Peter Bartlett wrote:
And that should make you sceptical of the results!
Indeed it should! I reran the test on VC9 SP1 (the VC8 in my post was a typo it was VC9). I got essentially identical results to my first run. I changed it from maximize speed to full optimizations and got much better results. I changed it to maximize speed and got similar results to full optimization. Confused? I was. Turns out the VC9 GUI lies. If I had optimization on inherit from project defaults it said Maximize Speed (/O2) but the /O2 switch did NOT appear on the command line. So my initial test was being run without the /O2 switch. Thanks Microsoft. VC9 SP1 with _SECURE_SCL=0 /O2 /Ob2 against boost trunk: Iterator accumulate took 0.109 seconds. Pointer accumulate took 0.11 seconds. Iterator for loop took 0.093 seconds. Pointer for loop took 0.094 seconds. Index for loop took 0.094 seconds. Iterator for_each took 0.093 seconds. Pointer for_each took 0.094 seconds. BOOST_FOREACH took 0.125 seconds. MSVC for each took 0.109 seconds. Press any key to continue . . . VC9 SP1 with _SECURE_SCL=0 /Ox /Ob2 against boost trunk: Iterator accumulate took 0.109 seconds. Pointer accumulate took 0.11 seconds. Iterator for loop took 0.109 seconds. Pointer for loop took 0.109 seconds. Index for loop took 0.11 seconds. Iterator for_each took 0.109 seconds. Pointer for_each took 0.109 seconds. BOOST_FOREACH took 0.11 seconds. MSVC for each took 0.109 seconds. Press any key to continue . . . I also corrected the "Index for_each" to read "Iterator for_each". -- Michael Marcin