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

Re: itk-4.5 and python



On Wed, 2014-03-26 at 11:51 -0400, Paul Novotny wrote:
> On Wed, 2014-03-26 at 12:10 +0100, Gert Wollny wrote:
> >    "-Wall -Wextra -Wdeprecated-declarations -O2 -g -funroll-loops 
> > -ftree-vectorize -pthread -march=native -mtune=native"
> > 
> > in the shell, and when I don't undefine these before running 
> > svn-buildpackage they are actually used, resulting in failing tests 
> > because the compiler issues instructions for aligned reads on un-aligned 
> > memory locations. Native in this case is AMD Phenom(tm) II X6 1035T.
> 
> Gert, out of curiosity, which tests fail because of this?
The tests  itkBoxSigmaImageFilterTest 3 and 10 fail in a very simple
loop. 
IIRC it was this loop in 

BoxSigmaImageFilter< TInputImage, TOutputImage >
::ThreadedGenerateData(const OutputImageRegionType &
outputRegionForThread, ThreadIdType threadId)

 ...
  typename TInputImage::SizeType internalRadius;
  for ( unsigned int i = 0; i < TInputImage::ImageDimension; i++ )
    {
    internalRadius[i] = this->GetRadius()[i] + 1;
    }
 ...


regards, 
Gert 

PS: Please put the list into the "To" field then my mail client offers
me directly "reply to the list", and there is no need to CC me.
thanks :)


Attachment: signature.asc
Description: This is a digitally signed message part


Reply to: