I'm working on a C++ rendering app with mostly single-precision
floating point calculations. Performance on a G5 is pretty decent,
but a simple recompile for i386 showed very poor performance on a
Core2 iMac -- it runs about half as fast on the Intel.
The only compiler flags that really made a difference were "-m64" or
"-arch x86_64". With these options the iMac was just as fast as the
G5. As I understand it, I can't have a GUI app (yet) with 64-bit
addressing turned on.
Can someone of you give me insight as to why the code is so much
faster with these options? I suspect that the memory addressing is
not the cause, but some Intel EM64T instructions are. Can I exploit
that in a 32-bit app?
By the way: I also tried the Intel compiler, and its output was about
ten times slower than with GCC! Maybe it is because my code makes use
of some heavy template expression magic which ICC doesn't like. Does
anyone share this experience?
Thanks in advance and best regards,
Do not post admin requests to the list. They will be ignored.
PerfOptimization-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden