
|
130 ;; Subscribe ls2_falu2_turn_enabled.
|
|
131 (define_insn "ls2_falu2_turn_enabled_insn"
|
|
132 [(unspec [(const_int 0)] UNSPEC_LOONGSON_FALU2_TURN_ENABLED_INSN)]
|
|
133 "TUNE_LOONGSON_2EF"
|
|
134 { gcc_unreachable (); }
|
|
135 [(set_attr "ls2_turn_type" "falu2")])
|
|
136
|
|
137 (define_insn_reservation "ls2_falu2_turn_enabled" 0
|
|
138 (eq_attr "ls2_turn_type" "falu2")
|
|
139 "ls2_falu2_turn_enabled")
|
|
140
|
|
141 ;; Automaton for memory operations.
|
|
142 (define_automaton "ls2_mem")
|
|
143
|
|
144 ;; Memory unit.
|
|
145 (define_query_cpu_unit "ls2_mem" "ls2_mem")
|
|
146
|
|
147 ;; Reservation for integer instructions.
|
|
148 (define_insn_reservation "ls2_alu" 2
|
|
149 (and (eq_attr "cpu" "loongson_2e,loongson_2f")
|
|
150 (eq_attr "type" "arith,condmove,const,logical,mfhilo,move,
|
|
151 mthilo,nop,shift,signext,slt"))
|
|
152 "ls2_alu")
|
|
153
|
|
154 ;; Reservation for branch instructions.
|
|
155 (define_insn_reservation "ls2_branch" 2
|
|
156 (and (eq_attr "cpu" "loongson_2e,loongson_2f")
|
|
157 (eq_attr "type" "branch,jump,call,trap"))
|
|
158 "ls2_alu1")
|
|
159
|
|
160 ;; Reservation for integer multiplication instructions.
|
|
161 (define_insn_reservation "ls2_imult" 5
|
|
162 (and (eq_attr "cpu" "loongson_2e,loongson_2f")
|
|
163 (eq_attr "type" "imul,imul3nc"))
|
|
164 "ls2_alu2,ls2_alu2_core")
|
|
165
|
|
166 ;; Reservation for integer division / remainder instructions.
|
|
167 ;; These instructions use the SRT algorithm and hence take 2-38 cycles.
|
|
168 (define_insn_reservation "ls2_idiv" 20
|
|
169 (and (eq_attr "cpu" "loongson_2e,loongson_2f")
|
|
170 (eq_attr "type" "idiv,idiv3"))
|
|
171 "ls2_alu2,ls2_alu2_core*18")
|
|
172
|
|
173 ;; Reservation for memory load instructions.
|
|
174 (define_insn_reservation "ls2_load" 5
|
|
175 (and (eq_attr "cpu" "loongson_2e,loongson_2f")
|
|
176 (eq_attr "type" "load,fpload,mfc,mtc"))
|
|
177 "ls2_mem")
|
|
178
|
|
179 (define_insn_reservation "ls2_prefetch" 0
|
|
180 (and (eq_attr "cpu" "loongson_2e,loongson_2f")
|
|
181 (eq_attr "type" "prefetch,prefetchx"))
|
|
182 "ls2_mem")
|
|
183
|
|
184 ;; Reservation for memory store instructions.
|
|
185 ;; With stores we assume they don't alias with dependent loads.
|
|
186 ;; Therefore we set the latency to zero.
|
|
187 (define_insn_reservation "ls2_store" 0
|
|
188 (and (eq_attr "cpu" "loongson_2e,loongson_2f")
|
|
189 (eq_attr "type" "store,fpstore"))
|
|
190 "ls2_mem")
|
|
191
|
|
192 ;; Reservation for floating-point instructions of latency 3.
|
|
193 (define_insn_reservation "ls2_fp3" 3
|
|
194 (and (eq_attr "cpu" "loongson_2e,loongson_2f")
|
|
195 (eq_attr "type" "fabs,fneg,fcmp,fmove"))
|
|
196 "ls2_falu1")
|
|
197
|
|
198 ;; Reservation for floating-point instructions of latency 5.
|
|
199 (define_insn_reservation "ls2_fp5" 5
|
|
200 (and (eq_attr "cpu" "loongson_2e,loongson_2f")
|
|
201 (eq_attr "type" "fcvt"))
|
|
202 "ls2_falu1")
|
|
203
|
|
204 ;; Reservation for floating-point instructions that can go
|
|
205 ;; to either of FALU1/2 units.
|
|
206 (define_insn_reservation "ls2_falu" 7
|
|
207 (and (eq_attr "cpu" "loongson_2e,loongson_2f")
|
|
208 (eq_attr "type" "fadd,fmul,fmadd"))
|
|
209 "ls2_falu")
|
|
210
|
|
211 ;; Reservation for floating-point division / remainder instructions.
|
|
212 ;; These instructions use the SRT algorithm and hence take a variable amount
|
|
213 ;; of cycles:
|
|
214 ;; div.s takes 5-11 cycles
|
|
215 ;; div.d takes 5-18 cycles
|
|
216 (define_insn_reservation "ls2_fdiv" 9
|
|
217 (and (eq_attr "cpu" "loongson_2e,loongson_2f")
|
|
218 (eq_attr "type" "fdiv"))
|
|
219 "ls2_falu2,ls2_falu2_core*7")
|
|
220
|
|
221 ;; Reservation for floating-point sqrt instructions.
|
|
222 ;; These instructions use the SRT algorithm and hence take a variable amount
|
|
223 ;; of cycles:
|
|
224 ;; sqrt.s takes 5-17 cycles
|
|
225 ;; sqrt.d takes 5-32 cycles
|
|
226 (define_insn_reservation "ls2_fsqrt" 15
|
|
227 (and (eq_attr "cpu" "loongson_2e,loongson_2f")
|
|
228 (eq_attr "type" "fsqrt"))
|
|
229 "ls2_falu2,ls2_falu2_core*13")
|
|
230
|
|
231 ;; Two consecutive ALU instructions.
|
|
232 (define_insn_reservation "ls2_multi" 4
|
|
233 (and (eq_attr "cpu" "loongson_2e,loongson_2f")
|
|
234 (eq_attr "type" "multi"))
|
|
235 "(ls2_alu1,ls2_alu2_core)|(ls2_alu2,ls2_alu1_core)")
|
|
236
|
|
237 ;; Reservation for everything else. Normally, this reservation
|
|
238 ;; will only be used to handle cases like compiling for non-loongson
|
|
239 ;; CPUs with -mtune=loongson2?.
|
|
240 ;;
|
|
241 ;; This reservation depends upon the fact that DFA will check
|
|
242 ;; reservations in the same order as they appear in the file.
|
|
243 (define_insn_reservation "ls2_unknown" 1
|
|
244 (eq_attr "cpu" "loongson_2e,loongson_2f")
|
|
245 "ls2_alu1_core+ls2_alu2_core+ls2_falu1_core+ls2_falu2_core+ls2_mem")
|