AMD 250 manual DirectPath Instructions, Optimization

Software Optimization Guide for AMD64 Processors 25112 Rev. 3.06 September 2005

4.1DirectPath Instructions

Optimization

❖Use DirectPath instructions rather than VectorPath instructions. (To determine the type of an instruction—either DirectPath or VectorPath—see Appendix C, “Instruction Latencies.”)

Application

This optimization applies to:

•32-bit software

•64-bit software

Rationale

DirectPath instructions minimize the number of operations per AMD64 instruction, thus providing for optimally efficient decode and execution. Up to three DirectPath Single instructions, or one and a half DirectPath Double instructions, can be decoded per cycle. VectorPath instructions block the decoding of DirectPath instructions.

The AMD Athlon 64 and AMD Opteron processors implement the majority of instructions used by a compiler as DirectPath Single and DirectPath Double instructions. However, assembly writers must still take into consideration the use of DirectPath versus VectorPath instructions.

72 Instruction-Decoding Optimizations Chapter 4