Multi-Threaded Split Command

[Semen Matusovski]

lance -
how greg uses threads is separate question.
when i wrote previous message, i didn't know dave's algorithm.
just now i found it.
like algorithm it's absolutelly ineffective
i guess, that dave wanted to demonstration pb functions only.

------------------

[Lance Edmonds]

Semen, Greg's code is using a thread in a very inefficient way... his preimary thread is locked in a loop waiting for the second thread to finish... in this instance it would be more efficient to place all the code in the main thread.

However, if Greg wishes to retain the multi-threaded approach for other reasons, he may do well to consider releasing time slices in his "wait for thread to finish" loop so as to cede processor time to the other thread... ie:

...
DO        
  THREAD STATUS MakeArray(0) TO ThreadResult    
  SLEEP 100  ' cede approx 100mSec to other threads & processes.
LOOP WHILE ThreadResult
...

------------------
Lance
PowerBASIC Support
mailto:[email protected][email protected]</A>

[Semen Matusovski]

Greg --
What do you name "slowly" ?
Sure, that on average PC it possible to convert tab-delimitted array 1-2 Mb to Asciiz array during 100-200ms maximum.
I don't know Dave's algorithm; I mean a method, posted above.

I cut a fragment from one utility (for internal usage; so, there is no deep control) and tested on my real array
(Rows = 8741; Columns = 9; SizeofFile = 535145 bytes)

Results (Win2000, P-III-550, fast SCSI HDD) - 20ms.

   #Compile Exe
   #Register None
   #Dim All
   #Include "Win32Api.Inc"

   %maxEl = 100000
   $FileNm = "nsiesr.txt"
      
   Function PbMain
      Dim t1 As Single, t2 As Single
      t1 = Timer
      
      Dim Info As String, SizeOfFile As Long
      Dim k1 As Byte Ptr, k2 As Byte Ptr, kk As Byte Ptr
      Dim nRow As Long, nCol As Long, nEl As Long
      Open $FileNm For Binary As #1
      SizeOfFile = Lof(1): Get$ #1, SizeOfFile, Info
      Close #1
  
      ReDim ArTmp(%maxEl) As Asciiz Ptr
      k1 = StrPtr(Info): kk = k1: k2 = k1 + SizeOfFile
      Do
         If k1 >= k2 Then Exit Do
         If @k1 = 9 Then
            ArTmp(nEl) = kk: kk = k1 + 1: @k1 = 0: Incr nEl
         ElseIf @k1 = 13 Then
            ArTmp(nEl) = kk: kk = k1 + 2: @k1 = 0: Incr nEl
            If nCol = 0 Then nCol = nEl
            Incr nRow: If (nRow * nCol) <> nEl Then Beep: Exit Function
         End If
         Incr k1
      Loop
      If (nRow * nCol) <> nEl Then Beep: Exit Function
  
      ReDim Ar(1 To nCol, 1 To nRow) As Asciiz Ptr At VarPtr(ArTmp(0))
      
      t2 = Timer
      MsgBox "nRow =" + Str$(nRow) + "  " + "  nCol =" + Str$(nCol) + _
         " SizeOfFile =" + Str$(SizeOfFile), , "Time = " + Format$(t2 -t1, "##.### ms")

      ' Usage @Ar(Column, Row) !!!
      MsgBox @Ar(2, 2)

   End Function

[This message has been edited by Semen Matusovski (edited July 12, 2000).]

[Gregery D Engle]

I found where the problem was when I used the command
THREAD STATUS it would take an extra 5 seconds or so to return a
value so instead I just check the ubound on the array and if
its > 0 then it exits the loop

Thanks anyways.

------------------
-Greg

[Gregery D Engle]

I am writing a dll function that is going to be performing
duty cycle tests on a square wave. I am using a program
called test point which is VERY slow for mathematical
operations so I am going to make this.

It first loads up a file called "c:\greg.dat" which is
tab delimited and CRLF after each row. There are
16 rows and 350 lines so it loads that up and what I'm trying
to do is populate an array matrix [Rows,Cols] the downfall is when
I use Dave's Split command, it seems that is takes very long
to get done with the split procedure. So at first I thought
maybe if I put it in a thread, well it still seems to take longer
then it should. What I am trying to accomplish is something
that takes only a second or so because after I put it in a [1,0]
array then I have to sort it out and put it in the [rows,cols]
array and then do all the math stuff. Any ideas? I was thinking
maybe if there was a may to actually thread the split command?

ah another idea I just thought of was splitting the buffer up in
multiple parts and then reassembling them and calling multiple
threads. I'll try and but any other ideas are welcome.

here is the code, to create the greg.dat file open up excel
and type random numbers in a column with 7 decimal places and
have it go down to 350 rows and then just copy over to fill
all 16 columns and then save as a TAB delimited file. Or
email me at [email protected] and I'll send you the file.

COMPILE EXE
GLOBAL StringArray() AS STRING
GLOBAL Taskbuffer AS STRING

FUNCTION GetRows(InBuffer AS STRING) AS LONG
END FUNCTION

SUB Split(BYVAL expression AS STRING, BYVAL delimeter AS STRING, StrArray() AS STRING)
  LOCAL c AS LONG
  LOCAL x AS LONG
  IF LEN(delimeter) = 0 THEN
    delimeter = " "
  END IF
  c = MAX(PARSECOUNT(expression, delimeter), 1)
  REDIM StrArray(0 TO c - 1) AS STRING
  FOR x = 1 TO c
    StrArray(x - 1) = PARSE$(expression, delimeter, x)
  NEXT
END SUB

FUNCTION MakeArray(InputId AS LONG) EXPORT AS LONG
   'MSGBOX sbuffer
   Split TaskBuffer, $TAB, StringArray()
'   MSGBOX STR$(UBOUND(StringArray()))
END FUNCTION


FUNCTION OutputDuty(InFilename AS STRING) AS LONG
    '//open InFilename and pump into buffer
DIM sbuffer AS STRING
    OPEN InFilename FOR BINARY AS #1
    sbuffer = STRING$(LOF(1), 0)
    GET 1,,sbuffer
    CLOSE #1
    '//Make temporary buffer for THREADED array function
    TaskBuffer = sbuffer
    '//Replace all the CRLF with tabs for easier importing
    REPLACE $CRLF WITH $TAB IN TaskBuffer
    '//Launch THREAD for [1,0] array
DIM ThreadID AS LONG
    THREAD CREATE MakeArray(0) TO ThreadID
    '//find first CRLF
DIM FirstCRLF AS LONG
    FirstCRLF = INSTR(1, sbuffer, $CRLF)
    '//put first line in buffer
DIM linebuffer AS STRING
    linebuffer = LEFT$(sbuffer, FirstCRLF)
    '//count how many fields are present and ADD + for CRLF
DIM NumOfCols AS LONG
    NumOfCols = TALLY(linebuffer, $TAB) + 1
    '//clear line buffer
    linebuffer = ""
    '//Get total number CRLF
DIM NumOfRows AS LONG
    NumOfRows = TALLY(sbuffer, $CRLF)
    '//Number of Records
DIM NumOfRecords AS LONG
    NumOfRecords = NumOfRows * NumOfCols
    '//Assemble a Matrix on [NumRows X NumCols]
DIM Matrix(NumofRows,NumOfCols) AS STRING
    '//wait for matrix to finish
DIM ThreadResult AS LONG
    DO
        THREAD STATUS MakeArray(0) TO ThreadResult
    LOOP WHILE ThreadResult


MSGBOX STR$(UBOUND(StringArray()))
END FUNCTION


FUNCTION PBMAIN() AS LONG
MSGBOX STR$(OutPutDuty("C:\greg.dat"))
'MSGBOX STR$(UBOUND(StringArray()))
END FUNCTION

------------------
-Greg

[This message has been edited by Gregery D Engle (edited July 12, 2000).]