Re: compound types on different machines - is this expected?

• To: netcdf-hdf@xxxxxxxxxxxxxxxx
• Subject: Re: compound types on different machines - is this expected?
• From: Quincey Koziol <koziol@xxxxxxxxxxxxx>
• Date: Sat, 30 Apr 2005 12:36:26 -0500 (CDT)

Hi Ed,

> Elena Pourmal <epourmal@xxxxxxxxxxxxx> writes:
> 
> > Here is an example how to use H5Tget_native instead of constructing memory 
> > type
> >
> > http://hdf.ncsa.uiuc.edu/training/other-ex5/h5nestc.c
> >
> 
> OK, this seems like magic!

    :-)

> I've changed by program to use the get_native_type function, and now
> everything works perfectly. I've read the documentation for
> H5Tget_native_type and it's really not clear to me how it manages to
> deal with the padding of a struct.
> 
> Can someone explain what's going on? What does get_native_type do for
> compound types?

    As part of the process of building the library, we generate information
about the compiler's padding for fields in compound types.  H5Tget_native_type
uses that information to translate a non-native compound type into a native
compound type.

    Quincey

P.S. - Sorry this is late - it was mis-filed by my procmail recipes and I just
    found it.

• Follow-Ups:
  • Re: compound types on different machines - is this expected?
    • From: Ed Hartnett

• References:
  • Re: compound types on different machines - is this expected?
    • From: Ed Hartnett