1;; Faraday FA526 Pipeline Description
2;; Copyright (C) 2010-2020 Free Software Foundation, Inc.
3;; Written by I-Jui Sung, based on ARM926EJ-S Pipeline Description.
4
5;; This file is part of GCC.
6;;
7;; GCC is free software; you can redistribute it and/or modify it under
8;; the terms of the GNU General Public License as published by the Free
9;; Software Foundation; either version 3, or (at your option) any later
10;; version.
11;;
12;; GCC is distributed in the hope that it will be useful, but WITHOUT ANY
13;; WARRANTY; without even the implied warranty of MERCHANTABILITY or
14;; FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
15;; for more details.
16;;
17;; You should have received a copy of the GNU General Public License
18;; along with GCC; see the file COPYING3.  If not see
19;; <http://www.gnu.org/licenses/>.  */
20
21;; These descriptions are based on the information contained in the
22;; FA526 Core Design Note, Copyright (c) 2010 Faraday Technology Corp.
23;;
24;; Modeled pipeline characteristics:
25;; LD -> any use: latency = 3 (2 cycle penalty).
26;; ALU -> any use: latency = 2 (1 cycle penalty).
27
28;; This automaton provides a pipeline description for the Faraday
29;; FA526 core.
30;;
31;; The model given here assumes that the condition for all conditional
32;; instructions is "true", i.e., that all of the instructions are
33;; actually executed.
34
35(define_automaton "fa526")
36
37;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
38;; Pipelines
39;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
40
41;; There is a single pipeline
42;;
43;;   The ALU pipeline has fetch, decode, execute, memory, and
44;;   write stages.  We only need to model the execute, memory and write
45;;   stages.
46
47;;      S      E      M      W
48
49(define_cpu_unit "fa526_core" "fa526")
50
51;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
52;; ALU Instructions
53;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
54
55;; ALU instructions require two cycles to execute, and use the ALU
56;; pipeline in each of the three stages.  The results are available
57;; after the execute stage has finished.
58;;
59;; If the destination register is the PC, the pipelines are stalled
60;; for several cycles.  That case is not modeled here.
61
62;; ALU operations
63(define_insn_reservation "526_alu_op" 1
64 (and (eq_attr "tune" "fa526")
65      (eq_attr "type" "alu_imm,alus_imm,logic_imm,logics_imm,\
66                       alu_sreg,alus_sreg,logic_reg,logics_reg,\
67                       adc_imm,adcs_imm,adc_reg,adcs_reg,\
68                       adr,bfm,rev,\
69                       shift_imm,shift_reg,\
70                       mov_imm,mov_reg,mvn_imm,mvn_reg,\
71                       mrs,multiple"))
72 "fa526_core")
73
74(define_insn_reservation "526_alu_shift_op" 2
75 (and (eq_attr "tune" "fa526")
76      (eq_attr "type" "extend,\
77                       alu_shift_imm,alus_shift_imm,\
78                       logic_shift_imm,logics_shift_imm,\
79                       alu_shift_reg,alus_shift_reg,\
80                       logic_shift_reg,logics_shift_reg,\
81                       mov_shift,mov_shift_reg,\
82                       mvn_shift,mvn_shift_reg"))
83 "fa526_core")
84
85;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
86;; Multiplication Instructions
87;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
88
89(define_insn_reservation "526_mult1" 2
90 (and (eq_attr "tune" "fa526")
91      (eq_attr "type" "smlalxy,smulxy,smlaxy,smlalxy"))
92 "fa526_core")
93
94(define_insn_reservation "526_mult2" 5
95 (and (eq_attr "tune" "fa526")
96      (eq_attr "type" "mul,mla,muls,mlas,umull,umlal,smull,smlal,umulls,\
97                       umlals,smulls,smlals,smlawx"))
98 "fa526_core*4")
99
100;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
101;; Load/Store Instructions
102;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
103
104;; The models for load/store instructions do not accurately describe
105;; the difference between operations with a base register writeback
106;; (such as "ldm!").  These models assume that all memory references
107;; hit in dcache.
108
109(define_insn_reservation "526_load1_op" 3
110 (and (eq_attr "tune" "fa526")
111      (eq_attr "type" "load_4,load_byte"))
112 "fa526_core")
113
114(define_insn_reservation "526_load2_op" 4
115 (and (eq_attr "tune" "fa526")
116      (eq_attr "type" "load_8"))
117 "fa526_core*2")
118
119(define_insn_reservation "526_load3_op" 5
120 (and (eq_attr "tune" "fa526")
121      (eq_attr "type" "load_12"))
122 "fa526_core*3")
123
124(define_insn_reservation "526_load4_op" 6
125 (and (eq_attr "tune" "fa526")
126      (eq_attr "type" "load_16"))
127 "fa526_core*4")
128
129(define_insn_reservation "526_store1_op" 0
130 (and (eq_attr "tune" "fa526")
131      (eq_attr "type" "store_4"))
132 "fa526_core")
133
134(define_insn_reservation "526_store2_op" 1
135 (and (eq_attr "tune" "fa526")
136      (eq_attr "type" "store_8"))
137 "fa526_core*2")
138
139(define_insn_reservation "526_store3_op" 2
140 (and (eq_attr "tune" "fa526")
141      (eq_attr "type" "store_12"))
142 "fa526_core*3")
143
144(define_insn_reservation "526_store4_op" 3
145 (and (eq_attr "tune" "fa526")
146      (eq_attr "type" "store_16"))
147 "fa526_core*4")
148
149;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
150;; Branch and Call Instructions
151;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
152
153;; Branch instructions are difficult to model accurately.  The FA526
154;; core can predict most branches.  If the branch is predicted
155;; correctly, and predicted early enough, the branch can be completely
156;; eliminated from the instruction stream.  Some branches can
157;; therefore appear to require zero cycle to execute.  We assume that
158;; all branches are predicted correctly, and that the latency is
159;; therefore the minimum value.
160
161(define_insn_reservation "526_branch_op" 0
162 (and (eq_attr "tune" "fa526")
163      (eq_attr "type" "branch"))
164 "fa526_core")
165
166;; The latency for a call is actually the latency when the result is available.
167;; i.e. R0 ready for int return value.  For most cases, the return value is set
168;; by a mov instruction, which has 1 cycle latency.
169(define_insn_reservation "526_call_op" 1
170 (and (eq_attr "tune" "fa526")
171      (eq_attr "type" "call"))
172 "fa526_core")
173
174