Short:        Src/executable for 040 amigas, shows use of Move16 instruction
Architecture: m68k-amigaos

Recent: This is the 2nd release, the 1st release generated extra instcn.

The 1st release added an extra dc.w 0 to the move16 opcode generated
by the assembler, this is incorrect, it created the following:

move16 (a0)+,(a1)+
ori.b #0,d0

Basically it was harmless but was generating an extra instruction
between move16;s this release fixes that by generating the
correct opcode in a macro

For use with SAS/C 6.3
2 source files, Move16.c and Move16_a.asm


This archive contains source and executable which demonstrates
the use of the 68040 Move16 instruction on an 040 amiga.
I am not sure if this instruction is safe for all amigas,
but it is the fastest way to copy mem blocks on an 040
machine.

You must align the src and dest on 16 byte boundaries, and
on some 040's you need a NOP before each block
of MOVE16's and on some 040's a page fault during a move16
can cause problems, but on my A4000/040 It seems to work
just fine.

If anyone knows of any problems with the Move16 instruction
on the Amiga, or would like to email me the
times for their systems, send me internet mail at

uzun@crash.cts.com
or on bix as 
ruzun

This pgm has 2 copy block loops, one uses Move16's the
other longword moves, it reports the time to execute
each block for each mem type.

SAS/C 6.3 ASM BUG,
the SAS/C Assembler cannot correctly generate the opcode
for a move16 (a0)+,(a1)+, 
The asm file uses a macro to generate the correct
opcode for this instruction.

Results on my A4000/040

**FAST TO FAST**
MoveMem   = 2.2000 secs
MoveMem16 = 1.5000 secs
**Fast to CHIP**
MoveMem   = 3.3000 secs
MoveMem16 = 3.2800 secs
**CHIP to Fast**
MoveMem   = 4.0600 secs
MoveMem16 = 3.2800 secs
**CHIP to CHIP**
MoveMem   = 5.1200 secs
MoveMem16 = 5.1200 secs