decode.cc revision 11567:560d7fbbddd1 
1/*
2 * Copyright (c) 2013-2014 ARM Limited
3 * All rights reserved
4 *
5 * The license below extends only to copyright in the software and shall
6 * not be construed as granting a license to any other intellectual
7 * property including but not limited to intellectual property relating
8 * to a hardware implementation of the functionality of the software
9 * licensed hereunder.  You may use the software subject to the license
10 * terms below provided that you ensure that this notice is replicated
11 * unmodified and in its entirety in all distributions of the software,
12 * modified or unmodified, in source code or in binary form.
13 *
14 * Redistribution and use in source and binary forms, with or without
15 * modification, are permitted provided that the following conditions are
16 * met: redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer;
18 * redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution;
21 * neither the name of the copyright holders nor the names of its
22 * contributors may be used to endorse or promote products derived from
23 * this software without specific prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
26 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
27 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
28 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
29 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
30 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
31 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
32 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
33 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
34 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
35 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36 *
37 * Authors: Andrew Bardsley
38 */
39
40#include "cpu/minor/decode.hh"
41#include "cpu/minor/pipeline.hh"
42#include "debug/Decode.hh"
43
44namespace Minor
45{
46
47Decode::Decode(const std::string &name,
48    MinorCPU &cpu_,
49    MinorCPUParams &params,
50    Latch<ForwardInstData>::Output inp_,
51    Latch<ForwardInstData>::Input out_,
52    std::vector<InputBuffer<ForwardInstData>> &next_stage_input_buffer) :
53    Named(name),
54    cpu(cpu_),
55    inp(inp_),
56    out(out_),
57    nextStageReserve(next_stage_input_buffer),
58    outputWidth(params.executeInputWidth),
59    processMoreThanOneInput(params.decodeCycleInput),
60    decodeInfo(params.numThreads),
61    threadPriority(0)
62{
63    if (outputWidth < 1)
64        fatal("%s: executeInputWidth must be >= 1 (%d)\n", name, outputWidth);
65
66    if (params.decodeInputBufferSize < 1) {
67        fatal("%s: decodeInputBufferSize must be >= 1 (%d)\n", name,
68        params.decodeInputBufferSize);
69    }
70
71    /* Per-thread input buffers */
72    for (ThreadID tid = 0; tid < params.numThreads; tid++) {
73        inputBuffer.push_back(
74            InputBuffer<ForwardInstData>(
75                name + ".inputBuffer" + std::to_string(tid), "insts",
76                params.decodeInputBufferSize));
77    }
78}
79
80const ForwardInstData *
81Decode::getInput(ThreadID tid)
82{
83    /* Get insts from the inputBuffer to work with */
84    if (!inputBuffer[tid].empty()) {
85        const ForwardInstData &head = inputBuffer[tid].front();
86
87        return (head.isBubble() ? NULL : &(inputBuffer[tid].front()));
88    } else {
89        return NULL;
90    }
91}
92
93void
94Decode::popInput(ThreadID tid)
95{
96    if (!inputBuffer[tid].empty())
97        inputBuffer[tid].pop();
98
99    decodeInfo[tid].inputIndex = 0;
100    decodeInfo[tid].inMacroop = false;
101}
102
103#if TRACING_ON
104/** Add the tracing data to an instruction.  This originates in
105 *  decode because this is the first place that execSeqNums are known
106 *  (these are used as the 'FetchSeq' in tracing data) */
107static void
108dynInstAddTracing(MinorDynInstPtr inst, StaticInstPtr static_inst,
109    MinorCPU &cpu)
110{
111    inst->traceData = cpu.getTracer()->getInstRecord(curTick(),
112        cpu.getContext(inst->id.threadId),
113        inst->staticInst, inst->pc, static_inst);
114
115    /* Use the execSeqNum as the fetch sequence number as this most closely
116     *  matches the other processor models' idea of fetch sequence */
117    if (inst->traceData)
118        inst->traceData->setFetchSeq(inst->id.execSeqNum);
119}
120#endif
121
122void
123Decode::evaluate()
124{
125    /* Push input onto appropriate input buffer */
126    if (!inp.outputWire->isBubble())
127        inputBuffer[inp.outputWire->threadId].setTail(*inp.outputWire);
128
129    ForwardInstData &insts_out = *out.inputWire;
130
131    assert(insts_out.isBubble());
132
133    for (ThreadID tid = 0; tid < cpu.numThreads; tid++)
134        decodeInfo[tid].blocked = !nextStageReserve[tid].canReserve();
135
136    ThreadID tid = getScheduledThread();
137
138    if (tid != InvalidThreadID) {
139        DecodeThreadInfo &decode_info = decodeInfo[tid];
140        const ForwardInstData *insts_in = getInput(tid);
141
142        unsigned int output_index = 0;
143
144        /* Pack instructions into the output while we can.  This may involve
145         * using more than one input line */
146        while (insts_in &&
147           decode_info.inputIndex < insts_in->width() && /* Still more input */
148           output_index < outputWidth /* Still more output to fill */)
149        {
150            MinorDynInstPtr inst = insts_in->insts[decode_info.inputIndex];
151
152            if (inst->isBubble()) {
153                /* Skip */
154                decode_info.inputIndex++;
155                decode_info.inMacroop = false;
156            } else {
157                StaticInstPtr static_inst = inst->staticInst;
158                /* Static inst of a macro-op above the output_inst */
159                StaticInstPtr parent_static_inst = NULL;
160                MinorDynInstPtr output_inst = inst;
161
162                if (inst->isFault()) {
163                    DPRINTF(Decode, "Fault being passed: %d\n",
164                        inst->fault->name());
165
166                    decode_info.inputIndex++;
167                    decode_info.inMacroop = false;
168                } else if (static_inst->isMacroop()) {
169                    /* Generate a new micro-op */
170                    StaticInstPtr static_micro_inst;
171
172                    /* Set up PC for the next micro-op emitted */
173                    if (!decode_info.inMacroop) {
174                        decode_info.microopPC = inst->pc;
175                        decode_info.inMacroop = true;
176                    }
177
178                    /* Get the micro-op static instruction from the
179                     * static_inst. */
180                    static_micro_inst =
181                        static_inst->fetchMicroop(
182                                decode_info.microopPC.microPC());
183
184                    output_inst = new MinorDynInst(inst->id);
185                    output_inst->pc = decode_info.microopPC;
186                    output_inst->staticInst = static_micro_inst;
187                    output_inst->fault = NoFault;
188
189                    /* Allow a predicted next address only on the last
190                     *  microop */
191                    if (static_micro_inst->isLastMicroop()) {
192                        output_inst->predictedTaken = inst->predictedTaken;
193                        output_inst->predictedTarget = inst->predictedTarget;
194                    }
195
196                    DPRINTF(Decode, "Microop decomposition inputIndex:"
197                        " %d output_index: %d lastMicroop: %s microopPC:"
198                        " %d.%d inst: %d\n",
199                        decode_info.inputIndex, output_index,
200                        (static_micro_inst->isLastMicroop() ?
201                            "true" : "false"),
202                        decode_info.microopPC.instAddr(),
203                        decode_info.microopPC.microPC(),
204                        *output_inst);
205
206                    /* Acknowledge that the static_inst isn't mine, it's my
207                     * parent macro-op's */
208                    parent_static_inst = static_inst;
209
210                    static_micro_inst->advancePC(decode_info.microopPC);
211
212                    /* Step input if this is the last micro-op */
213                    if (static_micro_inst->isLastMicroop()) {
214                        decode_info.inputIndex++;
215                        decode_info.inMacroop = false;
216                    }
217                } else {
218                    /* Doesn't need decomposing, pass on instruction */
219                    DPRINTF(Decode, "Passing on inst: %s inputIndex:"
220                        " %d output_index: %d\n",
221                        *output_inst, decode_info.inputIndex, output_index);
222
223                    parent_static_inst = static_inst;
224
225                    /* Step input */
226                    decode_info.inputIndex++;
227                    decode_info.inMacroop = false;
228                }
229
230                /* Set execSeqNum of output_inst */
231                output_inst->id.execSeqNum = decode_info.execSeqNum;
232                /* Add tracing */
233#if TRACING_ON
234                dynInstAddTracing(output_inst, parent_static_inst, cpu);
235#endif
236
237                /* Step to next sequence number */
238                decode_info.execSeqNum++;
239
240                /* Correctly size the output before writing */
241                if (output_index == 0) insts_out.resize(outputWidth);
242                /* Push into output */
243                insts_out.insts[output_index] = output_inst;
244                output_index++;
245            }
246
247            /* Have we finished with the input? */
248            if (decode_info.inputIndex == insts_in->width()) {
249                /* If we have just been producing micro-ops, we *must* have
250                 * got to the end of that for inputIndex to be pushed past
251                 * insts_in->width() */
252                assert(!decode_info.inMacroop);
253                popInput(tid);
254                insts_in = NULL;
255
256                if (processMoreThanOneInput) {
257                    DPRINTF(Decode, "Wrapping\n");
258                    insts_in = getInput(tid);
259                }
260            }
261        }
262
263        /* The rest of the output (if any) should already have been packed
264         *  with bubble instructions by insts_out's initialisation
265         *
266         *  for (; output_index < outputWidth; output_index++)
267         *      assert(insts_out.insts[output_index]->isBubble());
268         */
269    }
270
271    /* If we generated output, reserve space for the result in the next stage
272     *  and mark the stage as being active this cycle */
273    if (!insts_out.isBubble()) {
274        /* Note activity of following buffer */
275        cpu.activityRecorder->activity();
276        insts_out.threadId = tid;
277        nextStageReserve[tid].reserve();
278    }
279
280    /* If we still have input to process and somewhere to put it,
281     *  mark stage as active */
282    for (ThreadID i = 0; i < cpu.numThreads; i++)
283    {
284        if (getInput(i) && nextStageReserve[i].canReserve()) {
285            cpu.activityRecorder->activateStage(Pipeline::DecodeStageId);
286            break;
287        }
288    }
289
290    /* Make sure the input (if any left) is pushed */
291    if (!inp.outputWire->isBubble())
292        inputBuffer[inp.outputWire->threadId].pushTail();
293}
294
295inline ThreadID
296Decode::getScheduledThread()
297{
298    /* Select thread via policy. */
299    std::vector<ThreadID> priority_list;
300
301    switch (cpu.threadPolicy) {
302      case Enums::SingleThreaded:
303        priority_list.push_back(0);
304        break;
305      case Enums::RoundRobin:
306        priority_list = cpu.roundRobinPriority(threadPriority);
307        break;
308      case Enums::Random:
309        priority_list = cpu.randomPriority();
310        break;
311      default:
312        panic("Unknown fetch policy");
313    }
314
315    for (auto tid : priority_list) {
316        if (cpu.getContext(tid)->status() == ThreadContext::Active &&
317            getInput(tid) && !decodeInfo[tid].blocked) {
318            threadPriority = tid;
319            return tid;
320        }
321    }
322
323   return InvalidThreadID;
324}
325
326bool
327Decode::isDrained()
328{
329    for (const auto &buffer : inputBuffer) {
330        if (!buffer.empty())
331            return false;
332    }
333
334    return (*inp.outputWire).isBubble();
335}
336
337void
338Decode::minorTrace() const
339{
340    std::ostringstream data;
341
342    if (decodeInfo[0].blocked)
343        data << 'B';
344    else
345        (*out.inputWire).reportData(data);
346
347    MINORTRACE("insts=%s\n", data.str());
348    inputBuffer[0].minorTrace();
349}
350
351}
352