source: soft/giet_vm/convol/main.c @ 352

////////////////////////////////////////////////////////////////////////////////////////////
// File   : main.c   (for convol application)
// Date   : june 2014
// author : Alain Greiner
//
// The "convol" application implements a 2D convolution product.  
////////////////////////////////////////////////////////////////////////////////////////////

#include "hard_config.h"
#include "stdio.h"
#include "barrier.h"

#define VERBOSE                    1

#define INITIAL_DISPLAY_ENABLE     1
#define FINAL_DISPLAY_ENABLE       1

#define NB_CLUSTERS_MAX            (X_SIZE * Y_SIZE)
#define PIXEL_SIZE                 2
#define NL                         1024
#define NP                         1024
#define NB_PIXELS                  (NP * NL)
#define FRAME_SIZE                 (NB_PIXELS * PIXEL_SIZE)

#define TA(c,l,p)  (A[c][((NP) * (l)) + (p)])
#define TB(c,p,l)  (B[c][((NL) * (p)) + (l)])
#define TC(c,l,p)  (C[c][((NP) * (l)) + (p)])
#define TD(c,l,p)  (D[c][((NP) * (l)) + (p)])
#define TZ(c,l,p)  (Z[c][((NP) * (l)) + (p)])

#define max(x,y) ((x) > (y) ? (x) : (y))
#define min(x,y) ((x) < (y) ? (x) : (y))

// instrumentation counters indexed by (cluster_id, lpid]
unsigned int START[NB_CLUSTERS_MAX][NB_PROCS_MAX];
unsigned int H_BEG[NB_CLUSTERS_MAX][NB_PROCS_MAX];
unsigned int H_END[NB_CLUSTERS_MAX][NB_PROCS_MAX];
unsigned int V_BEG[NB_CLUSTERS_MAX][NB_PROCS_MAX];
unsigned int V_END[NB_CLUSTERS_MAX][NB_PROCS_MAX];
unsigned int D_BEG[NB_CLUSTERS_MAX][NB_PROCS_MAX];
unsigned int D_END[NB_CLUSTERS_MAX][NB_PROCS_MAX];

// synchronization barriers
giet_barrier_t barrier_1;
giet_barrier_t barrier_2;
giet_barrier_t barrier_3;
giet_barrier_t barrier_4;

volatile unsigned int init_ok = 0;
volatile unsigned int exit_ok = 0;

//////////////////////////////////////////////////////
// Required by GCC when initializing an array 
//////////////////////////////////////////////////////
static void *memcpy( void *        _dst, 
                     const void *  _src, 
                     unsigned int  size )
{
    unsigned int         *dst = _dst;
    const unsigned int   *src = _src;

    if (! ((unsigned int)dst & 3) && ! ((unsigned int)src & 3))
    {
        while (size > 3)
        {
            *dst++ = *src++;
            size -= 4;
        }
    }

    unsigned char *cdst = (unsigned char*)dst;
    unsigned char *csrc = (unsigned char*)src;

    while ( size-- )
    {
        *cdst++ = *csrc++;
    }
    return _dst;
}

///////////////////////////////////////////
__attribute__ ((constructor)) void main()
///////////////////////////////////////////
{
    //////////////////////////////////
    // convolution kernel parameters
    // The content of this section is
    // Philips proprietary information.
    ///////////////////////////////////

    int   vnorm  = 115;
    int   vf[35] = { 1, 1, 2, 2, 2,
                     2, 3, 3, 3, 4,
                     4, 4, 4, 5, 5,
                     5, 5, 5, 5, 5,
                     5, 5, 4, 4, 4,
                     4, 3, 3, 3, 2,
                     2, 2, 2, 1, 1 };

    int hrange = 100;
    int hnorm  = 201;

    unsigned int date = 0;

    int c; // cluster index for loops
    int l; // line index for loops
    int p; // pixel index for loops
    int z; // vertical filter index for loops

    int          file;                                        // file descriptor
    unsigned int pid         = giet_procid();                 // processor id
    unsigned int nprocs      = NB_PROCS_MAX;                  // procs per cluster
    unsigned int nclusters   = NB_CLUSTERS_MAX;               // number of clusters
    unsigned int lpid        = pid % nprocs;                  // local task id
    unsigned int cluster_xy  = pid / nprocs;                  // cluster index
    unsigned int x           = cluster_xy >> Y_WIDTH;         // x coordinate
    unsigned int y           = cluster_xy & ((1<<Y_WIDTH)-1); // y coordinate
    unsigned int cluster_id  = (x * Y_SIZE) + y;              // continuous cluster index
    unsigned int task_id     = (cluster_id * nprocs) + lpid;  // continuous task index
    unsigned int ntasks      = nclusters * nprocs;            // number of tasks
    unsigned int npixels     = NB_PIXELS;                     // pixels per frame
    unsigned int frame_size  = FRAME_SIZE;                    // total size (bytes)
    unsigned int nblocks     = frame_size / 512;              // number of blocks per frame

    unsigned int lines_per_task     = NL / ntasks;            // lines per task
    unsigned int lines_per_cluster  = NL / nclusters;         // lines per cluster
    unsigned int pixels_per_task    = NP / ntasks;            // columns per task
    unsigned int pixels_per_cluster = NP / nclusters;         // columns per cluster

    int first, last;

    date = giet_proctime();
    giet_shr_printf( "\n[CONVOL] task[%d,%d,%d] starts at cycle %d\n",
                     x,y,lpid, date );
    START[cluster_id][lpid] = date;

     // parameters checking 
   
    if ((nprocs != 1) && (nprocs != 2) && (nprocs != 4))
        giet_exit( "[CONVOL ERROR] NB_PROCS_MAX must be 1, 2 or 4\n");

    if ((X_SIZE!=1) && (X_SIZE!=2) && (X_SIZE!=4) && (X_SIZE!=8))
        giet_exit( "[CONVOL ERROR] X_SIZE must be 1, 2, 4, 8\n");
        
    if ((Y_SIZE!=1) && (Y_SIZE!=2) && (Y_SIZE!=4) && (Y_SIZE!=8))
        giet_exit( "[CONVOL ERROR] X_SIZE must be 1, 2, 4, 8\n");

    if ( NL % nclusters != 0 )
        giet_exit( "[CONVOL ERROR] NB_CLUSTERS must be a divider of NL");

    if ( NP % nclusters != 0 )
        giet_exit( "[CONVOL ERROR] NB_CLUSTERS must be a divider of NP");

    ///////////////////////////////////////////////////////////////////
    // All tasks initialise in their private stack five
    // arrays of pointers on the shared, distributed buffers,
    // containing the image. These buffers are indexed by cluster_id,
    // as there one set of 5 buffers per cluster.
    ///////////////////////////////////////////////////////////////////

    unsigned short * A[NB_CLUSTERS_MAX];
    int *            B[NB_CLUSTERS_MAX];
    int *            C[NB_CLUSTERS_MAX];
    int *            D[NB_CLUSTERS_MAX];
    unsigned char *  Z[NB_CLUSTERS_MAX];

    // get heap vaddr in cluster[0,0]
    unsigned int base;
    giet_vobj_get_vbase( "convol", "conv_heap_0_0", &base );

    // size allocated to each cluster depends on the number of clusters
    unsigned int size = 0x01000000 / nclusters;

    // initialise pointers on distributed buffers
    for (c = 0; c < nclusters; c++)
    {
        unsigned int cluster_offset = base + (size * c);
        A[c] = (unsigned short *) (cluster_offset);
        B[c] = (int *)            (cluster_offset + (frame_size * 1 / nclusters)); 
        C[c] = (int *)            (cluster_offset + (frame_size * 3 / nclusters)); 
        D[c] = (int *)            (cluster_offset + (frame_size * 5 / nclusters)); 
        Z[c] = (unsigned char *)  (cluster_offset + (frame_size * 7 / nclusters));

        if ( pid == 0  && VERBOSE )
        {
            giet_shr_printf( "\n########################################\n"
                             "### A[%d] = %x\n"
                             "### B[%d] = %x\n"               
                             "### C[%d] = %x\n"               
                             "### D[%d] = %x\n"               
                             "### Z[%d] = %x\n"               
                             "########################################\n",
                             c,A[c],c,B[c],c,C[c],c,D[c],c,Z[c] );
        }
    }

    ///////////////////////////////////////////////////////////////////////////
    // task[0,0,0] makes barrier initialisation,
    // open the file containing image, and load it from disk to A[c] buffers 
    // (nblocks / nclusters loaded in each cluster).
    // Other tasks are waiting on the init_ok condition.
    //////////////////////////////////////////////////////////////////////////
    if ( pid == 0 )
    {
        giet_shr_printf("\n[CONVOL] task[0,0,0] starts barrier init at cycle %d\n" 
                        "- NB_CLUSTERS     = %d\n"
                        "- NB_LOCAL_PROCS  = %d\n" 
                        "- NB_TASKS        = %d\n" 
                        "- NB_PIXELS       = %x\n"           
                        "- FRAME_SIZE      = %x\n"              
                        "- NB_BLOCKS       = %x\n",
                        giet_proctime(), nclusters, nprocs, ntasks, 
                        npixels, frame_size, nblocks );

        //  barriers initialization
        barrier_init( &barrier_1, ntasks );
        barrier_init( &barrier_2, ntasks );
        barrier_init( &barrier_3, ntasks );
        barrier_init( &barrier_4, ntasks );

        giet_shr_printf( "\n[CONVOL] task[0,0,0] completes barrier init at cycle %d\n", 
                giet_proctime() );

        // open file
        file = giet_fat_open("misc/philips_image.raw", 0 );
        if ( file < 0 ) giet_exit( "[CONVOL ERROR] task[0,0,0] cannot open"
                                   " file misc/philips_image.raw" );
 
        giet_shr_printf( "\n[CONVOL] task[0,0,0] open file misc/philips_image.raw"
                         " at cycle %d\n", giet_proctime() );

        for ( c = 0 ; c < nclusters ; c++ )
        {
            giet_shr_printf( "\n[CONVOL] task[%d,%d,%d] starts load"
                             " for cluster %d at cycle %d\n",
                             x, y, lpid, c, giet_proctime() );
        
            giet_fat_read( file,
                           A[c],
                           nblocks/nclusters,
                           (nblocks/nclusters)*c );

            giet_shr_printf( "[CONVOL] task[%d,%d,%d] completes load"
                             " for cluster %d at cycle %d\n",
                             x, y, lpid, c, giet_proctime() );
        }
        init_ok = 1;
    }
    else
    {
        while ( init_ok == 0 );
    }

    /////////////////////////////////////////////////////////////////////////////
    // Optionnal parallel display of the initial image stored in A[c] buffers.
    // Eah task displays (NL/ntasks) lines. (one byte per pixel).
    /////////////////////////////////////////////////////////////////////////////

    if ( INITIAL_DISPLAY_ENABLE )
    {
        date = giet_proctime();
        giet_shr_printf( "\n[CONVOL] task[%d,%d,%d] starts initial display at cycle %d\n",
                         x, y, lpid, date);

        unsigned int line;
        unsigned int offset = lines_per_task * lpid;

        for ( l = 0 ; l < lines_per_task ; l++ )
        {
            line = offset + l;

            for ( p = 0 ; p < NP ; p++ )
            {
                TZ(cluster_id, line, p) = (unsigned char)(TA(cluster_id, line, p) >> 8);
            }

            giet_fb_sync_write( NP*(l + (task_id * lines_per_task) ), 
                                &TZ(cluster_id, line, 0), 
                                NP);
        }

        date = giet_proctime();
        giet_shr_printf( "\n[CONVOL] task[%d,%d,%d] completes initial display at cycle %d\n",
                         x, y, lpid, date);

        ///////////////////////////
        barrier_wait( &barrier_1 );

    }

    ////////////////////////////////////////////////////////
    // parallel horizontal filter : 
    // B <= transpose(FH(A))
    // D <= A - FH(A)
    // Each task computes (NL/ntasks) lines 
    // The image must be extended :
    // if (z<0)    TA(cluster_id,l,z) == TA(cluster_id,l,0)
    // if (z>NP-1) TA(cluster_id,l,z) == TA(cluster_id,l,NP-1)
    ////////////////////////////////////////////////////////

    date  = giet_proctime();
    giet_shr_printf( "\n[CONVOL] task[%d,%d,%d] starts horizontal filter at cycle %d\n",
            x, y, lpid, date );
    H_BEG[cluster_id][lpid] = date;

    // l = absolute line index / p = absolute pixel index  
    // first & last define which lines are handled by a given task

    first = task_id * lines_per_task;
    last  = first + lines_per_task;

    for (l = first; l < last; l++)
    {
        // src_c and src_l are the cluster index and the line index for A & D
        int src_c = l / lines_per_cluster;
        int src_l = l % lines_per_cluster;

        // We use the specific values of the horizontal ep-filter for optimisation:
        // sum(p) = sum(p-1) + TA[p+hrange] - TA[p-hrange-1]
        // To minimize the number of tests, the loop on pixels is split in three domains 

        int sum_p = (hrange + 2) * TA(src_c, src_l, 0);
        for (z = 1; z < hrange; z++)
        {
            sum_p = sum_p + TA(src_c, src_l, z);
        }

        // first domain : from 0 to hrange
        for (p = 0; p < hrange + 1; p++)
        {
            // dst_c and dst_p are the cluster index and the pixel index for B
            int dst_c = p / pixels_per_cluster;
            int dst_p = p % pixels_per_cluster;
            sum_p = sum_p + (int) TA(src_c, src_l, p + hrange) - (int) TA(src_c, src_l, 0);
            TB(dst_c, dst_p, l) = sum_p / hnorm;
            TD(src_c, src_l, p) = (int) TA(src_c, src_l, p) - sum_p / hnorm;
        }
        // second domain : from (hrange+1) to (NP-hrange-1)
        for (p = hrange + 1; p < NP - hrange; p++)
        {
            // dst_c and dst_p are the cluster index and the pixel index for B
            int dst_c = p / pixels_per_cluster;
            int dst_p = p % pixels_per_cluster;
            sum_p = sum_p + (int) TA(src_c, src_l, p + hrange) 
                          - (int) TA(src_c, src_l, p - hrange - 1);
            TB(dst_c, dst_p, l) = sum_p / hnorm;
            TD(src_c, src_l, p) = (int) TA(src_c, src_l, p) - sum_p / hnorm;
        }
        // third domain : from (NP-hrange) to (NP-1)
        for (p = NP - hrange; p < NP; p++)
        {
            // dst_c and dst_p are the cluster index and the pixel index for B
            int dst_c = p / pixels_per_cluster;
            int dst_p = p % pixels_per_cluster;
            sum_p = sum_p + (int) TA(src_c, src_l, NP - 1) 
                          - (int) TA(src_c, src_l, p - hrange - 1);
            TB(dst_c, dst_p, l) = sum_p / hnorm;
            TD(src_c, src_l, p) = (int) TA(src_c, src_l, p) - sum_p / hnorm;
        }

        if ( VERBOSE )
        {
            giet_shr_printf(" - line %d computed at cycle %d\n", l, giet_proctime() );
        }
    }

    date  = giet_proctime();
    giet_shr_printf( "\n[CONVOL] task[%d,%d,%d] completes horizontal filter at cycle %d\n",
            x, y, lpid, date );
    H_END[cluster_id][lpid] = date;

    ///////////////////////////
    barrier_wait( &barrier_2 );

    //////////////////////////////////////////////////////////
    // parallel vertical filter : 
    // C <= transpose(FV(B))
    // Each task computes (NP/ntasks) columns
    // The image must be extended :
    // if (l<0)    TB(cluster_id,p,l) == TB(cluster_id,p,0)
    // if (l>NL-1)   TB(cluster_id,p,l) == TB(cluster_id,p,NL-1)
    //////////////////////////////////////////////////////////

    date  = giet_proctime();
    giet_shr_printf( "\n[CONVOL] task[%d,%d,%d] starts vertical filter at cycle %d\n",
            x, y, lpid, date );
    V_BEG[cluster_id][lpid] = date;

    // l = absolute line index / p = absolute pixel index
    // first & last define which pixels are handled by a given task

    first = task_id * pixels_per_task;
    last  = first + pixels_per_task;

    for (p = first; p < last; p++)
    {
        // src_c and src_p are the cluster index and the pixel index for B
        int src_c = p / pixels_per_cluster;
        int src_p = p % pixels_per_cluster;

        int sum_l;

        // We use the specific values of the vertical ep-filter
        // To minimize the number of tests, the NL lines are split in three domains 

        // first domain : explicit computation for the first 18 values
        for (l = 0; l < 18; l++)
        {
            // dst_c and dst_l are the cluster index and the line index for C
            int dst_c = l / lines_per_cluster;
            int dst_l = l % lines_per_cluster;

            for (z = 0, sum_l = 0; z < 35; z++)
            {
                sum_l = sum_l + vf[z] * TB(src_c, src_p, max(l - 17 + z,0) );
            }
            TC(dst_c, dst_l, p) = sum_l / vnorm;
        }
        // second domain
        for (l = 18; l < NL - 17; l++)
        {
            // dst_c and dst_l are the cluster index and the line index for C
            int dst_c = l / lines_per_cluster;
            int dst_l = l % lines_per_cluster;

            sum_l = sum_l + TB(src_c, src_p, l + 4)
                  + TB(src_c, src_p, l + 8)
                  + TB(src_c, src_p, l + 11)
                  + TB(src_c, src_p, l + 15)
                  + TB(src_c, src_p, l + 17)
                  - TB(src_c, src_p, l - 5)
                  - TB(src_c, src_p, l - 9)
                  - TB(src_c, src_p, l - 12)
                  - TB(src_c, src_p, l - 16)
                  - TB(src_c, src_p, l - 18);

            TC(dst_c, dst_l, p) = sum_l / vnorm;
        }
        // third domain
        for (l = NL - 17; l < NL; l++)
        {
            // dst_c and dst_l are the cluster index and the line index for C
            int dst_c = l / lines_per_cluster;
            int dst_l = l % lines_per_cluster;

            sum_l = sum_l + TB(src_c, src_p, min(l + 4, NL - 1))
                  + TB(src_c, src_p, min(l + 8, NL - 1))
                  + TB(src_c, src_p, min(l + 11, NL - 1))
                  + TB(src_c, src_p, min(l + 15, NL - 1))
                  + TB(src_c, src_p, min(l + 17, NL - 1))
                  - TB(src_c, src_p, l - 5)
                  - TB(src_c, src_p, l - 9)
                  - TB(src_c, src_p, l - 12)
                  - TB(src_c, src_p, l - 16)
                  - TB(src_c, src_p, l - 18);

            TC(dst_c, dst_l, p) = sum_l / vnorm;
        }

        if ( VERBOSE )
        {
            giet_shr_printf(" - column %d computed at cycle %d\n", p, giet_proctime());
        }
    }

    date  = giet_proctime();
    giet_shr_printf( "\n[CONVOL] task[%d,%d,%d] completes vertical filter at cycle %d\n",
            x, y, lpid, date );
    V_END[cluster_id][lpid] = date;

    ///////////////////////////
    barrier_wait( &barrier_3 );

    ////////////////////////////////////////////////////////////////
    // Optional parallel display of the final image Z <= D + C
    // Eah task displays (NL/ntasks) lines. (one byte per pixel).
    ////////////////////////////////////////////////////////////////

    if ( FINAL_DISPLAY_ENABLE )
    {
        date  = giet_proctime();
        giet_shr_printf( "\n[CONVOL] task[%d,%d,%d] starts final display at cycle %d\n",
                         x, y, lpid, date);
        D_BEG[cluster_id][lpid] = date;

        unsigned int line;
        unsigned int offset = lines_per_task * lpid;

        for ( l = 0 ; l < lines_per_task ; l++ )
        {
            line = offset + l;

            for ( p = 0 ; p < NP ; p++ )
            {
                TZ(cluster_id, line, p) = 
                   (unsigned char)( (TD(cluster_id, line, p) + 
                                     TC(cluster_id, line, p) ) >> 8 );
            }

            giet_fb_sync_write( NP*(l + (task_id * lines_per_task) ), 
                                &TZ(cluster_id, line, 0), 
                                NP);
        }

        date  = giet_proctime();
        giet_shr_printf( "\n[CONVOL] task[%d,%d,%d] completes final display at cycle %d\n",
            x, y, lpid, date);
        D_END[cluster_id][lpid] = date;
     
        ///////////////////////////
        barrier_wait( &barrier_4 );
    }

    /////////////////////////////////////////////////////////
    // Task[0,0,0] makes the instrumentation 
    /////////////////////////////////////////////////////////

    if ( pid == 0 )
    {
        date  = giet_proctime();
        giet_shr_printf("\n*** Starting Instrumentation at cycle %d\n\n", date);

        int cc, pp;

        unsigned int min_start = 0xFFFFFFFF;
        unsigned int max_start = 0;

        unsigned int min_h_beg = 0xFFFFFFFF;
        unsigned int max_h_beg = 0;

        unsigned int min_h_end = 0xFFFFFFFF;
        unsigned int max_h_end = 0;

        unsigned int min_v_beg = 0xFFFFFFFF;
        unsigned int max_v_beg = 0;

        unsigned int min_v_end = 0xFFFFFFFF;
        unsigned int max_v_end = 0;

        unsigned int min_d_beg = 0xFFFFFFFF;
        unsigned int max_d_beg = 0;

        unsigned int min_d_end = 0xFFFFFFFF;
        unsigned int max_d_end = 0;

        for (cc = 0; cc < nclusters; cc++)
        {
            for (pp = 0; pp < nprocs; pp++ )
            {
                if (START[cc][pp] < min_start) min_start = START[cc][pp];
                if (START[cc][pp] > max_start) max_start = START[cc][pp];

                if (H_BEG[cc][pp] < min_h_beg) min_h_beg = H_BEG[cc][pp];
                if (H_BEG[cc][pp] > max_h_beg) max_h_beg = H_BEG[cc][pp];

                if (H_END[cc][pp] < min_h_end) min_h_end = H_END[cc][pp];
                if (H_END[cc][pp] > max_h_end) max_h_end = H_END[cc][pp];

                if (V_BEG[cc][pp] < min_v_beg) min_v_beg = V_BEG[cc][pp];
                if (V_BEG[cc][pp] > max_v_beg) max_v_beg = V_BEG[cc][pp];

                if (V_END[cc][pp] < min_v_end) min_v_end = V_END[cc][pp];
                if (V_END[cc][pp] > max_v_end) max_v_end = V_END[cc][pp];

                if (D_BEG[cc][pp] < min_d_beg) min_d_beg = D_BEG[cc][pp];
                if (D_BEG[cc][pp] > max_d_beg) max_d_beg = D_BEG[cc][pp];

                if (D_END[cc][pp] < min_d_end) min_d_end = D_END[cc][pp];
                if (D_END[cc][pp] > max_d_end) max_d_end = D_END[cc][pp];
            }
        }

        giet_shr_printf(" - START : min = %d / max = %d / med = %d / delta = %d\n",
               min_start, max_start, (min_start+max_start)/2, max_start-min_start);

        giet_shr_printf(" - H_BEG : min = %d / max = %d / med = %d / delta = %d\n",
               min_h_beg, max_h_beg, (min_h_beg+max_h_beg)/2, max_h_beg-min_h_beg);

        giet_shr_printf(" - H_END   : min = %d / max = %d / med = %d / delta = %d\n",
               min_h_end, max_h_end, (min_h_end+max_h_end)/2, max_h_end-min_h_end);

        giet_shr_printf(" - V_BEG : min = %d / max = %d / med = %d / delta = %d\n",
               min_v_beg, max_v_beg, (min_v_beg+max_v_beg)/2, max_v_beg-min_v_beg);

        giet_shr_printf(" - V_END   : min = %d / max = %d / med = %d / delta = %d\n",
               min_v_end, max_v_end, (min_v_end+max_v_end)/2, max_v_end-min_v_end);

        giet_shr_printf(" - D_BEG : min = %d / max = %d / med = %d / delta = %d\n",
               min_d_beg, max_d_beg, (min_d_beg+max_d_beg)/2, max_d_beg-min_d_beg);

        giet_shr_printf(" - D_END   : min = %d / max = %d / med = %d / delta = %d\n",
               min_d_end, max_d_end, (min_d_end+max_d_end)/2, max_d_end-min_d_end);

        giet_shr_printf( "\n General Scenario (Kcycles for each step)\n" );
        giet_shr_printf( " - BOOT OS           = %d\n", min_start );
        giet_shr_printf( " - LOAD IMAGE        = %d\n", min_h_beg - min_start);
        giet_shr_printf( " - H_FILTER          = %d\n", max_h_end - min_h_beg );
        giet_shr_printf( " - BARRIER HORI/VERT = %d\n", min_v_beg - max_h_end);
        giet_shr_printf( " - V_FILTER          = %d\n", max_v_end - min_v_beg );
        giet_shr_printf( " - BARRIER VERT/DISP = %d\n", min_d_beg - max_v_end);
        giet_shr_printf( " - DISPLAY           = %d\n", max_d_end - min_d_beg );

        exit_ok = 1;
    }
    else
    {
        while ( exit_ok == 0 );
    }

    giet_exit( "completed");

} // end main()

// Local Variables:
// tab-width: 3
// c-basic-offset: 3
// c-file-offsets:((innamespace . 0)(inline-open . 0))
// indent-tabs-mode: nil
// End:

// vim: filetype=cpp:expandtab:shiftwidth=3:tabstop=3:softtabstop=3