topic enabling s/w prefetching in Software Tuning, Performance Optimization & Platform Monitoring

enabling s/w prefetching

ramacn — Tue, 03 Apr 2012 23:50:10 GMT

Hi all,

I was trying to enable software prefetching on Opteron by compiling an application using Intel compiler. But I am facing some problem.

I use the below compilation command.

-> icc -openmp -O3 -xsse4.2 -opt-prefetch=4 is.c

But when I run the program I get the below error.

"Fatal Error: This program was not built to run on the processor in your system.

The allowed processors are: Intel processors with SSE4.2 and POPCNT instructions support."

The CPU model is:

model name : Quad-Core AMD Opteron Processor 2354

Does anybody know what is the correct compiler option for icc to enable s/w prefetching? Please let me know.

Thank you

Regards,

Ram

enabling s/w prefetching

TimP — Wed, 04 Apr 2012 03:01:40 GMT

If you want sse4.2 code and also an Opteron code path, you will need something like -axsse4.2 -msse3 . Prefetch option doesn't over-ride your architecture option.

enabling s/w prefetching

ramacn — Wed, 04 Apr 2012 03:17:02 GMT

Hi Tim,

I tried this option, but I don't think the prefetch instruction is being generated.

I executed below command with the option you mentioned along with "-S" option to see the assembly file.

--> icc -c -openmp -O3 -axsse4.2 -msse3 -opt-prefetch=4 -S is.c

But the generated assembly file is.s doesn't seem to have any prefetch instruction inserted.

Between, I happened to read some where that '-xsse4.2' option is must to generate prefetch instruction, hence I was using it before.

Let me know what you think about this approach.

Regards,

Ram