problem using ctypes with MMX intrinsics

Robert Kern robert.kern at gmail.com
Wed Oct 15 19:23:39 EDT 2008


dcharno wrote:
> Robert Kern wrote:
>> Well, stack alignment would be a problem with how the shared library 
>> gets compiled, nothing to do with ctypes (I think). However, if you are 
> 
> The shared library is built with '-m32 -msse2' and works fine from a 
> C/C++ test harness, so I think its compiled okay.
> 
>  > passing in arrays from ctypes, *they* may also be misaligned. Try to
>  > check the addresses of the ctypes values you are passing in. I'm not
>  > entirely sure how to do that, though.
> 
> I'm only passing filenames and some string options through the API.  All 
> the data that gets processed with MMX/SSE is opened inside the shared 
> library.

Hmm. Okay. Start your program under gdb, find what data it's crashing on, check 
the alignment on it, and if it's misaligned, figure out where that data is 
coming from.

Is the ctypes driver doing the same exact operations as the C++ test harness? If 
not, it may be exposing a bug in your code. I've had a few occasions when using 
SSE2 intrinsics where I had a bug in my loop such that I was going off the end 
of the array due to a thinko.

-- 
Robert Kern

"I have come to believe that the whole world is an enigma, a harmless enigma
  that is made terrible by our own mad attempt to interpret it as though it had
  an underlying truth."
   -- Umberto Eco




More information about the Python-list mailing list