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
|
;; DFA scheduling description for EPIPHANY
;; Copyright (C) 2004-2015 Free Software Foundation, Inc.
;; Contributed by Embecosm on behalf of Adapteva, 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/>.
;; Two automata are defined to reduce number of states
;; which a single large automaton will have. (Factoring)
(define_automaton "inst_pipeline,fpu_pipe")
;; This unit is basically the decode unit of the processor.
;; Since epiphany is a dual issue machine, it is as if there are two
;; units so that any insn can be processed by either one
;; of the decoding unit.
(define_cpu_unit "pipe_01,pipe_02" "inst_pipeline")
;; The fixed point arithmetic unit.
(define_cpu_unit "int" "inst_pipeline")
;; The floating point unit.
(define_cpu_unit "F0" "fpu_pipe")
;; ----------------------------------------------------
;; This reservation is to simplify the dual issue description.
(define_reservation "issue" "pipe_01|pipe_02")
;; This is to express instructions that cannot be paired.
(define_reservation "d_lock" "pipe_01+pipe_02")
;; We don't model all pipeline stages; we model the issue stage
;; inasmuch as we allow only two instructions to issue simultaneously,
;; and flow instructions prevent any simultaneous issue of another instruction.
;; (This uses pipe_01 and pipe_02).
;; Double issue of 'other' insns is prevented by using the int unit in the
;; E1 stage.
;; Double issue of float instructions is prevented by using F0 in the E1 stage.
(define_insn_reservation "simple_arith" 2
(and (eq_attr "pipe_model" "epiphany")
(eq_attr "type" "move,cmove,compare,shift,misc,mul")
(eq_attr "length" "4"))
"issue,int")
; anything but fp / fp_int / v2fp has a bypass
(define_bypass 1 "simple_arith" "simple_arith,simple_arith_2,simple_arith_4,load,store,branch,call,flow")
(define_insn_reservation "simple_arith_2" 2
(and (eq_attr "pipe_model" "epiphany")
(eq_attr "type" "move,cmove,compare,shift,misc,mul")
(eq_attr "length" "8"))
"issue,issue+int,int")
(define_insn_reservation "simple_arith_4" 4
(and (eq_attr "pipe_model" "epiphany")
(eq_attr "type" "move,compare,shift,misc,mul")
(eq_attr "length" "12,16,20,24"))
"issue,issue+int,issue+int,issue+int,int")
;; Loads have a latency of two.
;; Note that we fix up the latency of post_modify in epiphany.c:epiphany_adjust_cost
(define_insn_reservation "load" 3
(and (eq_attr "pipe_model" "epiphany")
(eq_attr "type" "load"))
"issue,int")
; anything but fp / fp_int / v2fp has a bypass
(define_bypass 2 "load" "simple_arith,simple_arith_2,simple_arith_4,load,store,branch,call,flow")
(define_insn_reservation "store" 1
(and (eq_attr "pipe_model" "epiphany")
(eq_attr "type" "store"))
"issue,int")
;; Branch
;; Latency when taken: 3
;; Issue Rate: 1
;; The latency is 1 when the branch is not taken.
;; We can't really do much with the latency, even if we could express it,
;; but the pairing restrictions are useful to take into account.
(define_insn_reservation "branch" 1
(and (eq_attr "pipe_model" "epiphany")
(eq_attr "type" "branch,uncond_branch"))
"d_lock")
;; calls introduce a longisch delay that is likely to flush the pipelines
;; of the caller's instructions. Both the call instruction itself and
;; the rts at the end of the call / sfunc incurs a three cycle penalty,
;; thus also isolating the scheduling of caller and callee.
(define_insn_reservation "call" 8
(and (eq_attr "pipe_model" "epiphany")
(eq_attr "type" "call,sfunc,fp_sfunc"))
"d_lock*8")
(define_insn_reservation "flow" 1
(and (eq_attr "pipe_model" "epiphany")
(eq_attr "type" "flow"))
"d_lock")
(define_insn_reservation "fp_arith" 5
(and (eq_attr "pipe_model" "epiphany")
(eq_attr "type" "fp,fp_int"))
"issue,F0")
(define_bypass 4 "fp_arith" "store")
; There are two main consumers for v2fp:
; - other v2fp operation - in that case, the latencies can dovetail to
; save one cycle of latency.
; - 64 bit store operations - we need both registers, but OTOH the latency is
; one lower to start with.
; of the bypass saving one cyles then.
(define_insn_reservation "v2fp_arith" 5
(and (eq_attr "pipe_model" "epiphany")
(eq_attr "type" "v2fp"))
"issue,issue+F0,F0")
; A boolean attribute for use by peephole2 patterns that try to figure out
; if we overcommitted the FPU.
; This is notionally a numeric attribute to avoid dependency problems.
(define_attr "sched_use_fpu" ""
(cond [(eq_attr "type" "fp,fp_int,v2fp") (const_int 1)]
(const_int 0)))
|
