We have run into some really strange questions  in assembly class. Anyone know a site or two that would get into the specifics like finding average cost per bit, and potential opcodes and register amounts, how sec code works. Google isn't helping much, maybe you guys can give some sites. Thx
Posted on 2005-10-20 11:07:52 by grl_jst1otheguyz
About 10 years ago processor performance was specified in things like cycle count but the hardware has changed so much that this tecnology no longer applies. The mechanics are that some instructions are faster than others because of how they are built into the hardware and generally the most used lower level ones are a lot faster than the less used higher level types.

Basically the fast ones are directly built into the hardware and the slower ones are constructed in what is called microcode but note that this changes from processor to processor over time as newer versions are released. Getting some idea of how different instructions work comes from understanding how to schedule them in sequence so they can take advantage of fundamental processor architecture. Particular to this you need to learn some about how a pipeline works then the consideraions of multiple pipelines working together and you will get some idea of what instruction costs are.

Reference material is not for the faint of heart, the series of Intel manuals contain the technical reference for Intel hardware, there is specific data on AMD hardware available from their site ad its worth the effort to get Agner Fog's optimisation manual and you will be well on the way to understanding this stuff. The other useful thing to do is start writng the code and timing it as this is the only reliable way to determine if a particular instruction sequence is faster than another.
Posted on 2005-10-20 18:04:17 by hutch--