Mercurial > hg > CbC > CbC_gcc
comparison gcc/config/arm/fa526.md @ 68:561a7518be6b
update gcc-4.6
author | Nobuyasu Oshiro <dimolto@cr.ie.u-ryukyu.ac.jp> |
---|---|
date | Sun, 21 Aug 2011 07:07:55 +0900 |
parents | |
children | 04ced10e8804 |
comparison
equal
deleted
inserted
replaced
67:f6334be47118 | 68:561a7518be6b |
---|---|
1 ;; Faraday FA526 Pipeline Description | |
2 ;; Copyright (C) 2010 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 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")) | |
66 "fa526_core") | |
67 | |
68 (define_insn_reservation "526_alu_shift_op" 2 | |
69 (and (eq_attr "tune" "fa526") | |
70 (eq_attr "type" "alu_shift,alu_shift_reg")) | |
71 "fa526_core") | |
72 | |
73 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; | |
74 ;; Multiplication Instructions | |
75 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; | |
76 | |
77 (define_insn_reservation "526_mult1" 2 | |
78 (and (eq_attr "tune" "fa526") | |
79 (eq_attr "insn" "smlalxy,smulxy,smlaxy,smlalxy")) | |
80 "fa526_core") | |
81 | |
82 (define_insn_reservation "526_mult2" 5 | |
83 (and (eq_attr "tune" "fa526") | |
84 (eq_attr "insn" "mul,mla,muls,mlas,umull,umlal,smull,smlal,umulls,\ | |
85 umlals,smulls,smlals,smlawx")) | |
86 "fa526_core*4") | |
87 | |
88 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; | |
89 ;; Load/Store Instructions | |
90 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; | |
91 | |
92 ;; The models for load/store instructions do not accurately describe | |
93 ;; the difference between operations with a base register writeback | |
94 ;; (such as "ldm!"). These models assume that all memory references | |
95 ;; hit in dcache. | |
96 | |
97 (define_insn_reservation "526_load1_op" 3 | |
98 (and (eq_attr "tune" "fa526") | |
99 (eq_attr "type" "load1,load_byte")) | |
100 "fa526_core") | |
101 | |
102 (define_insn_reservation "526_load2_op" 4 | |
103 (and (eq_attr "tune" "fa526") | |
104 (eq_attr "type" "load2")) | |
105 "fa526_core*2") | |
106 | |
107 (define_insn_reservation "526_load3_op" 5 | |
108 (and (eq_attr "tune" "fa526") | |
109 (eq_attr "type" "load3")) | |
110 "fa526_core*3") | |
111 | |
112 (define_insn_reservation "526_load4_op" 6 | |
113 (and (eq_attr "tune" "fa526") | |
114 (eq_attr "type" "load4")) | |
115 "fa526_core*4") | |
116 | |
117 (define_insn_reservation "526_store1_op" 0 | |
118 (and (eq_attr "tune" "fa526") | |
119 (eq_attr "type" "store1")) | |
120 "fa526_core") | |
121 | |
122 (define_insn_reservation "526_store2_op" 1 | |
123 (and (eq_attr "tune" "fa526") | |
124 (eq_attr "type" "store2")) | |
125 "fa526_core*2") | |
126 | |
127 (define_insn_reservation "526_store3_op" 2 | |
128 (and (eq_attr "tune" "fa526") | |
129 (eq_attr "type" "store3")) | |
130 "fa526_core*3") | |
131 | |
132 (define_insn_reservation "526_store4_op" 3 | |
133 (and (eq_attr "tune" "fa526") | |
134 (eq_attr "type" "store4")) | |
135 "fa526_core*4") | |
136 | |
137 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; | |
138 ;; Branch and Call Instructions | |
139 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;; | |
140 | |
141 ;; Branch instructions are difficult to model accurately. The FA526 | |
142 ;; core can predict most branches. If the branch is predicted | |
143 ;; correctly, and predicted early enough, the branch can be completely | |
144 ;; eliminated from the instruction stream. Some branches can | |
145 ;; therefore appear to require zero cycle to execute. We assume that | |
146 ;; all branches are predicted correctly, and that the latency is | |
147 ;; therefore the minimum value. | |
148 | |
149 (define_insn_reservation "526_branch_op" 0 | |
150 (and (eq_attr "tune" "fa526") | |
151 (eq_attr "type" "branch")) | |
152 "fa526_core") | |
153 | |
154 ;; The latency for a call is actually the latency when the result is available. | |
155 ;; i.e. R0 ready for int return value. For most cases, the return value is set | |
156 ;; by a mov instruction, which has 1 cycle latency. | |
157 (define_insn_reservation "526_call_op" 1 | |
158 (and (eq_attr "tune" "fa526") | |
159 (eq_attr "type" "call")) | |
160 "fa526_core") | |
161 |