2011 Update 13: Fixes, Release Notes, vs. 2013?

2011 Update 13: Fixes, Release Notes, vs. 2013?

Imagen de Transpire

With the release of Update 13, I am trying to evaluate the fixes to determing if we need to upgrade, and ran into the following issues/questions:

1.) The Release Notes linked from the downloads page have apparently not yet been updated for Update 13 - They still have a date of 14-August, with no changes since Update 12:

http://registrationcenter-download.intel.com/akdlm/irc_nas/2836/w_fcompx...

2.) The fixes list (http://software.intel.com/en-us/articles/intel-composer-xe-2011-compiler...) lists DPD200236414, "Fortran Incorrect results in vectorized loop for the SSE level less than Intel® Advanced Vector Extensions (Intel® AVX)"  I have searched for DPD200236414, but cannot find it.  Where can I go to find out exactly what the bug was, how it manifested itself?

3.) Where can I find a comparison of the 2011 vs. 2013 versions of Visual Fortran (so I know what has been fixed/changed/enhanced)?

Thanks,

 - Rick Gould, rick@transpireinc.com

publicaciones de 6 / 0 nuevos
Último envío
Para obtener más información sobre las optimizaciones del compilador, consulte el aviso sobre la optimización.
Imagen de Steve Lionel (Intel)

The release notes were not updated for Update 13, and as far as I know there are no fixes to the Fortran compiler in that update. I suggest that you ignore update 13 to the 2011 product.

The 2013 release notes tell you what is new and what has changed. Bug fixes are listed in the "ReadMe" file that is linked on the same page where you download the new installer.

Steve
Imagen de Transpire

Thanks Steve, but according to the fixes list I referenced, there are two fixes to the Fortran compiler:
http://software.intel.com/en-us/articles/intel-composer-xe-2011-compiler...

Of particular interest to me is:
DPD200236414, "Fortran Incorrect results in vectorized loop for the SSE level less than Intel® Advanced Vector Extensions (Intel® AVX)"

I have searched for DPD200236414, but cannot find it. Where can I go to find out exactly what the bug was, how it manifested itself?

Thanks,

- Rick

Imagen de Steve Lionel (Intel)

That's an internal tracking ID. I will look tomorrow to see if it's anything I can describe - often this sort of issue is very specific to a particular program.

Steve
Imagen de Tim Prince

I was informed that update 13 fixed an issue with suppression of auto-vectorization in OpenMP parallel regions. That doesn't appear in the quoted list.
Generally speaking, new features introduced in ifort 13.x can't be expected to appear in 12.1 updates, while most fixes for 12.1 should appear soon in 13.0 updates.

Imagen de Steve Lionel (Intel)

I looked at the issue but, as I expected, there isn't any useful way I can describe it. It was a loop involving array operations that, when vectorized using SSE instructions, got a wrong answer. But it took a very peculiar combination of operations to reveal the problem.

Steve

Inicie sesión para dejar un comentario.