Fortran on Xeon X56xx vs Xeon E56xx

[hornos]

Solved: there was an optimization flag difference between one of the source files.

I have a fortran code which I compile with Intel Compiler 12.0.4 on two different machines with Xeon E5620 and X5650. The X machine produces a slightly bigger code. The code generated on the X machine runs well on both machines however the smaller code generated on the E machine produces wrong results on both. The problem is not presented if the code is built by an older 10.2 compiler.

OS: SLES 11.1

Compiler: 12.0.4 Fortran

Flags: -O2 -xSSE4.2 -ip -ftz

Is there any difference in code generation on various Xeon processors from the same family (56xx)?

[Patrick_F_Intel1]

Hello hornos,
Can you elaborate on 'produces wrong results on both'?
What are the errors?
I'm not a compiler guy but I'm sure this is one of the first questions the compiler folks will ask.
Thanks,
Pat

[TimP]

You could compare the compiler output files. There shouldn't be any difference, if the compiler installations and options chosen are the same. The -ip option is more likely to expose source code bugs such as data over-runs, as well as increasing the code size. -xSSE4.2 may produce smaller or slightly slower code than -xSSE4.1.