void atags_detect(uint32_t *atags, struct atag_info_t *info) {
/* each tag has at least 32-bit size and then 32-bit value */
/* some tags have multiple values */
uint32_t size,total_ram;
// int tag_value;
uint32_t *tags=atags;
char *cmdline;
/* clear out the info array */
memset(info,0,sizeof(struct atag_info_t));
while(1) {
size=tags[0];
switch (tags[1]) {
/* Start List */
case ATAG_CORE:
tags += size;
break;
/* Physical Memory */
case ATAG_MEM:
total_ram=tags[2];
if (tags[3]!=0) {
printk("Warning! We do not handle memory not starting at zero!\n");
}
info->ramsize=total_ram;
tags += size;
break;
/* VGA Text Display */
case ATAG_VIDEOTEXT:
tags += size;
break;
/* RAMDISK Use */
case ATAG_RAMDISK:
tags += size;
break;
/* INITRD Ramdisk */
case ATAG_INITRD2:
tags += size;
break;
/* 64-bit serial number */
case ATAG_SERIAL:
tags += size;
break;
/* Board Revision */
case ATAG_REVISION:
tags += size;
break;
/* VESA Framebuffer Info */
case ATAG_VIDEOLFB:
tags += size;
break;
case ATAG_CMDLINE:
cmdline = (char *)(&tags[2]);
info->framebuffer_x=parse_cmdline_int(cmdline,"fbwidth");
info->framebuffer_y=parse_cmdline_int(cmdline,"fbheight");
info->revision=parse_cmdline_int(cmdline,"rev");
/* http://elinux.org/RPi_HardwareHistory */
switch(info->revision) {
case 0x2:
case 0x3:
case 0x4:
case 0x5:
case 0x6:
case 0xd:
case 0xe:
case 0xf: info->hardware_type=RPI_MODEL_B;
break;
case 0x7:
case 0x8:
case 0x9: info->hardware_type=RPI_MODEL_A;
break;
case 0x10: info->hardware_type=RPI_MODEL_BPLUS;
break;
case 0x11: info->hardware_type=RPI_COMPUTE_NODE;
break;
case 0x12: info->hardware_type=RPI_MODEL_APLUS;
break;
default: info->hardware_type=RPI_UNKNOWN;
break;
}
tags += size;
break;
/* Empty tag to end list */
case ATAG_NONE:
return;
break;
default:
printk("ERROR! Unknown atag\n");
break;
}
}
}
int main( int argc, char** argv )
{
int ni;
long N;
pvo_cookie_t cookie;
pvo_float3_t* pts;
int* cia;
int* cja;
uint8_t* types;
pvo_vtu_file_t fh;
double* U;
float* V;
int* ranks;
double t0, t1, t2, t3;
long i, nnodes, ncells;
int N_u, N_v, N_r;
double r_min, r_max, gamma;
MPI_Init( &argc, &argv );
pvo_init( MPI_COMM_WORLD );
MPI_Barrier( MPI_COMM_WORLD );
t0 = MPI_Wtime();
// Minimal and maximal value for the radius
r_min = parse_cmdline_double();
r_max = parse_cmdline_double();
gamma = parse_cmdline_double();
// Number of grid cells for each direction
N_u = parse_cmdline_double();
N_v = parse_cmdline_double();
N_r = parse_cmdline_double();
/* The data is decomposed by slicing in the radial
direction
*/
decompose( &N_r, &r_min, &r_max );
nnodes = (N_u+1)*(N_v+1)*(N_r+1);
ncells = N_u * N_v * N_r ;
pts = malloc( nnodes*sizeof(pvo_float3_t) );
cia = malloc( (ncells+1)*sizeof(int) );
cja = malloc( 8*ncells*sizeof(int) );
U = malloc( nnodes*3*sizeof(double) );
V = malloc( ncells*sizeof(float) );
types = malloc( ncells*sizeof(uint8_t) );
ranks = malloc( ncells*sizeof(int) );
ni = parse_cmdline_int();
if( 0 == pvo_world_rank() )
{
printf( " +-----------------------------------------+\n" );
printf( " | PVO CHUNK VTU BENCHMARK |\n" );
printf( " +-----------------------------------------+\n" );
printf( "\n");
printf( " PVO_DEFAULT_LOW_IO_LAYER : \"%s\"\n", str_low_io_layer[PVO_DEFAULT_LOW_IO_LAYER] );
printf( " no of cores : %d\n", pvo_world_size() );
printf( " no of nodes : %d\n", nnodes );
printf( " no of cells : %d\n", ncells );
printf( " no of islands : %d\n", ni );
}
create_chunk( N_u, N_v, N_r, r_min, r_max, gamma, nnodes, pts, ncells, cia, cja, types, U, V );
for( i = 0; i < ncells; ++i )
ranks[i] = pvo_world_rank();
if( -1 == pvo_cookie_create( pvo_world_rank()%ni, &cookie ))
MPI_Abort( MPI_COMM_WORLD, __LINE__ );
if( -1 == pvo_cookie_insert_var( cookie, PVO_VAR_NODEDATA, PVO_VAR_FLOAT64, 3, "U", U ))
MPI_Abort( MPI_COMM_WORLD, __LINE__ );
if( -1 == pvo_cookie_insert_var( cookie, PVO_VAR_CELLDATA, PVO_VAR_FLOAT32, 1, "V", V ))
MPI_Abort( MPI_COMM_WORLD, __LINE__ );
if( -1 == pvo_cookie_insert_var( cookie, PVO_VAR_CELLDATA, PVO_VAR_INT32, 1, "ranks", ranks ))
MPI_Abort( MPI_COMM_WORLD, __LINE__ );
if( -1 == pvo_vtu_file_open( "chunk", cookie, nnodes, pts, ncells, cia, cja, types, &fh ))
MPI_Abort( MPI_COMM_WORLD, __LINE__ );
MPI_Barrier( MPI_COMM_WORLD );
t1 = MPI_Wtime();
pvo_file_write( (pvo_file_t )fh );
MPI_Barrier( MPI_COMM_WORLD );
t2 = MPI_Wtime();
if( 0 == pvo_world_rank() )
printf( " time [sec] : %f\n", t2-t1 );
N = (3*4/*pts*/ + 3*8/*U*/)*nnodes + (4/*cia*/ + 8*4/*cja*/ + 1/*types*/ + 4/*V*/ + 4/*ranks*/)*ncells;
MPI_Allreduce( MPI_IN_PLACE, &N, 1, MPI_INT, MPI_SUM, MPI_COMM_WORLD );
if( 0 == pvo_world_rank() )
printf( " bandwidth [MB/sec] : %f\n", (1.0*N)/(1048576*(t2-t1)) );
if( -1 == pvo_vtu_file_close( fh ))
MPI_Abort( MPI_COMM_WORLD, __LINE__ );
if( -1 == pvo_cookie_delete( cookie ))
MPI_Abort( MPI_COMM_WORLD, __LINE__ );
MPI_Barrier( MPI_COMM_WORLD );
t3 = MPI_Wtime();
if( 0 == pvo_world_rank() )
{
printf( " total execution time [sec] : %f\n", t3-t0 );
printf( "\n" );
}
pvo_quit();
return MPI_Finalize();
}
