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
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
|
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1//EN"
"http://www.w3.org/TR/xhtml11/DTD/xhtml11.dtd">
<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en">
<head>
<meta http-equiv="Content-Type" content="application/xhtml+xml; charset=UTF-8" />
<meta name="generator" content="AsciiDoc 10.2.0" />
<title>Parallel Checkout Design Notes</title>
<style type="text/css">
/* Shared CSS for AsciiDoc xhtml11 and html5 backends */
/* Default font. */
body {
font-family: Georgia,serif;
}
/* Title font. */
h1, h2, h3, h4, h5, h6,
div.title, caption.title,
thead, p.table.header,
#toctitle,
#author, #revnumber, #revdate, #revremark,
#footer {
font-family: Arial,Helvetica,sans-serif;
}
body {
margin: 1em 5% 1em 5%;
}
a {
color: blue;
text-decoration: underline;
}
a:visited {
color: fuchsia;
}
em {
font-style: italic;
color: navy;
}
strong {
font-weight: bold;
color: #083194;
}
h1, h2, h3, h4, h5, h6 {
color: #527bbd;
margin-top: 1.2em;
margin-bottom: 0.5em;
line-height: 1.3;
}
h1, h2, h3 {
border-bottom: 2px solid silver;
}
h2 {
padding-top: 0.5em;
}
h3 {
float: left;
}
h3 + * {
clear: left;
}
h5 {
font-size: 1.0em;
}
div.sectionbody {
margin-left: 0;
}
hr {
border: 1px solid silver;
}
p {
margin-top: 0.5em;
margin-bottom: 0.5em;
}
ul, ol, li > p {
margin-top: 0;
}
ul > li { color: #aaa; }
ul > li > * { color: black; }
.monospaced, code, pre {
font-family: "Courier New", Courier, monospace;
font-size: inherit;
color: navy;
padding: 0;
margin: 0;
}
pre {
white-space: pre-wrap;
}
#author {
color: #527bbd;
font-weight: bold;
font-size: 1.1em;
}
#email {
}
#revnumber, #revdate, #revremark {
}
#footer {
font-size: small;
border-top: 2px solid silver;
padding-top: 0.5em;
margin-top: 4.0em;
}
#footer-text {
float: left;
padding-bottom: 0.5em;
}
#footer-badges {
float: right;
padding-bottom: 0.5em;
}
#preamble {
margin-top: 1.5em;
margin-bottom: 1.5em;
}
div.imageblock, div.exampleblock, div.verseblock,
div.quoteblock, div.literalblock, div.listingblock, div.sidebarblock,
div.admonitionblock {
margin-top: 1.0em;
margin-bottom: 1.5em;
}
div.admonitionblock {
margin-top: 2.0em;
margin-bottom: 2.0em;
margin-right: 10%;
color: #606060;
}
div.content { /* Block element content. */
padding: 0;
}
/* Block element titles. */
div.title, caption.title {
color: #527bbd;
font-weight: bold;
text-align: left;
margin-top: 1.0em;
margin-bottom: 0.5em;
}
div.title + * {
margin-top: 0;
}
td div.title:first-child {
margin-top: 0.0em;
}
div.content div.title:first-child {
margin-top: 0.0em;
}
div.content + div.title {
margin-top: 0.0em;
}
div.sidebarblock > div.content {
background: #ffffee;
border: 1px solid #dddddd;
border-left: 4px solid #f0f0f0;
padding: 0.5em;
}
div.listingblock > div.content {
border: 1px solid #dddddd;
border-left: 5px solid #f0f0f0;
background: #f8f8f8;
padding: 0.5em;
}
div.quoteblock, div.verseblock {
padding-left: 1.0em;
margin-left: 1.0em;
margin-right: 10%;
border-left: 5px solid #f0f0f0;
color: #888;
}
div.quoteblock > div.attribution {
padding-top: 0.5em;
text-align: right;
}
div.verseblock > pre.content {
font-family: inherit;
font-size: inherit;
}
div.verseblock > div.attribution {
padding-top: 0.75em;
text-align: left;
}
/* DEPRECATED: Pre version 8.2.7 verse style literal block. */
div.verseblock + div.attribution {
text-align: left;
}
div.admonitionblock .icon {
vertical-align: top;
font-size: 1.1em;
font-weight: bold;
text-decoration: underline;
color: #527bbd;
padding-right: 0.5em;
}
div.admonitionblock td.content {
padding-left: 0.5em;
border-left: 3px solid #dddddd;
}
div.exampleblock > div.content {
border-left: 3px solid #dddddd;
padding-left: 0.5em;
}
div.imageblock div.content { padding-left: 0; }
span.image img { border-style: none; vertical-align: text-bottom; }
a.image:visited { color: white; }
dl {
margin-top: 0.8em;
margin-bottom: 0.8em;
}
dt {
margin-top: 0.5em;
margin-bottom: 0;
font-style: normal;
color: navy;
}
dd > *:first-child {
margin-top: 0.1em;
}
ul, ol {
list-style-position: outside;
}
ol.arabic {
list-style-type: decimal;
}
ol.loweralpha {
list-style-type: lower-alpha;
}
ol.upperalpha {
list-style-type: upper-alpha;
}
ol.lowerroman {
list-style-type: lower-roman;
}
ol.upperroman {
list-style-type: upper-roman;
}
div.compact ul, div.compact ol,
div.compact p, div.compact p,
div.compact div, div.compact div {
margin-top: 0.1em;
margin-bottom: 0.1em;
}
tfoot {
font-weight: bold;
}
td > div.verse {
white-space: pre;
}
div.hdlist {
margin-top: 0.8em;
margin-bottom: 0.8em;
}
div.hdlist tr {
padding-bottom: 15px;
}
dt.hdlist1.strong, td.hdlist1.strong {
font-weight: bold;
}
td.hdlist1 {
vertical-align: top;
font-style: normal;
padding-right: 0.8em;
color: navy;
}
td.hdlist2 {
vertical-align: top;
}
div.hdlist.compact tr {
margin: 0;
padding-bottom: 0;
}
.comment {
background: yellow;
}
.footnote, .footnoteref {
font-size: 0.8em;
}
span.footnote, span.footnoteref {
vertical-align: super;
}
#footnotes {
margin: 20px 0 20px 0;
padding: 7px 0 0 0;
}
#footnotes div.footnote {
margin: 0 0 5px 0;
}
#footnotes hr {
border: none;
border-top: 1px solid silver;
height: 1px;
text-align: left;
margin-left: 0;
width: 20%;
min-width: 100px;
}
div.colist td {
padding-right: 0.5em;
padding-bottom: 0.3em;
vertical-align: top;
}
div.colist td img {
margin-top: 0.3em;
}
@media print {
#footer-badges { display: none; }
}
#toc {
margin-bottom: 2.5em;
}
#toctitle {
color: #527bbd;
font-size: 1.1em;
font-weight: bold;
margin-top: 1.0em;
margin-bottom: 0.1em;
}
div.toclevel0, div.toclevel1, div.toclevel2, div.toclevel3, div.toclevel4 {
margin-top: 0;
margin-bottom: 0;
}
div.toclevel2 {
margin-left: 2em;
font-size: 0.9em;
}
div.toclevel3 {
margin-left: 4em;
font-size: 0.9em;
}
div.toclevel4 {
margin-left: 6em;
font-size: 0.9em;
}
span.aqua { color: aqua; }
span.black { color: black; }
span.blue { color: blue; }
span.fuchsia { color: fuchsia; }
span.gray { color: gray; }
span.green { color: green; }
span.lime { color: lime; }
span.maroon { color: maroon; }
span.navy { color: navy; }
span.olive { color: olive; }
span.purple { color: purple; }
span.red { color: red; }
span.silver { color: silver; }
span.teal { color: teal; }
span.white { color: white; }
span.yellow { color: yellow; }
span.aqua-background { background: aqua; }
span.black-background { background: black; }
span.blue-background { background: blue; }
span.fuchsia-background { background: fuchsia; }
span.gray-background { background: gray; }
span.green-background { background: green; }
span.lime-background { background: lime; }
span.maroon-background { background: maroon; }
span.navy-background { background: navy; }
span.olive-background { background: olive; }
span.purple-background { background: purple; }
span.red-background { background: red; }
span.silver-background { background: silver; }
span.teal-background { background: teal; }
span.white-background { background: white; }
span.yellow-background { background: yellow; }
span.big { font-size: 2em; }
span.small { font-size: 0.6em; }
span.underline { text-decoration: underline; }
span.overline { text-decoration: overline; }
span.line-through { text-decoration: line-through; }
div.unbreakable { page-break-inside: avoid; }
/*
* xhtml11 specific
*
* */
div.tableblock {
margin-top: 1.0em;
margin-bottom: 1.5em;
}
div.tableblock > table {
border: 3px solid #527bbd;
}
thead, p.table.header {
font-weight: bold;
color: #527bbd;
}
p.table {
margin-top: 0;
}
/* Because the table frame attribute is overridden by CSS in most browsers. */
div.tableblock > table[frame="void"] {
border-style: none;
}
div.tableblock > table[frame="hsides"] {
border-left-style: none;
border-right-style: none;
}
div.tableblock > table[frame="vsides"] {
border-top-style: none;
border-bottom-style: none;
}
/*
* html5 specific
*
* */
table.tableblock {
margin-top: 1.0em;
margin-bottom: 1.5em;
}
thead, p.tableblock.header {
font-weight: bold;
color: #527bbd;
}
p.tableblock {
margin-top: 0;
}
table.tableblock {
border-width: 3px;
border-spacing: 0px;
border-style: solid;
border-color: #527bbd;
border-collapse: collapse;
}
th.tableblock, td.tableblock {
border-width: 1px;
padding: 4px;
border-style: solid;
border-color: #527bbd;
}
table.tableblock.frame-topbot {
border-left-style: hidden;
border-right-style: hidden;
}
table.tableblock.frame-sides {
border-top-style: hidden;
border-bottom-style: hidden;
}
table.tableblock.frame-none {
border-style: hidden;
}
th.tableblock.halign-left, td.tableblock.halign-left {
text-align: left;
}
th.tableblock.halign-center, td.tableblock.halign-center {
text-align: center;
}
th.tableblock.halign-right, td.tableblock.halign-right {
text-align: right;
}
th.tableblock.valign-top, td.tableblock.valign-top {
vertical-align: top;
}
th.tableblock.valign-middle, td.tableblock.valign-middle {
vertical-align: middle;
}
th.tableblock.valign-bottom, td.tableblock.valign-bottom {
vertical-align: bottom;
}
/*
* manpage specific
*
* */
body.manpage h1 {
padding-top: 0.5em;
padding-bottom: 0.5em;
border-top: 2px solid silver;
border-bottom: 2px solid silver;
}
body.manpage h2 {
border-style: none;
}
body.manpage div.sectionbody {
margin-left: 3em;
}
@media print {
body.manpage div#toc { display: none; }
}
</style>
<script type="text/javascript">
/*<+'])');
// Function that scans the DOM tree for header elements (the DOM2
// nodeIterator API would be a better technique but not supported by all
// browsers).
var iterate = function (el) {
for (var i = el.firstChild; i != null; i = i.nextSibling) {
if (i.nodeType == 1 /* Node.ELEMENT_NODE */) {
var mo = re.exec(i.tagName);
if (mo && (i.getAttribute("class") || i.getAttribute("className")) != "float") {
result[result.length] = new TocEntry(i, getText(i), mo[1]-1);
}
iterate(i);
}
}
}
iterate(el);
return result;
}
var toc = document.getElementById("toc");
if (!toc) {
return;
}
// Delete existing TOC entries in case we're reloading the TOC.
var tocEntriesToRemove = [];
var i;
for (i = 0; i < toc.childNodes.length; i++) {
var entry = toc.childNodes[i];
if (entry.nodeName.toLowerCase() == 'div'
&& entry.getAttribute("class")
&& entry.getAttribute("class").match(/^toclevel/))
tocEntriesToRemove.push(entry);
}
for (i = 0; i < tocEntriesToRemove.length; i++) {
toc.removeChild(tocEntriesToRemove[i]);
}
// Rebuild TOC entries.
var entries = tocEntries(document.getElementById("content"), toclevels);
for (var i = 0; i < entries.length; ++i) {
var entry = entries[i];
if (entry.element.id == "")
entry.element.id = "_toc_" + i;
var a = document.createElement("a");
a.href = "#" + entry.element.id;
a.appendChild(document.createTextNode(entry.text));
var div = document.createElement("div");
div.appendChild(a);
div.className = "toclevel" + entry.toclevel;
toc.appendChild(div);
}
if (entries.length == 0)
toc.parentNode.removeChild(toc);
},
/////////////////////////////////////////////////////////////////////
// Footnotes generator
/////////////////////////////////////////////////////////////////////
/* Based on footnote generation code from:
* http://www.brandspankingnew.net/archive/2005/07/format_footnote.html
*/
footnotes: function () {
// Delete existing footnote entries in case we're reloading the footnodes.
var i;
var noteholder = document.getElementById("footnotes");
if (!noteholder) {
return;
}
var entriesToRemove = [];
for (i = 0; i < noteholder.childNodes.length; i++) {
var entry = noteholder.childNodes[i];
if (entry.nodeName.toLowerCase() == 'div' && entry.getAttribute("class") == "footnote")
entriesToRemove.push(entry);
}
for (i = 0; i < entriesToRemove.length; i++) {
noteholder.removeChild(entriesToRemove[i]);
}
// Rebuild footnote entries.
var cont = document.getElementById("content");
var spans = cont.getElementsByTagName("span");
var refs = {};
var n = 0;
for (i=0; i<spans.length; i++) {
if (spans[i].className == "footnote") {
n++;
var note = spans[i].getAttribute("data-note");
if (!note) {
// Use [\s\S] in place of . so multi-line matches work.
// Because JavaScript has no s (dotall) regex flag.
note = spans[i].innerHTML.match(/\s*\[([\s\S]*)]\s*/)[1];
spans[i].innerHTML =
"[<a id='_footnoteref_" + n + "' href='#_footnote_" + n +
"' title='View footnote' class='footnote'>" + n + "</a>]";
spans[i].setAttribute("data-note", note);
}
noteholder.innerHTML +=
"<div class='footnote' id='_footnote_" + n + "'>" +
"<a href='#_footnoteref_" + n + "' title='Return to text'>" +
n + "</a>. " + note + "</div>";
var id =spans[i].getAttribute("id");
if (id != null) refs["#"+id] = n;
}
}
if (n == 0)
noteholder.parentNode.removeChild(noteholder);
else {
// Process footnoterefs.
for (i=0; i<spans.length; i++) {
if (spans[i].className == "footnoteref") {
var href = spans[i].getElementsByTagName("a")[0].getAttribute("href");
href = href.match(/#.*/)[0]; // Because IE return full URL.
n = refs[href];
spans[i].innerHTML =
"[<a href='#_footnote_" + n +
"' title='View footnote' class='footnote'>" + n + "</a>]";
}
}
}
},
install: function(toclevels) {
var timerId;
function reinstall() {
asciidoc.footnotes();
if (toclevels) {
asciidoc.toc(toclevels);
}
}
function reinstallAndRemoveTimer() {
clearInterval(timerId);
reinstall();
}
timerId = setInterval(reinstall, 500);
if (document.addEventListener)
document.addEventListener("DOMContentLoaded", reinstallAndRemoveTimer, false);
else
window.onload = reinstallAndRemoveTimer;
}
}
asciidoc.install();
/*]]>*/
</script>
</head>
<body class="article">
<div id="header">
<h1>Parallel Checkout Design Notes</h1>
<span id="revdate">2023-05-09</span>
</div>
<div id="content">
<div id="preamble">
<div class="sectionbody">
<div class="paragraph"><p>The "Parallel Checkout" feature attempts to use multiple processes to
parallelize the work of uncompressing the blobs, applying in-core
filters, and writing the resulting contents to the working tree during a
checkout operation. It can be used by all checkout-related commands,
such as <code>clone</code>, <code>checkout</code>, <code>reset</code>, <code>sparse-checkout</code>, and others.</p></div>
<div class="paragraph"><p>These commands share the following basic structure:</p></div>
<div class="ulist"><ul>
<li>
<p>
Step 1: Read the current index file into memory.
</p>
</li>
<li>
<p>
Step 2: Modify the in-memory index based upon the command, and
temporarily mark all cache entries that need to be updated.
</p>
</li>
<li>
<p>
Step 3: Populate the working tree to match the new candidate index.
This includes iterating over all of the to-be-updated cache entries
and delete, create, or overwrite the associated files in the working
tree.
</p>
</li>
<li>
<p>
Step 4: Write the new index to disk.
</p>
</li>
</ul></div>
<div class="paragraph"><p>Step 3 is the focus of the "parallel checkout" effort described here.</p></div>
</div>
</div>
<div class="sect1">
<h2 id="_sequential_implementation">Sequential Implementation</h2>
<div class="sectionbody">
<div class="paragraph"><p>For the purposes of discussion here, the current sequential
implementation of Step 3 is divided in 3 parts, each one implemented in
its own function:</p></div>
<div class="ulist"><ul>
<li>
<p>
Step 3a: <code>unpack-trees.c:check_updates()</code> contains a series of
sequential loops iterating over the <code>cache_entry</code>'s array. The main
loop in this function calls the Step 3b function for each of the
to-be-updated entries.
</p>
</li>
<li>
<p>
Step 3b: <code>entry.c:checkout_entry()</code> examines the existing working tree
for file conflicts, collisions, and unsaved changes. It removes files
and creates leading directories as necessary. It calls the Step 3c
function for each entry to be written.
</p>
</li>
<li>
<p>
Step 3c: <code>entry.c:write_entry()</code> loads the blob into memory, smudges
it if necessary, creates the file in the working tree, writes the
smudged contents, calls <code>fstat()</code> or <code>lstat()</code>, and updates the
associated <code>cache_entry</code> struct with the stat information gathered.
</p>
</li>
</ul></div>
<div class="paragraph"><p>It wouldn’t be safe to perform Step 3b in parallel, as there could be
race conditions between file creations and removals. Instead, the
parallel checkout framework lets the sequential code handle Step 3b,
and uses parallel workers to replace the sequential
<code>entry.c:write_entry()</code> calls from Step 3c.</p></div>
</div>
</div>
<div class="sect1">
<h2 id="_rejected_multi_threaded_solution">Rejected Multi-Threaded Solution</h2>
<div class="sectionbody">
<div class="paragraph"><p>The most "straightforward" implementation would be to spread the set of
to-be-updated cache entries across multiple threads. But due to the
thread-unsafe functions in the object database code, we would have to use locks to
coordinate the parallel operation. An early prototype of this solution
showed that the multi-threaded checkout would bring performance
improvements over the sequential code, but there was still too much lock
contention. A <code>perf</code> profiling indicated that around 20% of the runtime
during a local Linux clone (on an SSD) was spent in locking functions.
For this reason this approach was rejected in favor of using multiple
child processes, which led to a better performance.</p></div>
</div>
</div>
<div class="sect1">
<h2 id="_multi_process_solution">Multi-Process Solution</h2>
<div class="sectionbody">
<div class="paragraph"><p>Parallel checkout alters the aforementioned Step 3 to use multiple
<code>checkout--worker</code> background processes to distribute the work. The
long-running worker processes are controlled by the foreground Git
command using the existing run-command API.</p></div>
<div class="sect2">
<h3 id="_overview">Overview</h3>
<div class="paragraph"><p>Step 3b is only slightly altered; for each entry to be checked out, the
main process performs the following steps:</p></div>
<div class="ulist"><ul>
<li>
<p>
M1: Check whether there is any untracked or unclean file in the
working tree which would be overwritten by this entry, and decide
whether to proceed (removing the file(s)) or not.
</p>
</li>
<li>
<p>
M2: Create the leading directories.
</p>
</li>
<li>
<p>
M3: Load the conversion attributes for the entry’s path.
</p>
</li>
<li>
<p>
M4: Check, based on the entry’s type and conversion attributes,
whether the entry is eligible for parallel checkout (more on this
later). If it is eligible, enqueue the entry and the loaded
attributes to later write the entry in parallel. If not, write the
entry right away, using the default sequential code.
</p>
</li>
</ul></div>
<div class="paragraph"><p>Note: we save the conversion attributes associated with each entry
because the workers don’t have access to the main process' index state,
so they can’t load the attributes by themselves (and the attributes are
needed to properly smudge the entry). Additionally, this has a positive
impact on performance as (1) we don’t need to load the attributes twice
and (2) the attributes machinery is optimized to handle paths in
sequential order.</p></div>
<div class="paragraph"><p>After all entries have passed through the above steps, the main process
checks if the number of enqueued entries is sufficient to spread among
the workers. If not, it just writes them sequentially. Otherwise, it
spawns the workers and distributes the queued entries uniformly in
continuous chunks. This aims to minimize the chances of two workers
writing to the same directory simultaneously, which could increase lock
contention in the kernel.</p></div>
<div class="paragraph"><p>Then, for each assigned item, each worker:</p></div>
<div class="ulist"><ul>
<li>
<p>
W1: Checks if there is any non-directory file in the leading part of
the entry’s path or if there already exists a file at the entry' path.
If so, mark the entry with <code>PC_ITEM_COLLIDED</code> and skip it (more on
this later).
</p>
</li>
<li>
<p>
W2: Creates the file (with O_CREAT and O_EXCL).
</p>
</li>
<li>
<p>
W3: Loads the blob into memory (inflating and delta reconstructing
it).
</p>
</li>
<li>
<p>
W4: Applies any required in-process filter, like end-of-line
conversion and re-encoding.
</p>
</li>
<li>
<p>
W5: Writes the result to the file descriptor opened at W2.
</p>
</li>
<li>
<p>
W6: Calls <code>fstat()</code> or lstat()` on the just-written path, and sends
the result back to the main process, together with the end status of
the operation and the item’s identification number.
</p>
</li>
</ul></div>
<div class="paragraph"><p>Note that, when possible, steps W3 to W5 are delegated to the streaming
machinery, removing the need to keep the entire blob in memory.</p></div>
<div class="paragraph"><p>If the worker fails to read the blob or to write it to the working tree,
it removes the created file to avoid leaving empty files behind. This is
the <strong>only</strong> time a worker is allowed to remove a file.</p></div>
<div class="paragraph"><p>As mentioned earlier, it is the responsibility of the main process to
remove any file that blocks the checkout operation (or abort if the
removal(s) would cause data loss and the user didn’t ask to <code>--force</code>).
This is crucial to avoid race conditions and also to properly detect
path collisions at Step W1.</p></div>
<div class="paragraph"><p>After the workers finish writing the items and sending back the required
information, the main process handles the results in two steps:</p></div>
<div class="ulist"><ul>
<li>
<p>
First, it updates the in-memory index with the <code>lstat()</code> information
sent by the workers. (This must be done first as this information
might me required in the following step.)
</p>
</li>
<li>
<p>
Then it writes the items which collided on disk (i.e. items marked
with <code>PC_ITEM_COLLIDED</code>). More on this below.
</p>
</li>
</ul></div>
</div>
</div>
</div>
<div class="sect1">
<h2 id="_path_collisions">Path Collisions</h2>
<div class="sectionbody">
<div class="paragraph"><p>Path collisions happen when two different paths correspond to the same
entry in the file system. E.g. the paths <em>a</em> and <em>A</em> would collide in a
case-insensitive file system.</p></div>
<div class="paragraph"><p>The sequential checkout deals with collisions in the same way that it
deals with files that were already present in the working tree before
checkout. Basically, it checks if the path that it wants to write
already exists on disk, makes sure the existing file doesn’t have
unsaved data, and then overwrites it. (To be more pedantic: it deletes
the existing file and creates the new one.) So, if there are multiple
colliding files to be checked out, the sequential code will write each
one of them but only the last will actually survive on disk.</p></div>
<div class="paragraph"><p>Parallel checkout aims to reproduce the same behavior. However, we
cannot let the workers racily write to the same file on disk. Instead,
the workers detect when the entry that they want to check out would
collide with an existing file, and mark it with <code>PC_ITEM_COLLIDED</code>.
Later, the main process can sequentially feed these entries back to
<code>checkout_entry()</code> without the risk of race conditions. On clone, this
also has the effect of marking the colliding entries to later emit a
warning for the user, like the classic sequential checkout does.</p></div>
<div class="paragraph"><p>The workers are able to detect both collisions among the entries being
concurrently written and collisions between a parallel-eligible entry
and an ineligible entry. The general idea for collision detection is
quite straightforward: for each parallel-eligible entry, the main
process must remove all files that prevent this entry from being written
(before enqueueing it). This includes any non-directory file in the
leading path of the entry. Later, when a worker gets assigned the entry,
it looks again for the non-directories files and for an already existing
file at the entry’s path. If any of these checks finds something, the
worker knows that there was a path collision.</p></div>
<div class="paragraph"><p>Because parallel checkout can distinguish path collisions from the case
where the file was already present in the working tree before checkout,
we could alternatively choose to skip the checkout of colliding entries.
However, each entry that doesn’t get written would have NULL <code>lstat()</code>
fields on the index. This could cause performance penalties for
subsequent commands that need to refresh the index, as they would have
to go to the file system to see if the entry is dirty. Thus, if we have
N entries in a colliding group and we decide to write and <code>lstat()</code> only
one of them, every subsequent <code>git-status</code> will have to read, convert,
and hash the written file N - 1 times. By checking out all colliding
entries (like the sequential code does), we only pay the overhead once,
during checkout.</p></div>
</div>
</div>
<div class="sect1">
<h2 id="_eligible_entries_for_parallel_checkout">Eligible Entries for Parallel Checkout</h2>
<div class="sectionbody">
<div class="paragraph"><p>As previously mentioned, not all entries passed to <code>checkout_entry()</code>
will be considered eligible for parallel checkout. More specifically, we
exclude:</p></div>
<div class="ulist"><ul>
<li>
<p>
Symbolic links; to avoid race conditions that, in combination with
path collisions, could cause workers to write files at the wrong
place. For example, if we were to concurrently check out a symlink
<em>a</em> → <em>b</em> and a regular file <em>A/f</em> in a case-insensitive file system,
we could potentially end up writing the file <em>A/f</em> at <em>a/f</em>, due to a
race condition.
</p>
</li>
<li>
<p>
Regular files that require external filters (either "one shot" filters
or long-running process filters). These filters are black-boxes to Git
and may have their own internal locking or non-concurrent assumptions.
So it might not be safe to run multiple instances in parallel.
</p>
<div class="paragraph"><p>Besides, long-running filters may use the delayed checkout feature to
postpone the return of some filtered blobs. The delayed checkout queue
and the parallel checkout queue are not compatible and should remain
separate.</p></div>
<div class="paragraph"><p>Note: regular files that only require internal filters, like end-of-line
conversion and re-encoding, are eligible for parallel checkout.</p></div>
</li>
</ul></div>
<div class="paragraph"><p>Ineligible entries are checked out by the classic sequential codepath
<strong>before</strong> spawning workers.</p></div>
<div class="paragraph"><p>Note: submodules’s files are also eligible for parallel checkout (as
long as they don’t fall into any of the excluding categories mentioned
above). But since each submodule is checked out in its own child
process, we don’t mix the superproject’s and the submodules' files in
the same parallel checkout process or queue.</p></div>
</div>
</div>
<div class="sect1">
<h2 id="_the_api">The API</h2>
<div class="sectionbody">
<div class="paragraph"><p>The parallel checkout API was designed with the goal of minimizing
changes to the current users of the checkout machinery. This means that
they don’t have to call a different function for sequential or parallel
checkout. As already mentioned, <code>checkout_entry()</code> will automatically
insert the given entry in the parallel checkout queue when this feature
is enabled and the entry is eligible; otherwise, it will just write the
entry right away, using the sequential code. In general, callers of the
parallel checkout API should look similar to this:</p></div>
<div class="listingblock">
<div class="content">
<pre><code>int pc_workers, pc_threshold, err = 0;
struct checkout state;
get_parallel_checkout_configs(&pc_workers, &pc_threshold);
/*
* This check is not strictly required, but it
* should save some time in sequential mode.
*/
if (pc_workers > 1)
init_parallel_checkout();
for (each cache_entry ce to-be-updated)
err |= checkout_entry(ce, &state, NULL, NULL);
err |= run_parallel_checkout(&state, pc_workers, pc_threshold, NULL, NULL);</code></pre>
</div></div>
</div>
</div>
</div>
<div id="footnotes"><hr /></div>
<div id="footer">
<div id="footer-text">
Last updated
2022-11-11 23:55:30 PST
</div>
</div>
</body>
</html>
|
