Hehe, now you will get PMC tester?
http://www.agner.org/optimize/testp.zip
Madis and i had gone through the same trouble
while optimizing our 3D-gfx library and we agreed on PMC tester.
It give same same result on any PC (if setup correctly)
The only drawback is its coded in C and a big project to convert asm.
But i prototype all my code in C first anyway so having fast C code is good start
then when convert to asm the optimizations are easy to spot and fix.