I have some problems with my source code, when I build it with Intel C++ Compiler 11 and Intel C++ Compiler 14. The difference is showed at the table below:
|Intel Compiler 11||Intel Compiler 14|
|Original source (ms)||120||110|
|Optimized source (ms)||100||110|
+ Original source: I use many variables to store data and calculate on it.
+ Optimized source: I use a struct to store variables on original source and vectorize my calculations. Moreover, I have removed some redundant calculations.
I don't know the reason for this difference between two sources above.
Please help me explain this difference ?
If you aren't willing to show the example, you may need to compare opt reports and asm files to see if there is a reason. Such small differences may be due to effects beyond your control.
i have the same situation under Linux , we use a testsuite from nearly 12 testprograms, dhrystone, whetstone, scimark2c , flops,
and so far and compile it with the gcc-4.9, gcc-4.8.3 , icc 15.0.0 (Cluster_Studio_XE 2015 Beta ), pathscale and pgi c/c++ compilers
for all we use -O3 -m64 -mtune=corei7 as compile options on dhrystone, whetstone, livermoore loops , flops was the intel compiler
absolut the fastest, only on the scimark2c benchmark was the gcc-4.9 absolut the fastest between 8 to 10 % faster then the
Intel C/C++ compiler, all on the same hardware, ASUS P6T6 WS Revolution Motherboard Intel i7-950 with stable at 4.00 Ghz
24 Gbyte RAM DDR3 XMS, two GeForce GTX560 Ti in SLI Mode, openSUSE 13.2 M0 64 Bit Linux Kernel 3.15.1, KDE 4.13.40,
actual NVIDIA Driver 334.17
Can you provide the sample code in question? And what's the specific versions and compilation options for both compilers?
Intel Developer Support
Please participate in our redesigned community support web site:
User forums: http://software.intel.com/en-us/forums/