Hello,

> The following commands for ex(1) will fix the asm:
> 
> g/;/ s//#/
> g/[^a-z]r[0-9]/ s/\([^a-z]\)r\([0-9]\)/\1%r\2/g
> g/[^a-z]cr[0-9]/ s/\([^a-z]\)cr\([0-9]\)/\1%cr\2/g
> g/[^a-z]v[0-9]/ s/\([^a-z]\)v\([0-9]\)/\1%v\2/g
> g/hi16(.*)/ s;hi16(\(.*\));\1@h;
> g/lo16(.*)/ s;lo16(\(.*\));\1@l;
> wq
> 
> and make it assemble with the NetBSD gas.  Feel free to adapt
> them to sed or whatever method make you happy. :)

Thanks, I'll try it as soon as possible ( machines are busy right now... ) - this could even work with AIX when using a gcc compiled to use gas ( IBM's assembler won't know anything about altivec, at least none that would work on my box ) - maybe I'll post some benchmark results with assembler vs. code generated by xlc :)

have fun
Michael