Inefficient Packing: Null Bytes - Analog Devices VisualDSP++ 3.5 Manual

LDF Commands

Inefficient Packing: Null Bytes

Given the target byte order requirements of packing in
page 3-35, it is much simpler to program memory access by directing the
linker to add unused (null) bytes to your executable file. This method sim-
plifies alignment and sequencing.
Consider an executable file that places bytes (in the following order) into
three 16-bit memory addresses (MSByte on the left).
B2, B1,
Suppose you want to load them as two 24-bit instructions.
B3, B2, B1,
Reordering them with
and programming overhead, as shown above. However, the addition of
null bytes after the third byte and the sixth byte simplifies things
considerably.
PACKING(6 B3 B2 B1 B0 B6 B5 B4 B0)
This order defines four 16-bit memory reads, which generate two 24-bit
addresses. Reads 1 and 3 are copied to the two MSBytes on the PMD.
Reads 2 and 4 are copied to the LSByte on the PMD. The lower byte is
ignored.
L
The same number of bytes are reordered as in the efficient packing
example. Hence, the byte count parameter (6) to the
command is the same.
The byte designated as
The linker sets that byte to zero in external memory.
3-36
B4, B3, B6, B5
B6, B5, B4
PACKING(6 B3 B2 B1 B6 B5 B4)
in the
B0 PACKING()
VisualDSP++ 3.5 Linker and Utilities Manual
Table 3-4 on
leaves alignment
PACKING()
syntax acts as a place holder.
for 16-Bit Processors