efulsang@ADS.COM (Ejner Fulsang) (04/25/91)
I've been following the suggestions for efficient compiling and
linking of C++ programs using Borland C++ while running under
Windows 3.0. The consensus on the net seemed to be to load
tkernel first and then run Windows in the standard mode as follows:
tkernel hi=yes kilos=1024
win /s
I tried this and obtained the following results:
1) Windows operation seemed to hesitate and stumble around
a bit but otherwise did not seem too slow;
2) Ordinary C++ programs, e.g., the circle example from the
Borland manuals seemed to complie fairly efficiently (~13 sec);
3) Windows applications (Petzold's hellowin program) really
bogged down (5:24 under Windows vs. 0:18 under DOS).
Is there something else I should be doing or, were the above results
as good as I can expect?
BTW - my machine is a Compaq 386/20e with 5 MB of RAM.
Ejner Fulsang
Sr. Systems Engineer
Advanced Decision Systems
1500 Plymouth Street
Mountain View, CA 94043
efulsang@ads.comsidney@borland.com (Sidney Markowitz) (04/26/91)
In article <{G0&GB_@ads.com> efulsang@ADS.COM (Ejner Fulsang) writes: > >I've been following the suggestions for efficient compiling and >linking of C++ programs using Borland C++ while running under >Windows 3.0. The consensus on the net seemed to be to load >tkernel first and then run Windows in the standard mode as follows: > > tkernel hi=yes kilos=1024 > win /s One more step, which is actually the whole reason for doing the above: run bcx instead of bc. My results building whello.prj from the BC++ examples directory on a Compaq Portable 386/20 with lots of memory: 1) using bc.exe under DOS: ~16 seconds 2) using bc.exe with tkernel loaded under win/s and no pif (which is a worst case - bc.exe doesn't make use of tkernel, which just takes memory away from it, and without a pif bc has no EMS memory available for fast swapping): ~5 minutes 3) using bcx.exe with tkernel and win/s as above, no pif: ~12 seconds -- sidney markowitz <sidney@borland.com>