Re: int vs binary questions

[Hans Boldt <boldt@xxxxxxxxxx>]

cozzi@xxxxxxxxx wrote:

>    Hans,
>    That's all well and good, but...
>    1) Why is DDS limited to 2 and 4 byte binaries and seems to use the
>    RPG-like declaration for them (or has that changed?)
>    2) Why would the timing of int->int copy be slower than Bin->Bin in RPG IV
>    as was reported here earlier today?
>     


1) DDS is NOT limited to 2 and 4 byte binaries. DDS also supports 8 byte 
binaries.

2) Why did Steve report faster timings on bin->bin copy compared to 
int->int copy?  First, on a simple "EVAL B1=B2;" copy from binary to 
binary where both variables have the same size, no conversion happens. 
In W-Code, we do just a LOD and STR. For an assignment "EVAL I1=I2;", it 
actually is a little bit more complicated. If I1 and I2 are defined as 
20I0, there's a LOD and STR which runs at around the same speed as the 
binary. However, if the source argument is not 20I0, the compiler emits 
a "convert to 8-byte integer" instruction first, and that tends to slow 
things down a little bit. The performance difference disappears with 
optimization, however.

Cheers! Hans