PDA

View Full Version : [SOLVED] Way to speed up this newbie's simple recalc/copy/paste loop?



kofgadol
03-05-2017, 11:39 AM
Hi all,

I've written a very basic macro (the only kind I know how!) which simulates the rolling of a die for a specified number of trials, recording each result as a hard-coded value in a table. It works fine but it is very slow. 1000 trials takes around 1.5 minutes. Yet I've seen other simulations in Excel go several times faster. Is there anything I can do to speed this up?

The code is below.

Basically, the user types in the number of trials in a cell named "Number_of_trials" and then runs the macro. Then, for each trial,

-- the die roll is simulated with a formula which uses the RAND function. This formula is in a cell named "ResultsLive". For each trial, the worksheet recalculates so that RAND generates a new value here.

-- the macro goes to this "ResultsLive" cell, copies the result, then goes to the top of the results table -- which is a cell named "FirstResults" -- and pastes it as a hard-coded value.

It then repeats as neccessary, storing each new result below the previous one, until the loop ends.

#######################################################
##################Code begins###########################

''' For the specified number of times, generates a random decimal between 0 and 1
''' and records it in a list

Dim i As Integer

For i = 0 To Range("Number_of_trials").Value
Calculate

Application.Goto Reference:="ResultsLive"
Selection.Copy
Application.Goto Reference:="FirstResults"
ActiveCell.Offset(i, 0).Select
Selection.PasteSpecial Paste:=xlPasteValues

Next i

##################Code ends############################
#######################################################

My guess is that it's slow because the macro is literally "travelling" here and there many times as result of my use of "Goto Reference". Is there a quicker way to get each new, recalculated value to the proper destination cell? :think:

Logit
03-05-2017, 02:53 PM
I suspect because you are copying and pasting is slowing things down.

See if this is faster for you. Takes a different approach.



Option Explicit


Sub RollDice()
Randomize
Dim NextRow As Long
Range("Die1") = Int(Rnd() * 6) + 1
Range("Die2") = Int(Rnd() * 6) + 1
NextRow = Range("A65536").End(xlUp).Row + 1
If NextRow = Range("HistoryHeader").Row + 1 Then Cells(NextRow, 1) = 1 Else Cells(NextRow, 1) = Cells(NextRow - 1, 1) + 1
Cells(NextRow, 2) = Range("Die1")
Cells(NextRow, 3) = Range("Die2")
Cells(NextRow, 4) = Range("Total")
End Sub


Sub ClearHistory()
Range(Range("HistoryHeader").Offset(1, 0), Range("D65536")).Clear
End Sub

mancubus
03-05-2017, 02:57 PM
welcome to the forum.
please take time to read the forum rules.
wrap your code with code tags (see my signature)

you don't need to select objects in order to work with them.

try this:



For i = 0 To Range("Number_of_trials").Value
Calculate
Range("FirstResults").Offset(i).Value = Range("ResultsLive").Value
Next i

mancubus
03-05-2017, 02:59 PM
ooops
the thread had a reply whilst i was arranging the ranges in the OP's code.
:dunno

:rofl:

kofgadol
03-05-2017, 03:24 PM
Thank you very much mancubus, and sorry for the transgression

Paul_Hossler
03-05-2017, 04:37 PM
I think you were over complicating it

This runs fast and I added a #Dice option to stress test it

As it is it returns 1 - 6, but change the ' comment line and it returns a floating point between 0 and 1

It's pretty straight forward and online help is good, but ask questions if you have any



Option Explicit

Sub RollDice()
Dim iNumberRolls As Long, iNumberDice As Long, iRoll As Long, iDice As Long
Dim rResultsStart As Range
Dim aRolls() As Double

iNumberRolls = ActiveSheet.Range("A2").Value
iNumberDice = ActiveSheet.Range("A5").Value
Set rResultsStart = ActiveSheet.Range("B2")

Application.ScreenUpdating = False

For iDice = 1 To iNumberDice
Range(rResultsStart, rResultsStart.End(xlDown)).Clear

ReDim aRolls(1 To iNumberRolls)

Randomize
For iRoll = 1 To iNumberRolls
' aRolls(iroll) = Rnd ' 0<=1
aRolls(iRoll) = Int(6# * Rnd) + 1# ' 1-6
Next iRoll

rResultsStart.Resize(iNumberRolls, 1).Value = Application.WorksheetFunction.Transpose(aRolls)

Set rResultsStart = rResultsStart.Offset(0, 1)
Next iDice
Application.ScreenUpdating = True
End Sub

kofgadol
03-05-2017, 06:34 PM
welcome to the forum.
please take time to read the forum rules.
wrap your code with code tags (see my signature)

you don't need to select objects in order to work with them.

try this:



For i = 0 To Range("Number_of_trials").Value
Calculate
Range("FirstResults").Offset(i).Value = Range("ResultsLive").Value
Next i


You,sir,are a genius. That was exactly it. I'd wondered if there were a quicker way than literally copying and pasting all the time. You showed me that basically I could just link the source and destination with an equals sign. Before using your code, it took 86 seconds to run 1,000 trials; after, 15 seconds! Awesome. :hi:Thanks very much again. :bow::beerchug::kiss::super::super::yay:bow:

kofgadol
03-05-2017, 06:36 PM
How do I flag this as solved?

Paul_Hossler
03-05-2017, 07:11 PM
Above post #1 is [Thread Tools] and one option is to mark SOLVED

As an aside, the macro in #6 did 10,000 'rolls' in < 1sec

The data might not have been in the same output format

kofgadol
03-05-2017, 07:33 PM
Above post #1 is [Thread Tools] and one option is to mark SOLVED

As an aside, the macro in #6 did 10,000 'rolls' in < 1sec

The data might not have been in the same output format

Ok, I've flagged this as solved.


About the tip in #6 -- 10,000 in less than a second? Wow. I didn't try it as -- something I didn't mention -- I'm actually using this as a monte carlo simulator which can handle other types of probabilty functions (normal, lognormal, triangular etc) which in turn rely either on bundled Excel or custom user functions, which I would not know how to write in VBA; rather I'd just let the sheet recalc and then compile the results of each "live" distribution output cell. Is there a way I could apply your method to these functions? : pray2:If so, that would be ... life-changing. :p Especially as this example is trivial, compared to the much larger models I do in real life.

Paul_Hossler
03-05-2017, 08:06 PM
You might want to start another thread if you have different questions -- It makes it easier for people to see the new questions

1. If you're doing a lot of computation, it's faster to use arrays instead of putting and getting a lot of WS data

2. Planning ahead, you can use Excel's display capability to present the data

3. Here's something to think about. This generates normally distributed random data in an array. It puts the array on the worksheet for you to look at (B1:B***), but if all you might be interested in is the mean, you can compute that totally internally (A10 and A12)

4. Col D is a sort of the random Normal data in col B in order to make the graph

18557





Option Explicit
Sub GenerateNormal()
Dim iNumSamples As Long, i As Long
Dim dMean As Double, dSigma As Double
Dim rSamples As Range
Dim a() As Double

With Worksheets("Sheet2")
iNumSamples = .Range("A2").Value
dMean = .Range("A5").Value
dSigma = .Range("A7").Value
Set rSamples = .Range("B1")
ReDim a(1 To iNumSamples)

For i = 1 To iNumSamples
a(i) = Application.WorksheetFunction.NormInv(Rnd, dMean, dSigma)
Next I

.Range("A10").Value = Application.WorksheetFunction.Average(a)
.Range("A12").Value = Application.WorksheetFunction.StDev(a)

rSamples.Resize(iNumSamples, 1).Value = Application.WorksheetFunction.Transpose(a)

End With

End Sub