COPYFRAME strategy is defined at compile time

[marcomontevechi1]

I believe it could be defined at runtime with a PV with no considerable performance costs or complexity added to the code.

For now what we have is:

#define MIN_DELAY       1e-5
#define USE_COPYFRAME   1    // 0: call dcambuf_lockframe to access image, 1: call dcambuf_copyframe to access image
#define OUTPUT_IMAGE    0    // 0: not output, 1: output the accessed image by raw data

(...)

#if USE_COPYFRAME
    // set user buffer information and copied ROI
    bufframe.buf        = buf;
    bufframe.rowbytes   = rowbytes;
    bufframe.left       = ox;
    bufframe.top        = oy;
    bufframe.width      = cx;
    bufframe.height     = cy;
    
    // access image
    err = dcambuf_copyframe( hdcam, &bufframe );
    if ( checkErrAndPrint(__FUNCTION__, err, "dcambuf_copyframe()") )
        return FALSE;
#else
    // access image
    err = dcambuf_lockframe( hdcam, &bufframe );
    if ( checkErrAndPrint(__FUNCTION__, err, "dcambuf_lockframe()") )
        return FALSE;

    if( bufframe.type != DCAM_PIXELTYPE_MONO16 )
    {
        printf( "not implement pixel type\n" );
        return FALSE;
    }

    // copy target ROI
    int32 copyrowbytes = cx * 2;
    char* pSrc = (char*)bufframe.buf + oy * bufframe.rowbytes + ox * 2;
    char* pDst = (char*)buf;

    int y;
    for( y = 0; y < cy; y++ )
    {
        memcpy_s( pDst, rowbytes, pSrc, copyrowbytes );

        pSrc += bufframe.rowbytes;
        pDst += rowbytes;
    }
#endif

[MarkRivers]

What are the pros and cons of using copyframe?

[marcomontevechi1]

Hm, after reading the API documentation I see no clear advantage or disadvantage other than just the buffer management mechanism. The documentation does mention that using dcambuf_attach() might be less performatic because of data types conversions, but it's a different thing.

I assume that dcambuf_copyframe is less performatic than dcambuf_lockframe but I would need to perform a few tests to be sure.

I created this ticket mostly just because I was doing some code cleanup and found this and thought it would be good to remember. Maybe this shouldn't be here, not sure.