Comments on: Efficient JNI programming IV: Wrapping native data objects http://kennke.org/blog/2007/07/30/efficient-jni-programming-iv-wrapping-native-data-objects/ Roman Kennke's ramblings Wed, 07 Jan 2009 04:23:20 +0000 http://wordpress.org/?v=2.5 By: Satyavathi http://kennke.org/blog/2007/07/30/efficient-jni-programming-iv-wrapping-native-data-objects/#comment-65010 Satyavathi Mon, 05 Jan 2009 07:06:13 +0000 http://kennke.org/blog/2007/07/30/efficient-jni-programming-iv-wrapping-native-data-objects/#comment-65010 Can DirectByteBuffer be used for C++ classes also, just like we use it for C Structures. Any pitfalls if it can be used? Can DirectByteBuffer be used for C++ classes also, just like we use it for C Structures. Any pitfalls if it can be used?

]]> By: Satyavathi http://kennke.org/blog/2007/07/30/efficient-jni-programming-iv-wrapping-native-data-objects/#comment-65007 Satyavathi Wed, 31 Dec 2008 11:10:54 +0000 http://kennke.org/blog/2007/07/30/efficient-jni-programming-iv-wrapping-native-data-objects/#comment-65007 there is a pointer that points to millions entries in the memory. It is very easy to get nextelement in c, by just incrementing the pointer. If you want to get the million entries that are there in memory to a java object, this can be done in a single JNI call, poplating the java object. To get into the java object you need to traverse over all the entries once, and if these entries are to be processed then one more scan is needed on the java object. Processing can be done only on the java side. Therefore 2 scans on millions of entries. In the other scenario, do a single scan over the entries, but call million times the JNI method. Do you still recommend to get everything on to the java side and process. there is a pointer that points to millions entries in the memory.
It is very easy to get nextelement in c, by just incrementing the pointer.
If you want to get the million entries that are there in memory to a java object, this can be done in a single JNI call, poplating the java object.
To get into the java object you need to traverse over all the entries once, and if these entries are to be processed then one more scan is needed on the java object. Processing can be done only on the java side.

Therefore 2 scans on millions of entries.

In the other scenario, do a single scan over the entries, but call million times the JNI method.

Do you still recommend to get everything on to the java side and process.

]]> By: Satyavathi http://kennke.org/blog/2007/07/30/efficient-jni-programming-iv-wrapping-native-data-objects/#comment-65006 Satyavathi Wed, 31 Dec 2008 10:23:50 +0000 http://kennke.org/blog/2007/07/30/efficient-jni-programming-iv-wrapping-native-data-objects/#comment-65006 There is native pointer that points to Huge memory. Will there be performance pitfall, if all the info in memory to a java object, and do processing on java code? or is it better to do processing on the fly by using the native ptr - by invoking the native methods many times. There is native pointer that points to Huge memory.
Will there be performance pitfall, if all the info in memory to a java object, and do processing on java code? or is it better to do processing on the fly by using the native ptr - by invoking the native methods many times.

]]> By: Olkenava » Efficient JNI programming IV: Wrapping native data objects http://kennke.org/blog/2007/07/30/efficient-jni-programming-iv-wrapping-native-data-objects/#comment-55410 Olkenava » Efficient JNI programming IV: Wrapping native data objects Mon, 05 Nov 2007 16:02:41 +0000 http://kennke.org/blog/2007/07/30/efficient-jni-programming-iv-wrapping-native-data-objects/#comment-55410 [...] read more | digg story [...] […] read more | digg story […]

]]>