Late-breaking updates to the Intel(R) C++ and Fortran Composer XE 2013 documentation specific to the Intel(R) Many Integrated Core (Intel(R) MIC) architecture will be shown here. Please subscribe to this post, add it to your RSS feed or otherwise monitor it regularly for future changes.
Runtime errors or crashes when running an application built with the initial Intel® Composer XE 2013 product release with the offload libraries from 2013 update 1
There is a breaking binary compatibility change in the offload libraries for Intel® Composer XE 2013 update 1 that will cause runtime errors or crashes if you use the update 1 libraries with a binary built with the previous Intel Composer XE 2013 compiler. Examples of the errors you may observe in this situation are:
***Warning: offload to device #0 : failed
Segmentation fault (core dumped)
terminate called after throwing an instance of 'COIRESULT'
terminate called recursively
CARD--ERROR:1 myoiPageFaultHandler: (nil) Out of Range!
CARD--ERROR:1 _myoiPageFaultHandler: (nil) switch to default signal handle
CARD--ERROR:1 Segment Fault!
HOST--ERROR:myoiScifGetRecvId: Call recv() Header Failed ! errno = 104
^CHOST--ERROR:myoiScifSend: Call send() Failed! errno = 104
HOST--ERROR:myoiSend: Fail to send message!
HOST--ERROR:myoiBcastToOthers: Fail to send message to 1!
HOST--ERROR:myoiBcast: Fail to send message to others!
To resolve these issues, you should recompile your application fully with the Intel Composer XE 2013 update 1 compiler in order to use the offload libraries included in the new package.
Default code generation no longer supports Intel® Xeon Phi™ coprocessor A0 stepping in Composer XE 2013 Update 1
Intel® Composer XE 2013 update 1 now generates new streaming store instructions that were introduced in the Intel® Xeon Phi™ coprocessor B0 stepping. These instructions are not supported on the A0 stepping, and will cause runtime errors. If you require your application to run on A0 steppings, use the option –opt-streaming-stores never to avoid generating these instructions. This may decrease performance on B0 or later steppings.