[vsipl++] Matlab IO Patch

[Jules Bergmann]

Mark Mitchell wrote:
> Jules Bergmann wrote:
> 
>>    int32_t pad[Dim%2];
>>    data_element array_name_header;
> 
> Zero-element arrays are not valid ISO C++.  However, if data_element is
> a 64-bit type, then it will be aligned to an 8-byte boundary on most
> systems.

Thanks Mark for pointing that out, I was worried that zero-element 
arrays might not work :(

data_element is a struct containing two 32-bit types, will that be 
aligned to 64-bits?

If we rely on compiler to do the padding for us, we definitely should 
capture this with an assertion so that we know if it ever breaks.

Alternatively, we could use a union to do the padding, something like:

   union
   {
     int32_t dim[Dim];
     int32_t pad[Dim + Dim%2];
   }

> 
>> For int, we need some support from configure.ac to determine whether int is 32 bits (miINT32) or 64 bits (miINT64). 
> 
> Why not just:
> 
> template <>
> struct Matlab_type_traits<int>
> {
>   static int const data_type = sizeof (int) == 4) ? miINT32 : miINT64;
> }

Good suggestion, thanks!

				-- Jules

-- 
Jules Bergmann
CodeSourcery
jules at codesourcery.com
(650) 331-3385 x705