Misc/abrash-black-book
Watch
1
0
Fork
You've already forked abrash-black-book
 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Replace invalid characters with HTML entities

Browse source 
— with —
’ with ’
+ with +
× with x
ç with ç
“ with “
” with ”
‘ with ‘
• with •
– with -
µ with µ
† with †
Fix C++
θ with θ
Yen symbol instead of times
Fix broken apos
Bullet again
E-circumflex
This commit is contained in:
James Gregory 2013-12-30 12:50:32 +11:00
commit 500e7f5654
353 changed files with 5091 additions and 5091 deletions
Download patch file Download diff file Expand all files Collapse all files

44
51-05.html
View file

@ -37,18 +37,18 @@
</CENTER>
<P><BR></P>
</P>
<P>With the above optimizations, the sample program is certainly adequately responsive on a 20 MHz 386 (sans 387; I&#146;m sure it&#146;s wonderfully responsive with a math coprocessor). Still, it couldn&#146;t quite keep up with the keyboard when I modified it to read only one key each time through the loop&#151;and we&#146;re talking about only eight vertices here. This indicates that we&#146;re already near the limit of animation complexity possible with our current approach. It&#146;s time to start rethinking that approach; over two-thirds of the overall time is spent in floating-point calculations, and it&#146;s there that we&#146;ll begin to attack the performance bottleneck we find ourselves up against.
<P>With the above optimizations, the sample program is certainly adequately responsive on a 20 MHz 386 (sans 387; I&rsquo;m sure it&rsquo;s wonderfully responsive with a math coprocessor). Still, it couldn&rsquo;t quite keep up with the keyboard when I modified it to read only one key each time through the loop&mdash;and we&rsquo;re talking about only eight vertices here. This indicates that we&rsquo;re already near the limit of animation complexity possible with our current approach. It&rsquo;s time to start rethinking that approach; over two-thirds of the overall time is spent in floating-point calculations, and it&rsquo;s there that we&rsquo;ll begin to attack the performance bottleneck we find ourselves up against.
</P>
<H3><A NAME="Heading5"></A><FONT COLOR="#000077">Incremental Transformation</FONT></H3>
<P>Listing 51.4 contains three functions; each concatenates an additional rotation around one of the three axes to an existing rotation. To improve performance, only the matrix entries that are affected in a rotation around each particular axis are recalculated (all but four of the entries in a single-axis rotation matrix are either 0 or 1, as shown in Chapter 50). This cuts the number of floating-point multiplies from the 64 required for the multiplication of two 4&#215;4 matrices to just 12, and floating point adds from 48 to 6.
<P>Listing 51.4 contains three functions; each concatenates an additional rotation around one of the three axes to an existing rotation. To improve performance, only the matrix entries that are affected in a rotation around each particular axis are recalculated (all but four of the entries in a single-axis rotation matrix are either 0 or 1, as shown in Chapter 50). This cuts the number of floating-point multiplies from the 64 required for the multiplication of two 4x4 matrices to just 12, and floating point adds from 48 to 6.
</P>
<P>Be aware that Listing 51.4 performs an incremental rotation on top of whatever rotation is already in the matrix. The cube may already have been turned left, right, up, down, and sideways; regardless, Listing 51.4 just tacks the specified rotation onto whatever already exists. In this way, the object-to-world transformation matrix contains a history of all the rotations ever specified by the user, concatenated one after another onto the original matrix. Potential loss of precision is a problem associated with using such an approach to represent a very long concatenation of transformations, especially with fixed-point arithmetic; that&#146;s not a problem for us yet, but we&#146;ll run into it eventually.</P>
<P>Be aware that Listing 51.4 performs an incremental rotation on top of whatever rotation is already in the matrix. The cube may already have been turned left, right, up, down, and sideways; regardless, Listing 51.4 just tacks the specified rotation onto whatever already exists. In this way, the object-to-world transformation matrix contains a history of all the rotations ever specified by the user, concatenated one after another onto the original matrix. Potential loss of precision is a problem associated with using such an approach to represent a very long concatenation of transformations, especially with fixed-point arithmetic; that&rsquo;s not a problem for us yet, but we&rsquo;ll run into it eventually.</P>
<P><B>LISTING 51.4 L51-4.C</B></P>
<!-- CODE //-->
<PRE>
/* Routines to perform incremental rotations around the three axes */
#include &ltmath.h&gt
#include &#147;polygon.h&#148;
#include &ldquo;polygon.h&rdquo;
/* Concatenate a rotation by Angle around the X axis to the transformation in
XformToChange, placing result back in XformToChange. */
@ -57,12 +57,12 @@
double Temp10, Temp11, Temp12, Temp20, Temp21, Temp22;
double CosTemp = cos(Angle), SinTemp = sin(Angle);
/* Calculate the new values of the four affected matrix entries */
Temp10 = CosTemp*XformToChange[1][0]&#43; -SinTemp*XformToChange[2][0];
Temp11 = CosTemp*XformToChange[1][1]&#43; -SinTemp*XformToChange[2][1];
Temp12 = CosTemp*XformToChange[1][2]&#43; -SinTemp*XformToChange[2][2];
Temp20 = SinTemp*XformToChange[1][0]&#43; CosTemp*XformToChange[2][0];
Temp21 = SinTemp*XformToChange[1][1]&#43; CosTemp*XformToChange[2][1];
Temp22 = SinTemp*XformToChange[1][2]&#43; CosTemp*XformToChange[2][2];
Temp10 = CosTemp*XformToChange[1][0]+ -SinTemp*XformToChange[2][0];
Temp11 = CosTemp*XformToChange[1][1]+ -SinTemp*XformToChange[2][1];
Temp12 = CosTemp*XformToChange[1][2]+ -SinTemp*XformToChange[2][2];
Temp20 = SinTemp*XformToChange[1][0]+ CosTemp*XformToChange[2][0];
Temp21 = SinTemp*XformToChange[1][1]+ CosTemp*XformToChange[2][1];
Temp22 = SinTemp*XformToChange[1][2]+ CosTemp*XformToChange[2][2];
/* Put the results back into XformToChange */
XformToChange[1][0] = Temp10; XformToChange[1][1] = Temp11;
XformToChange[1][2] = Temp12; XformToChange[2][0] = Temp20;
@ -77,12 +77,12 @@
double CosTemp = cos(Angle), SinTemp = sin(Angle);
/* Calculate the new values of the four affected matrix entries */
Temp00 = CosTemp*XformToChange[0][0]&#43; SinTemp*XformToChange[2][0];
Temp01 = CosTemp*XformToChange[0][1]&#43; SinTemp*XformToChange[2][1];
Temp02 = CosTemp*XformToChange[0][2]&#43; SinTemp*XformToChange[2][2];
Temp20 = -SinTemp*XformToChange[0][0]&#43; CosTemp*XformToChange[2][0];
Temp21 = -SinTemp*XformToChange[0][1]&#43; CosTemp*XformToChange[2][1];
Temp22 = -SinTemp*XformToChange[0][2]&#43; CosTemp*XformToChange[2][2];
Temp00 = CosTemp*XformToChange[0][0]+ SinTemp*XformToChange[2][0];
Temp01 = CosTemp*XformToChange[0][1]+ SinTemp*XformToChange[2][1];
Temp02 = CosTemp*XformToChange[0][2]+ SinTemp*XformToChange[2][2];
Temp20 = -SinTemp*XformToChange[0][0]+ CosTemp*XformToChange[2][0];
Temp21 = -SinTemp*XformToChange[0][1]+ CosTemp*XformToChange[2][1];
Temp22 = -SinTemp*XformToChange[0][2]+ CosTemp*XformToChange[2][2];
/* Put the results back into XformToChange */
XformToChange[0][0] = Temp00; XformToChange[0][1] = Temp01;
XformToChange[0][2] = Temp02; XformToChange[2][0] = Temp20;
@ -96,12 +96,12 @@
double Temp00, Temp01, Temp02, Temp10, Temp11, Temp12;
double CosTemp = cos(Angle), SinTemp = sin(Angle);
/* Calculate the new values of the four affected matrix entries */
Temp00 = CosTemp*XformToChange[0][0]&#43; -SinTemp*XformToChange[1][0];
Temp01 = CosTemp*XformToChange[0][1]&#43; -SinTemp*XformToChange[1][1];
Temp02 = CosTemp*XformToChange[0][2]&#43; -SinTemp*XformToChange[1][2];
Temp10 = SinTemp*XformToChange[0][0]&#43; CosTemp*XformToChange[1][0];
Temp11 = SinTemp*XformToChange[0][1]&#43; CosTemp*XformToChange[1][1];
Temp12 = SinTemp*XformToChange[0][2]&#43; CosTemp*XformToChange[1][2];
Temp00 = CosTemp*XformToChange[0][0]+ -SinTemp*XformToChange[1][0];
Temp01 = CosTemp*XformToChange[0][1]+ -SinTemp*XformToChange[1][1];
Temp02 = CosTemp*XformToChange[0][2]+ -SinTemp*XformToChange[1][2];
Temp10 = SinTemp*XformToChange[0][0]+ CosTemp*XformToChange[1][0];
Temp11 = SinTemp*XformToChange[0][1]+ CosTemp*XformToChange[1][1];
Temp12 = SinTemp*XformToChange[0][2]+ CosTemp*XformToChange[1][2];
/* Put the results back into XformToChange */
XformToChange[0][0] = Temp00; XformToChange[0][1] = Temp01;
XformToChange[0][2] = Temp02; XformToChange[1][0] = Temp10;