if processing 30000 by 700000 array how long labview will take?

[janki]

I was just wandering if i can use labview or not for the following as the last day is tuesday.

i have 700000 locations' latitude,longitude.and i have 30000 other locations'latitude,longitude.

now i will take one by one location from 30000 and want to find all the locations out of 700000 which are within the 20 miles radius of my location(which is one from the 30000).

i dont want to compare distance between one point and all the other 700000. instead can we do this comparison for some lesser values. i mean just a look at all 700000 and it may be reduced without comparing each and every value. is it possible?

if not then--- 

I have made a distance calculator using formula node and it works with 100000.( i have tested) now if i make it to calculate for 700000 and that also 30000 times, how long will it take, anybody has any idea? here time is restriction.

i am using labview 8.6 evaluation version.

[Joseph_Loo]

If you preporcess your 700,000 locations by sorting Lat and Long (does not matter which one you pick). Once you know the actual location you are trying to find, you can then look at the Lat and Long. You can create a rough box of the Lat and Longs that are acceptable. There you can then use your distance formula to locate your 20 mile radius. The second method probably is order nxm where this should be somewhere order m or near linear depending on your densities.

[altenbach]

---

janki wrote:

I was just wandering if i can use labview or not for the following as the last day is tuesday.

---

Of course you can use LabVIEW!

If the problem fits into the memory contraint of a 32 bit OS, LabVIEW is equally sutable as any other programming language.

As Joseph already mentioned, deciding on a good internal data structure is important for speed and efficiency. Make sure to keep the large data structures "in place" and avoid constantly making data copies in memory. I assume you have an efficient algorithm to calculate the distance between two points given their long/lat coordinates.

Why don't you start with a scaled down program containing small arrays of artificial data and work out the main algorithm. Feel free to post it here so we can discuss they layout possibly offer improvments. Good luck! 🙂

[altenbach]

---

janki wrote:

I have made a distance calculator using formula node and it works with 100000.( i have tested) now if i make it to calculate for 700000 and that also 30000 times, how long will it take, anybody has any idea? here time is restriction.

---

It also all depends on the typical data and now you want the output represented. If typically all (or most) points are within 20 miles of each other, you will run out of memory for the result, guranteed. If a match is rare, your success is more likely. 😉

[janki]

 hey, joseph

thanks a lot

I had thoght this option considering lat and long as x and y coordinates . i have only one doubt-- if i will sort w.r.t lat, i think i will miss some  points (which are actually within 20 miles of radius) because i am considering only horizontal points and creating rough box.

infact what i am considering is i should sort horizontally as well as vertically then only i can get the whole circle( 20 miles radius).

please reply soon.

and i am not getting--The second method probably is order nxm where this should be somewhere order m or near linear depending on your densities.

and i will surely post it soon if i complete it successfully.

[Joseph_Loo]

You ar missing the point of the Lat and Long. You need to set the Lat assuming it is within 20 miles, same long, and Long range with 20 miles same lat., This is equivalent of an L1 measure. You have eliminated all the points that is not within 20 miles. Thus you now have a more limited number of points to look at. Of course there other methods to lock it down even more.