Line data Source code
1 : // Copyright 2014 the V8 project authors. All rights reserved.
2 : // Use of this source code is governed by a BSD-style license that can be
3 : // found in the LICENSE file.
4 :
5 : #include "src/compiler/backend/gap-resolver.h"
6 :
7 : #include <algorithm>
8 : #include <set>
9 :
10 : #include "src/base/enum-set.h"
11 : #include "src/register-configuration.h"
12 :
13 : namespace v8 {
14 : namespace internal {
15 : namespace compiler {
16 :
17 : namespace {
18 :
19 : // Splits a FP move between two location operands into the equivalent series of
20 : // moves between smaller sub-operands, e.g. a double move to two single moves.
21 : // This helps reduce the number of cycles that would normally occur under FP
22 : // aliasing, and makes swaps much easier to implement.
23 : MoveOperands* Split(MoveOperands* move, MachineRepresentation smaller_rep,
24 : ParallelMove* moves) {
25 : DCHECK(!kSimpleFPAliasing);
26 : // Splitting is only possible when the slot size is the same as float size.
27 : DCHECK_EQ(kSystemPointerSize, kFloatSize);
28 : const LocationOperand& src_loc = LocationOperand::cast(move->source());
29 : const LocationOperand& dst_loc = LocationOperand::cast(move->destination());
30 : MachineRepresentation dst_rep = dst_loc.representation();
31 : DCHECK_NE(smaller_rep, dst_rep);
32 : auto src_kind = src_loc.location_kind();
33 : auto dst_kind = dst_loc.location_kind();
34 :
35 : int aliases =
36 : 1 << (ElementSizeLog2Of(dst_rep) - ElementSizeLog2Of(smaller_rep));
37 : int base = -1;
38 : USE(base);
39 : DCHECK_EQ(aliases, RegisterConfiguration::Default()->GetAliases(
40 : dst_rep, 0, smaller_rep, &base));
41 :
42 : int src_index = -1;
43 : int slot_size = (1 << ElementSizeLog2Of(smaller_rep)) / kSystemPointerSize;
44 : int src_step = 1;
45 : if (src_kind == LocationOperand::REGISTER) {
46 : src_index = src_loc.register_code() * aliases;
47 : } else {
48 : src_index = src_loc.index();
49 : // For operands that occupy multiple slots, the index refers to the last
50 : // slot. On little-endian architectures, we start at the high slot and use a
51 : // negative step so that register-to-slot moves are in the correct order.
52 : src_step = -slot_size;
53 : }
54 : int dst_index = -1;
55 : int dst_step = 1;
56 : if (dst_kind == LocationOperand::REGISTER) {
57 : dst_index = dst_loc.register_code() * aliases;
58 : } else {
59 : dst_index = dst_loc.index();
60 : dst_step = -slot_size;
61 : }
62 :
63 : // Reuse 'move' for the first fragment. It is not pending.
64 : move->set_source(AllocatedOperand(src_kind, smaller_rep, src_index));
65 : move->set_destination(AllocatedOperand(dst_kind, smaller_rep, dst_index));
66 : // Add the remaining fragment moves.
67 : for (int i = 1; i < aliases; ++i) {
68 : src_index += src_step;
69 : dst_index += dst_step;
70 : moves->AddMove(AllocatedOperand(src_kind, smaller_rep, src_index),
71 : AllocatedOperand(dst_kind, smaller_rep, dst_index));
72 : }
73 : // Return the first fragment.
74 : return move;
75 : }
76 :
77 : enum MoveOperandKind : uint8_t { kConstant, kGpReg, kFpReg, kStack };
78 :
79 : MoveOperandKind GetKind(const InstructionOperand& move) {
80 76551558 : if (move.IsConstant()) return kConstant;
81 : LocationOperand loc_op = LocationOperand::cast(move);
82 57487727 : if (loc_op.location_kind() != LocationOperand::REGISTER) return kStack;
83 42461519 : return IsFloatingPoint(loc_op.representation()) ? kFpReg : kGpReg;
84 : }
85 :
86 : } // namespace
87 :
88 48889807 : void GapResolver::Resolve(ParallelMove* moves) {
89 : base::EnumSet<MoveOperandKind, uint8_t> source_kinds;
90 : base::EnumSet<MoveOperandKind, uint8_t> destination_kinds;
91 :
92 : // Remove redundant moves, collect source kinds and destination kinds to
93 : // detect simple non-overlapping moves, and collect FP move representations if
94 : // aliasing is non-simple.
95 : int fp_reps = 0;
96 : size_t nmoves = moves->size();
97 106349330 : for (size_t i = 0; i < nmoves;) {
98 57459054 : MoveOperands* move = (*moves)[i];
99 57459054 : if (move->IsRedundant()) {
100 19183744 : nmoves--;
101 28735794 : if (i < nmoves) (*moves)[i] = (*moves)[nmoves];
102 : continue;
103 : }
104 38275779 : i++;
105 : source_kinds.Add(GetKind(move->source()));
106 : destination_kinds.Add(GetKind(move->destination()));
107 : if (!kSimpleFPAliasing && move->destination().IsFPRegister()) {
108 : fp_reps |= RepresentationBit(
109 : LocationOperand::cast(move->destination()).representation());
110 : }
111 : }
112 48890276 : if (nmoves != moves->size()) moves->resize(nmoves);
113 :
114 54791044 : if ((source_kinds & destination_kinds).empty() || moves->size() < 2) {
115 : // Fast path for non-conflicting parallel moves.
116 68702827 : for (MoveOperands* move : *moves) {
117 47715892 : assembler_->AssembleMove(&move->source(), &move->destination());
118 : }
119 : return;
120 : }
121 :
122 : if (!kSimpleFPAliasing) {
123 : if (fp_reps && !base::bits::IsPowerOfTwo(fp_reps)) {
124 : // Start with the smallest FP moves, so we never encounter smaller moves
125 : // in the middle of a cycle of larger moves.
126 : if ((fp_reps & RepresentationBit(MachineRepresentation::kFloat32)) != 0) {
127 : split_rep_ = MachineRepresentation::kFloat32;
128 : for (size_t i = 0; i < moves->size(); ++i) {
129 : auto move = (*moves)[i];
130 : if (!move->IsEliminated() && move->destination().IsFloatRegister())
131 : PerformMove(moves, move);
132 : }
133 : }
134 : if ((fp_reps & RepresentationBit(MachineRepresentation::kFloat64)) != 0) {
135 : split_rep_ = MachineRepresentation::kFloat64;
136 : for (size_t i = 0; i < moves->size(); ++i) {
137 : auto move = (*moves)[i];
138 : if (!move->IsEliminated() && move->destination().IsDoubleRegister())
139 : PerformMove(moves, move);
140 : }
141 : }
142 : }
143 : split_rep_ = MachineRepresentation::kSimd128;
144 : }
145 :
146 32889511 : for (size_t i = 0; i < moves->size(); ++i) {
147 14423000 : auto move = (*moves)[i];
148 14423000 : if (!move->IsEliminated()) PerformMove(moves, move);
149 : }
150 : }
151 :
152 14422982 : void GapResolver::PerformMove(ParallelMove* moves, MoveOperands* move) {
153 : // Each call to this function performs a move and deletes it from the move
154 : // graph. We first recursively perform any move blocking this one. We mark a
155 : // move as "pending" on entry to PerformMove in order to detect cycles in the
156 : // move graph. We use operand swaps to resolve cycles, which means that a
157 : // call to PerformMove could change any source operand in the move graph.
158 : DCHECK(!move->IsPending());
159 : DCHECK(!move->IsRedundant());
160 :
161 : // Clear this move's destination to indicate a pending move. The actual
162 : // destination is saved on the side.
163 14422982 : InstructionOperand source = move->source();
164 : DCHECK(!source.IsInvalid()); // Or else it will look eliminated.
165 14422982 : InstructionOperand destination = move->destination();
166 : move->SetPending();
167 :
168 : // We may need to split moves between FP locations differently.
169 : const bool is_fp_loc_move =
170 : !kSimpleFPAliasing && destination.IsFPLocationOperand();
171 :
172 : // Perform a depth-first traversal of the move graph to resolve dependencies.
173 : // Any unperformed, unpending move with a source the same as this one's
174 : // destination blocks this one so recursively perform all such moves.
175 145899838 : for (size_t i = 0; i < moves->size(); ++i) {
176 65738507 : auto other = (*moves)[i];
177 65738507 : if (other->IsEliminated()) continue;
178 40709316 : if (other->IsPending()) continue;
179 25610692 : if (other->source().InterferesWith(destination)) {
180 : if (is_fp_loc_move &&
181 : LocationOperand::cast(other->source()).representation() >
182 : split_rep_) {
183 : // 'other' must also be an FP location move. Break it into fragments
184 : // of the same size as 'move'. 'other' is set to one of the fragments,
185 : // and the rest are appended to 'moves'.
186 : other = Split(other, split_rep_, moves);
187 : // 'other' may not block destination now.
188 : if (!other->source().InterferesWith(destination)) continue;
189 : }
190 : // Though PerformMove can change any source operand in the move graph,
191 : // this call cannot create a blocking move via a swap (this loop does not
192 : // miss any). Assume there is a non-blocking move with source A and this
193 : // move is blocked on source B and there is a swap of A and B. Then A and
194 : // B must be involved in the same cycle (or they would not be swapped).
195 : // Since this move's destination is B and there is only a single incoming
196 : // edge to an operand, this move must also be involved in the same cycle.
197 : // In that case, the blocking move will be created but will be "pending"
198 : // when we return from PerformMove.
199 586541 : PerformMove(moves, other);
200 : }
201 : }
202 :
203 : // This move's source may have changed due to swaps to resolve cycles and so
204 : // it may now be the last move in the cycle. If so remove it.
205 14422903 : source = move->source();
206 14422782 : if (source.EqualsCanonicalized(destination)) {
207 : move->Eliminate();
208 14341695 : return;
209 : }
210 :
211 : // We are about to resolve this move and don't need it marked as pending, so
212 : // restore its destination.
213 : move->set_destination(destination);
214 :
215 : // The move may be blocked on a (at most one) pending move, in which case we
216 : // have a cycle. Search for such a blocking move and perform a swap to
217 : // resolve it.
218 : auto blocker =
219 : std::find_if(moves->begin(), moves->end(), [&](MoveOperands* move) {
220 104399044 : return !move->IsEliminated() &&
221 79039544 : move->source().InterferesWith(destination);
222 : });
223 14351900 : if (blocker == moves->end()) {
224 : // The easy case: This move is not blocked.
225 14270764 : assembler_->AssembleMove(&source, &destination);
226 : move->Eliminate();
227 : return;
228 : }
229 :
230 : // Ensure source is a register or both are stack slots, to limit swap cases.
231 157443 : if (source.IsStackSlot() || source.IsFPStackSlot()) {
232 : std::swap(source, destination);
233 : }
234 81136 : assembler_->AssembleSwap(&source, &destination);
235 : move->Eliminate();
236 :
237 : // Update outstanding moves whose source may now have been moved.
238 : if (is_fp_loc_move) {
239 : // We may have to split larger moves.
240 : for (size_t i = 0; i < moves->size(); ++i) {
241 : auto other = (*moves)[i];
242 : if (other->IsEliminated()) continue;
243 : if (source.InterferesWith(other->source())) {
244 : if (LocationOperand::cast(other->source()).representation() >
245 : split_rep_) {
246 : other = Split(other, split_rep_, moves);
247 : if (!source.InterferesWith(other->source())) continue;
248 : }
249 : other->set_source(destination);
250 : } else if (destination.InterferesWith(other->source())) {
251 : if (LocationOperand::cast(other->source()).representation() >
252 : split_rep_) {
253 : other = Split(other, split_rep_, moves);
254 : if (!destination.InterferesWith(other->source())) continue;
255 : }
256 : other->set_source(source);
257 : }
258 : }
259 : } else {
260 664509 : for (auto other : *moves) {
261 583373 : if (other->IsEliminated()) continue;
262 317821 : if (source.EqualsCanonicalized(other->source())) {
263 : other->set_source(destination);
264 306240 : } else if (destination.EqualsCanonicalized(other->source())) {
265 : other->set_source(source);
266 : }
267 : }
268 : }
269 : }
270 : } // namespace compiler
271 : } // namespace internal
272 121996 : } // namespace v8
|
