Lapack and GDL

[Robinajax]

Is it possible to integrate LAPACK (Linear Algebra Package) into the GDL compilation under Windows? I would like to write functions for automatic control and optimal filtering in GDL language, and this would be a great help!

[GillesDuvert]

Everything is possible with GDL 😄 ... Under Windows, a bit less, alas, as we have no Windows expert.

It depends what you seek. There are a number of LAPACK fuctions integrated into IDL (all the LA_XXXX commands) that are not already into GDL, perhaps you need one of them: in this case it would be good to ultimately insert them in GDL. At the moment a number of LA_XXX functions have been written in GDL but we preferred to use their equivalent in Eigen:: because we already used this library (and, apologies if this is not true, Eigen:: is just way faster) . I suggest to check Eigen:: capabilities first. (1)

You may consider using CALL_EXTERNAL, LINKIMAGE or DLM to connect GDL with an external code, a loadable library that uses LAPACK. (But currently this does not work well under Windows). Or directly integrate a set of LAPACK functions in GDL, and have GDL link with LAPACK, using the GDL internal C++ functions.

In summary this needs careful examination of the needs and the respective 'cost' of all possible solutions.

(1) In fact, should we add a LA_XXX function, we would always look fits at the possibility that a better (faster) code is available freely and does the same job.

[GillesDuvert]

You could start by cloning the GDL repo and try experiments?

[Robinajax]

Thanks for your answer, I will try to compile GDL + Lapack under Msys2 and test it! Best regards! Samuel Le lun. 28 juil. 2025 à 11:15, Giloo ***@***.***> a écrit :

…

You could start by cloning the GDL repo and try experiments? — Reply to this email directly, view it on GitHub <#2055 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AGBPYJEJBXW7G4SUQC3QK5D3KXS35AVCNFSM6AAAAACCQBZUVSVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTGOJQG44TOMA> . You are receiving this because you authored the thread.Message ID: ***@***.***>