New EA SyTLAS EA / Multi-currencies MT5 EA

And now in version 1.7, at what level is the stop loss?

 

Afar255
SL are dynamics for each pair. They constantly evolve according to price action.

 

Omg... that awsome... how u can able get to gpu instruction (except u are using native language dll)? I think in mql5 programming library cant do it...

And what mean apex scanner? Im too stupid for ur high level keyword...

 

myg3nx
You can easily find tutos on this matter...

Code:

//+------------------------------------------------------------------+
//| FFT_GPU                                                          |
//+------------------------------------------------------------------+
bool FFT_GPU(int direction,int power,float &data_real[],float &data_imag[],ulong &time_gpu)
  {
//--- calculate the number of points
   int num_points=1;
   for(int i=0;i<power;i++)
      num_points*=2;
//--- prepare data array for GPU calculation
   float data[];
   ArrayResize(data,2*num_points);
   for(int i=0; i<num_points; i++)
     {
      data[2*i]=data_real[i];
      data[2*i+1]=data_imag[i];
     }

   COpenCL OpenCL;
   if(!OpenCL.Initialize(cl_program,true))
     {
      PrintFormat("Error in OpenCL initialization. Error code=%d",GetLastError());
      return(false);
     }
//--- create kernels
   OpenCL.SetKernelsCount(3);
   OpenCL.KernelCreate(0,kernel_init);
   OpenCL.KernelCreate(1,kernel_stage);
   OpenCL.KernelCreate(2,kernel_scale);
//--- create buffers
   OpenCL.SetBuffersCount(2);
   if(!OpenCL.BufferFromArray(0,data,0,2*num_points,CL_MEM_READ_ONLY))
     {
      PrintFormat("Error in BufferFromArray for input buffer. Error code=%d",GetLastError());
      return(false);
     }
   if(!OpenCL.BufferCreate(1,2*num_points*sizeof(float),CL_MEM_READ_WRITE))
     {
      PrintFormat("Error in BufferCreate for data buffer. Error code=%d",GetLastError());
      return(false);
     }
//--- determine maximum work-group size
   int local_size=(int)CLGetInfoInteger(OpenCL.GetKernel(0),CL_KERNEL_WORK_GROUP_SIZE);
//--- determine local memory size
   uint local_mem_size=(uint)CLGetInfoInteger(OpenCL.GetContext(),CL_DEVICE_LOCAL_MEM_SIZE);
   int points_per_group=(int)local_mem_size/(2*sizeof(float));
   if(points_per_group>num_points)
      points_per_group=num_points;
//--- set kernel arguments
   OpenCL.SetArgumentBuffer(0,0,0);
   OpenCL.SetArgumentBuffer(0,1,1);
   OpenCL.SetArgumentLocalMemory(0,2,local_mem_size);
   OpenCL.SetArgument(0,3,points_per_group);
   OpenCL.SetArgument(0,4,num_points);
   OpenCL.SetArgument(0,5,direction);
//--- OpenCL execute settings
   int task_dimension=1;
   uint global_size=(uint)((num_points/points_per_group)*local_size);
   uint global_work_offset[1]={0};
   uint global_work_size[1];
   global_work_size[0]=global_size;
   uint local_work_size[1];
   local_work_size[0]=local_size;
//--- GPU calculation start
   time_gpu=GetMicrosecondCount();
//-- execute kernel fft_init
   if(!OpenCL.Execute(0,task_dimension,global_work_offset,global_work_size,local_work_size))
     {
      PrintFormat("fft_init: Error in CLExecute. Error code=%d",GetLastError());
      return(false);
     }
//-- further stages of the FFT
   if(num_points>points_per_group)
     {
      //--- set arguments for kernel 1
      OpenCL.SetArgumentBuffer(1,0,1);
      OpenCL.SetArgument(1,2,points_per_group);
      OpenCL.SetArgument(1,3,direction);
      for(int stage=2; stage<=num_points/points_per_group; stage<<=1)
        {
         OpenCL.SetArgument(1,1,stage);
         //-- execute kernel fft_stage
         if(!OpenCL.Execute(1,task_dimension,global_work_offset,global_work_size,local_work_size))
           {
            PrintFormat("fft_stage: Error in CLExecute. Error code=%d",GetLastError());
            return(false);
           }
        }
     }
//--- scale values if performing the inverse FFT
   if(direction<0)
     {
      OpenCL.SetArgumentBuffer(2,0,1);
      OpenCL.SetArgument(2,1,points_per_group);
      OpenCL.SetArgument(2,2,num_points);
      //-- execute kernel fft_scale
      if(!OpenCL.Execute(2,task_dimension,global_work_offset,global_work_size,local_work_size))
        {
         PrintFormat("fft_scale: Error in CLExecute. Error code=%d",GetLastError());
         return(false);
        }
     }
//--- read the results from GPU memory
   if(!OpenCL.BufferRead(1,data,0,0,2*num_points))
     {
      PrintFormat("Error in BufferRead for data_buffer2. Error code=%d",GetLastError());
      return(false);
     }
//--- GPU calculation finished
   time_gpu=ulong((GetMicrosecondCount()-time_gpu));
//--- copy calculated data and release OpenCL handles
   for(int i=0; i<num_points; i++)
     {
      data_real[i]=data[2*i];
      data_imag[i]=data[2*i+1];
     }
   OpenCL.Shutdown();
//---
   return(true);
  }

APEX scanner is just some code to detect extremes to the upside or the downside...

 

ça va être incroyable ! sais tu quand sera disponible cette version ?

 

Why do u use gpu instead using cpu core to make it work? And im sure 98% vps in this world doesnt have cuda driver installed... and if exist, for renting is soooo expensive like hell...

 

That's why I've made this option switchable

I'm learning while giving to this community, nothing wrong with that...

 

As long as we keep learning life stays interesting . You are doing an amazing job and love the idea of using more CPU/GPU for improved pattern detection.

 

This is my review for this thread:
FULL STAR

 

Hello, the expert is free? Is there limitation? thank you.

 

ho_analista
Hello !
The only limitation of the version posted on this thread is the lots size...
Best regards

 

Symphoenix, can not a little more about the stop loss? And in absolute numbers?

 

Thanks Symphoenix, can I do backtesting?

 

Awesome work here. We hope to try this.

 

Hi Symphoenix

Correct me if I am wrong, it seems EA has more profits if running on account with bigger spread than with smaller spread

 

This is my review for this thread:
Promising EA!!! Thanks for the work Symphoenix

 

Afar255
There is a stealth stop loss that is created when the drawdown exceeds Emergency_Limit (in points) value. When this happens, the EA tracks price action in order to close a batch of orders to compensate.

ho_analista
No backtesting possible, it crashes MT5.

Tavamanya
Are you talking about the differences of results between a standard account and an ECN one ?

 

Yes Symphoenix

 

Looks good