-
Notifications
You must be signed in to change notification settings - Fork 104
/
Copy pathindex.html
244 lines (238 loc) · 7.75 KB
/
index.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
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<link rel="stylesheet" href="../common-revealjs/css/reveal.css">
<link rel="stylesheet" href="../common-revealjs/css/theme/white.css">
<link rel="stylesheet" href="../common-revealjs/css/custom.css">
<script>
// This is needed when printing the slides to pdf
var link = document.createElement( 'link' );
link.rel = 'stylesheet';
link.type = 'text/css';
link.href = window.location.search.match( /print-pdf/gi ) ? '../common-revealjs/css/print/pdf.css' : '../common-revealjs/css/print/paper.css';
document.getElementsByTagName( 'head' )[0].appendChild( link );
</script>
<script>
// This is used to display the static images on each slide,
// See global-images in this html file and custom.css
(function() {
if(window.addEventListener) {
window.addEventListener('load', () => {
let slides = document.getElementsByClassName("slide-background");
if (slides.length === 0) {
slides = document.getElementsByClassName("pdf-page")
}
// Insert global images on each slide
for(let i = 0, max = slides.length; i < max; i++) {
let cln = document.getElementById("global-images").cloneNode(true);
cln.removeAttribute("id");
slides[i].appendChild(cln);
}
// Remove top level global images
let elem = document.getElementById("global-images");
elem.parentElement.removeChild(elem);
}, false);
}
})();
</script>
</head>
<body>
<div class="reveal">
<div class="slides">
<div id="global-images" class="global-images">
<img src="../common-revealjs/images/sycl_academy.png" />
<img src="../common-revealjs/images/sycl_logo.png" />
<img src="../common-revealjs/images/trademarks.png" />
<img src="../common-revealjs/images/codeplay.png" />
</div>
<!--Slide 1-->
<section class="hbox">
<div class="hbox" data-markdown>
## Queues (In and Out of Order)
</div>
</section>
<!--Slide 2-->
<section class="hbox" data-markdown>
## Learning Objectives
* Learn about out-of-order and in-order execution
* Learn about how to use in-order queues for maximal performance
</section>
<!--Slide 3-->
<section>
<div class="hbox" data-markdown>
#### Out-of-order execution
</div>
<div class="container">
<div class="col" data-markdown>
</div>
<div class="col" data-markdown>
* SYCL `queue`s are by default out-of-order.
* This means commands are allowed to be overlapped, re-ordered, and executed concurrently, providing dependencies are honoured to ensure consistency.
</div>
</div>
</section>
<!--Slide 4-->
<section>
<div class="hbox" data-markdown>
#### In-of-order execution
</div>
<div class="container">
<div class="col" data-markdown>
</div>
<div class="col" data-markdown>
* SYCL `queue`s can be configured to be in-order.
* This mean commands must execute strictly in the order they were enqueued.
</div>
</div>
</section>
<!--Slide 7-->
<section>
<div class="hbox" data-markdown>
#### Using an out-of-order queue (USM)
</div>
<div class="container">
<div class="col">
<code class="code-100pc"><pre>
sycl::queue Q;
for (std::function task: tasks)
Q.submit(task);
Q.wait();
</code></pre>
</div>
<div class="col" data-markdown>
* All commands can execute concurrently if hardware allows it
* See exercise for real world data
</div>
</div>
</section>
<section>
<div class="hbox" data-markdown>
#### Using an out-of-order queue (USM): how to handle dependencies?
</div>
<div class="container">
<div class="col">
<code class="code-100pc"><pre>
sycl::queue Q;
auto e1 = Q.submit(taskA);
auto e2 = Q.submit([&](sycl::handle &cgh) {
cgh.depends_on(e1);
taskB();
}
auto e3 = Q.submit([&](sycl::handle &cgh) {
cgh.depends_on(e1);
taskC();
}
Q.submit([&](sycl::handle &cgh) {
cgh.depends_on({e2,e3});
taskD();
}.wait();
</code></pre>
</div>
<div class="col" data-markdown>
* Define an ordering manually (error prone)
* Scheduling done automatically (maximize concurrency)
* Dynamic scheduling (potentially higher latency overhead than manually crafted scheduling)
</div>
</div>
</section>
<!--Slide 7-->
<section>
<div class="hbox" data-markdown>
#### Using an in-order queue (USM)
</div>
<div class="container">
<div class="col">
<code class="code-100pc"><pre>
sycl::queue Q{sycl::property::queue::in_order};
for (std::function task: tasks)
Q.submit(task);
Q.wait();
</code></pre>
</div>
<div class="col" data-markdown>
* All commands are executed serially
* Ease of programming (no race conditions can occur)
* Potentialy lower-latency than out-of-order queues
* Doesn't allow concurrency, potentially suboptimal performance
</div>
</div>
</section>
<section>
<div class="hbox" data-markdown>
#### Using an out-of-order queue to enable concurrency in kernel execution (USM)
</div>
<div class="container">
<div class="col">
<code class="code-100pc"><pre>
std::vector<std::pair<sycl::queue, std::function>> queues_tasks;
for (auto [Q, task]: queues_tasks)
Q.submit(task);
for (auto [Q, _]: queues_tasks)
Q.wait();
</code></pre>
</div>
<div class="col" data-markdown>
* Manual Scheduling (tasks need to be mapped to queues)
* Allow concurency between queues
* Painful to extract full concurrency
</div>
</div>
</section>
<!--Slide 8-->
<section>
<div class="hbox" data-markdown>
## Note on buffer/accessors
* Best of Both worlds
* Buffer / Accessors automatically handle dependencies for you (no need for `depends_on`)
* Hence, using in or out-of-order queues with buffer/accessors will not change the semantics of your program!
* Can use out-of-order without any drawbacks.
</div>
</section>
<!--Slide 8-->
<section>
<div class="hbox" data-markdown>
## Questions
</div>
</section>
<!--Slide 17-->
<section>
<div class="hbox" data-markdown>
#### Exercise
</div>
<div class="container" data-markdown>
Code_Exercises/In_Order_Queue/source_queue_benchmarking.cpp
</div>
<div class="container" data-markdown>
- Transform the serial in-order scheduling to allow concurrent execution (using an out-of-order queue or multiple in-order queues)
- Mesure speedup
</div>
</section>
<section>
<div class="hbox" data-markdown>
#### Exercise
</div>
<div class="container" data-markdown>
Code_Exercises/In_Order_Queue/source_vector_add.cpp
</div>
<div class="container" data-markdown>
</div>
<div class="container" data-markdown>
Take the diamond data flow graph we implemented in the last exercise and convert it to use an in-order `queue`.
</div>
</section>
</div>
</div>
<script src="../common-revealjs/js/reveal.js"></script>
<script src="../common-revealjs/plugin/markdown/marked.js"></script>
<script src="../common-revealjs/plugin/markdown/markdown.js"></script>
<script src="../common-revealjs/plugin/notes/notes.js"></script>
<script>
Reveal.initialize({mouseWheel: true, defaultNotes: true});
Reveal.configure({ slideNumber: true });
</script>
</body>
</html>