inorder: update support for branch delay slots
[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 InstSeqNum seq_num = inst->seqNum;
82
83 DPRINTF(InOrderFetchSeq, "[tid:%i]: Current PC is %s\n", tid,
84 pc[tid]);
85
86 switch (fs_req->cmd)
87 {
88 case AssignNextPC:
89 {
90 if (pcValid[tid]) {
91 inst->pcState(pc[tid]);
92 inst->setMemAddr(pc[tid].instAddr());
93
94 // Advance to next PC (typically PC + 4)
95 pc[tid].advance();
96
97 inst->setSeqNum(cpu->getAndIncrementInstSeq(tid));
98
99 DPRINTF(InOrderFetchSeq, "[tid:%i]: Assigning [sn:%i] to "
100 "PC %s\n", tid, inst->seqNum, inst->pcState());
101
102 fs_req->done();
103 } else {
104 DPRINTF(InOrderStall, "STALL: [tid:%i]: NPC not valid\n", tid);
105 fs_req->done(false);
106 }
107 }
108 break;
109
110 case UpdateTargetPC:
111 {
112 if (inst->isControl()) {
113 // If it's a return, then we must wait for resolved address.
114 // The Predictor will mark a return a false as "not taken"
115 // if there is no RAS entry
116 if (inst->isReturn() && !inst->predTaken()) {
117 cpu->pipelineStage[stage_num]->
118 toPrevStages->stageBlock[stage_num][tid] = true;
119 pcValid[tid] = false;
120 pcBlockStage[tid] = stage_num;
121 } else if (inst->isCondDelaySlot() && !inst->predTaken()) {
122 assert(0 && "Not Handling Conditional Delay Slot");
123 // Not-Taken AND Conditional Control
124 DPRINTF(InOrderFetchSeq, "[tid:%i]: [sn:%i]: [PC:%s] "
125 "Predicted Not-Taken Cond. Delay inst. Skipping "
126 "delay slot and Updating PC to %s\n",
127 tid, inst->seqNum, inst->pcState(),
128 inst->readPredTarg());
129
130 DPRINTF(InOrderFetchSeq, "[tid:%i] Setting up squash to "
131 "start from stage %i, after [sn:%i].\n", tid,
132 stage_num, seq_num);
133
134 inst->bdelaySeqNum = seq_num;
135 inst->squashingStage = stage_num;
136
137 squashAfterInst(inst, stage_num, tid);
138 } else if (!inst->isCondDelaySlot() && !inst->predTaken()) {
139 // Not-Taken Control
140 DPRINTF(InOrderFetchSeq, "[tid:%i]: [sn:%i]: Predicted "
141 "Not-Taken Control "
142 "inst. updating PC to %s\n", tid, inst->seqNum,
143 inst->readPredTarg());
144 } else if (inst->predTaken()) {
145 // Taken Control
146 inst->bdelaySeqNum = seq_num;
147 inst->squashingStage = stage_num;
148 DPRINTF(InOrderFetchSeq, "[tid:%i] Setting up squash to "
149 "start from stage %i, after [sn:%i].\n",
150 tid, stage_num, inst->bdelaySeqNum);
151
152 // Do Squashing
153 squashAfterInst(inst, stage_num, tid);
154 }
155 } else {
156 DPRINTF(InOrderFetchSeq, "[tid:%i]: [sn:%i]: Ignoring branch "
157 "target update since then is not a control "
158 "instruction.\n", tid, inst->seqNum);
159 }
160
161 fs_req->done();
162 }
163 break;
164
165 default:
166 fatal("Unrecognized command to %s", resName);
167 }
168 }
169
170 inline void
171 FetchSeqUnit::squashAfterInst(DynInstPtr inst, int stage_num, ThreadID tid)
172 {
173 // Squash In Pipeline Stage
174 cpu->pipelineStage[stage_num]->squashDueToBranch(inst, tid);
175
176 // Schedule Squash Through-out Resource Pool
177 cpu->resPool->scheduleEvent(
178 (InOrderCPU::CPUEventType)ResourcePool::SquashAll, inst, 0);
179 }
180
181 void
182 FetchSeqUnit::squash(DynInstPtr inst, int squash_stage,
183 InstSeqNum squash_seq_num, ThreadID tid)
184 {
185 DPRINTF(InOrderFetchSeq, "[tid:%i]: Updating due to squash from %s (%s) "
186 "stage %i.\n", tid, inst->instName(), inst->pcState(),
187 squash_stage);
188 assert(squash_seq_num == inst->seqNum);
189
190 TheISA::PCState nextPC = inst->pcState();
191 assert(inst->staticInst);
192 advancePC(nextPC, inst->staticInst);
193
194 #if ISA_HAS_DELAY_SLOT
195 if (inst->isControl()) {
196 if (inst->onInstList) {
197 ListIt inst_it = inst->getInstListIt();
198 inst_it++;
199 if (inst_it != cpu->instList[tid].end()) {
200 DynInstPtr delaySlotInst = (*inst_it);
201 if (delaySlotInst->pcState() != nextPC)
202 squash_seq_num = delaySlotInst->seqNum;
203 }
204 }
205 }
206 #endif
207
208 if (squashSeqNum[tid] <= squash_seq_num &&
209 lastSquashCycle[tid] == curTick()) {
210 DPRINTF(InOrderFetchSeq, "[tid:%i]: Ignoring squash from stage %i, "
211 "since there is an outstanding squash that is older.\n",
212 tid, squash_stage);
213 } else {
214 squashSeqNum[tid] = squash_seq_num;
215 lastSquashCycle[tid] = curTick();
216
217 DPRINTF(InOrderFetchSeq, "[tid:%i]: Setting PC to %s.\n",
218 tid, nextPC);
219 pc[tid] = nextPC;
220
221 // Unblock Any Stages Waiting for this information to be updated ...
222 if (!pcValid[tid]) {
223 cpu->pipelineStage[pcBlockStage[tid]]->
224 toPrevStages->stageUnblock[pcBlockStage[tid]][tid] = true;
225 }
226
227 pcValid[tid] = true;
228 }
229
230 Resource::squash(inst, squash_stage, squash_seq_num, tid);
231 }
232
233 FetchSeqUnit::FetchSeqEvent::FetchSeqEvent()
234 : ResourceEvent()
235 { }
236
237 void
238 FetchSeqUnit::FetchSeqEvent::process()
239 {
240 FetchSeqUnit* fs_res = dynamic_cast<FetchSeqUnit*>(resource);
241 assert(fs_res);
242
243 for (int i = 0; i < MaxThreads; i++) {
244 fs_res->pc[i] = fs_res->cpu->pcState(i);
245 DPRINTF(InOrderFetchSeq, "[tid:%i]: Setting PC: %s.\n",
246 fs_res->pc[i]);
247
248 fs_res->pcValid[i] = true;
249 }
250 }
251
252
253 void
254 FetchSeqUnit::activateThread(ThreadID tid)
255 {
256 pcValid[tid] = true;
257
258 pc[tid] = cpu->pcState(tid);
259
260 cpu->fetchPriorityList.push_back(tid);
261
262 DPRINTF(InOrderFetchSeq, "[tid:%i]: Reading PC: %s.\n",
263 tid, pc[tid]);
264 }
265
266 void
267 FetchSeqUnit::deactivateThread(ThreadID tid)
268 {
269 pcValid[tid] = false;
270 pcBlockStage[tid] = 0;
271
272 squashSeqNum[tid] = (InstSeqNum)-1;
273 lastSquashCycle[tid] = 0;
274
275 list<ThreadID>::iterator thread_it = find(cpu->fetchPriorityList.begin(),
276 cpu->fetchPriorityList.end(),
277 tid);
278
279 if (thread_it != cpu->fetchPriorityList.end())
280 cpu->fetchPriorityList.erase(thread_it);
281 }
282
283 void
284 FetchSeqUnit::suspendThread(ThreadID tid)
285 {
286 deactivateThread(tid);
287 }
288
289 void
290 FetchSeqUnit::updateAfterContextSwitch(DynInstPtr inst, ThreadID tid)
291 {
292 pcValid[tid] = true;
293
294 if (cpu->thread[tid]->lastGradIsBranch) {
295 /** This function assumes that the instruction causing the context
296 * switch was right after the branch. Thus, if it's not, then
297 * we are updating incorrectly here
298 */
299 assert(cpu->nextInstAddr(tid) == inst->instAddr());
300 pc[tid] = cpu->thread[tid]->lastBranchPC;
301 } else {
302 pc[tid] = inst->pcState();
303 }
304 assert(inst->staticInst);
305 advancePC(pc[tid], inst->staticInst);
306
307 DPRINTF(InOrderFetchSeq, "[tid:%i]: Updating PCs due to Context Switch."
308 "Assigning PC: %s.\n", tid, pc[tid]);
309 }