1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
|
;; DFA-based pipeline descriptions for MIPS32 5K processor family
;; Contributed by David Ung (davidu@mips.com)
;; and Nigel Stephens (nigel@mips.com)
;;
;; References:
;; "MIPS64 5K Processor Core Family Software User's Manual,
;; Doc no: MD00012, Rev 2.09, Jan 28, 2005."
;;
;; 5Kc - Single issue with no floating point unit.
;; 5kf - Separate floating point pipe which can dual-issue with the
;; integer pipe.
;;
;; Copyright (C) 2005-2016 Free Software Foundation, Inc.
;;
;; This file is part of GCC.
;;
;; GCC is free software; you can redistribute it and/or modify it
;; under the terms of the GNU General Public License as published
;; by the Free Software Foundation; either version 3, or (at your
;; option) any later version.
;; GCC is distributed in the hope that it will be useful, but WITHOUT
;; ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
;; or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public
;; License for more details.
;; You should have received a copy of the GNU General Public License
;; along with GCC; see the file COPYING3. If not see
;; <http://www.gnu.org/licenses/>.
(define_automaton "r5k_cpu, r5k_mdu, r5k_fpu")
;; Integer execution unit.
(define_cpu_unit "r5k_ixu_arith" "r5k_cpu")
(define_cpu_unit "r5k_ixu_mpydiv" "r5k_mdu")
(define_cpu_unit "r5kf_fpu_arith" "r5k_fpu")
(define_insn_reservation "r5k_int_load" 2
(and (eq_attr "cpu" "5kc,5kf")
(eq_attr "type" "load"))
"r5k_ixu_arith")
(define_insn_reservation "r5k_int_prefetch" 1
(and (eq_attr "cpu" "5kc,5kf")
(eq_attr "type" "prefetch,prefetchx"))
"r5k_ixu_arith")
(define_insn_reservation "r5k_int_store" 1
(and (eq_attr "cpu" "5kc,5kf")
(eq_attr "type" "store"))
"r5k_ixu_arith")
;; Divides
(define_insn_reservation "r5k_int_divsi" 34
(and (eq_attr "cpu" "5kc,5kf")
(and (eq_attr "type" "idiv")
(eq_attr "mode" "!DI")))
"r5k_ixu_arith+(r5k_ixu_mpydiv*34)")
(define_insn_reservation "r5k_int_divdi" 66
(and (eq_attr "cpu" "5kc,5kf")
(and (eq_attr "type" "idiv")
(eq_attr "mode" "DI")))
"r5k_ixu_arith+(r5k_ixu_mpydiv*66)")
;; 32x32 multiply
;; 32x16 is faster, but there's no way to detect this
(define_insn_reservation "r5k_int_mult" 2
(and (eq_attr "cpu" "5kc,5kf")
(and (eq_attr "type" "imul,imadd")
(eq_attr "mode" "SI")))
"r5k_ixu_arith+(r5k_ixu_mpydiv*2)")
;; 64x64 multiply
(define_insn_reservation "r5k_int_mult_64" 9
(and (eq_attr "cpu" "5kc,5kf")
(and (eq_attr "type" "imul,imadd")
(eq_attr "mode" "DI")))
"r5k_ixu_arith+(r5k_ixu_mpydiv*2)")
;; 3 operand MUL 32x32
(define_insn_reservation "r5k_int_mul" 4
(and (eq_attr "cpu" "5kc,5kf")
(and (eq_attr "type" "imul3")
(eq_attr "mode" "SI")))
"r5k_ixu_arith+(r5k_ixu_mpydiv*2)")
;; Move to HI/LO -> MADD/MSUB,MFHI/MFLO has a 1 cycle latency.
(define_insn_reservation "r5k_int_mthilo" 1
(and (eq_attr "cpu" "5kc,5kf")
(eq_attr "type" "mthi,mtlo"))
"r5k_ixu_arith+r5k_ixu_mpydiv")
;; Move from HI/LO -> integer operation has a 2 cycle latency.
(define_insn_reservation "r5k_int_mfhilo" 2
(and (eq_attr "cpu" "5kc,5kf")
(eq_attr "type" "mfhi,mflo"))
"r5k_ixu_arith+r5k_ixu_mpydiv")
;; All other integer insns.
(define_insn_reservation "r5k_int_alu" 1
(and (eq_attr "cpu" "5kc,5kf")
(eq_attr "type" "arith,condmove,const,logical,move,nop,shift,signext,slt"))
"r5k_ixu_arith")
(define_insn_reservation "r5k_int_branch" 1
(and (eq_attr "cpu" "5kc,5kf")
(eq_attr "type" "branch"))
"r5k_ixu_arith")
;; JR/JALR always cause one pipeline bubble because of interlock.
(define_insn_reservation "r5k_int_jump" 2
(and (eq_attr "cpu" "5kc,5kf")
(eq_attr "type" "jump,call"))
"r5k_ixu_arith")
;; Any -> JR/JALR (without dependency) : 1 clock issue delay
;; Any -> JR/JALR (with dependency) : 2 clock issue delay
;; load -> JR/JALR (with dependency) : 3 clock issue delay
;; mfhilo -> JR/JALR (with dependency) : 3 clock issue delay
;; mul -> JR/JALR (with dependency) : 3 clock issue delay
(define_bypass 2 "r5k_int_alu" "r5k_int_jump")
(define_bypass 3 "r5k_int_load" "r5k_int_jump")
(define_bypass 3 "r5k_int_mfhilo" "r5k_int_jump")
(define_bypass 3 "r5k_int_mul" "r5k_int_jump")
;; Unknown or multi - single issue
(define_insn_reservation "r5k_int_unknown" 1
(and (eq_attr "cpu" "5kc,5kf")
(eq_attr "type" "unknown,multi,atomic,syncloop"))
"r5k_ixu_arith+r5k_ixu_mpydiv")
;; Floating Point Instructions
;; The 5Kf is a partial dual-issue cpu which can dual issue an integer
;; and floating-point instruction in the same cycle.
;; fadd, fabs, fneg
(define_insn_reservation "r5kf_fadd" 4
(and (eq_attr "cpu" "5kf")
(eq_attr "type" "fadd,fabs,fneg"))
"r5kf_fpu_arith")
;; fmove, fcmove
(define_insn_reservation "r5kf_fmove" 4
(and (eq_attr "cpu" "5kf")
(eq_attr "type" "fmove"))
"r5kf_fpu_arith")
;; fload
(define_insn_reservation "r5kf_fload" 3
(and (eq_attr "cpu" "5kf")
(eq_attr "type" "fpload,fpidxload"))
"r5kf_fpu_arith")
;; fstore
(define_insn_reservation "r5kf_fstore" 1
(and (eq_attr "cpu" "5kf")
(eq_attr "type" "fpstore"))
"r5kf_fpu_arith")
;; fmul, fmadd
(define_insn_reservation "r5kf_fmul_sf" 4
(and (eq_attr "cpu" "5kf")
(and (eq_attr "type" "fmul,fmadd")
(eq_attr "mode" "SF")))
"r5kf_fpu_arith")
(define_insn_reservation "r5kf_fmul_df" 5
(and (eq_attr "cpu" "5kf")
(and (eq_attr "type" "fmul,fmadd")
(eq_attr "mode" "DF")))
"r5kf_fpu_arith*2")
;; fdiv, fsqrt, frsqrt
(define_insn_reservation "r5kf_fdiv_sf" 17
(and (eq_attr "cpu" "5kf")
(and (eq_attr "type" "fdiv,fsqrt,frsqrt")
(eq_attr "mode" "SF")))
"r5kf_fpu_arith*14")
(define_insn_reservation "r5kf_fdiv_df" 32
(and (eq_attr "cpu" "5kf")
(and (eq_attr "type" "fdiv,fsqrt")
(eq_attr "mode" "DF")))
"r5kf_fpu_arith*29")
;; frsqrt
(define_insn_reservation "r5kf_frsqrt_df" 35
(and (eq_attr "cpu" "5kf")
(and (eq_attr "type" "frsqrt")
(eq_attr "mode" "DF")))
"r5kf_fpu_arith*31")
;; fcmp
(define_insn_reservation "r5kf_fcmp" 2
(and (eq_attr "cpu" "5kf")
(eq_attr "type" "fcmp"))
"r5kf_fpu_arith")
;; fcmp -> movf.fmt & movt.fmt bypass (dependency must be on condition)
(define_bypass 1 "r5kf_fcmp" "r5kf_fmove")
;; fcvt (cvt.d.s, cvt.[sd].[wl]
(define_insn_reservation "r5kf_fcvt_d2s" 4
(and (eq_attr "cpu" "5kf")
(and (eq_attr "type" "fcvt")
(eq_attr "cnv_mode" "I2S,I2D,S2D")))
"r5kf_fpu_arith")
;; fcvt (cvt.s.d)
(define_insn_reservation "r5kf_fcvt_s2d" 6
(and (eq_attr "cpu" "5kc")
(and (eq_attr "type" "fcvt")
(eq_attr "cnv_mode" "D2S")))
"r5kf_fpu_arith")
;; fcvt (cvt.[wl].[sd], etc)
(define_insn_reservation "r5kf_fcvt_f2i" 5
(and (eq_attr "cpu" "5kf")
(and (eq_attr "type" "fcvt")
(eq_attr "cnv_mode" "S2I,D2I")))
"r5kf_fpu_arith")
;; fxfer (mfc1, mfhc1, mtc1, mthc1) - single issue
(define_insn_reservation "r5kf_fxfer" 2
(and (eq_attr "cpu" "5kf")
(eq_attr "type" "mfc,mtc"))
"r5k_ixu_arith+r5kf_fpu_arith")
|