Hello,
I've a dual Pentium III (with Streaming SIMD Extensions)and I've tried compiling code with these instructions enabled using the Intel C Compiler v 5.01. The optimizations string is:
/G6 /Gr /MT /W4 /Ob2 /I /QxK
The odd thing, it runs slower (!). Considerably slower than when I leave out the /QxK.
Clearly I'm going wrong somewhere. Any pointers?
Hello,
It has been a while since version 5.0 ...
I dont remember if there were any obvious gotchas.
When you looked at the vectorization report, did you see quite a few loops get vectorized? You might want to check to see if one of these is getting hit.
Also you might want to explicitly try setting Qfp_port-.
Dis that help?
Please open this in premier.intel.com when you isolate a testcase.
Cheers,
Ganesh
Addionally you might want to consider trying version 7.0 (trial download at http://intel.com/software/products/compilers). You can install 7.0 in addition to 5.0.
There have been quite a few improvements to the vectorizer.
Take care,
Ganesh