-
Notifications
You must be signed in to change notification settings - Fork 0
/
index-prev.html
1739 lines (1638 loc) · 59.1 KB
/
index-prev.html
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
<!doctype html>
<html>
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0, maximum-scale=1.0, user-scalable=no">
<title>Compiler Explorer</title>
<link rel="stylesheet" href="reveal.js/css/reveal.css">
<link rel="stylesheet" href="drw.css">
<style>
.faded {
opacity: 0.7;
}
div.footnote {
margin-top: 4em;
font-size: smaller;
font-style: italic;
}
#sum td {
border: 1px solid black;
}
#sum tr {
border: 1px solid black;
}
#sum .filled {
background: rgb(141, 211, 199);
}
#registers th {
font-size: smaller;
text-align: center;
}
#registers td {
text-align: center;
}
#registers td.register {
border: 1px solid black;
}
#registers td.rax {
background: rgb(141, 211, 199);
}
#registers td.eax {
background: rgb(255, 255, 179);
}
#registers td.ax {
background: rgb(190, 186, 218);
}
#registers td.ah {
background: rgb(251, 128, 114);
}
#registers td.al {
background: rgb(128, 177, 211);
}
#registers td.regnote {
font-size: smaller;
font-style: italic;
}
.soh {
color: rgb(27, 145, 33);
}
a.view-button {
font-size: 12px;
}
div.left-pane {
width: 50%;
float: left;
}
div.right-pane {
width: 50%;
float: right;
}
.fix-key {
color: #03330c;
}
.fix-value {
color: #1f1f1f;
}
</style>
<link rel="stylesheet" href="reveal.js/lib/css/zenburn.css">
<script>
var link = document.createElement('link');
link.rel = 'stylesheet';
link.type = 'text/css';
link.href = window.location.search.match(/print-pdf/gi) ? 'reveal.js/css/print/pdf.css' : 'reveal.js/css/print/paper.css';
document.getElementsByTagName('head')[0].appendChild(link);
</script>
</head>
<body>
<div class="reveal">
<!-- TODO compile all with -Wall -Wextra -->
<!-- TODO unify indentation -->
<section class="slides">
<section id="title">
<em>September 2017</em>
<!-- TODO, better title format -->
<h3>Unbolting the Compiler's Lid:</h3>
<h5>What Has My Compiler Done for Me Lately</h5>
<em>Matt Godbolt, DRW Trading</em>
<br>
<em>
<a href="https://twitter.com/mattgodbolt">@mattgodbolt</a>
<a href="mailto:[email protected]">[email protected]</a>
</em>
<br>
<em><a href="https://drw.com/careers">drw.com</a></em>
</section>
<!------------------------------------------------------>
<section id="why">
<section>
<h3>Why am I here?</h3>
<aside class="notes">
<p>I'm wondering that too. If I explain a bit about myself perhaps it'll help.</p>
<p>Passion for performance, mechanical sympathy</p>
</aside>
</section>
<section>
<h3>My background</h3>
<div class="fragment">
<img src="images/6502_z80.jpg" width="200">
<!-- TODO photo of ARM here -->
<img src="images/red_dog.png" width="200">
</div>
<div class="fragment">
<img src="images/profactor.jpg" width="100">
<img src="images/google.png" width="100">
<img src="images/DRWSmallLogo.png" width="100">
</div>
<aside class="notes">
<h4>Potted history:</h4>
<ul>
<li>Learned Z80, 6502 assembly in the 80s to write my own games</li>
<li>Moved on to ARM assembly</li>
<li>Wrote full functioned IRC client in ARM asm</li>
<li>Dismissed C as just a macro assembler</li>
<li>Learned C to write a MUD at Uni</li>
<li>Eventually moved on to C++</li>
<li>Got a job making games, high performance stuff</li>
<li>Co-ran C++ tools company</li>
<li>Google; occasional performance stuff for cellphones</li>
<li>Now 6 years at DRW doing performance stuff for trading systems</li>
<li>Still an asm hacker at heart though don't touch it so more</li>
</ul>
<p>Amazing opportunity to talk to lots of C++ programmers at all levels. My passion.</p>
</aside>
</section>
<section>
<h3>My goals</h3>
<ul>
<li>Un-scary-fy assembler</li>
<li>Compilers are amazing</li>
<li class="fragment">...but always worth checking</li>
</ul>
</section>
</section>
<!------------------------------------------------------>
<section id="backstory">
<section>
<h3>Backstory</h3>
<pre class="fragment"><code data-trim class="cpp">
int sum(const vector<int> &v) {
int result = 0;
for (size_t i = 0; i < v.size(); ++i)
result += v[i];
return result;
}</code></pre>
<pre class="fragment"><code data-trim class="cpp">
int sum(const vector<int> &v) {
int result = 0;
for (int x : v) result += x;
return result;
}</code></pre>
<div class="fragment">Which is better?</div>
<div class="fragment">Why care?</div>
<aside class="notes">
<ul>
<li>2012, C++11isms, range-for</li>
</ul>
</aside>
</section>
<section>
<h3>WARNING</h3>
<ul>
<li>Reading assembly alone is bad</li>
<li><em>Always</em> measure too</li>
</ul>
<aside class="notes">
Shout out to google benchmarking tool. Microbenchmarks, perils.
But always measure end-to-end perf if you can too.
</aside>
</section>
</section>
<!------------------------------------------------------->
<section id="asm101">
<!-- reference "just enough asm to be dangerous talk -->
<section>
<h3>x86 Assembly 101</h3>
</section>
<section>
<h3>Registers</h3>
<ul>
<li>rax, rbx, rcx, rdx, rsp, rbp, rsi, rdi, r8-r15</li>
<li>xmm0-15, ymm0-15, zmm0-15</li>
<li>cs, ds, ss, es, fs, gs</li>
<li>flags</li>
</ul>
</section>
<section>
<h3>Registers</h3>
<table id="registers">
<thead>
<tr>
<th>63...56</th>
<th>55...48</th>
<th>47...40</th>
<th>39...32</th>
<th>31...24</th>
<th>23...16</th>
<th>15...8</th>
<th>7...0</th>
</tr>
</thead>
<tbody>
<tr>
<td colspan="8" class="register rax">rax</td>
</tr>
<tr>
<td colspan="4" class="regnote">(zeroed on write)</td>
<td colspan="4" class="register eax">eax</td>
</tr>
<tr>
<td colspan="6"></td>
<td colspan="2" class="register ax">ax</td>
</tr>
<tr>
<td colspan="6"></td>
<td colspan="1" class="register ah">ah</td>
<td colspan="1"></td>
</tr>
<tr>
<td colspan="7"></td>
<td colspan="1" class="register al">al</td>
</tr>
</tbody>
</table>
</section>
<section>
<h3>Calling convention</h3>
<pre><code class="cpp">long func(int a, short b, char c);</code></pre>
<ul class="fragment">
<li>Params: rdi, rsi, rdx, rcx, r8, r9; xmm0-7</li>
<li>Return: rax / xmm0</li>
<li>Callee-save: rbp, rbx, r12–r15</li>
</ul>
<div class="fragment">
<pre><code class="cpp">rax func(edi a, si b, dl c);</code></pre>
</div>
<div class="footnote">(System V AMD64 ABI)</div>
</section>
<section>
<h3>Instructions</h3>
<pre><code data-trim class="x86asm">
op
op dest
op dest, src
</code></pre>
<ul>
<li><code>op</code> is e.g. <code>add</code>, <code>sub</code>, <code>sub</code></li>
<li><code>src</code>, <code>dest</code> is a register or a memory reference</li>
</ul>
<div class="footnote">(Intel asm syntax)</div>
</section>
<section>
<h3>Instructions</h3>
<pre><code data-trim class="x86asm">
mov eax, 1234 ; eax = 1234
mov ecx, DWORD PTR g[rip] ; ecx = contents of global variable g
add eax, ecx ; eax += ecx
cmp eax, 6912 ; eax == 6912 ?
je all_ok ; then goto all_ok
call report_error ; else report error
all_ok:
ret ; and return
</code></pre>
</section>
<section>
<h3>Instructions</h3>
<pre><code data-trim class="x86asm">
mov eax, DWORD PTR [r14] ; eax = *(uint32_t *)r14
add eax, DWORD PTR [r14 + 4] ; eax += *(uint32_t *)(r14 + 4)
sub eax, DWORD PTR [r14 + 4 * rbx]
; eax -= *(uint32_t *)(r14 + 4 * rbx)
lea rax, [r14 + 4 * rbx] ; eax = r14 + 4 * rbx
</code></pre>
<pre class="fragment"><code data-trim class="cpp">
int eax = *r14; // int *r14;
eax += r14[1];
eax -= r14[rbx];
int *rax = &r14[rbx];
</code></pre>
</section>
<section>
<h3>Instructions</h3>
<pre><code data-trim class="x86asm">
movsx rax, di
push rax
vpaddd ymm0, ymm0, ymmword ptr [rax]
popcnt rax
lzcnt eax, edi
bsf rax
int 3
</code></pre>
</section>
<section>
<h3>Summary</h3>
<ul>
<li>Registers: <code>rax</code>, <code>rbx</code>, <code>rcx</code> ...</li>
<li>Size: <code>rax</code>, <code>eax</code>, <code>ax</code> ...</li>
<li>Params: <code>rdi</code>, <code>rsi</code>, <code>rdx</code>, <code>rcx</code> ...</li>
<li>Result: <code>rax</code></li>
<li><code>op dest, src</code></li>
<li><code>dest</code>, <code>src</code> are registers or memory</li>
</ul>
</section>
</section>
<!------------------------------------------------------->
<section id="back2backstory">
<section>
<h3>Where were we?</h3>
<pre><code data-trim class="cpp">
int sum(const vector<int> &v) {
int result = 0;
for (size_t i = 0; i < v.size(); ++i)
result += v[i];
return result;
}</code></pre>
<pre><code data-trim class="cpp">
int sum(const vector<int> &v) {
int result = 0;
for (int x : v) result += x;
return result;
}</code></pre>
</section>
<section>
<h3>Compiler Explorer v0.1</h3>
<pre><code data-noescape data-trim class="bash">
$ g++ /tmp/test.cc -O1 -c -S -o -masm=intel -
</code></pre>
<pre class="fragment"><code class="x86asm" data-trim>
.file "test.cc"
.text
.globl _Z3sumRKSt6vectorIiSaIiEE
.type _Z3sumRKSt6vectorIiSaIiEE, @function
_Z3sumRKSt6vectorIiSaIiEE:
.LFB786:
.cfi_startproc
mov rcx, QWORD PTR [rdi]
mov rax, QWORD PTR 8[rdi]
sub rax, rcx
...</code></pre>
</section>
<section>
<h3>Compiler Explorer v0.1</h3>
<pre><code data-noescape data-trim class="bash">
$ g++ /tmp/test.cc -O1 -c -S -o -masm=intel - | c++filt
</code></pre>
<pre><code class="x86asm" data-trim>
.file "test.cc"
.text
.globl sum(std::vector<int, std::allocator<int> > const&)
.type sum(std::vector<int, std::allocator<int> > const&), @function
sum(std::vector<int, std::allocator<int> > const&):
.LFB786:
.cfi_startproc
mov rcx, QWORD PTR [rdi]
mov rax, QWORD PTR 8[rdi]
sub rax, rcx
...</code></pre>
</section>
<section>
<h3>Compiler Explorer v0.1</h3>
<pre><code data-noescape data-trim class="bash">
$ g++ /tmp/test.cc -O1 -c -S -o -masm=intel - | c++filt \
| grep -vE '\s+\.'
</code></pre>
<pre><code class="x86asm" data-trim>
sum(std::vector<int, std::allocator<int> > const&):
.LFB786:
mov rcx, QWORD PTR [rdi]
mov rax, QWORD PTR 8[rdi]
sub rax, rcx
sar rax, 2
mov rsi, rax
test rax, rax
mov edx, 0
mov eax, 0
...</code></pre>
</section>
<section><h3>Compiler Explorer v0.1</h3>
<div>Not very pretty</div>
<h5 class="fragment">To the web!</h5>
</section>
</section>
<!------------------------------------------------------->
<section id="andNow">
<section>
<h3>Demo</h3>
<pre class="ce">
/// g71:-O2 -std=c++1z
// setup
#include <numeric>
#include <vector>
using namespace std;
int sum(const vector<int> &v) {
int result = 0;
for (size_t i = 0; i < v.size(); ++i)
result += v[i];
return result;
}</pre>
<aside class="notes">
<ul>
<li>Pop out</li>
<li>Walk through code (using following slides), note optimizer is on</li>
<li>Show optimizer off</li>
</ul>
</aside>
</section>
<section>
<h3>Walkthrough</h3>
<pre><code data-trim class="x86asm">
; rdi = const vector<int> *
mov rdx, QWORD PTR [rdi] ; rdx = *rdi ≡ rdi->begin
mov rcx, QWORD PTR [rdi+8] ; rcx = *(rdi + 8) ≡ rdi->end
</code></pre>
<pre class="fragment"><code data-trim class="cpp">
template<typename T> struct _Vector_impl {
T *_M_start;
T *_M_finish;
T *_M_end_of_storage;
};
</code></pre>
</section>
<section>
<h3>Walkthrough</h3>
<pre><code data-trim class="x86asm faded">
; rdi = const vector<int> *
mov rdx, QWORD PTR [rdi] ; rdx = *rdi ≡ rdi->begin
mov rcx, QWORD PTR [rdi+8] ; rcx = *(rdi + 8) ≡ rdi->end
</code></pre>
<pre><code data-trim class="x86asm">
sub rcx, rdx ; rcx -= rdx ≡ end - begin
mov rax, rcx ; rax = rcx ≡ end - begin
sar rax, 2 ; rax = rax / 4
; ≡ (end - begin) / sizeof(int)
</code></pre>
<pre class="fragment"><code data-trim class="cpp">
size_t size() const noexcept {
return _M_finish - _M_start;
} </code></pre>
</section>
<section>
<h3>Walkthrough</h3>
<pre><code data-trim class="x86asm faded">
; rdi = const vector<int> *
mov rdx, QWORD PTR [rdi] ; rdx = *rdi ≡ rdi->begin
mov rcx, QWORD PTR [rdi+8] ; rcx = *(rdi + 8) ≡ rdi->end
sub rcx, rdx ; rcx -= rdx ≡ end - begin
mov rax, rcx ; rax = rcx ≡ end - begin
sar rax, 2 ; rax = rax / 4
; ≡ (end - begin) / sizeof(int)
</code></pre>
<pre><code data-trim class="x86asm">
test rax, rax ; is rax == 0?
je .L4 ; if so, goto L4
</code></pre>
</section>
<section>
<h3>Walkthrough</h3>
<pre><code data-trim class="x86asm faded">
; rdi = const vector<int> *
mov rdx, QWORD PTR [rdi] ; rdx = *rdi ≡ rdi->begin
mov rcx, QWORD PTR [rdi+8] ; rcx = *(rdi + 8) ≡ rdi->end
sub rcx, rdx ; rcx -= rdx ≡ end - begin
mov rax, rcx ; rax = rcx ≡ end - begin
sar rax, 2 ; rax = rax / 4
; ≡ (end - begin) / sizeof(int)
test rax, rax ; is rax == 0?
je .L4 ; if so, goto L4
</code></pre>
<pre><code data-trim class="x86asm">
add rcx, rdx ; rcx += rdx
; ≡ (end - begin) + begin
; ≡ end
xor eax, eax ; eax = 0
</code></pre>
</section>
<section>
<h3>Walkthrough</h3>
<pre><code data-trim class="x86asm">
; rcx ≡ end, rdx = begin, eax = 0
.L3:
add eax, DWORD PTR [rdx] ; eax += *rdx
add rdx, 4 ; rdx += sizeof(int)
cmp rdx, rcx ; is rdx == end?
jne .L3 ; if not, loop
ret ; we're done
</code></pre>
</section>
<section>
<h3>Walkthrough</h3>
<pre><code data-trim class="x86asm">
.L4:
xor eax, eax ; return 0
ret
</code></pre>
</section>
<section><h3>Backstory</h3>
<h5>So, which approach is best?</h5></section>
<section>
<h3>Also</h3>
<ul>
<li>Optimizer settings are important</li>
<li><code>std::accumulate</code></li>
</ul>
</section>
</section>
<!------------------------------------------------------->
<section id="compilersAreSuperSmart">
<section><h3>Compilers are super-smart!</h3></section>
<section>
<h3>Multiplication</h3>
<div class="left-pane">
<pre><code class="cpp ce" data-trim>
/// g71:-O2 -std=c++1z
int mulByY(int x, int y) {
return x * y;
}
</code></pre>
</div>
<div class="right-pane">
<pre><code class="x86asm" data-trim>
mulByY(int, int):
mov eax, edi
imul eax, esi
ret
</code></pre>
</div>
</section>
<section>
<h3>Multiplication</h3>
<pre>
<span
data-fragment-index="4"
class="fragment highlight-current-red">1</span><span
data-fragment-index="3" class="fragment highlight-current-red">1</span><span
data-fragment-index="2" class="fragment highlight-current-red">0</span><span
data-fragment-index="1" class="fragment highlight-current-red">1</span> (13)
x 0101 (5)
--------
<span data-fragment-index="1" class="fragment">1101</span>
<span data-fragment-index="2" class="fragment">0000</span>
<span data-fragment-index="3" class="fragment">1101</span>
<span data-fragment-index="4" class="fragment">+ 0000</span>
<span data-fragment-index="5" class="fragment">--------
01000001 (65)</span></pre>
<div class="fragment">That's a lot of additions!</div>
<div class="fragment">Haswell 32-bit multiply - 4 cycles</div>
</section>
<section>
<h3>Multiplication</h3>
<pre><code class="cpp ce">
/// g71:-O2 -std=c++1z
int mulBy3(int x) { return x * 3; }
int mulBy5(int x) { return x * 5; }
int mulBy7(int x) { return x * 7; }
int mulBy9(int x) { return x * 9; }
</code></pre>
</section>
<section>
<h3>Multiplication</h3>
<pre><code class="cpp ce">
/// g71:-O2 -std=c++1z
int mulBy65599(int a) {
return (a << 16) + (a << 6) - a;
// ^ ^
// a * 65536 |
// a * 64
// 65536a + 64a - 1a = 65599a
} </code></pre>
<aside class="notes">
<pre>-march=i486 -m32</pre>
shows up what you asked
</aside>
</section>
<section>
<h3>Division</h3>
<div class="left-pane">
<!---/// g71:-O2 -std=c++1z -->
<pre><code class="cpp ce" data-trim>
int divByY(int x, int y) {
return x / y;
}
int modByY(int x, int y) {
return x % y;
}
</code></pre>
</div>
<div class="right-pane">
<pre><code class="x86asm" data-trim>
divByY(int, int):
mov eax, edi
cdq
idiv esi
ret
modByY(int, int):
mov eax, edi
cdq
idiv esi
mov eax, edx
ret
</code></pre>
</div>
<div class="fragment">Haswell 32-bit divide - 22-29 cycles!</div>
</section>
<section>
<h3>Division</h3>
<pre><code class="cpp ce">
/// g71:-O2 -std=c++1z
int divBy3(int x) { return x / 3; }
</code></pre>
<aside class="notes">
<code>cdq</code> does sign expansion into registers. <code>div</code> divides eax:edx with the
operand, results in
eax (divisor) and edx (dividend).
<code>cdq</code> sign extends eax into edx, ready for a div
</aside>
</section>
<section>
<h3>Division</h3>
<pre><code class="x86asm" data-trim>
divBy3(int):
mov eax, edi ; eax = x
mov edx, 1431655766
sar edi, 31 ; edi = x>>31
; ≡ x < 0 ? -1 : 0
imul edx ; eax:edx = x * 1431655766
mov eax, edx ; (x * 1431655766) >> 32
; 1431655766 / 4294967296 = 0.3333333335
sub eax, edi ; ... - (x < 0 ? -1 : 0)
ret
</code></pre>
<!--TODO explain more, improve, and make reveal-y-->
</section>
<section>
<h3>Modulus</h3>
<div class="left-pane">
<!---/// g71:-O2 -std=c++1z -->
<pre><code class="cpp ce" data-trim>
int modBy3(int x) {
return x % 3;
}
</code></pre>
</div>
<div class="right-pane">
<pre><code class="x86asm" data-noescape data-trim>
<span class="faded">modBy3(int):
mov eax, edi
mov edx, 1431655766
imul edx
mov eax, edi
sar eax, 31
sub edx, eax
</span> lea eax, [rdx+rdx*2]
sub edi, eax
mov eax, edi
ret
</code></pre>
</div>
</section>
<!-- TODO perf measurements -->
<section><h3>Ternary operator</h3>
<div class="left-pane">
<!---/// g71:-O2 -std=c++1z -->
<pre><code class="cpp ce" data-trim>
int func(int i) {
return (i < 0) ? 1234 : 5678;
}
</code></pre>
</div>
<div class="right-pane">
<pre><code class="x86asm" data-noescape data-trim>
func(int):
mov eax, edi
sar eax, 31
and eax, -4444
add eax, 5678
ret
</code></pre>
</div>
</section>
<section><h3>Counting bits</h3>
<pre><code class="ce">
/// g71:-O2 -std=c++1z -march=haswell
int countSetBits(int a) {
int count = 0;
while (a) {
count++;
a &= (a-1);
}
return count;
}
</code></pre>
<aside class="notes">
Explain a & (a - 1)
</aside>
</section>
<section>
<h3>Summary</h3>
Compilers are pretty smart at cute arithmetic tricks.
</section>
</section>
<!------------------------------------------------------->
<section id="butNotClairvoyant">
<section>
<h3>Compilers are super-smart!</h3>
<h5 class="fragment">...but aren't clairvoyant</h5>
</section>
<section>
<h3>Functions</h3>
<pre><code class="cpp ce">
/// g71:-O2 -std=c++1z
// setup
#include <vector>
using namespace std;
int someFunc(int);
int sum(const vector<int> &v) {
int result = 0;
for (size_t i = 0; i < v.size(); ++i)
result += someFunc(v[i]);
return result;
}</code></pre>
<aside class="notes">
Compiler can't make any assumptions about someFunc. So has to reload size.
Even though vector is const, compiler can't assume it's not going to change.
Using range-for fixes issue as range-for is defined to copy the extents of the iterand outside of
loop.
</aside>
</section>
<section><h3>Virtual methods</h3>
<pre><code class="cpp ce">
/// g71:-O2 -std=c++1z
// setup
#include <vector>
using namespace std;
struct Coster { virtual int costFor(int x) = 0; };
int totalCost(Coster &coster,
const vector<int> &v) {
int result = 0;
for (auto i : v)
result += coster.costFor(i);
return result;
}
</code></pre>
</section>
<!-- TODO WORK HERE
needs to have a better "theme". fit in with story?
vector is good but complicates code?
nextCost?
even this generates tons of code
// setup
#include <vector>
using namespace std;
struct Coster { virtual int nextCost() = 0; };
int costForX(Coster &c, int x) {
int totalCost = 0;
for (int i = 0; i < x; ++i)
totalCost += c.nextCost();
return totalCost;
} ?
void costForX(Coster &c, int x, int *totalCost) {
for (int i = 0; i < x; ++i)
*totalCost += c.nextCost();
}
-->
<section><h3>Virtual methods</h3>
<pre><code class="cpp ce">
/// g71:-O2 -std=c++1z
// setup
#include <vector>
using namespace std;
struct Coster { virtual int costFor(int x) = 0; };
struct StaticCost : Coster {
virtual int costFor(int x) { return 1; }
};
int totalCost(Coster &coster,
const vector<int> &v) {
int result = 0;
for (auto i : v)
result += coster.costFor(i);
return result;
}
</code></pre>
</section>
<section><h3>Virtual methods</h3>
<pre><code class="cpp ce">
/// g71:-O2 -std=c++1z
// setup
#include <vector>
using namespace std;
struct Coster { virtual int costFor(int x) = 0; };
struct StaticCost : Coster {
virtual int costFor(int x) { return x; }
};
/// TODO!
</code></pre>
</section>
<section>
<h3>Aliasing</h3>
<pre><code class="cpp ce">
/// g71:-O2 -std=c++1z
// setup
#include <functional>
#include <cstddef>
void sum(int *first, int num,
int *out_sum) {
*out_sum = 0;
for (int i = 0; i < num; ++i)
*out_sum += first[i];
}
</code></pre>
</section>
<section>
<h3>Likelihood analysis</h3>
<!-- TODO -->
<!--clang vs gcc difference here -->
<!-- likely doesn't work here in clang :'( -->
<pre><code class="cpp ce">
/// g71:-O2 -std=c++1z
// setup
extern void allZero();
extern void notAllZero();
void testZeroFirst(int a, int b, int c) {
if (a == 0 && b == 0 && c == 0) allZero();
else notAllZero();
}
void testNotZeroFirst(int a, int b, int c) {
if (!(a == 0 && b == 0 && c == 0)) notAllZero();
else allZero();
}
</code></pre>
</section>
<section>
<h3>Helping the compiler out</h3>
<pre><code class="cpp ce">
/// g71:-O3 -std=c++1z -march=haswell
// setup
#include <cstddef>
int sum(const int *input, size_t length) {
input = static_cast<decltype(input)>(
__builtin_assume_aligned(input, 64));
if (length & 63) __builtin_unreachable();
if (length == 0) __builtin_unreachable();
int sum = 0;
for (int i = 0; i < length; ++i) {
sum += input[i];
}
return sum;
}
</code></pre>
</section>
<section>
<h3>Helping the compiler out?</h3>
<pre><code class="cpp ce">
/// g71:-O2 -std=c++1z
// setup
#include <vector>
using namespace std;
struct ImageHolder {
const vector<int> &intensities() const;
};
double average(const ImageHolder &ih) {
auto data = ih.intensities();
if (data.empty()) return 0;
long sum = 0;
for (auto &&v : data) sum += v;
return sum / static_cast<double>(data.size());
}
</code></pre>
</section>
<!-- https://github.com/apple/swift/commit/21ab99bf80b91a025405d9d59277839074a840d7 -->
</section>
<!------------------------------------------------------->
<section id="ubIsYourFriend">
<section>
<h3>Undefined behaviour</h3>
<h5 class="fragment">Can be your friend</h5>
</section>
<section>
<h3>Infinite loops</h3>
<!-- TODO get C++ standards quote -->
<pre><code data-trim class="cpp ce">
/// g71:-O2 -DNDEBUG -std=c++1z
// setup
#include <cassert>
struct Foo {
Foo *next{nullptr};
bool isOk() const;
};
static void debugCheck(Foo *first) {
for (auto p = first; p; p = p->next) {
assert(p->isOk());
}
}
bool test(Foo *f) {
debugCheck(f);
return f != nullptr;
}
</code></pre>
</section>
<!-- <section><h3>index signed vs unsigned</h3>
<pre><code data-trim class="cpp ce">
/// g71:-O2 -std=c++1z
// setup
#include <cstdint>
int sum1000(const char *a, uint32_t offset) {
int res = 0;
auto end = offset + 1000;
while (offset < end)
res += a[offset++];
return res;
}
</code></pre>
<aside class="notes">
show it uint32_t and int32_t
TODO still weak
</aside>
</section>-->
<section><h3>Heap elision</h3>
<pre><code data-trim class="cpp ce">
/// g71:-O2 -std=c++1z
// setup
#include <memory>
using namespace std;
int func() {
auto a = make_unique<int>(42);
auto b = make_unique<int>(24);
return *a + *b;