Overview
Examples
Screenshots
Comparisons
Applications
Download
Documentation
Tutorials
Bazaar
Status & Roadmap
FAQ
Authors & License
Forums
Funding Ultimate++
Search on this site
Search in forums












SourceForge.net Logo
Home » Community » Coffee corner » CoWork usage question
CoWork usage question [message #51091] Tue, 22 January 2019 14:59 Go to next message
Tom1
Messages: 659
Registered: March 2007
Contributor
Hi,

As I need more than coffee to solve this out, I decided to put it here Smile

The question is: What am I doing wrong, as the following code does not work as I expect, i.e. multiply the items by two and show the results:
#include <Core/Core.h>

using namespace Upp;
#define ICOUNT 15
class A{
public:
	Buffer<int> ib;
	
	A(){
		ib.Alloc(ICOUNT,0);
		for(int i=0;i<ICOUNT;i++) ib[i]=i+1;
	}
	
	void operation(int &x){
		x*=2;
	}

	void operationp(int *x){
		*x*=2;
	}
	
	void Run(){
		Cout() << "ib = ";
		for(int i=0;i<ICOUNT;i++) Cout() << ib[i] << ", ";

		CoWork co;
		for(int i=0;i<ICOUNT;i++){
			//co & [&] { operation(ib[i]); }; // Reference variant
			co & [&] { operationp(&ib[i]); }; // Pointer variant
		}
		co.Finish();

		Sleep(400);
		Cout() << "\n\nib*2 = ";
		for(int i=0;i<ICOUNT;i++) Cout() << ib[i] << ", ";
		Cout() << "\n\n";
	}
};


CONSOLE_APP_MAIN
{
	A ac;
	ac.Run();
}


Mostly the results are 'exciting' to say the least.

Best regards,

Tom
Re: CoWork usage question [message #51092 is a reply to message #51091] Tue, 22 January 2019 15:15 Go to previous messageGo to next message
Tom1
Messages: 659
Registered: March 2007
Contributor
Well, the second cup of coffee solved it for me:
		CoWork co;
		for(int i=0;i<ICOUNT;i++){
			//co & [&, i] { operation(ib[i]); }; // Reference variant
			co & [&, i] { operationp(&ib[i]); }; // Pointer variant
		}
		co.Finish();


Must have 'copy capture' for i, or otherwise the index i will run away before used.

Sorry for bothering with such trivial issue.

Best regards,

Tom
Re: CoWork usage question [message #51098 is a reply to message #51092] Wed, 23 January 2019 15:32 Go to previous messageGo to next message
mirek is currently offline  mirek
Messages: 12116
Registered: November 2005
Ultimate Member
Tom1 wrote on Tue, 22 January 2019 15:15
Well, the second cup of coffee solved it for me:
		CoWork co;
		for(int i=0;i<ICOUNT;i++){
			//co & [&, i] { operation(ib[i]); }; // Reference variant
			co & [&, i] { operationp(&ib[i]); }; // Pointer variant
		}
		co.Finish();


Must have 'copy capture' for i, or otherwise the index i will run away before used.

Sorry for bothering with such trivial issue.

Best regards,

Tom


Yeah, got caught there too...

BTW, have you noticed the looper variant?

CoWork co;
co & [&] {
   for(;;) {
      int i = co.GetNext();
      if(i >= ICOUNT) break;
      operationp(&ib[i]);
   }
}


It is faster and more resilient w.r.t. those issues.

Mirek
Re: CoWork usage question [message #51101 is a reply to message #51098] Thu, 24 January 2019 09:35 Go to previous message
Tom1
Messages: 659
Registered: March 2007
Contributor
Hi Mirek,

Well, that was interesting. I did some benchmarking and the looper is indeed a clear winner when there are a lot of small jobs to handle.

I further simplified the syntax for a "i = 0 .. n-1" -type of loop with a simple macro:

#define PARALLEL_LOOP(_ICOUNT_,_OPERATION_) { CoWork _co_; _co_ * [&] { for(int i = _co_.Next(); i < _ICOUNT_ ; i = _co_.Next()) { _OPERATION_; } };}


So, I can run the loop with:

PARALLEL_LOOP(n, operation(ib[i]));


Thanks and best regards,

Tom
Previous Topic: Does anyone have experience with dual booting on a notebook with recovery partition?
Next Topic: The eternal quarrel between Python and C
Goto Forum:
  


Current Time: Fri Dec 06 05:05:06 CET 2019

Total time taken to generate the page: 0.02429 seconds