🚀 go-pugleaf

In article <1991May9.025504.16646@iitmax.iit.edu>,
wallace@iitmax.iit.edu (Wallace) writes:
|> In article <3225@spim.mips.COM> cprice@mips.com (Charlie Price) writes:

|> >>>All current MIPS chips can flip byte order dynamically.  ARC-compliant
|> >If a program is not the native endianess, the OS flips the bytes in
|> >a word end-for-end on transfers to/from the kernel.
|> >A non-native text editor can create a text file that the native
|> >cat program can display.
|> >
|> >The cost of doing the data transform for all I/O is not prohibitive.
|>                                           ^^^^^^^
|> My kind of guy!
|> This can be pipelined in the cpu, so, the net cost is one cycle
|> per flip. Not bad!
|> 

As Charley Price said, this has to be done by the OS.  In a test recently,
I saw about 150 nanoseconds per byte for swapping.  For programs that are
completely I/O bound (cp for example) this doubles the kernel's copy loop
time but only amounts to 20% of systems time.  It has almost no effect on
wall clock or user time (which is determined by physical disk or network
timing).  For non I/O bound programs the time penalty was negligible.

-----------------------------------------------------------------------------
==	jeff kenton		Consulting at kenton@decvax.dec.com        ==

				-- Not the official word of Digital --

==	(617) 894-4508			(603) 881-0011			   ==
-----------------------------------------------------------------------------