What are real performance gain? Doubt that it will be significant. 26.12.2016 2:59 PM "Colin Close" <[email protected]> napisał(a):
> The X32_ABI looks interesting and I think that is what D21 was referring > to at the end of his post. > It does look like this would offer improved performance for 32bit apps > running on 64bit boxes. > What is more the SSE instructions become available. > I wonder what is involved in creating such a release? > Best, > Colin > > On Monday, 26 December 2016 07:53:05 GMT Paulo César Pereira de Andrade > wrote: > > 2016-12-25 16:33 GMT-05:00 His Shadow <[email protected]>: > > > I thought that only addresses doubled, not the instruction opcodes. > > > > This is mostly right. Instructions that use 64 bit operands, or > > registers with absolute value larger than 7 also need a prefix > > https://en.wikipedia.org/wiki/VEX_prefix#REX > > > > Running in 32 bit mode also uses only 8 from the 16 registers. > > > > It is possible to have a 32 bit distro enabled, just that some > > directories will conflict, e.g. there is no /bin32 or /bin64. > > > > About the upper 32 bits, some instructions do not touch the upper > > 32 bit, some zero fill and other sign fill the result, and most have > > some encoding to tell if using 32 or 64 input operands. > > > > Some distros have a x32 version: > > https://en.wikipedia.org/wiki/X32_ABI > > that uses 32 bit pointers. > > > > Thanks, > > Paulo > > > > _______________________________________________ > > OM-Cooker mailing list > > [email protected] > > http://ml.openmandriva.org/mailman/listinfo/om-cooker_ml. > openmandriva.org > > > > > > _______________________________________________ > OM-Cooker mailing list > [email protected] > http://ml.openmandriva.org/mailman/listinfo/om-cooker_ml.openmandriva.org >
_______________________________________________ OM-Cooker mailing list [email protected] http://ml.openmandriva.org/mailman/listinfo/om-cooker_ml.openmandriva.org
