🚀 go-pugleaf

vhs@rhein-main.de (Volker Herminghaus-Shirai) writes:

>Subject line almost says it all. Many modern processors have an
>instruction to swap the bytes in a word. Can one quantify the
>win of using this instruction vs. the equivalent sequence of
>instructions? How much silicon is needed? What applications
>win the most and how much? There must be *some* resaon why the
>RISC folks put it in, right?

This is mostly so big-endian machines can simulate little-endian
behaviour and vice versa (see discussion elsewhere in this group).

The savings depend on how long it takes to swap the bytes using other
instructions. Here is an example from an ond posting by
RWilson@acorn.co.uk (one of the designers of the ARM processor).

>ARM's byte sex is as VAX and NS32000 (little endian). The byte sex of
>a 32 bit word can be changed in 4 clock ticks by:
>      EOR R1,R0,R0,R0R #16
>      BIC R1,R1,#&FF0000
>      MOV R0,R0,ROR #8
>      EOR R0,R0,R1,LSR #8
>which reverses R0's bytes. Shifting and operating in one instruction
>is fun. 

Newer processors (including the new ARMs) tend to have selectable
endianness instead of byte swap instructions.

	Torben Mogensen (torbenm@diku.dk)