<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<div class="moz-cite-prefix">Hey,<br>
<br>
</div>
<blockquote cite="mid:5363976E.9000603@gmail.com" type="cite">
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
<blockquote cite="mid:536236D9.6070300@yahoo.de" type="cite">When
you try RAM_STORE on Android this probably won't happen.</blockquote>
<br>
Actually I see the same behavior even with RAM_STORE.<br>
The route calculations after the 1st one are more quick and seem
to getting quicker over the time for the first ones.<br>
I measured the time also by using your StopWatch method from
Android sample.<br>
</blockquote>
<br>
With the dalvik hotspot in mind that is not too strange. Although I
expected the loading from disk a lot more important<br>
<br>
<blockquote cite="mid:5363976E.9000603@gmail.com" type="cite"> <br>
<blockquote cite="mid:536236D9.6070300@yahoo.de" type="cite">a
random query through the area is probably more efficient and
simpler.<br>
</blockquote>
<br>
This approach could work if the random query has valid from / to
so that the route calculation can finish,<br>
but with random graphs from the disk how can we get valid from /
to locations?<br>
</blockquote>
<br>
e.g. instead of using the lat,lon you can use two random node ids
for querying the algorithm. But this requires using low level API so
maybe I should introduce a graphHopper.warmUp method. Feel free to
open an issue for this or even better a pull request :)<br>
<br>
Regards,<br>
Peter.<br>
<br>
</body>
</html>