On Fri, 27 Nov 2009, David Arturo Macias Corona wrote:

Hi,

> Here is the summary
> If you want detailed results for each case let me know
> Harbour 13035
> OpenWatcom 1.8
> Below are results using:
> ------------------------
>    hbmk2 -m -n -w -es2 -l -kmo -gc3 speedtst.prg
>       speedtst.exe
>    hbmk2 -m -n -w -es2 -l -kmo -gc3 -mt speedtst.prg
>       speedtst.exe
>       speedtst.exe --thread
>       speedtst.exe --thread=2 --scale
> ------------------------
> and in same order
> Changes of switchs in os2\watcom.mk and hbmk2.prg
> We can see immediatly:
> - speedtst complete all test
> - Are slow
> As conclusion things are changed
> Przemek, what should we do ?
[...]
> with -DHB_FM_DL_ALLOC, -s removed
> 
> [ total application time: ]....................................56.55
> [ total real time: ]...........................................56.33
> 
> [ total application time: ]....................................70.03
> [ total real time: ]...........................................69.96
> 
> [ total application time: ].....................................0.06
> [ total real time: ]...........................................69.75
> 
> [ total application time: ].....................................0.06
> [ total real time: ]..........................................277.91

Thank you very much. So now we know that we can remove -s as workaround.
Unfortunately it causes very big speed overhead. Much bigger then in
Linux where removing -s reduce the speed only ~24 percent. It's 363%
what is even bigger then in Windows (295% in my test).

Now I would like to check if -s is necessary as global switch or it
can be used locally only for GT2OS2 code.
Please restore -s switch (just like in clean SVN code) and then
add to src/rtl/gtos2/Makefile[5] this line:

   HB_BUILD_OPTIM := no

It should disable optimization switches.
Then check if speedtst works in MT mode.

best regards,
Przemek
_______________________________________________
Harbour mailing list (attachment size limit: 40KB)
Harbour@harbour-project.org
http://lists.harbour-project.org/mailman/listinfo/harbour

Reply via email to