progress 4gl - How to handle numbers of prowin32.exe -

i've been doing os-command execute -p "procedure.p" -param inside loop.

my question how can handle number of prowin32.exe not spawn won't consume processor , not less make fast?

note: version 8.3c

sample code

define variable filter character no-undo.  each item no-lock:     assign filter = "item.item = " + item.item no-error.     os-command no-wait     value("c:\dlcs\83c\bin\prowin32 " +           "-p c:\nanox\syte_server5\atms-configured-icons\live\nanox_utility\procedure.p " +           "-pf c:\nanox\syte_server5\atms-configured-icons\live\nanox_utility\conf\pilot.pf " +           "-param " + filter). end. quit. 

i bit differently. decide ahead of time how many threads , divvy work so:

/* worker.p  *  * i.e.:  *   mbpro dbname -p worker.p -param "0,5"  *   mbpro dbname -p worker.p -param "1,5"  *   mbpro dbname -p worker.p -param "2,5"  *   mbpro dbname -p worker.p -param "3,5"  *   mbpro dbname -p worker.p -param "4,5"  *  * in sample threadnum starts @ 0 -- end @ numthreads - 1  *  */   define variable threadnum  integer no-undo. define variable numthreads integer no-undo.  assign   threadnum  = integer( entry( 1, session:parameter ))   numthreads = integer( entry( 2, session:parameter )) .  each item no-lock (( item.item modulo numthreads ) = threadnum ):    /* whatever */  end.  return. 

for starters suggest numthreads number of cores available in server.

also, headless batch processes shouldn't using prowin32.exe. should use _progres.exe.

8.3 unspeakably ancient, obsolete , unsupported. system running on single core pentium windows 3.11? that's relevant because 8.3 "workgroup edition" not behave if multi-core system. if "enterprise" won't bad. you'd advised upgrade current release (11.3 of writing) if performance @ important you.