1 | //===- LiveRegMatrix.cpp - Track register interference --------------------===// |
2 | // |
3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
4 | // See https://llvm.org/LICENSE.txt for license information. |
5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
6 | // |
7 | //===----------------------------------------------------------------------===// |
8 | // |
9 | // This file defines the LiveRegMatrix analysis pass. |
10 | // |
11 | //===----------------------------------------------------------------------===// |
12 | |
13 | #include "llvm/CodeGen/LiveRegMatrix.h" |
14 | #include "RegisterCoalescer.h" |
15 | #include "llvm/ADT/Statistic.h" |
16 | #include "llvm/CodeGen/LiveInterval.h" |
17 | #include "llvm/CodeGen/LiveIntervalUnion.h" |
18 | #include "llvm/CodeGen/LiveIntervals.h" |
19 | #include "llvm/CodeGen/MachineFunction.h" |
20 | #include "llvm/CodeGen/TargetRegisterInfo.h" |
21 | #include "llvm/CodeGen/TargetSubtargetInfo.h" |
22 | #include "llvm/CodeGen/VirtRegMap.h" |
23 | #include "llvm/InitializePasses.h" |
24 | #include "llvm/MC/LaneBitmask.h" |
25 | #include "llvm/MC/MCRegisterInfo.h" |
26 | #include "llvm/Pass.h" |
27 | #include "llvm/Support/Debug.h" |
28 | #include "llvm/Support/raw_ostream.h" |
29 | #include <cassert> |
30 | |
31 | using namespace llvm; |
32 | |
33 | #define DEBUG_TYPE "regalloc" |
34 | |
35 | STATISTIC(NumAssigned , "Number of registers assigned" ); |
36 | STATISTIC(NumUnassigned , "Number of registers unassigned" ); |
37 | |
38 | char LiveRegMatrixWrapperLegacy::ID = 0; |
39 | INITIALIZE_PASS_BEGIN(LiveRegMatrixWrapperLegacy, "liveregmatrix" , |
40 | "Live Register Matrix" , false, false) |
41 | INITIALIZE_PASS_DEPENDENCY(LiveIntervalsWrapperPass) |
42 | INITIALIZE_PASS_DEPENDENCY(VirtRegMapWrapperLegacy) |
43 | INITIALIZE_PASS_END(LiveRegMatrixWrapperLegacy, "liveregmatrix" , |
44 | "Live Register Matrix" , false, true) |
45 | |
46 | void LiveRegMatrixWrapperLegacy::getAnalysisUsage(AnalysisUsage &AU) const { |
47 | AU.setPreservesAll(); |
48 | AU.addRequiredTransitive<LiveIntervalsWrapperPass>(); |
49 | AU.addRequiredTransitive<VirtRegMapWrapperLegacy>(); |
50 | MachineFunctionPass::getAnalysisUsage(AU); |
51 | } |
52 | |
53 | bool LiveRegMatrixWrapperLegacy::runOnMachineFunction(MachineFunction &MF) { |
54 | auto &LIS = getAnalysis<LiveIntervalsWrapperPass>().getLIS(); |
55 | auto &VRM = getAnalysis<VirtRegMapWrapperLegacy>().getVRM(); |
56 | LRM.init(MF, LIS, VRM); |
57 | return false; |
58 | } |
59 | |
60 | void LiveRegMatrix::init(MachineFunction &MF, LiveIntervals &pLIS, |
61 | VirtRegMap &pVRM) { |
62 | TRI = MF.getSubtarget().getRegisterInfo(); |
63 | LIS = &pLIS; |
64 | VRM = &pVRM; |
65 | |
66 | unsigned NumRegUnits = TRI->getNumRegUnits(); |
67 | if (NumRegUnits != Matrix.size()) |
68 | Queries.reset(p: new LiveIntervalUnion::Query[NumRegUnits]); |
69 | Matrix.init(*LIUAlloc, Size: NumRegUnits); |
70 | |
71 | // Make sure no stale queries get reused. |
72 | invalidateVirtRegs(); |
73 | } |
74 | |
75 | void LiveRegMatrixWrapperLegacy::releaseMemory() { LRM.releaseMemory(); } |
76 | |
77 | void LiveRegMatrix::releaseMemory() { |
78 | for (unsigned i = 0, e = Matrix.size(); i != e; ++i) { |
79 | Matrix[i].clear(); |
80 | // No need to clear Queries here, since LiveIntervalUnion::Query doesn't |
81 | // have anything important to clear and LiveRegMatrix's runOnFunction() |
82 | // does a std::unique_ptr::reset anyways. |
83 | } |
84 | } |
85 | |
86 | template <typename Callable> |
87 | static bool foreachUnit(const TargetRegisterInfo *TRI, |
88 | const LiveInterval &VRegInterval, MCRegister PhysReg, |
89 | Callable Func) { |
90 | if (VRegInterval.hasSubRanges()) { |
91 | for (MCRegUnitMaskIterator Units(PhysReg, TRI); Units.isValid(); ++Units) { |
92 | unsigned Unit = (*Units).first; |
93 | LaneBitmask Mask = (*Units).second; |
94 | for (const LiveInterval::SubRange &S : VRegInterval.subranges()) { |
95 | if ((S.LaneMask & Mask).any()) { |
96 | if (Func(Unit, S)) |
97 | return true; |
98 | break; |
99 | } |
100 | } |
101 | } |
102 | } else { |
103 | for (MCRegUnit Unit : TRI->regunits(Reg: PhysReg)) { |
104 | if (Func(Unit, VRegInterval)) |
105 | return true; |
106 | } |
107 | } |
108 | return false; |
109 | } |
110 | |
111 | void LiveRegMatrix::assign(const LiveInterval &VirtReg, MCRegister PhysReg) { |
112 | LLVM_DEBUG(dbgs() << "assigning " << printReg(VirtReg.reg(), TRI) << " to " |
113 | << printReg(PhysReg, TRI) << ':'); |
114 | assert(!VRM->hasPhys(VirtReg.reg()) && "Duplicate VirtReg assignment" ); |
115 | VRM->assignVirt2Phys(virtReg: VirtReg.reg(), physReg: PhysReg); |
116 | |
117 | foreachUnit( |
118 | TRI, VRegInterval: VirtReg, PhysReg, Func: [&](unsigned Unit, const LiveRange &Range) { |
119 | LLVM_DEBUG(dbgs() << ' ' << printRegUnit(Unit, TRI) << ' ' << Range); |
120 | Matrix[Unit].unify(VirtReg, Range); |
121 | return false; |
122 | }); |
123 | |
124 | ++NumAssigned; |
125 | LLVM_DEBUG(dbgs() << '\n'); |
126 | } |
127 | |
128 | void LiveRegMatrix::unassign(const LiveInterval &VirtReg) { |
129 | Register PhysReg = VRM->getPhys(virtReg: VirtReg.reg()); |
130 | LLVM_DEBUG(dbgs() << "unassigning " << printReg(VirtReg.reg(), TRI) |
131 | << " from " << printReg(PhysReg, TRI) << ':'); |
132 | VRM->clearVirt(virtReg: VirtReg.reg()); |
133 | |
134 | foreachUnit(TRI, VRegInterval: VirtReg, PhysReg, |
135 | Func: [&](unsigned Unit, const LiveRange &Range) { |
136 | LLVM_DEBUG(dbgs() << ' ' << printRegUnit(Unit, TRI)); |
137 | Matrix[Unit].extract(VirtReg, Range); |
138 | return false; |
139 | }); |
140 | |
141 | ++NumUnassigned; |
142 | LLVM_DEBUG(dbgs() << '\n'); |
143 | } |
144 | |
145 | bool LiveRegMatrix::isPhysRegUsed(MCRegister PhysReg) const { |
146 | for (MCRegUnit Unit : TRI->regunits(Reg: PhysReg)) { |
147 | if (!Matrix[Unit].empty()) |
148 | return true; |
149 | } |
150 | return false; |
151 | } |
152 | |
153 | bool LiveRegMatrix::checkRegMaskInterference(const LiveInterval &VirtReg, |
154 | MCRegister PhysReg) { |
155 | // Check if the cached information is valid. |
156 | // The same BitVector can be reused for all PhysRegs. |
157 | // We could cache multiple VirtRegs if it becomes necessary. |
158 | if (RegMaskVirtReg != VirtReg.reg() || RegMaskTag != UserTag) { |
159 | RegMaskVirtReg = VirtReg.reg(); |
160 | RegMaskTag = UserTag; |
161 | RegMaskUsable.clear(); |
162 | LIS->checkRegMaskInterference(LI: VirtReg, UsableRegs&: RegMaskUsable); |
163 | } |
164 | |
165 | // The BitVector is indexed by PhysReg, not register unit. |
166 | // Regmask interference is more fine grained than regunits. |
167 | // For example, a Win64 call can clobber %ymm8 yet preserve %xmm8. |
168 | return !RegMaskUsable.empty() && |
169 | (!PhysReg || !RegMaskUsable.test(Idx: PhysReg.id())); |
170 | } |
171 | |
172 | bool LiveRegMatrix::checkRegUnitInterference(const LiveInterval &VirtReg, |
173 | MCRegister PhysReg) { |
174 | if (VirtReg.empty()) |
175 | return false; |
176 | CoalescerPair CP(VirtReg.reg(), PhysReg, *TRI); |
177 | |
178 | bool Result = foreachUnit(TRI, VRegInterval: VirtReg, PhysReg, Func: [&](unsigned Unit, |
179 | const LiveRange &Range) { |
180 | const LiveRange &UnitRange = LIS->getRegUnit(Unit); |
181 | return Range.overlaps(Other: UnitRange, CP, *LIS->getSlotIndexes()); |
182 | }); |
183 | return Result; |
184 | } |
185 | |
186 | LiveIntervalUnion::Query &LiveRegMatrix::query(const LiveRange &LR, |
187 | MCRegUnit RegUnit) { |
188 | LiveIntervalUnion::Query &Q = Queries[RegUnit]; |
189 | Q.init(NewUserTag: UserTag, NewLR: LR, NewLiveUnion: Matrix[RegUnit]); |
190 | return Q; |
191 | } |
192 | |
193 | LiveRegMatrix::InterferenceKind |
194 | LiveRegMatrix::checkInterference(const LiveInterval &VirtReg, |
195 | MCRegister PhysReg) { |
196 | if (VirtReg.empty()) |
197 | return IK_Free; |
198 | |
199 | // Regmask interference is the fastest check. |
200 | if (checkRegMaskInterference(VirtReg, PhysReg)) |
201 | return IK_RegMask; |
202 | |
203 | // Check for fixed interference. |
204 | if (checkRegUnitInterference(VirtReg, PhysReg)) |
205 | return IK_RegUnit; |
206 | |
207 | // Check the matrix for virtual register interference. |
208 | bool Interference = foreachUnit(TRI, VRegInterval: VirtReg, PhysReg, |
209 | Func: [&](MCRegUnit Unit, const LiveRange &LR) { |
210 | return query(LR, RegUnit: Unit).checkInterference(); |
211 | }); |
212 | if (Interference) |
213 | return IK_VirtReg; |
214 | |
215 | return IK_Free; |
216 | } |
217 | |
218 | bool LiveRegMatrix::checkInterference(SlotIndex Start, SlotIndex End, |
219 | MCRegister PhysReg) { |
220 | // Construct artificial live range containing only one segment [Start, End). |
221 | VNInfo valno(0, Start); |
222 | LiveRange::Segment Seg(Start, End, &valno); |
223 | LiveRange LR; |
224 | LR.addSegment(S: Seg); |
225 | |
226 | // Check for interference with that segment |
227 | for (MCRegUnit Unit : TRI->regunits(Reg: PhysReg)) { |
228 | // LR is stack-allocated. LiveRegMatrix caches queries by a key that |
229 | // includes the address of the live range. If (for the same reg unit) this |
230 | // checkInterference overload is called twice, without any other query() |
231 | // calls in between (on heap-allocated LiveRanges) - which would invalidate |
232 | // the cached query - the LR address seen the second time may well be the |
233 | // same as that seen the first time, while the Start/End/valno may not - yet |
234 | // the same cached result would be fetched. To avoid that, we don't cache |
235 | // this query. |
236 | // |
237 | // FIXME: the usability of the Query API needs to be improved to avoid |
238 | // subtle bugs due to query identity. Avoiding caching, for example, would |
239 | // greatly simplify things. |
240 | LiveIntervalUnion::Query Q; |
241 | Q.reset(NewUserTag: UserTag, NewLR: LR, NewLiveUnion: Matrix[Unit]); |
242 | if (Q.checkInterference()) |
243 | return true; |
244 | } |
245 | return false; |
246 | } |
247 | |
248 | LaneBitmask LiveRegMatrix::checkInterferenceLanes(SlotIndex Start, |
249 | SlotIndex End, |
250 | MCRegister PhysReg) { |
251 | // Construct artificial live range containing only one segment [Start, End). |
252 | VNInfo valno(0, Start); |
253 | LiveRange::Segment Seg(Start, End, &valno); |
254 | LiveRange LR; |
255 | LR.addSegment(S: Seg); |
256 | |
257 | LaneBitmask InterferingLanes; |
258 | |
259 | // Check for interference with that segment |
260 | for (MCRegUnitMaskIterator MCRU(PhysReg, TRI); MCRU.isValid(); ++MCRU) { |
261 | auto [Unit, Lanes] = *MCRU; |
262 | // LR is stack-allocated. LiveRegMatrix caches queries by a key that |
263 | // includes the address of the live range. If (for the same reg unit) this |
264 | // checkInterference overload is called twice, without any other query() |
265 | // calls in between (on heap-allocated LiveRanges) - which would invalidate |
266 | // the cached query - the LR address seen the second time may well be the |
267 | // same as that seen the first time, while the Start/End/valno may not - yet |
268 | // the same cached result would be fetched. To avoid that, we don't cache |
269 | // this query. |
270 | // |
271 | // FIXME: the usability of the Query API needs to be improved to avoid |
272 | // subtle bugs due to query identity. Avoiding caching, for example, would |
273 | // greatly simplify things. |
274 | LiveIntervalUnion::Query Q; |
275 | Q.reset(NewUserTag: UserTag, NewLR: LR, NewLiveUnion: Matrix[Unit]); |
276 | if (Q.checkInterference()) |
277 | InterferingLanes |= Lanes; |
278 | } |
279 | |
280 | return InterferingLanes; |
281 | } |
282 | |
283 | Register LiveRegMatrix::getOneVReg(unsigned PhysReg) const { |
284 | const LiveInterval *VRegInterval = nullptr; |
285 | for (MCRegUnit Unit : TRI->regunits(Reg: PhysReg)) { |
286 | if ((VRegInterval = Matrix[Unit].getOneVReg())) |
287 | return VRegInterval->reg(); |
288 | } |
289 | |
290 | return MCRegister::NoRegister; |
291 | } |
292 | |
293 | AnalysisKey LiveRegMatrixAnalysis::Key; |
294 | |
295 | LiveRegMatrix LiveRegMatrixAnalysis::run(MachineFunction &MF, |
296 | MachineFunctionAnalysisManager &MFAM) { |
297 | auto &LIS = MFAM.getResult<LiveIntervalsAnalysis>(IR&: MF); |
298 | auto &VRM = MFAM.getResult<VirtRegMapAnalysis>(IR&: MF); |
299 | LiveRegMatrix LRM; |
300 | LRM.init(MF, pLIS&: LIS, pVRM&: VRM); |
301 | return LRM; |
302 | } |
303 | |