inorder: use setupSquash for misspeculation
[gem5.git] / src / cpu / inorder / resources / fetch_seq_unit.cc
1 /*
2 * Copyright (c) 2007 MIPS Technologies, Inc.
3 * All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions are
7 * met: redistributions of source code must retain the above copyright
8 * notice, this list of conditions and the following disclaimer;
9 * redistributions in binary form must reproduce the above copyright
10 * notice, this list of conditions and the following disclaimer in the
11 * documentation and/or other materials provided with the distribution;
12 * neither the name of the copyright holders nor the names of its
13 * contributors may be used to endorse or promote products derived from
14 * this software without specific prior written permission.
15 *
16 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
17 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
18 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
19 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
20 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
21 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
22 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
23 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
24 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
25 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
26 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
27 *
28 * Authors: Korey Sewell
29 *
30 */
31
32 #include "config/the_isa.hh"
33 #include "cpu/inorder/resources/fetch_seq_unit.hh"
34 #include "cpu/inorder/resource_pool.hh"
35 #include "debug/InOrderFetchSeq.hh"
36 #include "debug/InOrderStall.hh"
37
38 using namespace std;
39 using namespace TheISA;
40 using namespace ThePipeline;
41
42 FetchSeqUnit::FetchSeqUnit(std::string res_name, int res_id, int res_width,
43 int res_latency, InOrderCPU *_cpu,
44 ThePipeline::Params *params)
45 : Resource(res_name, res_id, res_width, res_latency, _cpu),
46 instSize(sizeof(MachInst))
47 {
48 for (ThreadID tid = 0; tid < ThePipeline::MaxThreads; tid++) {
49 pcValid[tid] = false;
50 pcBlockStage[tid] = 0;
51
52 squashSeqNum[tid] = (InstSeqNum)-1;
53 lastSquashCycle[tid] = 0;
54 }
55 }
56
57 FetchSeqUnit::~FetchSeqUnit()
58 {
59 delete [] resourceEvent;
60 }
61
62 void
63 FetchSeqUnit::init()
64 {
65 resourceEvent = new FetchSeqEvent[width];
66
67 for (int i = 0; i < width; i++) {
68 reqs[i] = new ResourceRequest(this);
69 }
70
71 initSlots();
72 }
73
74 void
75 FetchSeqUnit::execute(int slot_num)
76 {
77 ResourceRequest* fs_req = reqs[slot_num];
78 DynInstPtr inst = fs_req->inst;
79 ThreadID tid = inst->readTid();
80 int stage_num = fs_req->getStageNum();
81
82 DPRINTF(InOrderFetchSeq, "[tid:%i]: Current PC is %s\n", tid,
83 pc[tid]);
84
85 switch (fs_req->cmd)
86 {
87 case AssignNextPC:
88 {
89 if (pcValid[tid]) {
90 inst->pcState(pc[tid]);
91 inst->setMemAddr(pc[tid].instAddr());
92
93 // Advance to next PC (typically PC + 4)
94 pc[tid].advance();
95
96 inst->setSeqNum(cpu->getAndIncrementInstSeq(tid));
97
98 DPRINTF(InOrderFetchSeq, "[tid:%i]: Assigning [sn:%i] to "
99 "PC %s\n", tid, inst->seqNum, inst->pcState());
100
101 fs_req->done();
102 } else {
103 DPRINTF(InOrderStall, "STALL: [tid:%i]: NPC not valid\n", tid);
104 fs_req->done(false);
105 }
106 }
107 break;
108
109 case UpdateTargetPC:
110 {
111 assert(!inst->isCondDelaySlot() &&
112 "Not Handling Conditional Delay Slot");
113
114 if (inst->isControl()) {
115 if (inst->isReturn() && !inst->predTaken()) {
116 // If it's a return, then we must wait for resolved address.
117 // The Predictor will mark a return a false as "not taken"
118 // if there is no RAS entry
119 cpu->pipelineStage[stage_num]->
120 toPrevStages->stageBlock[stage_num][tid] = true;
121 pcValid[tid] = false;
122 pcBlockStage[tid] = stage_num;
123 } else if (inst->predTaken()) {
124 // Taken Control
125 inst->setSquashInfo(stage_num);
126 setupSquash(inst, stage_num, tid);
127
128 DPRINTF(InOrderFetchSeq, "[tid:%i] Setting up squash to "
129 "start from stage %i, after [sn:%i].\n",
130 tid, stage_num, inst->bdelaySeqNum);
131 }
132 } else {
133 DPRINTF(InOrderFetchSeq, "[tid:%i]: [sn:%i]: Ignoring branch "
134 "target update since then is not a control "
135 "instruction.\n", tid, inst->seqNum);
136 }
137
138 fs_req->done();
139 }
140 break;
141
142 default:
143 fatal("Unrecognized command to %s", resName);
144 }
145 }
146
147 void
148 FetchSeqUnit::squash(DynInstPtr inst, int squash_stage,
149 InstSeqNum squash_seq_num, ThreadID tid)
150 {
151 DPRINTF(InOrderFetchSeq, "[tid:%i]: Updating due to squash from %s (%s) "
152 "stage %i.\n", tid, inst->instName(), inst->pcState(),
153 squash_stage);
154
155 if (squashSeqNum[tid] <= squash_seq_num &&
156 lastSquashCycle[tid] == curTick()) {
157 DPRINTF(InOrderFetchSeq, "[tid:%i]: Ignoring squash from stage %i, "
158 "since there is an outstanding squash that is older.\n",
159 tid, squash_stage);
160 } else {
161 squashSeqNum[tid] = squash_seq_num;
162 lastSquashCycle[tid] = curTick();
163
164 TheISA::PCState nextPC;
165 assert(inst->staticInst);
166 if (inst->isControl()) {
167 nextPC = inst->readPredTarg();
168
169 // If we are already fetching this PC then advance to next PC
170 // =======
171 // This should handle ISAs w/delay slots and annulled delay
172 // slots to figure out which is the next PC to fetch after
173 // a mispredict
174 DynInstPtr bdelay_inst = NULL;
175 ListIt bdelay_it;
176 if (inst->onInstList) {
177 bdelay_it = inst->getInstListIt();
178 bdelay_it++;
179 } else {
180 InstSeqNum branch_delay_num = inst->seqNum + 1;
181 bdelay_it = cpu->findInst(branch_delay_num, tid);
182 }
183
184 if (bdelay_it != cpu->instList[tid].end()) {
185 bdelay_inst = (*bdelay_it);
186 }
187
188 if (bdelay_inst) {
189 DPRINTF(Resource, "Evaluating %s v. %s\n",
190 bdelay_inst->pc, nextPC);
191
192 if (bdelay_inst->pc.instAddr() == nextPC.instAddr()) {
193 advancePC(nextPC, inst->staticInst);
194 DPRINTF(Resource, "Advanced PC to %s\n", nextPC);
195 }
196 }
197 } else {
198 nextPC = inst->pcState();
199 advancePC(nextPC, inst->staticInst);
200 }
201
202
203 DPRINTF(InOrderFetchSeq, "[tid:%i]: Setting PC to %s.\n",
204 tid, nextPC);
205 pc[tid] = nextPC;
206
207 // Unblock Any Stages Waiting for this information to be updated ...
208 if (!pcValid[tid]) {
209 cpu->pipelineStage[pcBlockStage[tid]]->
210 toPrevStages->stageUnblock[pcBlockStage[tid]][tid] = true;
211 }
212
213 pcValid[tid] = true;
214 }
215
216 Resource::squash(inst, squash_stage, squash_seq_num, tid);
217 }
218
219 FetchSeqUnit::FetchSeqEvent::FetchSeqEvent()
220 : ResourceEvent()
221 { }
222
223 void
224 FetchSeqUnit::FetchSeqEvent::process()
225 {
226 FetchSeqUnit* fs_res = dynamic_cast<FetchSeqUnit*>(resource);
227 assert(fs_res);
228
229 for (int i = 0; i < MaxThreads; i++) {
230 fs_res->pc[i] = fs_res->cpu->pcState(i);
231 DPRINTF(InOrderFetchSeq, "[tid:%i]: Setting PC: %s.\n",
232 fs_res->pc[i]);
233
234 fs_res->pcValid[i] = true;
235 }
236 }
237
238
239 void
240 FetchSeqUnit::activateThread(ThreadID tid)
241 {
242 pcValid[tid] = true;
243
244 pc[tid] = cpu->pcState(tid);
245
246 cpu->fetchPriorityList.push_back(tid);
247
248 DPRINTF(InOrderFetchSeq, "[tid:%i]: Reading PC: %s.\n",
249 tid, pc[tid]);
250 }
251
252 void
253 FetchSeqUnit::deactivateThread(ThreadID tid)
254 {
255 pcValid[tid] = false;
256 pcBlockStage[tid] = 0;
257
258 squashSeqNum[tid] = (InstSeqNum)-1;
259 lastSquashCycle[tid] = 0;
260
261 list<ThreadID>::iterator thread_it = find(cpu->fetchPriorityList.begin(),
262 cpu->fetchPriorityList.end(),
263 tid);
264
265 if (thread_it != cpu->fetchPriorityList.end())
266 cpu->fetchPriorityList.erase(thread_it);
267 }
268
269 void
270 FetchSeqUnit::suspendThread(ThreadID tid)
271 {
272 deactivateThread(tid);
273 }
274
275 void
276 FetchSeqUnit::updateAfterContextSwitch(DynInstPtr inst, ThreadID tid)
277 {
278 pcValid[tid] = true;
279
280 if (cpu->thread[tid]->lastGradIsBranch) {
281 /** This function assumes that the instruction causing the context
282 * switch was right after the branch. Thus, if it's not, then
283 * we are updating incorrectly here
284 */
285 assert(cpu->nextInstAddr(tid) == inst->instAddr());
286 pc[tid] = cpu->thread[tid]->lastBranchPC;
287 } else {
288 pc[tid] = inst->pcState();
289 }
290 assert(inst->staticInst);
291 advancePC(pc[tid], inst->staticInst);
292
293 DPRINTF(InOrderFetchSeq, "[tid:%i]: Updating PCs due to Context Switch."
294 "Assigning PC: %s.\n", tid, pc[tid]);
295 }