Project

General

Profile

Forums » Support »

[Suggestion] Add in Math intrinsics

Added by Romain LE LAMER over 4 years ago

Hi,
I have a proposal for you, that of adding trunc(x,n) where n is the desired number of digits.

The idea came to me because I need to find a correspondence.
The DataType of a classic grib seems to be P17 during my calculation with exprf

10u = ((10u > 0) ? 1 : -1) * (sqr(nint(sqrt(abs(10u * 230.4)))) / 230.4);
10v = ((10v > 0) ? 1 : -1) * (sqr(nint(sqrt(abs(10v * 230.4)))) / 230.4);

I have to set the DataType to P24 to find an almost match.
With trunc(x,n) I wouldn't have this problem and I would match 100% which would be very interesting ;)
10u = trunc(((10u > 0) ? 1 : -1) * (sqr(nint(sqrt(abs(trunc(10u, 6) * 230.4)))) / 230.4), 6);
10v = trunc(((10v > 0) ? 1 : -1) * (sqr(nint(sqrt(abs(trunc(10v, 6) * 230.4)))) / 230.4), 6);


Replies (4)

RE: [Suggestion] Add in Math intrinsics - Added by Romain LE LAMER over 4 years ago

DataType F32 matches even better but weighs down the file even more ...

RE: [Suggestion] Add in Math intrinsics - Added by Ralf Mueller over 4 years ago

hi!

couldn't you use floor or ceil from libm?

Another option could be to substract both fields and work with a threashold:

cdo -gtc,0.0001 -sub infile1 infile2 outfile

cheers
ralf

RE: [Suggestion] Add in Math intrinsics - Added by Ralf Mueller over 4 years ago

libm offers a trunc which could be added to CDO's expr opertator (https://www.gnu.org/software/libc/manual/html_node/Rounding-Functions.html)

But I am not sure, that is really does, what you want ...

RE: [Suggestion] Add in Math intrinsics - Added by Romain LE LAMER over 4 years ago

Hi Ralf,

Indeed you are right on 2 points:
- I can use floor (x * 1000000) / 1000000 to get the truncation of my value with 6 digits.
- The function trunc (x) does not correspond to my expectations.

Thanks for the tip ;)

    (1-4/4)