On Sat, Dec 15, 2012 at 10:29 PM, Wolfgang Strobl <ne...@mystrobl.de> wrote: > Christian Heimes <christ...@python.org>: > >>To be fair, memcpy() is a pretty simple function. > > This, of course, depends on the target architecture. See for example > <http://www.microchip.com/forums/m120261.aspx> or > <http://www.piclist.com/techref/member/IMB-yahoo-J86/memcpy.asm.htm> > >>It can be implemented >>in just about two lines of C code plus five lines of boiler plate. > > Or in a single line of IBM 360 ASM/F Code, plus boiler plate. :-) > <http://www.bixoft.nl/english/opcd/mvcl.htm> > > In 1979, I implemented various block move code snippets for three > different eight bit microprocessors of that time. The parameterized > variant took between 20 and >30 instructions.
Same in 8086 - you just do a REP MOVSB (after setting CX to the length, DS:SI to the source, ES:DI to the destination). But it's slower than the equivalent with more explicit register action. ChrisA -- http://mail.python.org/mailman/listinfo/python-list
