1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
|
/*
* Copyright (c) 2015-2017, Intel Corporation
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* * Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
* * Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* * Neither the name of Intel Corporation nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
* LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
* CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
* SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
* INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
* CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
* POSSIBILITY OF SUCH DAMAGE.
*/
#include "goughcompile.h"
#include "goughcompile_dump.h"
#include "goughcompile_internal.h"
#include "gough_internal.h"
#include "grey.h"
#include "util/container.h"
#include "util/flat_containers.h"
#include "util/graph.h"
#include "util/graph_range.h"
#include "util/order_check.h"
#include "ue2common.h"
#include <algorithm>
#include <boost/graph/depth_first_search.hpp>
#include <boost/range/adaptor/map.hpp>
using namespace std;
using boost::adaptors::map_values;
namespace ue2 {
template<typename VarP, typename VarQ>
void push_back_all_raw(vector<VarP> *out, const vector<VarQ> &in) {
for (const auto &var : in) {
out->push_back(var.get());
}
}
static
void all_vars(const GoughGraph &g, vector<GoughSSAVar *> *out) {
for (auto v : vertices_range(g)) {
push_back_all_raw(out, g[v].vars);
}
for (const auto &e : edges_range(g)) {
push_back_all_raw(out, g[e].vars);
}
}
namespace {
struct GoughGraphAux {
map<const GoughSSAVar *, GoughVertex> containing_v;
map<const GoughSSAVar *, GoughEdge> containing_e;
map<const GoughSSAVar *, set<GoughVertex> > reporters;
};
}
static never_inline
void fill_aux(const GoughGraph &g, GoughGraphAux *aux) {
for (auto v : vertices_range(g)) {
for (const auto &var : g[v].vars) {
aux->containing_v[var.get()] = v;
DEBUG_PRINTF("%u is on vertex %u\n", var->slot, g[v].state_id);
}
for (GoughSSAVar *var : g[v].reports | map_values) {
aux->reporters[var].insert(v);
}
for (GoughSSAVar *var : g[v].reports_eod | map_values) {
aux->reporters[var].insert(v);
}
}
for (const auto &e : edges_range(g)) {
for (const auto &var : g[e].vars) {
aux->containing_e[var.get()] = e;
DEBUG_PRINTF("%u is on edge %u->%u\n", var->slot,
g[source(e, g)].state_id, g[target(e, g)].state_id);
}
}
}
static
bool is_block_local(const GoughGraph &cfg, GoughSSAVar *var,
const GoughGraphAux &aux) {
/* if var used as a report, it cannot be considered block local */
if (contains(aux.reporters, var)) {
return false;
}
/* (useful) vertex/join vars never local - they are terminal in blocks
* and so should be read by another block. */
if (!contains(aux.containing_e, var)) {
return false;
}
/* for other cases, require that all uses of var are later in the same edge
* or on the target AND if on target it is sole on flow coming from the
* edge in question. */
const GoughEdge &e = aux.containing_e.at(var);
GoughVertex t = target(e, cfg);
size_t seen_outputs = 0;
const flat_set<GoughSSAVarWithInputs *> &out = var->get_outputs();
bool seen_var = false;
for (const auto &e_var : cfg[e].vars) {
if (seen_var) {
GoughSSAVarWithInputs *w
= dynamic_cast<GoughSSAVarWithInputs *>(e_var.get());
if (contains(out, w)) {
seen_outputs++;
}
} else {
seen_var = var == e_var.get();
}
}
assert(seen_var);
for (const auto &t_var : cfg[t].vars) {
if (contains(out, t_var.get())) {
seen_outputs++;
const flat_set<GoughEdge> &flow = t_var->get_edges_for_input(var);
if (flow.size() != 1 || *flow.begin() != e) {
/* this var is used by the target join var BUT on a different
* flow, so this is not a block local variable */
return false;
}
}
}
assert(seen_outputs <= out.size());
return seen_outputs == out.size();
}
static
void handle_pending_edge(const GoughGraph &g, const GoughEdge &e,
GoughSSAVar *start, set<GoughVertex> &pending_vertex,
set<const GoughSSAVar *> &rv) {
const vector<shared_ptr<GoughSSAVar> > &vars = g[e].vars;
bool marking = !start;
DEBUG_PRINTF(" ---checking edge %u->%u %s %zu\n", g[source(e, g)].state_id,
g[target(e, g)].state_id, marking ? "full" : "partial",
vars.size());
for (auto it = vars.rbegin(); it != vars.rend(); ++it) {
GoughSSAVar *var = it->get();
if (contains(rv, var)) {
DEBUG_PRINTF("somebody has already processed this vertex [%u]\n",
var->slot);
return;
}
if (var == start) {
assert(!marking);
marking = true;
continue;
}
if (marking) {
rv.insert(var);
}
}
assert(marking);
GoughVertex s = source(e, g);
for (const auto &var : g[s].vars) {
DEBUG_PRINTF("interferes %u\n", var->slot);
rv.insert(var.get());
}
pending_vertex.insert(s);
}
static
void handle_pending_vars(GoughSSAVar *def, const GoughGraph &g,
const GoughGraphAux &aux,
const flat_set<GoughSSAVarWithInputs *> &pending_var,
set<GoughVertex> &pending_vertex,
set<const GoughSSAVar *> &rv) {
for (GoughSSAVarWithInputs *var : pending_var) {
if (contains(aux.containing_v, var)) {
/* def is used by join vertex, value only needs to be live on some
* incoming edges */
GoughSSAVarJoin *vj = (GoughSSAVarJoin *)var;
const flat_set<GoughEdge> &live_edges
= vj->get_edges_for_input(def);
for (const auto &e : live_edges) {
handle_pending_edge(g, e, nullptr, pending_vertex, rv);
}
continue;
}
const GoughEdge &e = aux.containing_e.at(var);
handle_pending_edge(g, e, var, pending_vertex, rv);
}
}
static
void handle_pending_vertex(GoughVertex def_v, const GoughGraph &g,
GoughVertex current,
set<GoughVertex> &pending_vertex,
set<const GoughSSAVar *> &rv) {
DEBUG_PRINTF("---checking vertex %u\n", g[current].state_id);
if (def_v == current) {
DEBUG_PRINTF("contains target vertex\n");
return; /* we have reached def */
}
for (const auto &e : in_edges_range(current, g)) {
handle_pending_edge(g, e, nullptr, pending_vertex, rv);
}
}
static
void handle_pending_vertices(GoughSSAVar *def, const GoughGraph &g,
const GoughGraphAux &aux,
set<GoughVertex> &pending_vertex,
set<const GoughSSAVar *> &rv) {
if (pending_vertex.empty()) {
return;
}
GoughVertex def_v = GoughGraph::null_vertex();
if (contains(aux.containing_v, def)) {
def_v = aux.containing_v.at(def);
}
unordered_set<GoughVertex> done;
while (!pending_vertex.empty()) {
GoughVertex current = *pending_vertex.begin();
pending_vertex.erase(current);
if (contains(done, current)) {
continue;
}
done.insert(current);
handle_pending_vertex(def_v, g, current, pending_vertex, rv);
}
}
/* returns set of labels that the given def is live at */
static never_inline
set<const GoughSSAVar *> live_during(GoughSSAVar *def, const GoughGraph &g,
const GoughGraphAux &aux) {
DEBUG_PRINTF("checking who is defined during %u lifetime\n", def->slot);
set<GoughVertex> pending_vertex;
set<const GoughSSAVar *> rv;
rv.insert(def);
if (contains(aux.reporters, def)) {
DEBUG_PRINTF("--> gets reported\n");
const set<GoughVertex> &reporters = aux.reporters.at(def);
for (auto v : reporters) {
pending_vertex.insert(v);
for (const auto &var : g[v].vars) {
DEBUG_PRINTF("interferes %u\n", var->slot);
rv.insert(var.get());
}
}
}
handle_pending_vars(def, g, aux, def->get_outputs(), pending_vertex, rv);
handle_pending_vertices(def, g, aux, pending_vertex, rv);
rv.erase(def);
return rv;
}
template<typename VarP>
void set_initial_slots(const vector<VarP> &vars, u32 *next_slot) {
for (auto &var : vars) {
assert(var->slot == INVALID_SLOT);
var->slot = (*next_slot)++;
}
}
/* crude, deterministic assignment of symbolic register slots.
* returns number of slots given out
*/
static
u32 initial_slots(const GoughGraph &g) {
u32 next_slot = 0;
for (auto v : vertices_range(g)) {
set_initial_slots(g[v].vars, &next_slot);
}
for (const auto &e : edges_range(g)) {
set_initial_slots(g[e].vars, &next_slot);
}
return next_slot;
}
#define NO_COLOUR (~0U)
static
u32 available_colour(const flat_set<u32> &bad_colours) {
u32 rv = 0;
for (const u32 &colour : bad_colours) {
if (colour != rv) {
assert(colour > rv);
break;
}
rv = colour + 1;
}
assert(rv != NO_COLOUR);
return rv;
}
static
void poison_colours(const set<const GoughSSAVar *> &live, u32 c,
const vector<u32> &colour_map,
vector<flat_set<u32> > *bad_colour) {
for (const GoughSSAVar *var : live) {
u32 var_index = var->slot;
if (colour_map[var_index] != NO_COLOUR) {
assert(c != colour_map[var_index]);
} else {
(*bad_colour)[var_index].insert(c);
}
}
}
static
void find_bad_due_to_live(const set<const GoughSSAVar *> &live,
const vector<u32> &colour_map, flat_set<u32> *out) {
for (const GoughSSAVar *var : live) {
u32 var_index = var->slot;
if (colour_map[var_index] != NO_COLOUR) {
out->insert(colour_map[var_index]);
}
}
}
static
void sequential_vertex_colouring(const GoughGraph &g, const GoughGraphAux &aux,
const vector<GoughSSAVar *> &order,
vector<u32> &colour_map) {
assert(order.size() < NO_COLOUR);
colour_map.clear();
colour_map.resize(order.size(), NO_COLOUR);
vector<u32> temp(order.size(), ~0U);
vector<flat_set<u32> > bad_colour(order.size());
for (GoughSSAVar *var : order) {
u32 var_index = var->slot;
if (is_block_local(g, var, aux)) {
DEBUG_PRINTF("%u is block local\n", var_index);
/* ignore variable whose lifetime is limited to their local block
* there is no need to assign stream state to these variables */
continue;
}
assert(colour_map[var_index] == NO_COLOUR);
set<const GoughSSAVar *> live = live_during(var, g, aux);
flat_set<u32> &local_bad = bad_colour[var_index];
find_bad_due_to_live(live, colour_map, &local_bad);
DEBUG_PRINTF("colouring %u\n", var_index);
u32 c = available_colour(local_bad);
colour_map[var_index] = c;
assert(!contains(bad_colour[var_index], c));
poison_colours(live, c, colour_map, &bad_colour);
flat_set<u32> temp_set;
local_bad.swap(temp_set);
DEBUG_PRINTF(" %u coloured %u\n", var_index, c);
}
}
template<typename VarP>
void add_to_dom_ordering(const vector<VarP> &vars,
vector<GoughSSAVar *> *out) {
for (const auto &var : vars) {
out->push_back(var.get());
}
}
namespace {
class FinishVisitor : public boost::default_dfs_visitor {
public:
explicit FinishVisitor(vector<GoughVertex> *o) : out(o) {}
void finish_vertex(const GoughVertex v, const GoughGraph &) {
out->push_back(v);
}
vector<GoughVertex> *out;
};
}
static
void find_dom_ordering(const GoughGraph &cfg, vector<GoughSSAVar *> *out) {
vector<GoughVertex> g_order;
/* due to construction quirks, default vertex order provides entry points */
depth_first_search(cfg, visitor(FinishVisitor(&g_order))
.root_vertex(cfg[boost::graph_bundle].initial_vertex));
for (auto it = g_order.rbegin(); it != g_order.rend(); ++it) {
add_to_dom_ordering(cfg[*it].vars, out);
for (const auto &e : out_edges_range(*it, cfg)) {
add_to_dom_ordering(cfg[e].vars, out);
}
}
}
static
void create_slot_mapping(const GoughGraph &cfg, UNUSED u32 old_slot_count,
vector<u32> *old_new) {
/* Interference graphs from SSA form are chordal -> optimally colourable in
* poly time.
*
* Chordal graphs can be coloured by walking in perfect elimination order.
* If the SSA CFG is iterated over in a way that respects dominance
* relationship, the interference graph will be iterated in a perfect
* elimination order.
*
* We can avoid creating the full interference graph and use liveness
* information as we iterate over the definitions to perform the colouring.
*
* See S Hack various 2006-
*/
vector<GoughSSAVar *> dom_order;
GoughGraphAux aux;
fill_aux(cfg, &aux);
find_dom_ordering(cfg, &dom_order);
assert(dom_order.size() == old_slot_count);
sequential_vertex_colouring(cfg, aux, dom_order, *old_new);
}
static
void update_local_slots(GoughGraph &g, set<GoughSSAVar *> &locals,
u32 local_base) {
DEBUG_PRINTF("%zu local variables\n", locals.size());
/* local variables only occur on edges (joins are never local) */
u32 allocated_count = 0;
for (const auto &e : edges_range(g)) {
u32 next_slot = local_base;
for (auto &var : g[e].vars) {
if (contains(locals, var.get())) {
DEBUG_PRINTF("updating slot %u using local %u\n", var->slot,
next_slot);
var->slot = next_slot++;
allocated_count++;
}
}
}
assert(allocated_count == locals.size());
}
static never_inline
u32 update_slots(GoughGraph &g, const vector<u32> &old_new,
UNUSED u32 old_slot_count) {
vector<GoughSSAVar *> vars;
set<GoughSSAVar *> locals;
all_vars(g, &vars);
u32 slot_count = 0;
for (GoughSSAVar *v : vars) {
assert(v->slot < old_new.size());
DEBUG_PRINTF("updating slot %u to %u\n", v->slot, old_new[v->slot]);
if (old_new[v->slot] != NO_COLOUR) { /* not local, assign final slot */
v->slot = old_new[v->slot];
ENSURE_AT_LEAST(&slot_count, v->slot + 1);
} else {
locals.insert(v);
}
}
assert(slot_count <= old_slot_count);
DEBUG_PRINTF("reduce stream slots from %u to %u\n", old_slot_count,
slot_count);
update_local_slots(g, locals, slot_count);
return slot_count;
}
u32 assign_slots(GoughGraph &cfg, const Grey &grey) {
u32 slot_count = initial_slots(cfg);
if (!grey.goughRegisterAllocate) {
return slot_count;
}
dump(cfg, "slots_pre", grey);
vector<u32> old_new;
create_slot_mapping(cfg, slot_count, &old_new);
slot_count = update_slots(cfg, old_new, slot_count);
return slot_count;
}
} // namespace ue2
|