| 19 Dec 2012 01:43 AM |
|
|
Hello,
It is not a first time when i am asking about IDL CPU parameters. You suggested me a help article from exelis home page nr. 3259. Actually it is not exist!
Could you suggest something else, please. Or explain by your self. I cant handle ENVI help.
ENVI help says, FLAASH works on every size of image. Firs error is non blank pixels. I increase the tile size every time up to 100 mb at the end it cant allocate the memory. I have 8 CPUs and 16 GB of RAM, i think it not so bad parameters. I think i need to change CPU parameters in ENVI! I have 6000000000 byte Unsigned Int image. The main question is, how many elements it have. 6000000000?
Sincerely, skywalker.
|
|
|
|
MariM
Veteran Member
Posts:2396  
| 19 Dec 2012 07:42 AM |
|
|
The memory allocation error is not affected by the CPU. It is related to the RAM on the system. Your image is very large and has substantial background so in order to access a tile of data large enough to actually contain some pixel values you need to set the tile size to be fairly large. Unfortunately, your system is running out of memory doing this. The alternative at this point is to subset the image and process them separately then mosaic them back together. Or you can use Quac. This issue with the background pixels should be fixed soon - possibly with the next release or service pack.
|
|
|
|
Deleted User
New Member
Posts: 
| 20 Dec 2012 12:37 AM |
|
|
I checked in Task Manager FLAASH action. It is using only one CPU and 20 % of the RAM memory.Why?
Could you answer please about array elements. It is the same amounth of bytes? Or it is something else?
We could i see which envi routine or model or algorithm is multithreated?
For example, could i perform K-means unsclassification with multi threading. If yes, how it happends? image 474256704 bytes, unsigned integer.
|
|
|
|
MariM
Veteran Member
Posts:2396
| 20 Dec 2012 07:47 AM |
|
|
Not all processes are multi-threaded. The RAM that ENVI/IDL needs to be contiguous ram, not just available ram. Since ram is quickly fragmented, you can end up with much smaller contiguous available RAM.
The number of bytes and elements aren't always the same. It depends on the data type. If the data is 16-bit integer, and you have a 1000(ns) x1000(nl)*5(nb)*2(byte) = 10,000,000 bytes.
|
|
|
|