openclmql5metatrader5

Using OpenCL in MQL5 - in a Custom Indicator code:


I am trying to modify an indicator file using the OpenCL library.

But I couldn't understand how I can make that possible. I tried to read and understand the Documentation but could not.

Here is my code:

#include <MovingAverages.mqh>

#property indicator_separate_window
#property indicator_buffers 6
#property indicator_plots   3
#property indicator_type1   DRAW_LINE
#property indicator_color1  Yellow//LightSeaGreen
#property indicator_style1  STYLE_SOLID
#property indicator_width1  1
#property indicator_type2   DRAW_LINE
#property indicator_color2  Lime//YellowGreen
#property indicator_style2  STYLE_SOLID//STYLE_DOT
#property indicator_width2  1
#property indicator_type3   DRAW_LINE
#property indicator_color3  Red//Wheat
#property indicator_style3  STYLE_SOLID//STYLE_DOT
#property indicator_width3  1
#property indicator_label1  "ADX"
#property indicator_label2  "+DI"
#property indicator_label3  "-DI"
//--- input parameters
input int InpPeriodADX=14; // Period
//---- buffers
double    ExtADXBuffer[];
double    ExtPDIBuffer[];
double    ExtNDIBuffer[];
double    ExtPDBuffer[];
double    ExtNDBuffer[];
double    ExtTmpBuffer[];
//--- global variables
int       ExtADXPeriod;

//+------------------------------------------------------------------+
//| Custom indicator initialization function                         |
//+------------------------------------------------------------------+
void OnInit()
  {
//--- check for input parameters
   if(InpPeriodADX>=100 || InpPeriodADX<=0)
     {
      ExtADXPeriod=14;
      printf("Incorrect value for input variable Period_ADX=%d. Indicator will use value=%d for calculations.",InpPeriodADX,ExtADXPeriod);
     }
   else ExtADXPeriod=InpPeriodADX;
//---- indicator buffers
   SetIndexBuffer(0,ExtADXBuffer);
   SetIndexBuffer(1,ExtPDIBuffer);
   SetIndexBuffer(2,ExtNDIBuffer);
   SetIndexBuffer(3,ExtPDBuffer,INDICATOR_CALCULATIONS);
   SetIndexBuffer(4,ExtNDBuffer,INDICATOR_CALCULATIONS);
   SetIndexBuffer(5,ExtTmpBuffer,INDICATOR_CALCULATIONS);
//--- indicator digits
   IndicatorSetInteger(INDICATOR_DIGITS,2);
//--- set draw begin
   PlotIndexSetInteger(0,PLOT_DRAW_BEGIN,ExtADXPeriod<<1);
   PlotIndexSetInteger(1,PLOT_DRAW_BEGIN,ExtADXPeriod);
   PlotIndexSetInteger(2,PLOT_DRAW_BEGIN,ExtADXPeriod);
//--- indicator short name
   string short_name="ADX("+string(ExtADXPeriod)+")";
   IndicatorSetString(INDICATOR_SHORTNAME,short_name);
//--- change 1-st index label
   PlotIndexSetString(0,PLOT_LABEL,short_name);
//---- end of initialization function
  }
//+------------------------------------------------------------------+
//| Custom indicator iteration function                              |
//+------------------------------------------------------------------+
int OnCalculate(const int rates_total,
                const int prev_calculated,
                const datetime &time[],
                const double &open[],
                const double &high[],
                const double &low[],
                const double &close[],
                const long &tick_volume[],
                const long &volume[],
                const int &spread[])
  {
//--- checking for bars count
   if(rates_total<ExtADXPeriod)
      return(0);
//--- detect start position
   int start;
   if(prev_calculated>1) start=prev_calculated-1;
   else
     {
      start=1;
      ExtPDIBuffer[0]=0.0;
      ExtNDIBuffer[0]=0.0;
      ExtADXBuffer[0]=0.0;
     }
//--- main cycle
   for(int i=start;i<rates_total && !IsStopped();i++)
     {
      //--- get some data
      double Hi    =high[i];
      double prevHi=high[i-1];
      double Lo    =low[i];
      double prevLo=low[i-1];
      double prevCl=close[i-1];
      //--- fill main positive and main negative buffers
      double dTmpP=Hi-prevHi;
      double dTmpN=prevLo-Lo;
      if(dTmpP<0.0)   dTmpP=0.0;
      if(dTmpN<0.0)   dTmpN=0.0;
      if(dTmpP>dTmpN) dTmpN=0.0;
      else
        {
         if(dTmpP<dTmpN) dTmpP=0.0;
         else
           {
            dTmpP=0.0;
            dTmpN=0.0;
           }
        }
      //--- define TR
      double tr=MathMax(MathMax(MathAbs(Hi-Lo),MathAbs(Hi-prevCl)),MathAbs(Lo-prevCl));
      //---
      if(tr!=0.0)
        {
         ExtPDBuffer[i]=100.0*dTmpP/tr;
         ExtNDBuffer[i]=100.0*dTmpN/tr;
        }
      else
        {
         ExtPDBuffer[i]=0.0;
         ExtNDBuffer[i]=0.0;
        }
      //--- fill smoothed positive and negative buffers
      ExtPDIBuffer[i]=ExponentialMA(i,ExtADXPeriod,ExtPDIBuffer[i-1],ExtPDBuffer);
      ExtNDIBuffer[i]=ExponentialMA(i,ExtADXPeriod,ExtNDIBuffer[i-1],ExtNDBuffer);
      //--- fill ADXTmp buffer
      double dTmp=ExtPDIBuffer[i]+ExtNDIBuffer[i];
      if(dTmp!=0.0)
         dTmp=100.0*MathAbs((ExtPDIBuffer[i]-ExtNDIBuffer[i])/dTmp);
      else
         dTmp=0.0;
      ExtTmpBuffer[i]=dTmp;
      //--- fill smoothed ADX buffer
      ExtADXBuffer[i]=ExponentialMA(i,ExtADXPeriod,ExtADXBuffer[i-1],ExtTmpBuffer);



     }

//---- OnCalculate done. Return new prev_calculated.

   return(rates_total);
  }
//+------------------------------------------------------------------+

Kindly, help me modify the above with OpenCL, so that it will help me understand what I can do with it.


Solution

  • A Principal Note:

    No, as was posted in many MQL4/5 posts already, [ CustomIndicator ] type-of-MQL4/5 code is no place for anything else but ultimately as synchronously responsive as possible external QUOTE-stream event processing, so no blocking, no GUI-MMI-interactions, no Print()-ing, nothing that may anyhow slow down, the less block a solo-thread ( SPOF ) of all [ CustomIndicator ] data-refresh / event synchronous-processing.

    OpenCL is not a tool for this ( if for no other reason, for it adds latency ).


    OpenCL opens gates to harness heterogeneous computing resources
    CPU + GPU + manyCore-device(s) and organises ComputingJOBs' workflows among them ...

    OpenCL programs are used for performing computations on video cards that support OpenCL 1.1 or higher. Modern video cards contain hundreds of small specialized processors that can simultaneously perform simple mathematical operations with incoming data streams. The OpenCL language organizes parallel computing and provides greater speed for a certain class of tasks.

    If going into OpenCL territory, best start with simple, non-sequential ( so most of the Technical Indicators, that convolve in TimeDOMAIN are out of the game anyways ) concurrent-computing, as asynchronous, distributed co-operation requires mastering multiple domains:


    Always first test the present <COMPUTING RESOURCES> available:

    void OnStart() {
    
         uchar cDATA[1024];
         uint  aSIZE;
         int   aDevCOUNT = CLGetInfoInteger( 0, CL_DEVICE_COUNT );
    
         for ( int aDEVICE  = 0;
                   aDEVICE <  aDevCOUNT;
                   aDEVICE++
                   )
         {
             int clCONTEXT =  CLContextCreate( aDEVICE );
             if (clCONTEXT == -1 ) Print( "ERROR in CLContextCreate() for aDEVICE#", aDEVICE );
             string aDeviceNAME;
             CLGetInfoString( clCONTEXT,
                              CL_DEVICE_NAME,
                              aDeviceNAME
                              );   Print( "INF[", aDEVICE, "].NAME::  ", aDeviceNAME );
             CLGetDeviceInfo( clCONTEXT,
                              CL_DEVICE_VENDOR,
                              cDATA,
                              aSIZE
                              );   Print( "INF[", aDEVICE, "].VENDOR:: ", CharArrayToString( cDATA ) );
         }
      }