-
Notifications
You must be signed in to change notification settings - Fork 102
/
index.html
289 lines (276 loc) · 9.26 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
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
<!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" />
</div>
<!--Slide 1-->
<section class="hbox" data-markdown>
## Data Parallelism
</section>
<!--Slide 2-->
<section class="hbox" data-markdown>
## Learning Objectives
* Learn about task parallelism and data parallelism
* Learn about the SPMD model for describing data parallelism
* Learn about SYCL execution and memory models
* Learn about enqueuing kernel functions with `parallel_for`
</section>
<!--Slide 3-->
<section>
<div class="hbox" data-markdown>
#### Task vs data parallelism
</div>
<div class="container" data-markdown>
</div>
<div class="container" data-markdown>
* **Task parallelism** is where you have several,
possibly distinct tasks executing in parallel.
* In task parallelism you optimize for latency.
* **Data parallelism** is where you have the same
task being performed on multiple elements of data.
* In data parallelism you optimize for throughput.
</div>
</section>
<!--Slide 4-->
<section>
<div class="hbox" data-markdown>
#### Vector processors
</div>
<div class="container" data-markdown>
* Many processors are vector processors, which means
they can naturally perform data parallelism.
* GPUs are designed to be parallel.
* CPUs have SIMD instructions which perform the
same instruction on a number elements of data.
</div>
</section>
<!--Slide 5-->
<section>
<div class="hbox" data-markdown>
#### SPMD model for describing data parallelism
</div>
<div class="container">
<div class="col">
Sequential CPU code
<code><pre>
void calc(const int in[], int out[]) {
// all iterations are run in the same
// thread in a loop
for (int i = 0; i < 1024; i++){
out[i] = in[i] * in[i];
}
}
// calc is invoked just once and all
// iterations are performed inline
calc(in, out);
</code></pre>
</div>
<div class="col">
Parallel SPMD code
<code><pre>
void calc(const int in[], int out[], int id) {
// function is described in terms of
// a single iteration
out[id] = in[id] * in[id];
}
// parallel_for invokes calc multiple
// times in parallel
parallel_for(calc, in, out, 1024);
</code></pre>
</div>
</div>
</section>
<!--Slide 6-->
<section>
<div class="hbox" data-markdown>
#### SYCL execution model
</div>
<div class="container">
<div class="col-left" data-markdown>
* In SYCL kernel functions are executed by
**work- items**.
* You can think of a work-item as a thread of
execution.
* Each work-item will execute a SYCL kernel function from start to end.
* A work-item can run on CPU threads, SIMD lanes,
GPU threads, or any other kind of processing
element.
</div>
<div class="col-right" data-markdown>
</div>
</div>
</section>
<!--Slide 7-->
<section>
<div class="hbox" data-markdown>
#### SYCL execution model
</div>
<div class="container">
<div class="col" data-markdown>
* Work-items are launched in parallel in a `sycl::range`.
* In order to maximize parallelism, the range should correspond to the problem size.
</div>
<div class="col" data-markdown>
</div>
</div>
</section>
<!--Slide 14-->
<section>
<div class="hbox" data-markdown>
#### Parallel_for
</div>
<div class="container">
<div class="col">
<code><pre>
cgh.<mark>parallel_for</mark><my_kernel>(<mark>range{64, 64}</mark>,
[=](id<2> idx){
// SYCL kernel function is executed
// on a range of work-items
});
</code></pre>
</div>
</div>
<div class="container" data-markdown>
* In SYCL, kernel functions can be enqueued to execute
over a range of work-items using `parallel_for`.
* When using `parallel_for` you must also pass `range`
which describes the iteration space over which the kernel
is to be executed.
</div>
</section>
<!--Slide 15-->
<section>
<div class="hbox" data-markdown>
#### Parallel_for
</div>
<div class="container">
<div class="col">
<code><pre>
cgh.parallel_for<my_kernel>(range{64, 64},
[=](<mark>id<2> idx</mark>){
// SYCL kernel function is executed
// on a range of work-items
});
</code></pre>
</div>
</div>
<div class="container" data-markdown>
* When using `parallel_for` you must also have the
function object which represents the kernel function take
an `id`.
* This represents the current work-item being executed and
its position within the iteration space.
</div>
</section>
<!--Slide 16-->
<section>
<div class="hbox" data-markdown>
#### Expressing parallelism
</div>
<div class="container">
<div class="col">
<code><pre>
cgh.parallel_for<kernel>(<mark>range<1>(1024)</mark>,
[=](<mark>id<1> idx</mark>){
/* kernel function code */
});
</code></pre>
<code><pre>
cgh.parallel_for<kernel>(<mark>range<1>(1024)</mark>,
[=](<mark>item<1> item</mark>){
/* kernel function code */
});
</code></pre>
<code><pre>
cgh.parallel_for<kernel>(nd_range<1>(<mark>range<1>(1024),
range<1>(32))</mark>,[=](<mark>nd_item<1> ndItem</mark>){
/* kernel function code */
});
</code></pre>
</div>
<div class="col" data-markdown>
* Overload taking a **range** object specifies the global range, runtime decides local range
* An **id** parameter represents the index within the global range
____________________________________________________________________________________________
* Overload taking a **range** object specifies the global range, runtime decides local range
* An **item** parameter represents the global range and the index within the global range
____________________________________________________________________________________________
* Overload taking an **nd_range** object specifies the global and local range
* An **nd_item** parameter represents the global and local range and index
</div>
</div>
</section>
<!--Slide 17-->
<section class="hbox" data-markdown>
## Questions
</section>
<!--Slide 18-->
<section>
<div class="hbox" data-markdown>
#### Exercise
</div>
<div class="container" data-markdown>
Code_Exercises/Data_Parallelism/source.cpp
</div>
<div class="container" data-markdown>
Implement a SYCL application that adds two arrays of
values together in parallel using `parallel_for`.
</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();
Reveal.configure({ slideNumber: true });
</script>
</body>
</html>