<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<font face="Georgia">I thought I'd extend the test to an older
squeak&vm<br>
<br>
---interpreter vm (3.8.12) squeak 3.2 ---------<br>
Float compareTiming: 1 to: 1.0 ==> 496<br>
Float compareTiming: 1.0 to: 1.0 ==> 761<br>
Float compareTiming: 1 to: 1 ==> 225<br>
Float compareTiming: 1.0 to: 1 ==> 494<br>
---interpreter vm (5.7.4.1) squeak 4.5 ---------<br>
Float compareTiming: 1 to: 1.0 ==> 719<br>
Float compareTiming: 1.0 to: 1.0 ==> 991<br>
Float compareTiming: 1 to: 1 ==> 312<br>
Float compareTiming: 1.0 to: 1 ==> 686<br>
<br>
The old one wins.<br>
<br>
Cheers,<br>
Bob</font>
<div class="moz-cite-prefix">On 9/27/13 2:12 PM, Bob Arning wrote:<br>
</div>
<blockquote cite="mid:5245CA86.1040400@comcast.net" type="cite">
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
Following up Herbert's question, I wrote:<br>
<br>
compareTiming: a to: b<br>
"<br>
---interpreter vm (5.7.4.1) ---------<br>
Float compareTiming: 1 to: 1.0 ==> 719<br>
Float compareTiming: 1.0 to: 1.0 ==> 991<br>
Float compareTiming: 1 to: 1 ==> 312<br>
Float compareTiming: 1.0 to: 1 ==> 686<br>
---cog vm (2749)---------<br>
<b>Float compareTiming: 1 to: 1.0 ==> 28162</b><br>
Float compareTiming: 1.0 to: 1.0 ==> 148<br>
Float compareTiming: 1 to: 1 ==> 97<br>
Float compareTiming: 1.0 to: 1 ==> 179<br>
"<br>
<br>
^[1000000 timesRepeat: [ <br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
a < b.<br>
]] timeToRun<br>
<br>
So, some obvious questions:<br>
- in the interpreter vm, why is comparing two floats slower that
an integer and a float?<br>
- what's up with Cog? Three tests are faster by the expected ratio
and one took a real dive.<br>
<br>
Also, the profiler Herbert used could use some attention to the
NewFloat printing:<br>
<br>
99.9 (898,069) Multiline1 doOneStep:<br>
<b>77.30000000000001</b> (694,902) Multiline1 [HkTestLine]
doTests:<br>
73.8 (663,438) HkRunin [HkTestStation] canStartNextTest:<br>
<br>
Cheers,<br>
Bob<br>
<br>
</blockquote>
<br>
</body>
</html>