1 ;; Faraday FA626TE Pipeline Description
2 ;; Copyright (C) 2010-2013 Free Software Foundation, Inc.
3 ;; Written by Mingfeng Wu, based on ARM926EJ-S Pipeline Description.
5 ;; This file is part of GCC.
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
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
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/>. */
21 ;; These descriptions are based on the information contained in the
22 ;; FMP626 Core Design Note, Copyright (c) 2010 Faraday Technology Corp.
24 ;; Pipeline architecture
26 ;; ___________________________________________
29 ;; ld/st1 ld/st2 ld/st3 ld/st4 ld/st5
31 ;; This automaton provides a pipeline description for the Faraday
34 ;; The model given here assumes that the condition for all conditional
35 ;; instructions is "true", i.e., that all of the instructions are
38 (define_automaton "fmp626")
40 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
42 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
44 ;; There is a single pipeline
46 ;; The ALU pipeline has fetch, decode, execute, memory, and
47 ;; write stages. We only need to model the execute, memory and write
50 (define_cpu_unit "fmp626_core" "fmp626")
52 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
54 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
56 ;; ALU instructions require two cycles to execute, and use the ALU
57 ;; pipeline in each of the three stages. The results are available
58 ;; after the execute stage stage has finished.
60 ;; If the destination register is the PC, the pipelines are stalled
61 ;; for several cycles. That case is not modeled here.
64 (define_insn_reservation "mp626_alu_op" 1
65 (and (eq_attr "tune" "fmp626")
66 (eq_attr "type" "arlo_imm,arlo_reg,shift,shift_reg,\
67 mov_imm,mov_reg,mvn_imm,mvn_reg"))
70 (define_insn_reservation "mp626_alu_shift_op" 2
71 (and (eq_attr "tune" "fmp626")
72 (eq_attr "type" "extend,arlo_shift,arlo_shift_reg,\
73 mov_shift,mov_shift_reg,\
74 mvn_shift,mvn_shift_reg"))
77 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
78 ;; Multiplication Instructions
79 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
81 (define_insn_reservation "mp626_mult1" 2
82 (and (eq_attr "tune" "fmp626")
83 (eq_attr "type" "smulwy,smlawy,smulxy,smlaxy"))
86 (define_insn_reservation "mp626_mult2" 2
87 (and (eq_attr "tune" "fmp626")
88 (eq_attr "type" "mul,mla"))
91 (define_insn_reservation "mp626_mult3" 3
92 (and (eq_attr "tune" "fmp626")
93 (eq_attr "type" "muls,mlas,smull,smlal,umull,umlal,smlalxy,smlawx"))
96 (define_insn_reservation "mp626_mult4" 4
97 (and (eq_attr "tune" "fmp626")
98 (eq_attr "type" "smulls,smlals,umulls,umlals"))
101 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
102 ;; Load/Store Instructions
103 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
105 ;; The models for load/store instructions do not accurately describe
106 ;; the difference between operations with a base register writeback
107 ;; (such as "ldm!"). These models assume that all memory references
110 (define_insn_reservation "mp626_load1_op" 5
111 (and (eq_attr "tune" "fmp626")
112 (eq_attr "type" "load1,load_byte"))
115 (define_insn_reservation "mp626_load2_op" 6
116 (and (eq_attr "tune" "fmp626")
117 (eq_attr "type" "load2,load3"))
120 (define_insn_reservation "mp626_load3_op" 7
121 (and (eq_attr "tune" "fmp626")
122 (eq_attr "type" "load4"))
125 (define_insn_reservation "mp626_store1_op" 0
126 (and (eq_attr "tune" "fmp626")
127 (eq_attr "type" "store1"))
130 (define_insn_reservation "mp626_store2_op" 1
131 (and (eq_attr "tune" "fmp626")
132 (eq_attr "type" "store2,store3"))
135 (define_insn_reservation "mp626_store3_op" 2
136 (and (eq_attr "tune" "fmp626")
137 (eq_attr "type" "store4"))
140 (define_bypass 1 "mp626_load1_op,mp626_load2_op,mp626_load3_op"
141 "mp626_store1_op,mp626_store2_op,mp626_store3_op"
142 "arm_no_early_store_addr_dep")
143 (define_bypass 1 "mp626_alu_op,mp626_alu_shift_op,mp626_mult1,mp626_mult2,\
144 mp626_mult3,mp626_mult4" "mp626_store1_op"
145 "arm_no_early_store_addr_dep")
146 (define_bypass 1 "mp626_alu_shift_op" "mp626_alu_op")
147 (define_bypass 1 "mp626_alu_shift_op" "mp626_alu_shift_op"
148 "arm_no_early_alu_shift_dep")
149 (define_bypass 1 "mp626_mult1,mp626_mult2" "mp626_alu_shift_op"
150 "arm_no_early_alu_shift_dep")
151 (define_bypass 2 "mp626_mult3" "mp626_alu_shift_op"
152 "arm_no_early_alu_shift_dep")
153 (define_bypass 3 "mp626_mult4" "mp626_alu_shift_op"
154 "arm_no_early_alu_shift_dep")
155 (define_bypass 1 "mp626_mult1,mp626_mult2" "mp626_alu_op")
156 (define_bypass 2 "mp626_mult3" "mp626_alu_op")
157 (define_bypass 3 "mp626_mult4" "mp626_alu_op")
158 (define_bypass 4 "mp626_load1_op" "mp626_alu_op")
159 (define_bypass 5 "mp626_load2_op" "mp626_alu_op")
160 (define_bypass 6 "mp626_load3_op" "mp626_alu_op")
162 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
163 ;; Branch and Call Instructions
164 ;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;
166 ;; Branch instructions are difficult to model accurately. The FMP626
167 ;; core can predict most branches. If the branch is predicted
168 ;; correctly, and predicted early enough, the branch can be completely
169 ;; eliminated from the instruction stream. Some branches can
170 ;; therefore appear to require zero cycle to execute. We assume that
171 ;; all branches are predicted correctly, and that the latency is
172 ;; therefore the minimum value.
174 (define_insn_reservation "mp626_branch_op" 0
175 (and (eq_attr "tune" "fmp626")
176 (eq_attr "type" "branch"))
179 ;; The latency for a call is actually the latency when the result is available.
180 ;; i.e. R0 ready for int return value.
181 (define_insn_reservation "mp626_call_op" 1
182 (and (eq_attr "tune" "fmp626")
183 (eq_attr "type" "call"))