2 * Copyright (c) 2007 MIPS Technologies, Inc.
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.
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.
28 * Authors: Korey Sewell
32 #include "cpu/inorder/resources/fetch_seq_unit.hh"
33 #include "cpu/inorder/resource_pool.hh"
36 using namespace TheISA
;
37 using namespace ThePipeline
;
39 FetchSeqUnit::FetchSeqUnit(std::string res_name
, int res_id
, int res_width
,
40 int res_latency
, InOrderCPU
*_cpu
, ThePipeline::Params
*params
)
41 : Resource(res_name
, res_id
, res_width
, res_latency
, _cpu
),
42 instSize(sizeof(MachInst
))
44 for (int tid
= 0; tid
< ThePipeline::MaxThreads
; tid
++) {
45 delaySlotInfo
[tid
].numInsts
= 0;
46 delaySlotInfo
[tid
].targetReady
= false;
49 pcBlockStage
[tid
] = 0;
51 squashSeqNum
[tid
] = (InstSeqNum
)-1;
52 lastSquashCycle
[tid
] = 0;
59 resourceEvent
= new FetchSeqEvent
[width
];
65 FetchSeqUnit::execute(int slot_num
)
67 // After this is working, change this to a reinterpret cast
68 // for performance considerations
69 ResourceRequest
* fs_req
= reqMap
[slot_num
];
70 DynInstPtr inst
= fs_req
->inst
;
71 int tid
= inst
->readTid();
72 int stage_num
= fs_req
->getStageNum();
73 int seq_num
= inst
->seqNum
;
75 fs_req
->fault
= NoFault
;
83 if (delaySlotInfo
[tid
].targetReady
&&
84 delaySlotInfo
[tid
].numInsts
== 0) {
86 PC
[tid
] = delaySlotInfo
[tid
].targetAddr
; //next_PC
87 nextPC
[tid
] = PC
[tid
] + instSize
; //next_NPC
88 nextNPC
[tid
] = PC
[tid
] + (2 * instSize
);
90 delaySlotInfo
[tid
].targetReady
= false;
92 DPRINTF(InOrderFetchSeq
, "[tid:%i]: Setting PC to delay slot target\n",tid
);
96 inst
->setNextPC(PC
[tid
] + instSize
);
97 inst
->setNextNPC(PC
[tid
] + (instSize
* 2));
99 inst
->setPredTarg(inst
->readNextNPC());
101 inst
->setMemAddr(PC
[tid
]);
102 inst
->setSeqNum(cpu
->getAndIncrementInstSeq(tid
));
104 DPRINTF(InOrderFetchSeq
, "[tid:%i]: Assigning [sn:%i] to PC %08p\n", tid
,
105 inst
->seqNum
, inst
->readPC());
107 if (delaySlotInfo
[tid
].numInsts
> 0) {
108 --delaySlotInfo
[tid
].numInsts
;
110 // It's OK to set PC to target of branch
111 if (delaySlotInfo
[tid
].numInsts
== 0) {
112 delaySlotInfo
[tid
].targetReady
= true;
115 DPRINTF(InOrderFetchSeq
, "[tid:%i]: %i delay slot inst(s) left to"
116 " process.\n", tid
, delaySlotInfo
[tid
].numInsts
);
119 PC
[tid
] = nextPC
[tid
];
120 nextPC
[tid
] = nextNPC
[tid
];
121 nextNPC
[tid
] += instSize
;
125 DPRINTF(InOrderStall
, "STALL: [tid:%i]: NPC not valid\n", tid
);
126 fs_req
->setCompleted(false);
133 if (inst
->isControl()) {
134 // If it's a return, then we must wait for resolved address.
135 if (inst
->isReturn() && !inst
->predTaken()) {
136 cpu
->pipelineStage
[stage_num
]->toPrevStages
->stageBlock
[stage_num
][tid
] = true;
137 pcValid
[tid
] = false;
138 pcBlockStage
[tid
] = stage_num
;
139 } else if (inst
->isCondDelaySlot() && !inst
->predTaken()) {
140 // Not-Taken AND Conditional Control
141 DPRINTF(InOrderFetchSeq
, "[tid:%i]: [sn:%i]: [PC:%08p] Predicted Not-Taken Cond. "
142 "Delay inst. Skipping delay slot and Updating PC to %08p\n",
143 tid
, inst
->seqNum
, inst
->readPC(), inst
->readPredTarg());
145 DPRINTF(InOrderFetchSeq
, "[tid:%i] Setting up squash to start from stage %i, after [sn:%i].\n",
146 tid
, stage_num
, seq_num
);
148 inst
->bdelaySeqNum
= seq_num
;
149 inst
->squashingStage
= stage_num
;
151 squashAfterInst(inst
, stage_num
, tid
);
152 } else if (!inst
->isCondDelaySlot() && !inst
->predTaken()) {
154 DPRINTF(InOrderFetchSeq
, "[tid:%i]: [sn:%i]: Predicted Not-Taken Control "
155 "inst. updating PC to %08p\n", tid
, inst
->seqNum
,
158 ++delaySlotInfo
[tid
].numInsts
;
159 delaySlotInfo
[tid
].targetReady
= false;
160 delaySlotInfo
[tid
].targetAddr
= inst
->readNextNPC();
162 } else if (inst
->predTaken()) {
164 ++delaySlotInfo
[tid
].numInsts
;
165 delaySlotInfo
[tid
].targetReady
= false;
166 delaySlotInfo
[tid
].targetAddr
= inst
->readPredTarg();
168 DPRINTF(InOrderFetchSeq
, "[tid:%i]: [sn:%i] Updating delay slot target "
169 "to PC %08p\n", tid
, inst
->seqNum
, inst
->readPredTarg());
171 // Set-Up Squash Through-Out Pipeline
172 DPRINTF(InOrderFetchSeq
, "[tid:%i] Setting up squash to start from stage %i, after [sn:%i].\n",
173 tid
, stage_num
, seq_num
+ 1);
174 inst
->bdelaySeqNum
= seq_num
+ 1;
175 inst
->squashingStage
= stage_num
;
178 squashAfterInst(inst
, stage_num
, tid
);
181 DPRINTF(InOrderFetchSeq
, "[tid:%i]: [sn:%i]: Ignoring branch target update "
182 "since then is not a control instruction.\n", tid
, inst
->seqNum
);
190 fatal("Unrecognized command to %s", resName
);
195 FetchSeqUnit::squashAfterInst(DynInstPtr inst
, int stage_num
, unsigned tid
)
197 // Squash In Pipeline Stage
198 cpu
->pipelineStage
[stage_num
]->squashDueToBranch(inst
, tid
);
200 // Squash inside current resource, so if there needs to be fetching on same cycle
201 // the fetch information will be correct.
202 // squash(inst, stage_num, inst->bdelaySeqNum, tid);
204 // Schedule Squash Through-out Resource Pool
205 cpu
->resPool
->scheduleEvent((InOrderCPU::CPUEventType
)ResourcePool::SquashAll
, inst
, 0);
208 FetchSeqUnit::squash(DynInstPtr inst
, int squash_stage
,
209 InstSeqNum squash_seq_num
, unsigned tid
)
211 DPRINTF(InOrderFetchSeq
, "[tid:%i]: Updating due to squash from stage %i.\n",
214 InstSeqNum done_seq_num
= inst
->bdelaySeqNum
;
215 Addr new_PC
= inst
->readPredTarg();
217 if (squashSeqNum
[tid
] <= done_seq_num
&&
218 lastSquashCycle
[tid
] == curTick
) {
219 DPRINTF(InOrderFetchSeq
, "[tid:%i]: Ignoring squash from stage %i, since"
220 "there is an outstanding squash that is older.\n",
223 squashSeqNum
[tid
] = done_seq_num
;
224 lastSquashCycle
[tid
] = curTick
;
226 // If The very next instruction number is the done seq. num,
227 // then we haven't seen the delay slot yet ... if it isn't
228 // the last done_seq_num then this is the delay slot inst.
229 if (cpu
->nextInstSeqNum(tid
) != done_seq_num
&&
230 !inst
->procDelaySlotOnMispred
) {
231 delaySlotInfo
[tid
].numInsts
= 0;
232 delaySlotInfo
[tid
].targetReady
= false;
236 nextPC
[tid
] = new_PC
+ instSize
;
237 nextNPC
[tid
] = new_PC
+ (2 * instSize
);
239 DPRINTF(InOrderFetchSeq
, "[tid:%i]: Setting PC to %08p.\n",
242 delaySlotInfo
[tid
].numInsts
= 1;
243 delaySlotInfo
[tid
].targetReady
= false;
244 delaySlotInfo
[tid
].targetAddr
= (inst
->procDelaySlotOnMispred
) ? inst
->branchTarget() : new_PC
;
246 // Reset PC to Delay Slot Instruction
247 if (inst
->procDelaySlotOnMispred
) {
249 nextPC
[tid
] = new_PC
+ instSize
;
250 nextNPC
[tid
] = new_PC
+ (2 * instSize
);
255 // Unblock Any Stages Waiting for this information to be updated ...
257 cpu
->pipelineStage
[pcBlockStage
[tid
]]->toPrevStages
->stageUnblock
[pcBlockStage
[tid
]][tid
] = true;
263 Resource::squash(inst
, squash_stage
, squash_seq_num
, tid
);
266 FetchSeqUnit::FetchSeqEvent::FetchSeqEvent()
271 FetchSeqUnit::FetchSeqEvent::process()
273 FetchSeqUnit
* fs_res
= dynamic_cast<FetchSeqUnit
*>(resource
);
276 for (int i
=0; i
< MaxThreads
; i
++) {
277 fs_res
->PC
[i
] = fs_res
->cpu
->readPC(i
);
278 fs_res
->nextPC
[i
] = fs_res
->cpu
->readNextPC(i
);
279 fs_res
->nextNPC
[i
] = fs_res
->cpu
->readNextNPC(i
);
280 DPRINTF(InOrderFetchSeq
, "[tid:%i]: Setting PC:%08p NPC:%08p NNPC:%08p.\n",
281 fs_res
->PC
[i
], fs_res
->nextPC
[i
], fs_res
->nextNPC
[i
]);
283 fs_res
->pcValid
[i
] = true;
286 //cpu->fetchPriorityList.push_back(tid);
291 FetchSeqUnit::activateThread(unsigned tid
)
295 PC
[tid
] = cpu
->readPC(tid
);
296 nextPC
[tid
] = cpu
->readNextPC(tid
);
297 nextNPC
[tid
] = cpu
->readNextNPC(tid
);
299 cpu
->fetchPriorityList
.push_back(tid
);
301 DPRINTF(InOrderFetchSeq
, "[tid:%i]: Reading PC:%08p NPC:%08p NNPC:%08p.\n",
302 tid
, PC
[tid
], nextPC
[tid
], nextNPC
[tid
]);
306 FetchSeqUnit::deactivateThread(unsigned tid
)
308 delaySlotInfo
[tid
].numInsts
= 0;
309 delaySlotInfo
[tid
].targetReady
= false;
311 pcValid
[tid
] = false;
312 pcBlockStage
[tid
] = 0;
314 squashSeqNum
[tid
] = (InstSeqNum
)-1;
315 lastSquashCycle
[tid
] = 0;
317 std::list
<unsigned>::iterator thread_it
= find(cpu
->fetchPriorityList
.begin(),
318 cpu
->fetchPriorityList
.end(),
321 if (thread_it
!= cpu
->fetchPriorityList
.end())
322 cpu
->fetchPriorityList
.erase(thread_it
);