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
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
|
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8" />
<meta name="viewport" content="width=device-width, initial-scale=1.0" />
<title>MergeFunctions pass, how it works — LLVM 13 documentation</title>
<link rel="stylesheet" href="_static/pygments.css" type="text/css" />
<link rel="stylesheet" href="_static/llvm-theme.css" type="text/css" />
<script id="documentation_options" data-url_root="./" src="_static/documentation_options.js"></script>
<script src="_static/jquery.js"></script>
<script src="_static/underscore.js"></script>
<script src="_static/doctools.js"></script>
<link rel="index" title="Index" href="genindex.html" />
<link rel="search" title="Search" href="search.html" />
<link rel="next" title="MCJIT Design and Implementation" href="MCJITDesignAndImplementation.html" />
<link rel="prev" title="MemorySSA" href="MemorySSA.html" />
<style type="text/css">
table.right { float: right; margin-left: 20px; }
table.right td { border: 1px solid #ccc; }
</style>
</head><body>
<div class="logo">
<a href="index.html">
<img src="_static/logo.png"
alt="LLVM Logo" width="250" height="88"/></a>
</div>
<div class="related" role="navigation" aria-label="related navigation">
<h3>Navigation</h3>
<ul>
<li class="right" style="margin-right: 10px">
<a href="genindex.html" title="General Index"
accesskey="I">index</a></li>
<li class="right" >
<a href="MCJITDesignAndImplementation.html" title="MCJIT Design and Implementation"
accesskey="N">next</a> |</li>
<li class="right" >
<a href="MemorySSA.html" title="MemorySSA"
accesskey="P">previous</a> |</li>
<li><a href="https://llvm.org/">LLVM Home</a> | </li>
<li><a href="index.html">Documentation</a>»</li>
<li class="nav-item nav-item-1"><a href="UserGuides.html" accesskey="U">User Guides</a> »</li>
<li class="nav-item nav-item-this"><a href="">MergeFunctions pass, how it works</a></li>
</ul>
</div>
<div class="sphinxsidebar" role="navigation" aria-label="main navigation">
<div class="sphinxsidebarwrapper">
<h3>Documentation</h3>
<ul class="want-points">
<li><a href="https://llvm.org/docs/GettingStartedTutorials.html">Getting Started/Tutorials</a></li>
<li><a href="https://llvm.org/docs/UserGuides.html">User Guides</a></li>
<li><a href="https://llvm.org/docs/Reference.html">Reference</a></li>
</ul>
<h3>Getting Involved</h3>
<ul class="want-points">
<li><a href="https://llvm.org/docs/Contributing.html">Contributing to LLVM</a></li>
<li><a href="https://llvm.org/docs/HowToSubmitABug.html">Submitting Bug Reports</a></li>
<li><a href="https://llvm.org/docs/GettingInvolved.html#mailing-lists">Mailing Lists</a></li>
<li><a href="https://llvm.org/docs/GettingInvolved.html#irc">IRC</a></li>
<li><a href="https://llvm.org/docs/GettingInvolved.html#meetups-and-social-events">Meetups and Social Events</a></li>
</ul>
<h3>Additional Links</h3>
<ul class="want-points">
<li><a href="https://llvm.org/docs/FAQ.html">FAQ</a></li>
<li><a href="https://llvm.org/docs/Lexicon.html">Glossary</a></li>
<li><a href="https://llvm.org/pubs">Publications</a></li>
<li><a href="https://github.com/llvm/llvm-project//">Github Repository</a></li>
</ul>
<div role="note" aria-label="source link">
<h3>This Page</h3>
<ul class="this-page-menu">
<li><a href="_sources/MergeFunctions.rst.txt"
rel="nofollow">Show Source</a></li>
</ul>
</div>
<div id="searchbox" style="display: none" role="search">
<h3 id="searchlabel">Quick search</h3>
<div class="searchformwrapper">
<form class="search" action="search.html" method="get">
<input type="text" name="q" aria-labelledby="searchlabel" />
<input type="submit" value="Go" />
</form>
</div>
</div>
<script>$('#searchbox').show(0);</script>
</div>
</div>
<div class="document">
<div class="documentwrapper">
<div class="bodywrapper">
<div class="body" role="main">
<div class="section" id="mergefunctions-pass-how-it-works">
<h1>MergeFunctions pass, how it works<a class="headerlink" href="#mergefunctions-pass-how-it-works" title="Permalink to this headline">¶</a></h1>
<div class="contents local topic" id="contents">
<ul class="simple">
<li><p><a class="reference internal" href="#introduction" id="id1">Introduction</a></p>
<ul>
<li><p><a class="reference internal" href="#what-should-i-know-to-be-able-to-follow-along-with-this-document" id="id2">What should I know to be able to follow along with this document?</a></p></li>
<li><p><a class="reference internal" href="#narrative-structure" id="id3">Narrative structure</a></p></li>
</ul>
</li>
<li><p><a class="reference internal" href="#basics" id="id4">Basics</a></p>
<ul>
<li><p><a class="reference internal" href="#how-to-do-it" id="id5">How to do it?</a></p>
<ul>
<li><p><a class="reference internal" href="#possible-solutions" id="id6">Possible solutions</a></p>
<ul>
<li><p><a class="reference internal" href="#random-access" id="id7">Random-access</a></p></li>
<li><p><a class="reference internal" href="#logarithmical-search" id="id8">Logarithmical search</a></p></li>
<li><p><a class="reference internal" href="#present-state" id="id9">Present state</a></p></li>
</ul>
</li>
<li><p><a class="reference internal" href="#mergefunctions-main-fields-and-runonmodule" id="id10">MergeFunctions, main fields and runOnModule</a></p>
<ul>
<li><p><a class="reference internal" href="#runonmodule" id="id11">runOnModule</a></p></li>
<li><p><a class="reference internal" href="#comparison-and-logarithmical-search" id="id12">Comparison and logarithmical search</a></p></li>
</ul>
</li>
</ul>
</li>
</ul>
</li>
<li><p><a class="reference internal" href="#functions-comparison" id="id13">Functions comparison</a></p>
<ul>
<li><p><a class="reference internal" href="#functioncomparator-compare-void" id="id14">FunctionComparator::compare(void)</a></p></li>
<li><p><a class="reference internal" href="#functioncomparator-cmptype" id="id15">FunctionComparator::cmpType</a></p></li>
<li><p><a class="reference internal" href="#cmpvalues-const-value-const-value" id="id16">cmpValues(const Value*, const Value*)</a></p>
<ul>
<li><p><a class="reference internal" href="#what-we-associate-in-cmpvalues" id="id17">What we associate in cmpValues?</a></p></li>
<li><p><a class="reference internal" href="#how-to-implement-cmpvalues" id="id18">How to implement cmpValues?</a></p></li>
</ul>
</li>
<li><p><a class="reference internal" href="#cmpconstants" id="id19">cmpConstants</a></p></li>
<li><p><a class="reference internal" href="#compare-const-basicblock-const-basicblock" id="id20">compare(const BasicBlock*, const BasicBlock*)</a></p></li>
<li><p><a class="reference internal" href="#cmpgep" id="id21">cmpGEP</a></p></li>
<li><p><a class="reference internal" href="#cmpoperation" id="id22">cmpOperation</a></p></li>
<li><p><a class="reference internal" href="#o-log-n" id="id23">O(log(N))</a></p></li>
</ul>
</li>
<li><p><a class="reference internal" href="#merging-process-mergetwofunctions" id="id24">Merging process, mergeTwoFunctions</a></p>
<ul>
<li><p><a class="reference internal" href="#if-f-may-be-overridden" id="id25">If “F” may be overridden</a></p>
<ul>
<li><p><a class="reference internal" href="#hasglobalaliases-removeusers" id="id26">HasGlobalAliases, removeUsers</a></p></li>
<li><p><a class="reference internal" href="#no-global-aliases-replacedirectcallers" id="id27">No global aliases, replaceDirectCallers</a></p>
<ul>
<li><p><a class="reference internal" href="#if-f-could-not-be-overridden-fix-it" id="id28">If “F” could not be overridden, fix it!</a></p></li>
</ul>
</li>
</ul>
</li>
</ul>
</li>
</ul>
</div>
<div class="section" id="introduction">
<h2><a class="toc-backref" href="#id1">Introduction</a><a class="headerlink" href="#introduction" title="Permalink to this headline">¶</a></h2>
<p>Sometimes code contains equal functions, or functions that does exactly the same
thing even though they are non-equal on the IR level (e.g.: multiplication on 2
and ‘shl 1’). It could happen due to several reasons: mainly, the usage of
templates and automatic code generators. Though, sometimes the user itself could
write the same thing twice :-)</p>
<p>The main purpose of this pass is to recognize such functions and merge them.</p>
<p>This document is the extension to pass comments and describes the pass logic. It
describes the algorithm that is used in order to compare functions and
explains how we could combine equal functions correctly to keep the module
valid.</p>
<p>Material is brought in a top-down form, so the reader could start to learn pass
from high level ideas and end with low-level algorithm details, thus preparing
him or her for reading the sources.</p>
<p>The main goal is to describe the algorithm and logic here and the concept. If
you <em>don’t want</em> to read the source code, but want to understand pass
algorithms, this document is good for you. The author tries not to repeat the
source-code and covers only common cases to avoid the cases of needing to
update this document after any minor code changes.</p>
<div class="section" id="what-should-i-know-to-be-able-to-follow-along-with-this-document">
<h3><a class="toc-backref" href="#id2">What should I know to be able to follow along with this document?</a><a class="headerlink" href="#what-should-i-know-to-be-able-to-follow-along-with-this-document" title="Permalink to this headline">¶</a></h3>
<p>The reader should be familiar with common compile-engineering principles and
LLVM code fundamentals. In this article, we assume the reader is familiar with
<a class="reference external" href="http://en.wikipedia.org/wiki/Static_single_assignment_form">Single Static Assignment</a>
concept and has an understanding of
<a class="reference external" href="https://llvm.org/docs/LangRef.html#high-level-structure">IR structure</a>.</p>
<p>We will use terms such as
“<a class="reference external" href="https://llvm.org/docs/LangRef.html#high-level-structure">module</a>”,
“<a class="reference external" href="https://llvm.org/docs/ProgrammersManual.html#the-function-class">function</a>”,
“<a class="reference external" href="http://en.wikipedia.org/wiki/Basic_block">basic block</a>”,
“<a class="reference external" href="https://llvm.org/docs/ProgrammersManual.html#the-user-class">user</a>”,
“<a class="reference external" href="https://llvm.org/docs/ProgrammersManual.html#the-value-class">value</a>”,
“<a class="reference external" href="https://llvm.org/docs/ProgrammersManual.html#the-instruction-class">instruction</a>”.</p>
<p>As a good starting point, the Kaleidoscope tutorial can be used:</p>
<p><a class="reference internal" href="tutorial/index.html"><span class="doc">LLVM Tutorial: Table of Contents</span></a></p>
<p>It’s especially important to understand chapter 3 of tutorial:</p>
<p><a class="reference internal" href="tutorial/LangImpl03.html"><span class="doc">Kaleidoscope Tutorial</span></a></p>
<p>The reader should also know how passes work in LLVM. They could use this
article as a reference and start point here:</p>
<p><a class="reference internal" href="WritingAnLLVMPass.html"><span class="doc">Writing an LLVM Pass</span></a></p>
<p>What else? Well perhaps the reader should also have some experience in LLVM pass
debugging and bug-fixing.</p>
</div>
<div class="section" id="narrative-structure">
<h3><a class="toc-backref" href="#id3">Narrative structure</a><a class="headerlink" href="#narrative-structure" title="Permalink to this headline">¶</a></h3>
<p>The article consists of three parts. The first part explains pass functionality
on the top-level. The second part describes the comparison procedure itself.
The third part describes the merging process.</p>
<p>In every part, the author tries to put the contents in the top-down form.
The top-level methods will first be described followed by the terminal ones at
the end, in the tail of each part. If the reader sees the reference to the
method that wasn’t described yet, they will find its description a bit below.</p>
</div>
</div>
<div class="section" id="basics">
<h2><a class="toc-backref" href="#id4">Basics</a><a class="headerlink" href="#basics" title="Permalink to this headline">¶</a></h2>
<div class="section" id="how-to-do-it">
<h3><a class="toc-backref" href="#id5">How to do it?</a><a class="headerlink" href="#how-to-do-it" title="Permalink to this headline">¶</a></h3>
<p>Do we need to merge functions? The obvious answer is: Yes, that is quite a
possible case. We usually <em>do</em> have duplicates and it would be good to get rid
of them. But how do we detect duplicates? This is the idea: we split functions
into smaller bricks or parts and compare the “bricks” amount. If equal,
we compare the “bricks” themselves, and then do our conclusions about functions
themselves.</p>
<p>What could the difference be? For example, on a machine with 64-bit pointers
(let’s assume we have only one address space), one function stores a 64-bit
integer, while another one stores a pointer. If the target is the machine
mentioned above, and if functions are identical, except the parameter type (we
could consider it as a part of function type), then we can treat a <code class="docutils literal notranslate"><span class="pre">uint64_t</span></code>
and a <code class="docutils literal notranslate"><span class="pre">void*</span></code> as equal.</p>
<p>This is just an example; more possible details are described a bit below.</p>
<p>As another example, the reader may imagine two more functions. The first
function performs a multiplication by 2, while the second one performs an
logical left shift by 1.</p>
<div class="section" id="possible-solutions">
<h4><a class="toc-backref" href="#id6">Possible solutions</a><a class="headerlink" href="#possible-solutions" title="Permalink to this headline">¶</a></h4>
<p>Let’s briefly consider possible options about how and what we have to implement
in order to create full-featured functions merging, and also what it would
mean for us.</p>
<p>Equal function detection obviously supposes that a “detector” method to be
implemented and latter should answer the question “whether functions are equal”.
This “detector” method consists of tiny “sub-detectors”, which each answers
exactly the same question, but for function parts.</p>
<p>As the second step, we should merge equal functions. So it should be a “merger”
method. “Merger” accepts two functions <em>F1</em> and <em>F2</em>, and produces <em>F1F2</em>
function, the result of merging.</p>
<p>Having such routines in our hands, we can process a whole module, and merge all
equal functions.</p>
<p>In this case, we have to compare every function with every another function. As
the reader may notice, this way seems to be quite expensive. Of course we could
introduce hashing and other helpers, but it is still just an optimization, and
thus the level of O(N*N) complexity.</p>
<p>Can we reach another level? Could we introduce logarithmical search, or random
access lookup? The answer is: “yes”.</p>
<div class="section" id="random-access">
<h5><a class="toc-backref" href="#id7">Random-access</a><a class="headerlink" href="#random-access" title="Permalink to this headline">¶</a></h5>
<p>How it could this be done? Just convert each function to a number, and gather
all of them in a special hash-table. Functions with equal hashes are equal.
Good hashing means, that every function part must be taken into account. That
means we have to convert every function part into some number, and then add it
into the hash. The lookup-up time would be small, but such an approach adds some
delay due to the hashing routine.</p>
</div>
<div class="section" id="logarithmical-search">
<h5><a class="toc-backref" href="#id8">Logarithmical search</a><a class="headerlink" href="#logarithmical-search" title="Permalink to this headline">¶</a></h5>
<p>We could introduce total ordering among the functions set, once ordered we
could then implement a logarithmical search. Lookup time still depends on N,
but adds a little of delay (<em>log(N)</em>).</p>
</div>
<div class="section" id="present-state">
<h5><a class="toc-backref" href="#id9">Present state</a><a class="headerlink" href="#present-state" title="Permalink to this headline">¶</a></h5>
<p>Both of the approaches (random-access and logarithmical) have been implemented
and tested and both give a very good improvement. What was most
surprising is that logarithmical search was faster; sometimes by up to 15%. The
hashing method needs some extra CPU time, which is the main reason why it works
slower; in most cases, total “hashing” time is greater than total
“logarithmical-search” time.</p>
<p>So, preference has been granted to the “logarithmical search”.</p>
<p>Though in the case of need, <em>logarithmical-search</em> (read “total-ordering”) could
be used as a milestone on our way to the <em>random-access</em> implementation.</p>
<p>Every comparison is based either on the numbers or on the flags comparison. In
the <em>random-access</em> approach, we could use the same comparison algorithm.
During comparison, we exit once we find the difference, but here we might have
to scan the whole function body every time (note, it could be slower). Like in
“total-ordering”, we will track every number and flag, but instead of
comparison, we should get the numbers sequence and then create the hash number.
So, once again, <em>total-ordering</em> could be considered as a milestone for even
faster (in theory) random-access approach.</p>
</div>
</div>
<div class="section" id="mergefunctions-main-fields-and-runonmodule">
<h4><a class="toc-backref" href="#id10">MergeFunctions, main fields and runOnModule</a><a class="headerlink" href="#mergefunctions-main-fields-and-runonmodule" title="Permalink to this headline">¶</a></h4>
<p>There are two main important fields in the class:</p>
<p><code class="docutils literal notranslate"><span class="pre">FnTree</span></code> – the set of all unique functions. It keeps items that couldn’t be
merged with each other. It is defined as:</p>
<p><code class="docutils literal notranslate"><span class="pre">std::set<FunctionNode></span> <span class="pre">FnTree;</span></code></p>
<p>Here <code class="docutils literal notranslate"><span class="pre">FunctionNode</span></code> is a wrapper for <code class="docutils literal notranslate"><span class="pre">llvm::Function</span></code> class, with
implemented “<” operator among the functions set (below we explain how it works
exactly; this is a key point in fast functions comparison).</p>
<p><code class="docutils literal notranslate"><span class="pre">Deferred</span></code> – merging process can affect bodies of functions that are in
<code class="docutils literal notranslate"><span class="pre">FnTree</span></code> already. Obviously, such functions should be rechecked again. In this
case, we remove them from <code class="docutils literal notranslate"><span class="pre">FnTree</span></code>, and mark them to be rescanned, namely
put them into <code class="docutils literal notranslate"><span class="pre">Deferred</span></code> list.</p>
<div class="section" id="runonmodule">
<h5><a class="toc-backref" href="#id11">runOnModule</a><a class="headerlink" href="#runonmodule" title="Permalink to this headline">¶</a></h5>
<p>The algorithm is pretty simple:</p>
<ol class="arabic simple">
<li><p>Put all module’s functions into the <em>worklist</em>.</p></li>
</ol>
<p>2. Scan <em>worklist</em>’s functions twice: first enumerate only strong functions and
then only weak ones:</p>
<blockquote>
<div><p>2.1. Loop body: take a function from <em>worklist</em> (call it <em>FCur</em>) and try to
insert it into <em>FnTree</em>: check whether <em>FCur</em> is equal to one of functions
in <em>FnTree</em>. If there <em>is</em> an equal function in <em>FnTree</em>
(call it <em>FExists</em>): merge function <em>FCur</em> with <em>FExists</em>. Otherwise add
the function from the <em>worklist</em> to <em>FnTree</em>.</p>
</div></blockquote>
<p>3. Once the <em>worklist</em> scanning and merging operations are complete, check the
<em>Deferred</em> list. If it is not empty: refill the <em>worklist</em> contents with
<em>Deferred</em> list and redo step 2, if the <em>Deferred</em> list is empty, then exit
from method.</p>
</div>
<div class="section" id="comparison-and-logarithmical-search">
<h5><a class="toc-backref" href="#id12">Comparison and logarithmical search</a><a class="headerlink" href="#comparison-and-logarithmical-search" title="Permalink to this headline">¶</a></h5>
<p>Let’s recall our task: for every function <em>F</em> from module <em>M</em>, we have to find
equal functions <em>F`</em> in the shortest time possible , and merge them into a
single function.</p>
<p>Defining total ordering among the functions set allows us to organize
functions into a binary tree. The lookup procedure complexity would be
estimated as O(log(N)) in this case. But how do we define <em>total-ordering</em>?</p>
<p>We have to introduce a single rule applicable to every pair of functions, and
following this rule, then evaluate which of them is greater. What kind of rule
could it be? Let’s declare it as the “compare” method that returns one of 3
possible values:</p>
<p>-1, left is <em>less</em> than right,</p>
<p>0, left and right are <em>equal</em>,</p>
<p>1, left is <em>greater</em> than right.</p>
<p>Of course it means, that we have to maintain
<em>strict and non-strict order relation properties</em>:</p>
<ul class="simple">
<li><p>reflexivity (<code class="docutils literal notranslate"><span class="pre">a</span> <span class="pre"><=</span> <span class="pre">a</span></code>, <code class="docutils literal notranslate"><span class="pre">a</span> <span class="pre">==</span> <span class="pre">a</span></code>, <code class="docutils literal notranslate"><span class="pre">a</span> <span class="pre">>=</span> <span class="pre">a</span></code>),</p></li>
<li><p>antisymmetry (if <code class="docutils literal notranslate"><span class="pre">a</span> <span class="pre"><=</span> <span class="pre">b</span></code> and <code class="docutils literal notranslate"><span class="pre">b</span> <span class="pre"><=</span> <span class="pre">a</span></code> then <code class="docutils literal notranslate"><span class="pre">a</span> <span class="pre">==</span> <span class="pre">b</span></code>),</p></li>
<li><p>transitivity (<code class="docutils literal notranslate"><span class="pre">a</span> <span class="pre"><=</span> <span class="pre">b</span></code> and <code class="docutils literal notranslate"><span class="pre">b</span> <span class="pre"><=</span> <span class="pre">c</span></code>, then <code class="docutils literal notranslate"><span class="pre">a</span> <span class="pre"><=</span> <span class="pre">c</span></code>)</p></li>
<li><p>asymmetry (if <code class="docutils literal notranslate"><span class="pre">a</span> <span class="pre"><</span> <span class="pre">b</span></code>, then <code class="docutils literal notranslate"><span class="pre">a</span> <span class="pre">></span> <span class="pre">b</span></code> or <code class="docutils literal notranslate"><span class="pre">a</span> <span class="pre">==</span> <span class="pre">b</span></code>).</p></li>
</ul>
<p>As mentioned before, the comparison routine consists of
“sub-comparison-routines”, with each of them also consisting of
“sub-comparison-routines”, and so on. Finally, it ends up with primitive
comparison.</p>
<p>Below, we will use the following operations:</p>
<ol class="arabic simple">
<li><p><code class="docutils literal notranslate"><span class="pre">cmpNumbers(number1,</span> <span class="pre">number2)</span></code> is a method that returns -1 if left is less
than right; 0, if left and right are equal; and 1 otherwise.</p></li>
<li><p><code class="docutils literal notranslate"><span class="pre">cmpFlags(flag1,</span> <span class="pre">flag2)</span></code> is a hypothetical method that compares two flags.
The logic is the same as in <code class="docutils literal notranslate"><span class="pre">cmpNumbers</span></code>, where <code class="docutils literal notranslate"><span class="pre">true</span></code> is 1, and
<code class="docutils literal notranslate"><span class="pre">false</span></code> is 0.</p></li>
</ol>
<p>The rest of the article is based on <em>MergeFunctions.cpp</em> source code
(found in <em><llvm_dir>/lib/Transforms/IPO/MergeFunctions.cpp</em>). We would like
to ask reader to keep this file open, so we could use it as a reference
for further explanations.</p>
<p>Now, we’re ready to proceed to the next chapter and see how it works.</p>
</div>
</div>
</div>
</div>
<div class="section" id="functions-comparison">
<h2><a class="toc-backref" href="#id13">Functions comparison</a><a class="headerlink" href="#functions-comparison" title="Permalink to this headline">¶</a></h2>
<p>At first, let’s define how exactly we compare complex objects.</p>
<p>Complex object comparison (function, basic-block, etc) is mostly based on its
sub-object comparison results. It is similar to the next “tree” objects
comparison:</p>
<ol class="arabic simple">
<li><p>For two trees <em>T1</em> and <em>T2</em> we perform <em>depth-first-traversal</em> and have
two sequences as a product: “<em>T1Items</em>” and “<em>T2Items</em>”.</p></li>
<li><p>We then compare chains “<em>T1Items</em>” and “<em>T2Items</em>” in
the most-significant-item-first order. The result of items comparison
would be the result of <em>T1</em> and <em>T2</em> comparison itself.</p></li>
</ol>
<div class="section" id="functioncomparator-compare-void">
<h3><a class="toc-backref" href="#id14">FunctionComparator::compare(void)</a><a class="headerlink" href="#functioncomparator-compare-void" title="Permalink to this headline">¶</a></h3>
<p>A brief look at the source code tells us that the comparison starts in the
“<code class="docutils literal notranslate"><span class="pre">int</span> <span class="pre">FunctionComparator::compare(void)</span></code>” method.</p>
<p>1. The first parts to be compared are the function’s attributes and some
properties that is outside the “attributes” term, but still could make the
function different without changing its body. This part of the comparison is
usually done within simple <em>cmpNumbers</em> or <em>cmpFlags</em> operations (e.g.
<code class="docutils literal notranslate"><span class="pre">cmpFlags(F1->hasGC(),</span> <span class="pre">F2->hasGC())</span></code>). Below is a full list of function’s
properties to be compared on this stage:</p>
<blockquote>
<div><ul class="simple">
<li><p><em>Attributes</em> (those are returned by <code class="docutils literal notranslate"><span class="pre">Function::getAttributes()</span></code>
method).</p></li>
<li><p><em>GC</em>, for equivalence, <em>RHS</em> and <em>LHS</em> should be both either without
<em>GC</em> or with the same one.</p></li>
<li><p><em>Section</em>, just like a <em>GC</em>: <em>RHS</em> and <em>LHS</em> should be defined in the
same section.</p></li>
<li><p><em>Variable arguments</em>. <em>LHS</em> and <em>RHS</em> should be both either with or
without <em>var-args</em>.</p></li>
<li><p><em>Calling convention</em> should be the same.</p></li>
</ul>
</div></blockquote>
<p>2. Function type. Checked by <code class="docutils literal notranslate"><span class="pre">FunctionComparator::cmpType(Type*,</span> <span class="pre">Type*)</span></code>
method. It checks return type and parameters type; the method itself will be
described later.</p>
<p>3. Associate function formal parameters with each other. Then comparing function
bodies, if we see the usage of <em>LHS</em>’s <em>i</em>-th argument in <em>LHS</em>’s body, then,
we want to see usage of <em>RHS</em>’s <em>i</em>-th argument at the same place in <em>RHS</em>’s
body, otherwise functions are different. On this stage we grant the preference
to those we met later in function body (value we met first would be <em>less</em>).
This is done by “<code class="docutils literal notranslate"><span class="pre">FunctionComparator::cmpValues(const</span> <span class="pre">Value*,</span> <span class="pre">const</span> <span class="pre">Value*)</span></code>”
method (will be described a bit later).</p>
<ol class="arabic simple" start="4">
<li><p>Function body comparison. As it written in method comments:</p></li>
</ol>
<p>“We do a CFG-ordered walk since the actual ordering of the blocks in the linked
list is immaterial. Our walk starts at the entry block for both functions, then
takes each block from each terminator in order. As an artifact, this also means
that unreachable blocks are ignored.”</p>
<p>So, using this walk we get BBs from <em>left</em> and <em>right</em> in the same order, and
compare them by “<code class="docutils literal notranslate"><span class="pre">FunctionComparator::compare(const</span> <span class="pre">BasicBlock*,</span> <span class="pre">const</span>
<span class="pre">BasicBlock*)</span></code>” method.</p>
<p>We also associate BBs with each other, like we did it with function formal
arguments (see <code class="docutils literal notranslate"><span class="pre">cmpValues</span></code> method below).</p>
</div>
<div class="section" id="functioncomparator-cmptype">
<h3><a class="toc-backref" href="#id15">FunctionComparator::cmpType</a><a class="headerlink" href="#functioncomparator-cmptype" title="Permalink to this headline">¶</a></h3>
<p>Consider how type comparison works.</p>
<p>1. Coerce pointer to integer. If left type is a pointer, try to coerce it to the
integer type. It could be done if its address space is 0, or if address spaces
are ignored at all. Do the same thing for the right type.</p>
<p>2. If left and right types are equal, return 0. Otherwise we need to give
preference to one of them. So proceed to the next step.</p>
<p>3. If types are of different kind (different type IDs). Return result of type
IDs comparison, treating them as numbers (use <code class="docutils literal notranslate"><span class="pre">cmpNumbers</span></code> operation).</p>
<p>4. If types are vectors or integers, return result of their pointers comparison,
comparing them as numbers.</p>
<ol class="arabic" start="5">
<li><p>Check whether type ID belongs to the next group (call it equivalent-group):</p>
<ul class="simple">
<li><p>Void</p></li>
<li><p>Float</p></li>
<li><p>Double</p></li>
<li><p>X86_FP80</p></li>
<li><p>FP128</p></li>
<li><p>PPC_FP128</p></li>
<li><p>Label</p></li>
<li><p>Metadata.</p></li>
</ul>
<p>If ID belongs to group above, return 0. Since it’s enough to see that
types has the same <code class="docutils literal notranslate"><span class="pre">TypeID</span></code>. No additional information is required.</p>
</li>
</ol>
<p>6. Left and right are pointers. Return result of address space comparison
(numbers comparison).</p>
<p>7. Complex types (structures, arrays, etc.). Follow complex objects comparison
technique (see the very first paragraph of this chapter). Both <em>left</em> and
<em>right</em> are to be expanded and their element types will be checked the same
way. If we get -1 or 1 on some stage, return it. Otherwise return 0.</p>
<p>8. Steps 1-6 describe all the possible cases, if we passed steps 1-6 and didn’t
get any conclusions, then invoke <code class="docutils literal notranslate"><span class="pre">llvm_unreachable</span></code>, since it’s quite an
unexpectable case.</p>
</div>
<div class="section" id="cmpvalues-const-value-const-value">
<h3><a class="toc-backref" href="#id16">cmpValues(const Value*, const Value*)</a><a class="headerlink" href="#cmpvalues-const-value-const-value" title="Permalink to this headline">¶</a></h3>
<p>Method that compares local values.</p>
<p>This method gives us an answer to a very curious question: whether we could
treat local values as equal, and which value is greater otherwise. It’s
better to start from example:</p>
<p>Consider the situation when we’re looking at the same place in left
function “<em>FL</em>” and in right function “<em>FR</em>”. Every part of <em>left</em> place is
equal to the corresponding part of <em>right</em> place, and (!) both parts use
<em>Value</em> instances, for example:</p>
<div class="highlight-text notranslate"><div class="highlight"><pre><span></span>instr0 i32 %LV ; left side, function FL
instr0 i32 %RV ; right side, function FR
</pre></div>
</div>
<p>So, now our conclusion depends on <em>Value</em> instances comparison.</p>
<p>The main purpose of this method is to determine relation between such values.</p>
<p>What can we expect from equal functions? At the same place, in functions
“<em>FL</em>” and “<em>FR</em>” we expect to see <em>equal</em> values, or values <em>defined</em> at
the same place in “<em>FL</em>” and “<em>FR</em>”.</p>
<p>Consider a small example here:</p>
<div class="highlight-text notranslate"><div class="highlight"><pre><span></span>define void %f(i32 %pf0, i32 %pf1) {
instr0 i32 %pf0 instr1 i32 %pf1 instr2 i32 123
}
</pre></div>
</div>
<div class="highlight-text notranslate"><div class="highlight"><pre><span></span>define void %g(i32 %pg0, i32 %pg1) {
instr0 i32 %pg0 instr1 i32 %pg0 instr2 i32 123
}
</pre></div>
</div>
<p>In this example, <em>pf0</em> is associated with <em>pg0</em>, <em>pf1</em> is associated with
<em>pg1</em>, and we also declare that <em>pf0</em> < <em>pf1</em>, and thus <em>pg0</em> < <em>pf1</em>.</p>
<p>Instructions with opcode “<em>instr0</em>” would be <em>equal</em>, since their types and
opcodes are equal, and values are <em>associated</em>.</p>
<p>Instructions with opcode “<em>instr1</em>” from <em>f</em> is <em>greater</em> than instructions
with opcode “<em>instr1</em>” from <em>g</em>; here we have equal types and opcodes, but
“<em>pf1</em> is greater than “<em>pg0</em>”.</p>
<p>Instructions with opcode “<em>instr2</em>” are equal, because their opcodes and
types are equal, and the same constant is used as a value.</p>
<div class="section" id="what-we-associate-in-cmpvalues">
<h4><a class="toc-backref" href="#id17">What we associate in cmpValues?</a><a class="headerlink" href="#what-we-associate-in-cmpvalues" title="Permalink to this headline">¶</a></h4>
<ul class="simple">
<li><p>Function arguments. <em>i</em>-th argument from left function associated with
<em>i</em>-th argument from right function.</p></li>
<li><p>BasicBlock instances. In basic-block enumeration loop we associate <em>i</em>-th
BasicBlock from the left function with <em>i</em>-th BasicBlock from the right
function.</p></li>
<li><p>Instructions.</p></li>
<li><p>Instruction operands. Note, we can meet <em>Value</em> here we have never seen
before. In this case it is not a function argument, nor <em>BasicBlock</em>, nor
<em>Instruction</em>. It is a global value. It is a constant, since it’s the only
supposed global here. The method also compares: Constants that are of the
same type and if right constant can be losslessly bit-casted to the left
one, then we also compare them.</p></li>
</ul>
</div>
<div class="section" id="how-to-implement-cmpvalues">
<h4><a class="toc-backref" href="#id18">How to implement cmpValues?</a><a class="headerlink" href="#how-to-implement-cmpvalues" title="Permalink to this headline">¶</a></h4>
<p><em>Association</em> is a case of equality for us. We just treat such values as equal,
but, in general, we need to implement antisymmetric relation. As mentioned
above, to understand what is <em>less</em>, we can use order in which we
meet values. If both values have the same order in a function (met at the same
time), we then treat values as <em>associated</em>. Otherwise – it depends on who was
first.</p>
<p>Every time we run the top-level compare method, we initialize two identical
maps (one for the left side, another one for the right side):</p>
<p><code class="docutils literal notranslate"><span class="pre">map<Value,</span> <span class="pre">int></span> <span class="pre">sn_mapL,</span> <span class="pre">sn_mapR;</span></code></p>
<p>The key of the map is the <em>Value</em> itself, the <em>value</em> – is its order (call it
<em>serial number</em>).</p>
<p>To add value <em>V</em> we need to perform the next procedure:</p>
<p><code class="docutils literal notranslate"><span class="pre">sn_map.insert(std::make_pair(V,</span> <span class="pre">sn_map.size()));</span></code></p>
<p>For the first <em>Value</em>, map will return <em>0</em>, for the second <em>Value</em> map will
return <em>1</em>, and so on.</p>
<p>We can then check whether left and right values met at the same time with
a simple comparison:</p>
<p><code class="docutils literal notranslate"><span class="pre">cmpNumbers(sn_mapL[Left],</span> <span class="pre">sn_mapR[Right]);</span></code></p>
<p>Of course, we can combine insertion and comparison:</p>
<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="n">std</span><span class="o">::</span><span class="n">pair</span><span class="o"><</span><span class="n">iterator</span><span class="p">,</span> <span class="kt">bool</span><span class="o">></span>
<span class="n">LeftRes</span> <span class="o">=</span> <span class="n">sn_mapL</span><span class="p">.</span><span class="n">insert</span><span class="p">(</span><span class="n">std</span><span class="o">::</span><span class="n">make_pair</span><span class="p">(</span><span class="n">Left</span><span class="p">,</span> <span class="n">sn_mapL</span><span class="p">.</span><span class="n">size</span><span class="p">())),</span> <span class="n">RightRes</span>
<span class="o">=</span> <span class="n">sn_mapR</span><span class="p">.</span><span class="n">insert</span><span class="p">(</span><span class="n">std</span><span class="o">::</span><span class="n">make_pair</span><span class="p">(</span><span class="n">Right</span><span class="p">,</span> <span class="n">sn_mapR</span><span class="p">.</span><span class="n">size</span><span class="p">()));</span>
<span class="k">return</span> <span class="nf">cmpNumbers</span><span class="p">(</span><span class="n">LeftRes</span><span class="p">.</span><span class="n">first</span><span class="o">-></span><span class="n">second</span><span class="p">,</span> <span class="n">RightRes</span><span class="p">.</span><span class="n">first</span><span class="o">-></span><span class="n">second</span><span class="p">);</span>
</pre></div>
</div>
<p>Let’s look, how whole method could be implemented.</p>
<p>1. We have to start with the bad news. Consider function self and
cross-referencing cases:</p>
<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="c1">// self-reference unsigned fact0(unsigned n) { return n > 1 ? n</span>
<span class="o">*</span> <span class="n">fact0</span><span class="p">(</span><span class="n">n</span><span class="mi">-1</span><span class="p">)</span> <span class="o">:</span> <span class="mi">1</span><span class="p">;</span> <span class="p">}</span> <span class="kt">unsigned</span> <span class="n">fact1</span><span class="p">(</span><span class="kt">unsigned</span> <span class="n">n</span><span class="p">)</span> <span class="p">{</span> <span class="k">return</span> <span class="n">n</span> <span class="o">></span> <span class="mi">1</span> <span class="o">?</span> <span class="n">n</span> <span class="o">*</span>
<span class="n">fact1</span><span class="p">(</span><span class="n">n</span><span class="mi">-1</span><span class="p">)</span> <span class="o">:</span> <span class="mi">1</span><span class="p">;</span> <span class="p">}</span>
<span class="c1">// cross-reference unsigned ping(unsigned n) { return n!= 0 ? pong(n-1) : 0;</span>
<span class="p">}</span> <span class="kt">unsigned</span> <span class="n">pong</span><span class="p">(</span><span class="kt">unsigned</span> <span class="n">n</span><span class="p">)</span> <span class="p">{</span> <span class="k">return</span> <span class="n">n</span><span class="o">!=</span> <span class="mi">0</span> <span class="o">?</span> <span class="n">ping</span><span class="p">(</span><span class="n">n</span><span class="mi">-1</span><span class="p">)</span> <span class="o">:</span> <span class="mi">0</span><span class="p">;</span> <span class="p">}</span>
</pre></div>
</div>
<blockquote>
<div><p>This comparison has been implemented in initial <em>MergeFunctions</em> pass
version. But, unfortunately, it is not transitive. And this is the only case
we can’t convert to less-equal-greater comparison. It is a seldom case, 4-5
functions of 10000 (checked in test-suite), and, we hope, the reader would
forgive us for such a sacrifice in order to get the O(log(N)) pass time.</p>
</div></blockquote>
<p>2. If left/right <em>Value</em> is a constant, we have to compare them. Return 0 if it
is the same constant, or use <code class="docutils literal notranslate"><span class="pre">cmpConstants</span></code> method otherwise.</p>
<p>3. If left/right is <em>InlineAsm</em> instance. Return result of <em>Value</em> pointers
comparison.</p>
<p>4. Explicit association of <em>L</em> (left value) and <em>R</em> (right value). We need to
find out whether values met at the same time, and thus are <em>associated</em>. Or we
need to put the rule: when we treat <em>L</em> < <em>R</em>. Now it is easy: we just return
the result of numbers comparison:</p>
<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="n">std</span><span class="o">::</span><span class="n">pair</span><span class="o"><</span><span class="n">iterator</span><span class="p">,</span> <span class="kt">bool</span><span class="o">></span>
<span class="n">LeftRes</span> <span class="o">=</span> <span class="n">sn_mapL</span><span class="p">.</span><span class="n">insert</span><span class="p">(</span><span class="n">std</span><span class="o">::</span><span class="n">make_pair</span><span class="p">(</span><span class="n">Left</span><span class="p">,</span> <span class="n">sn_mapL</span><span class="p">.</span><span class="n">size</span><span class="p">())),</span>
<span class="n">RightRes</span> <span class="o">=</span> <span class="n">sn_mapR</span><span class="p">.</span><span class="n">insert</span><span class="p">(</span><span class="n">std</span><span class="o">::</span><span class="n">make_pair</span><span class="p">(</span><span class="n">Right</span><span class="p">,</span> <span class="n">sn_mapR</span><span class="p">.</span><span class="n">size</span><span class="p">()));</span>
<span class="k">if</span> <span class="p">(</span><span class="n">LeftRes</span><span class="p">.</span><span class="n">first</span><span class="o">-></span><span class="n">second</span> <span class="o">==</span> <span class="n">RightRes</span><span class="p">.</span><span class="n">first</span><span class="o">-></span><span class="n">second</span><span class="p">)</span> <span class="k">return</span> <span class="mi">0</span><span class="p">;</span>
<span class="k">if</span> <span class="p">(</span><span class="n">LeftRes</span><span class="p">.</span><span class="n">first</span><span class="o">-></span><span class="n">second</span> <span class="o"><</span> <span class="n">RightRes</span><span class="p">.</span><span class="n">first</span><span class="o">-></span><span class="n">second</span><span class="p">)</span> <span class="k">return</span> <span class="mi">-1</span><span class="p">;</span>
<span class="k">return</span> <span class="mi">1</span><span class="p">;</span>
</pre></div>
</div>
<p>Now when <em>cmpValues</em> returns 0, we can proceed the comparison procedure.
Otherwise, if we get (-1 or 1), we need to pass this result to the top level,
and finish comparison procedure.</p>
</div>
</div>
<div class="section" id="cmpconstants">
<h3><a class="toc-backref" href="#id19">cmpConstants</a><a class="headerlink" href="#cmpconstants" title="Permalink to this headline">¶</a></h3>
<p>Performs constants comparison as follows:</p>
<p>1. Compare constant types using <code class="docutils literal notranslate"><span class="pre">cmpType</span></code> method. If the result is -1 or 1,
goto step 2, otherwise proceed to step 3.</p>
<p>2. If types are different, we still can check whether constants could be
losslessly bitcasted to each other. The further explanation is modification of
<code class="docutils literal notranslate"><span class="pre">canLosslesslyBitCastTo</span></code> method.</p>
<blockquote>
<div><p>2.1 Check whether constants are of the first class types
(<code class="docutils literal notranslate"><span class="pre">isFirstClassType</span></code> check):</p>
<p>2.1.1. If both constants are <em>not</em> of the first class type: return result
of <code class="docutils literal notranslate"><span class="pre">cmpType</span></code>.</p>
<p>2.1.2. Otherwise, if left type is not of the first class, return -1. If
right type is not of the first class, return 1.</p>
<p>2.1.3. If both types are of the first class type, proceed to the next step
(2.1.3.1).</p>
<p>2.1.3.1. If types are vectors, compare their bitwidth using the
<em>cmpNumbers</em>. If result is not 0, return it.</p>
<p>2.1.3.2. Different types, but not a vectors:</p>
<ul class="simple">
<li><p>if both of them are pointers, good for us, we can proceed to step 3.</p></li>
<li><p>if one of types is pointer, return result of <em>isPointer</em> flags
comparison (<em>cmpFlags</em> operation).</p></li>
<li><p>otherwise we have no methods to prove bitcastability, and thus return
result of types comparison (-1 or 1).</p></li>
</ul>
</div></blockquote>
<p>Steps below are for the case when types are equal, or case when constants are
bitcastable:</p>
<p>3. One of constants is a “<em>null</em>” value. Return the result of
<code class="docutils literal notranslate"><span class="pre">cmpFlags(L->isNullValue,</span> <span class="pre">R->isNullValue)</span></code> comparison.</p>
<ol class="arabic simple" start="4">
<li><p>Compare value IDs, and return result if it is not 0:</p></li>
</ol>
<div class="highlight-c++ notranslate"><div class="highlight"><pre><span></span><span class="k">if</span> <span class="p">(</span><span class="kt">int</span> <span class="n">Res</span> <span class="o">=</span> <span class="n">cmpNumbers</span><span class="p">(</span><span class="n">L</span><span class="o">-></span><span class="n">getValueID</span><span class="p">(),</span> <span class="n">R</span><span class="o">-></span><span class="n">getValueID</span><span class="p">()))</span>
<span class="k">return</span> <span class="n">Res</span><span class="p">;</span>
</pre></div>
</div>
<p>5. Compare the contents of constants. The comparison depends on the kind of
constants, but on this stage it is just a lexicographical comparison. Just see
how it was described in the beginning of “<em>Functions comparison</em>” paragraph.
Mathematically, it is equal to the next case: we encode left constant and right
constant (with similar way <em>bitcode-writer</em> does). Then compare left code
sequence and right code sequence.</p>
</div>
<div class="section" id="compare-const-basicblock-const-basicblock">
<h3><a class="toc-backref" href="#id20">compare(const BasicBlock*, const BasicBlock*)</a><a class="headerlink" href="#compare-const-basicblock-const-basicblock" title="Permalink to this headline">¶</a></h3>
<p>Compares two <em>BasicBlock</em> instances.</p>
<p>It enumerates instructions from left <em>BB</em> and right <em>BB</em>.</p>
<p>1. It assigns serial numbers to the left and right instructions, using
<code class="docutils literal notranslate"><span class="pre">cmpValues</span></code> method.</p>
<p>2. If one of left or right is <em>GEP</em> (<code class="docutils literal notranslate"><span class="pre">GetElementPtr</span></code>), then treat <em>GEP</em> as
greater than other instructions. If both instructions are <em>GEPs</em> use <code class="docutils literal notranslate"><span class="pre">cmpGEP</span></code>
method for comparison. If result is -1 or 1, pass it to the top-level
comparison (return it).</p>
<blockquote>
<div><p>3.1. Compare operations. Call <code class="docutils literal notranslate"><span class="pre">cmpOperation</span></code> method. If result is -1 or
1, return it.</p>
<p>3.2. Compare number of operands, if result is -1 or 1, return it.</p>
<p>3.3. Compare operands themselves, use <code class="docutils literal notranslate"><span class="pre">cmpValues</span></code> method. Return result
if it is -1 or 1.</p>
<p>3.4. Compare type of operands, using <code class="docutils literal notranslate"><span class="pre">cmpType</span></code> method. Return result if
it is -1 or 1.</p>
<p>3.5. Proceed to the next instruction.</p>
</div></blockquote>
<ol class="arabic" start="4">
<li><p>We can finish instruction enumeration in 3 cases:</p>
<p>4.1. We reached the end of both left and right basic-blocks. We didn’t
exit on steps 1-3, so contents are equal, return 0.</p>
<p>4.2. We have reached the end of the left basic-block. Return -1.</p>
<p>4.3. Return 1 (we reached the end of the right basic block).</p>
</li>
</ol>
</div>
<div class="section" id="cmpgep">
<h3><a class="toc-backref" href="#id21">cmpGEP</a><a class="headerlink" href="#cmpgep" title="Permalink to this headline">¶</a></h3>
<p>Compares two GEPs (<code class="docutils literal notranslate"><span class="pre">getelementptr</span></code> instructions).</p>
<p>It differs from regular operations comparison with the only thing: possibility
to use <code class="docutils literal notranslate"><span class="pre">accumulateConstantOffset</span></code> method.</p>
<p>So, if we get constant offset for both left and right <em>GEPs</em>, then compare it as
numbers, and return comparison result.</p>
<p>Otherwise treat it like a regular operation (see previous paragraph).</p>
</div>
<div class="section" id="cmpoperation">
<h3><a class="toc-backref" href="#id22">cmpOperation</a><a class="headerlink" href="#cmpoperation" title="Permalink to this headline">¶</a></h3>
<p>Compares instruction opcodes and some important operation properties.</p>
<ol class="arabic simple">
<li><p>Compare opcodes, if it differs return the result.</p></li>
<li><p>Compare number of operands. If it differs – return the result.</p></li>
</ol>
<p>3. Compare operation types, use <em>cmpType</em>. All the same – if types are
different, return result.</p>
<p>4. Compare <em>subclassOptionalData</em>, get it with <code class="docutils literal notranslate"><span class="pre">getRawSubclassOptionalData</span></code>
method, and compare it like a numbers.</p>
<ol class="arabic simple" start="5">
<li><p>Compare operand types.</p></li>
</ol>
<p>6. For some particular instructions, check equivalence (relation in our case) of
some significant attributes. For example, we have to compare alignment for
<code class="docutils literal notranslate"><span class="pre">load</span></code> instructions.</p>
</div>
<div class="section" id="o-log-n">
<h3><a class="toc-backref" href="#id23">O(log(N))</a><a class="headerlink" href="#o-log-n" title="Permalink to this headline">¶</a></h3>
<p>Methods described above implement order relationship. And latter, could be used
for nodes comparison in a binary tree. So we can organize functions set into
the binary tree and reduce the cost of lookup procedure from
O(N*N) to O(log(N)).</p>
</div>
</div>
<div class="section" id="merging-process-mergetwofunctions">
<h2><a class="toc-backref" href="#id24">Merging process, mergeTwoFunctions</a><a class="headerlink" href="#merging-process-mergetwofunctions" title="Permalink to this headline">¶</a></h2>
<p>Once <em>MergeFunctions</em> detected that current function (<em>G</em>) is equal to one that
were analyzed before (function <em>F</em>) it calls <code class="docutils literal notranslate"><span class="pre">mergeTwoFunctions(Function*,</span>
<span class="pre">Function*)</span></code>.</p>
<p>Operation affects <code class="docutils literal notranslate"><span class="pre">FnTree</span></code> contents with next way: <em>F</em> will stay in
<code class="docutils literal notranslate"><span class="pre">FnTree</span></code>. <em>G</em> being equal to <em>F</em> will not be added to <code class="docutils literal notranslate"><span class="pre">FnTree</span></code>. Calls of
<em>G</em> would be replaced with something else. It changes bodies of callers. So,
functions that calls <em>G</em> would be put into <code class="docutils literal notranslate"><span class="pre">Deferred</span></code> set and removed from
<code class="docutils literal notranslate"><span class="pre">FnTree</span></code>, and analyzed again.</p>
<p>The approach is next:</p>
<p>1. Most wished case: when we can use alias and both of <em>F</em> and <em>G</em> are weak. We
make both of them with aliases to the third strong function <em>H</em>. Actually <em>H</em>
is <em>F</em>. See below how it’s made (but it’s better to look straight into the
source code). Well, this is a case when we can just replace <em>G</em> with <em>F</em>
everywhere, we use <code class="docutils literal notranslate"><span class="pre">replaceAllUsesWith</span></code> operation here (<em>RAUW</em>).</p>
<p>2. <em>F</em> could not be overridden, while <em>G</em> could. It would be good to do the
next: after merging the places where overridable function were used, still use
overridable stub. So try to make <em>G</em> alias to <em>F</em>, or create overridable tail
call wrapper around <em>F</em> and replace <em>G</em> with that call.</p>
<p>3. Neither <em>F</em> nor <em>G</em> could be overridden. We can’t use <em>RAUW</em>. We can just
change the callers: call <em>F</em> instead of <em>G</em>. That’s what
<code class="docutils literal notranslate"><span class="pre">replaceDirectCallers</span></code> does.</p>
<p>Below is a detailed body description.</p>
<div class="section" id="if-f-may-be-overridden">
<h3><a class="toc-backref" href="#id25">If “F” may be overridden</a><a class="headerlink" href="#if-f-may-be-overridden" title="Permalink to this headline">¶</a></h3>
<p>As follows from <code class="docutils literal notranslate"><span class="pre">mayBeOverridden</span></code> comments: “whether the definition of this
global may be replaced by something non-equivalent at link time”. If so, that’s
ok: we can use alias to <em>F</em> instead of <em>G</em> or change call instructions itself.</p>
<div class="section" id="hasglobalaliases-removeusers">
<h4><a class="toc-backref" href="#id26">HasGlobalAliases, removeUsers</a><a class="headerlink" href="#hasglobalaliases-removeusers" title="Permalink to this headline">¶</a></h4>
<p>First consider the case when we have global aliases of one function name to
another. Our purpose is make both of them with aliases to the third strong
function. Though if we keep <em>F</em> alive and without major changes we can leave it
in <code class="docutils literal notranslate"><span class="pre">FnTree</span></code>. Try to combine these two goals.</p>
<p>Do stub replacement of <em>F</em> itself with an alias to <em>F</em>.</p>
<p>1. Create stub function <em>H</em>, with the same name and attributes like function
<em>F</em>. It takes maximum alignment of <em>F</em> and <em>G</em>.</p>
<p>2. Replace all uses of function <em>F</em> with uses of function <em>H</em>. It is the two
steps procedure instead. First of all, we must take into account, all functions
from whom <em>F</em> is called would be changed: since we change the call argument
(from <em>F</em> to <em>H</em>). If so we must to review these caller functions again after
this procedure. We remove callers from <code class="docutils literal notranslate"><span class="pre">FnTree</span></code>, method with name
<code class="docutils literal notranslate"><span class="pre">removeUsers(F)</span></code> does that (don’t confuse with <code class="docutils literal notranslate"><span class="pre">replaceAllUsesWith</span></code>):</p>
<blockquote>
<div><p>2.1. <code class="docutils literal notranslate"><span class="pre">Inside</span> <span class="pre">removeUsers(Value*</span>
<span class="pre">V)</span></code> we go through the all values that use value <em>V</em> (or <em>F</em> in our context).
If value is instruction, we go to function that holds this instruction and
mark it as to-be-analyzed-again (put to <code class="docutils literal notranslate"><span class="pre">Deferred</span></code> set), we also remove
caller from <code class="docutils literal notranslate"><span class="pre">FnTree</span></code>.</p>
<p>2.2. Now we can do the replacement: call <code class="docutils literal notranslate"><span class="pre">F->replaceAllUsesWith(H)</span></code>.</p>
</div></blockquote>
<p>3. <em>H</em> (that now “officially” plays <em>F</em>’s role) is replaced with alias to <em>F</em>.
Do the same with <em>G</em>: replace it with alias to <em>F</em>. So finally everywhere <em>F</em>
was used, we use <em>H</em> and it is alias to <em>F</em>, and everywhere <em>G</em> was used we
also have alias to <em>F</em>.</p>
<ol class="arabic simple" start="4">
<li><p>Set <em>F</em> linkage to private. Make it strong :-)</p></li>
</ol>
</div>
<div class="section" id="no-global-aliases-replacedirectcallers">
<h4><a class="toc-backref" href="#id27">No global aliases, replaceDirectCallers</a><a class="headerlink" href="#no-global-aliases-replacedirectcallers" title="Permalink to this headline">¶</a></h4>
<p>If global aliases are not supported. We call <code class="docutils literal notranslate"><span class="pre">replaceDirectCallers</span></code>. Just
go through all calls of <em>G</em> and replace it with calls of <em>F</em>. If you look into
the method you will see that it scans all uses of <em>G</em> too, and if use is callee
(if user is call instruction and <em>G</em> is used as what to be called), we replace
it with use of <em>F</em>.</p>
<div class="section" id="if-f-could-not-be-overridden-fix-it">
<h5><a class="toc-backref" href="#id28">If “F” could not be overridden, fix it!</a><a class="headerlink" href="#if-f-could-not-be-overridden-fix-it" title="Permalink to this headline">¶</a></h5>
<p>We call <code class="docutils literal notranslate"><span class="pre">writeThunkOrAlias(Function</span> <span class="pre">*F,</span> <span class="pre">Function</span> <span class="pre">*G)</span></code>. Here we try to replace
<em>G</em> with alias to <em>F</em> first. The next conditions are essential:</p>
<ul class="simple">
<li><p>target should support global aliases,</p></li>
<li><p>the address itself of <em>G</em> should be not significant, not named and not
referenced anywhere,</p></li>
<li><p>function should come with external, local or weak linkage.</p></li>
</ul>
<p>Otherwise we write thunk: some wrapper that has <em>G’s</em> interface and calls <em>F</em>,
so <em>G</em> could be replaced with this wrapper.</p>
<p><em>writeAlias</em></p>
<p>As follows from <em>llvm</em> reference:</p>
<p>“Aliases act as <em>second name</em> for the aliasee value”. So we just want to create
a second name for <em>F</em> and use it instead of <em>G</em>:</p>
<ol class="arabic">
<li><p>create global alias itself (<em>GA</em>),</p></li>
<li><p>adjust alignment of <em>F</em> so it must be maximum of current and <em>G’s</em> alignment;</p></li>
<li><p>replace uses of <em>G</em>:</p>
<p>3.1. first mark all callers of <em>G</em> as to-be-analyzed-again, using
<code class="docutils literal notranslate"><span class="pre">removeUsers</span></code> method (see chapter above),</p>
<p>3.2. call <code class="docutils literal notranslate"><span class="pre">G->replaceAllUsesWith(GA)</span></code>.</p>
</li>
<li><p>Get rid of <em>G</em>.</p></li>
</ol>
<p><em>writeThunk</em></p>
<p>As it written in method comments:</p>
<p>“Replace G with a simple tail call to bitcast(F). Also replace direct uses of G
with bitcast(F). Deletes G.”</p>
<p>In general it does the same as usual when we want to replace callee, except the
first point:</p>
<p>1. We generate tail call wrapper around <em>F</em>, but with interface that allows use
it instead of <em>G</em>.</p>
<ol class="arabic simple" start="2">
<li><p>“As-usual”: <code class="docutils literal notranslate"><span class="pre">removeUsers</span></code> and <code class="docutils literal notranslate"><span class="pre">replaceAllUsesWith</span></code> then.</p></li>
<li><p>Get rid of <em>G</em>.</p></li>
</ol>
</div>
</div>
</div>
</div>
</div>
<div class="clearer"></div>
</div>
</div>
</div>
<div class="clearer"></div>
</div>
<div class="related" role="navigation" aria-label="related navigation">
<h3>Navigation</h3>
<ul>
<li class="right" style="margin-right: 10px">
<a href="genindex.html" title="General Index"
>index</a></li>
<li class="right" >
<a href="MCJITDesignAndImplementation.html" title="MCJIT Design and Implementation"
>next</a> |</li>
<li class="right" >
<a href="MemorySSA.html" title="MemorySSA"
>previous</a> |</li>
<li><a href="https://llvm.org/">LLVM Home</a> | </li>
<li><a href="index.html">Documentation</a>»</li>
<li class="nav-item nav-item-1"><a href="UserGuides.html" >User Guides</a> »</li>
<li class="nav-item nav-item-this"><a href="">MergeFunctions pass, how it works</a></li>
</ul>
</div>
<div class="footer" role="contentinfo">
© Copyright 2003-2021, LLVM Project.
Last updated on 2021-09-18.
Created using <a href="https://www.sphinx-doc.org/">Sphinx</a> 3.5.4.
</div>
</body>
</html>
|
