Skip to content

Instantly share code, notes, and snippets.

@xatier

xatier/slides.html

Last active December 26, 2015 02:29
Show Gist options
  • Save xatier/7078807 to your computer and use it in GitHub Desktop.
Save xatier/7078807 to your computer and use it in GitHub Desktop.
Download ZIP
the slides for 10/21 meeting
Raw
slides.html
<!DOCTYPE html>
<html>
<head>
<title>Meeting II</title>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8"/>
<style type="text/css" media="screen">
/* Slideshow styles */
@import url(http://fonts.googleapis.com/css?family=Droid+Serif);
@import url(http://fonts.googleapis.com/css?family=Yanone+Kaffeesatz);
body {
font-family: 'Droid Serif';
font-size: 20px;
}
h1, h2, h3 {
font-family: 'Yanone Kaffeesatz';
font-weight: 400;
margin-bottom: 0;
}
h1 { font-size: 4em; }
h2 { font-size: 2em; }
h3 { font-size: 1.6em; }
.footnote {
position: absolute;
bottom: 3em;
}
li p { line-height: 1.25em; }
.red { color: #fa0000; }
.large { font-size: 2em; }
a, a > code {
color: rgb(249, 38, 114);
text-decoration: none;
}
code {
-moz-border-radius: 5px;
-web-border-radius: 5px;
background: #e7e8e2;
border-radius: 5px;
font-size: 16px;
}
.pull-left {
float: left;
width: 47%;
}
.pull-right {
float: right;
width: 47%;
}
.pull-right ~ p {
clear: both;
}
#slideshow .slide .content code {
font-size: 0.8em;
}
#slideshow .slide .content pre code {
font-size: 0.9em;
padding: 15px;
}
.inverse {
background: #272822;
color: #777872;
text-shadow: 0 0 20px #333;
}
.inverse h1, .inverse h2 {
color: #f3f3f3;
line-height: 0.8em;
}
/* Slide-specific styling */
#slide-inverse .footnote {
bottom: 12px;
left: 20px;
}
#slide-how .slides {
font-size: 0.9em;
position: absolute;
top: 151px;
right: 140px;
}
#slide-how .slides h3 {
margin-top: 0.2em;
}
#slide-how .slides .first, #slide-how .slides .second {
padding: 1px 20px;
height: 90px;
width: 120px;
-moz-box-shadow: 0 0 10px #777;
-webkit-box-shadow: 0 0 10px #777;
box-shadow: 0 0 10px #777;
}
#slide-how .slides .first {
background: #fff;
position: absolute;
top: 20%;
left: 20%;
z-index: 1;
}
#slide-how .slides .second {
position: relative;
background: #fff;
z-index: 0;
}
/* Two-column layout */
.left-column {
color: #777;
width: 20%;
height: 92%;
float: left;
}
.left-column h2:last-of-type, .left-column h3:last-child {
color: #000;
}
.right-column {
width: 75%;
float: right;
padding-top: 2em;
}
</style>
</head>
<body>
<!-- slides start from here --->
<textarea id="source">
name: inverse
class: center, middle, inverse
Meeting II
=======================
##### Yan-Ming Li #####
.footnote[Chap. 7, Chap. 8]
---
Chap 7. OpenCL Numerical Compliance
----------------------------------
- How OpenCL deal with floating point numbers?
---
Rounding Modes
--------------
- As defined in IEEE 754, we have 4 rounding modes:
+ round to nearest even (*)
+ round toward +INF
+ round toward -INF
+ round toward zero
---
INF, NaN, Denormalized Numbers
-----------------------------
- INF and NaN must be supported
- Denormalized numbers is optional
---
Floating-Point Exceptions
-------------------------
- disabled in OpenCL
- just follow the IEEE 754 spec (not enabled cases)
- floating-point flags and raising floating-point exceptions are implementation defined
---
Relative Errors as ULPs
-----------------------
- ULP: units in the last place
If x is a real number that lies between two finite
consecutive floating-point numbers a and b, without being
equal to one of them, then ulp(x) = |b - a|, otherwise
ulp(x) is the distance between the two non-equal finite
floating-point numbers nearest x. Moreover, ulp(NaN) is
NaN,
---
Edge case Behavior
------------------
- NaN cases
- flush behavior
- rounding error
---
Chap. 8 Image Addressing and Filtering
------------------------------
- How to address the pixels?
---
Image Coordinates
----------------------------------
- Related to `read_image{f|i|ui}`
- Let `(s, t, r)` to be the width, height and depth of the image in pixels
+ `coord.xy -> (s, t)`
+ `coord.xyz -> (s, t, r)`
---
Addressing and Filter Modes
----------------------------------
- `CLK_FILTER_NEAREST`
- `CLK_FILTER_LINEAR`
---
Conversion Rules
----------------------------------
- Normalized integer data types -> floating point values
- Floating point values -> normalized integer data types
- half precision floating point data types
- sRGBA images
read_imgef
write_imagef
---
Select an Image from an Image Array
----------------------------------
- Let `(u,v,w)` represent the unnormalized image coordinate values for reading from / writing to a 2D image in a 2D image array.
layer = clamp(rint(w), 0, dt – 1)
- Let `(u,v)` represent the unnormalized image coordinate values for reading from / writing to a 1D image in a 1D image array.
layer = clamp(rint(v), 0, ht – 1)
---
name: inverse
class: center, middle, inverse
Thanks
======
</textarea>
<script src="http://gnab.github.io/remark/downloads/remark-0.6.0.min.js" type="text/javascript">
</script>
<script type="text/javascript">
var slideshow = remark.create({
highlightStyle: 'arta',
highlightLanguage: 'bash'
}) ;
</script>
</body>
</html>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment