gm2
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Gm2] Update on numerical library


From: Michael Riedl
Subject: Re: [Gm2] Update on numerical library
Date: Thu, 14 Feb 2019 19:45:08 +0100
User-agent: Mozilla/5.0 (X11; Linux i686; rv:60.0) Gecko/20100101 Thunderbird/60.4.0

Gaius,

having switching float-sizes would be great !!! Will do tests whenever that will help.

At the moment I would need to review all the code and the interfaces (e.g to BLAS routines) which would most probably end up with a lot of compiler directives - not a nice solution at all. And for any legacy code I would assume that such an option would make it much easier to get them compiled and running with GM2.

Michael


Am 11.02.19 um 13:38 schrieb Gaius Mulley:
Michael Riedl <address@hidden> writes:

Gaius,

with the latest GM2 version I was able to compile

- all files for the InOut library expect FormAusLib.mod (error in
PreParseFormat)

- all but 4 files in the Numeric library

   - LibDBlasM2.mod & LibDBlas.mod : Compiler complains about missing
RETURN - I cannot find an issue (maybe blind on my own code ...)

   - SpezFunkt3.mod : Here I made a lot of tests to comment out part of
code etc. etc. but cannot isolate the issue

     I attached a file where most code is commented out, maybe helpful
in isolating the crash reason ...

  - Differ.mod : Isolated the error (see comment around line 74) - can
be compiled now but issue should be fixed ;-)

The preprocessor can be build without issue and works fine. If you
still have issues let me know and I will send you pre-processed files
separately, but just also creating some tar files which I will also
put in the repository ... (prepocessedfile.GM2.tar,
preprocessedfiles.XDS.tar)


Updated all code to "https://opensourceprojects.eu/p/modula2/";.

So we are getting close ...

Michael


PS : I will write another mail soon as I would like to re-open the old
discussion about the "default" sizes of floating point variables
... it's causing some pain :-(
Hi Michael,

many thanks for the email - I will look into your code!  Yes, ok, it
might be possible to have switchable default floating point sizes.  It
would require building the libraries multiple times and also require
changing some of the compiler building blocks.  But in principle I think
it is possible,


regards,
Gaius



reply via email to

[Prev in Thread] Current Thread [Next in Thread]