12-23-2008 08:42 AM
I'm using the griddata function in Mathscript as a direct replacement of a Matlab function we were using in the past.
I discovered a bug with griddata which was fixed in Labview 8.5. The funciton executed very well in that version.
I've upgraded the code to LV8.6, and now when executing griddata on the same data set in either 'linear' or 'cubic' mode, the function is MUCH slower than when using 'v4' or 'nearest'. All four methods had the same execution time in LV8.5.1
I have searched many topics covering issues with griddata and 'cubic', but I'm pretty sure this is a new bug.
I've attached a VI containing my data set. Running this VI in 8.5.1 and using 'linear' mode is much faster than in 8.6.
I'll be going back to Matlab, since the rest of the code has already been upgraded to 8.6 and this makes regression it unusable for my purposes.
Any ideas?
Thanks.
12-23-2008 09:12 AM
I should correct one detail in my message. The girddata bug was originally fixed for 8.5.1, not 8.5, per the following CAR: 41OHD4VQ
The attached VI should be run in 8.5.1 and then in 8.6 ('linear' method) to see the performance decrease.
12-24-2008 06:42 AM
Hi,
Thanks for letting us know this problem. I will look into it.
12-28-2008 11:34 PM
Hi,
Yes, the performance decreases in LabVIEW 8.6. The fix for original bug might cause the degradation. I have filed another bug report about this problem and will fix it later. Thanks again!