With moderate optimizations (just the -ox switch), and with debugging 
info left out, OW generates an executable of 190K for my modeler, while 
Dev-Cpp generates a 217K executable.

The only real speed bump was the lack of single-precision functions in 
the math library that ships with OW; but a bunch of macros to rename the 
double precision versions fixed that.  Then I took the time to comment 
out all of the local variables that weren't referenced.

It took me a while to figure out that when the linker says "files.obj: 
undefined void near Model::validate()", it means that the function 
Model::validate() was not found in any of the .obj files (which means 
that I forgot to add two sources to the project when I copied the files 
over).

Once everything compiled and linked, the resulting executable ran just 
fine, which is not surprising, since the same code resulted in an 
executable that ran fine with Dev-Cpp.  Dunno how fast it all runs, 
because the app isn't particularly demanding on the CPU, and I took out 
the O(N^2) situation that was dragging performance on earlier versions 
of the app.

Now if only the UI for OpenWatcom didn't look like something cobbled 
together for Win 3.x.  Dev-Cpp is much prettier...

Regards,
John

Post a reply to this message

> With moderate optimizations (just the -ox switch), and with debugging 
> info left out, OW generates an executable of 190K for my modeler, while 
> Dev-Cpp generates a 217K executable.

Try the -s switch on gcc (Dev-C++ shows it on the UI as "strip all 
symbols").

Post a reply to this message

Nicolas Alvarez wrote:

> 
>> With moderate optimizations (just the -ox switch), and with debugging 
>> info left out, OW generates an executable of 190K for my modeler, 
>> while Dev-Cpp generates a 217K executable.
> 
> 
> Try the -s switch on gcc (Dev-C++ shows it on the UI as "strip all 
> symbols").

I already had that option going in Dev-C++).  Without that option the 
executable is about 319K.

Regards,
John

Post a reply to this message