void generate_MCMC_proposition(MCMC_info *MCMC, int flag_chain_init) {
// Set the parameters for first trial parameters for the proposition
switch(flag_chain_init) {
case GBP_TRUE:
memcpy(MCMC->P_new, MCMC->P_init, (size_t)MCMC->n_P * sizeof(double));
memcpy(MCMC->P_last, MCMC->P_init, (size_t)MCMC->n_P * sizeof(double));
memcpy(MCMC->P_chain, MCMC->P_init, (size_t)MCMC->n_P * sizeof(double));
MCMC->flag_init_chain = GBP_FALSE;
MCMC->n_success = 0;
MCMC->n_fail = 0;
MCMC->n_propositions = 0;
break;
case GBP_FALSE:
generate_MCMC_parameters(MCMC);
break;
}
// Keep generating new parameter sets until the mapping function is satisfied
while(MCMC->map_P_to_M(MCMC->P_new, MCMC, MCMC->M_new)) {
MCMC->n_map_calls++;
MCMC->first_map_call = GBP_FALSE;
generate_MCMC_parameters(MCMC);
}
MCMC->first_map_call = GBP_FALSE;
MCMC->n_map_calls++;
// Produce likelihood for this proposition
MCMC->compute_MCMC_ln_likelihood(
MCMC, MCMC->M_new, MCMC->P_new, MCMC->ln_likelihood_DS, MCMC->n_DoF_DS, &(MCMC->ln_likelihood_new), &(MCMC->n_DoF));
if(!SID_CHECK_BITFIELD_SWITCH(MCMC->mode, MCMC_MODE_PARALLEL)) {
SID_Bcast(MCMC->ln_likelihood_DS, MCMC->n_DS, SID_DOUBLE, SID_MASTER_RANK, MCMC->comm);
SID_Bcast(MCMC->n_DoF_DS, MCMC->n_DS, SID_INT, SID_MASTER_RANK, MCMC->comm);
SID_Bcast(&(MCMC->ln_likelihood_new), 1, SID_DOUBLE, SID_MASTER_RANK, MCMC->comm);
SID_Bcast(&(MCMC->n_DoF), 1, SID_INT, SID_MASTER_RANK, MCMC->comm);
}
MCMC->first_likelihood_call = GBP_FALSE;
if(flag_chain_init) {
MCMC->ln_likelihood_chain = MCMC->ln_likelihood_new;
MCMC->ln_likelihood_best = MCMC->ln_likelihood_new;
memcpy(MCMC->P_best, MCMC->P_new, (size_t)MCMC->n_P * sizeof(double));
MCMC->ln_Pr_chain = 0.;
MCMC->ln_Pr_new = 0.;
}
// Keep track of the best proposition
else if(MCMC->ln_likelihood_new > MCMC->ln_likelihood_best) {
MCMC->ln_likelihood_best = MCMC->ln_likelihood_new;
memcpy(MCMC->P_best, MCMC->P_new, (size_t)MCMC->n_P * sizeof(double));
}
}
void init_field(int n_d, int *n, double *L, field_info *FFT) {
ptrdiff_t n_x_local;
ptrdiff_t i_x_start_local;
ptrdiff_t n_y_transpose_local;
ptrdiff_t i_y_start_transpose_local;
ptrdiff_t *n_x_rank;
int flag_active;
int n_active;
int min_size, max_size;
SID_log("Initializing ", SID_LOG_OPEN);
for(ptrdiff_t i_d = 0; i_d < n_d; i_d++) {
if(i_d < (n_d - 1))
SID_log("%dx", SID_LOG_CONTINUE, n[i_d]);
else
SID_log("%d element %d-d FFT ", SID_LOG_CONTINUE, n[i_d], n_d);
}
SID_log("(%d byte precision)...", SID_LOG_CONTINUE, (int)sizeof(GBPREAL));
// Initialize FFT sizes
FFT->n_d = n_d;
FFT->n = (ptrdiff_t *)SID_calloc(sizeof(ptrdiff_t) * FFT->n_d);
FFT->L = (double *)SID_calloc(sizeof(double) * FFT->n_d);
FFT->n_k_local = (ptrdiff_t *)SID_calloc(sizeof(ptrdiff_t) * FFT->n_d);
FFT->n_R_local = (ptrdiff_t *)SID_calloc(sizeof(ptrdiff_t) * FFT->n_d);
FFT->i_R_start_local = (ptrdiff_t *)SID_calloc(sizeof(ptrdiff_t) * FFT->n_d);
FFT->i_k_start_local = (ptrdiff_t *)SID_calloc(sizeof(ptrdiff_t) * FFT->n_d);
FFT->i_R_stop_local = (ptrdiff_t *)SID_calloc(sizeof(ptrdiff_t) * FFT->n_d);
FFT->i_k_stop_local = (ptrdiff_t *)SID_calloc(sizeof(ptrdiff_t) * FFT->n_d);
for(ptrdiff_t i_d = 0; i_d < FFT->n_d; i_d++) {
FFT->n[i_d] = n[i_d];
FFT->L[i_d] = L[i_d];
FFT->i_R_start_local[i_d] = 0;
FFT->i_k_start_local[i_d] = 0;
FFT->n_R_local[i_d] = FFT->n[i_d];
FFT->n_k_local[i_d] = FFT->n[i_d];
}
FFT->n_k_local[FFT->n_d - 1] = FFT->n[FFT->n_d - 1] / 2 + 1;
// Initialize FFTW
// Create an integer version of FFT->n[] to pass to ..._create_plan
int *n_int=(int *)SID_malloc(sizeof(int)*FFT->n_d);
for(int i_d=0;i_d<FFT->n_d;i_d++)
n_int[i_d]=(int)FFT->n[i_d];
#if FFTW_V2
#if USE_MPI
int total_local_size_int;
int n_x_local_int;
int i_x_start_local_int;
int n_y_transpose_local_int;
int i_y_start_transpose_local_int;
FFT->plan = rfftwnd_mpi_create_plan(SID.COMM_WORLD->comm, FFT->n_d, n_int, FFTW_REAL_TO_COMPLEX, FFTW_ESTIMATE);
FFT->iplan = rfftwnd_mpi_create_plan(SID.COMM_WORLD->comm, FFT->n_d, n_int, FFTW_COMPLEX_TO_REAL, FFTW_ESTIMATE);
rfftwnd_mpi_local_sizes(FFT->plan,
&(n_x_local_int),
&(i_x_start_local_int),
&(n_y_transpose_local_int),
&(i_y_start_transpose_local_int),
&total_local_size_int);
n_x_local = (ptrdiff_t)n_x_local_int;
i_x_start_local = (ptrdiff_t)i_x_start_local_int;
n_y_transpose_local = (ptrdiff_t)n_y_transpose_local_int;
i_y_start_transpose_local = (ptrdiff_t)i_y_start_transpose_local_int;
FFT->total_local_size = (size_t)total_local_size_int;
#else
FFT->total_local_size = 1;
for(ptrdiff_t i_d = 0; i_d < FFT->n_d; i_d++) {
if(i_d < FFT->n_d - 1)
FFT->total_local_size *= FFT->n[i_d];
else
FFT->total_local_size *= 2 * (FFT->n[i_d] / 2 + 1);
}
#if USE_DOUBLE
FFT->plan = fftwnd_create_plan(FFT->n_d, n_int, FFTW_REAL_TO_COMPLEX, FFTW_ESTIMATE | FFTW_IN_PLACE);
FFT->iplan = fftwnd_create_plan(FFT->n_d, n_int, FFTW_COMPLEX_TO_REAL, FFTW_ESTIMATE | FFTW_IN_PLACE);
#else
FFT->plan = rfftwnd_create_plan(FFT->n_d, n_int, FFTW_REAL_TO_COMPLEX, FFTW_ESTIMATE | FFTW_IN_PLACE);
FFT->iplan = rfftwnd_create_plan(FFT->n_d, n_int, FFTW_COMPLEX_TO_REAL, FFTW_ESTIMATE | FFTW_IN_PLACE);
#endif
#endif
#else
#if USE_MPI
#if USE_DOUBLE
fftw_mpi_init();
FFT->total_local_size = fftw_mpi_local_size_many_transposed(FFT->n_d,
FFT->n,
1,
FFTW_MPI_DEFAULT_BLOCK,
FFTW_MPI_DEFAULT_BLOCK,
SID_COMM_WORLD->comm,
&(n_x_local),
&(i_x_start_local),
&(n_y_transpose_local),
&(i_y_start_transpose_local));
FFT->plan = fftw_mpi_plan_dft_r2c(FFT->n_d, FFT->n, FFT->field_local, FFT->cfield_local, SID_COMM_WORLD->comm, FFTW_ESTIMATE);
FFT->iplan = fftw_mpi_plan_dft_c2r(FFT->n_d, FFT->n, FFT->cfield_local, FFT->field_local, SID_COMM_WORLD->comm, FFTW_ESTIMATE);
#else
fftwf_mpi_init();
FFT->total_local_size = fftwf_mpi_local_size_many_transposed(FFT->n_d,
FFT->n,
1,
FFTW_MPI_DEFAULT_BLOCK,
FFTW_MPI_DEFAULT_BLOCK,
SID_COMM_WORLD->comm,
&(n_x_local),
&(i_x_start_local),
&(n_y_transpose_local),
&(i_y_start_transpose_local));
FFT->plan = fftwf_mpi_plan_dft_r2c(FFT->n_d, FFT->n, FFT->field_local, FFT->cfield_local, SID_COMM_WORLD->comm, FFTW_ESTIMATE);
FFT->iplan = fftwf_mpi_plan_dft_c2r(FFT->n_d, FFT->n, FFT->cfield_local, FFT->field_local, SID_COMM_WORLD->comm, FFTW_ESTIMATE);
#endif
#else
FFT->total_local_size = 1;
for(ptrdiff_t i_d=0; i_d < FFT->n_d; i_d++) {
if(i_d < FFT->n_d - 1)
FFT->total_local_size *= FFT->n[i_d];
else
FFT->total_local_size *= 2 * (FFT->n[i_d] / 2 + 1);
}
#if USE_DOUBLE
FFT->plan = fftw_plan_dft_r2c(FFT->n_d, FFT->n, FFT->field_local, FFT->cfield_local, FFTW_ESTIMATE);
FFT->iplan = fftw_plan_dft_c2r(FFT->n_d, FFT->n, FFT->cfield_local, FFT->field_local, FFTW_ESTIMATE);
#else
FFT->plan = fftwf_plan_dft_r2c(FFT->n_d, FFT->n, FFT->field_local, FFT->cfield_local, FFTW_ESTIMATE);
FFT->iplan = fftwf_plan_dft_c2r(FFT->n_d, FFT->n, FFT->cfield_local, FFT->field_local, FFTW_ESTIMATE);
#endif
#endif
#endif
SID_free(SID_FARG n_int);
// Set empty slabs to start at 0 to make ignoring them simple.
if(n_x_local == 0)
i_x_start_local = 0;
if(n_y_transpose_local == 0)
i_y_start_transpose_local = 0;
// Modify the local slab dimensions according to what FFTW chose.
FFT->i_R_start_local[0] = i_x_start_local;
FFT->n_R_local[0] = n_x_local;
if(FFT->n_d > 1) {
FFT->i_k_start_local[1] = i_y_start_transpose_local;
FFT->n_k_local[1] = n_y_transpose_local;
}
// Allocate field
#if USE_FFTW3
FFT->field_local = (gbpFFT_real *)fftwf_alloc_real(FFT->total_local_size);
#else
FFT->field_local = (gbpFFT_real *)SID_malloc(sizeof(gbpFFT_real)*FFT->total_local_size);
#endif
FFT->cfield_local = (gbpFFT_complex *)FFT->field_local;
// Upper limits of slab decomposition
for(ptrdiff_t i_d = 0; i_d < FFT->n_d; i_d++) {
FFT->i_R_stop_local[i_d] = FFT->i_R_start_local[i_d] + FFT->n_R_local[i_d] - 1;
FFT->i_k_stop_local[i_d] = FFT->i_k_start_local[i_d] + FFT->n_k_local[i_d] - 1;
}
// FFTW padding sizes
if(FFT->n_d > 1) {
FFT->pad_size_R = 2 * (FFT->n_R_local[FFT->n_d - 1] / 2 + 1) - FFT->n_R_local[FFT->n_d - 1];
FFT->pad_size_k = 0;
} else {
FFT->pad_size_R = 0;
FFT->pad_size_k = 0;
}
// Number of elements (global and local) in the FFT
ptrdiff_t i_d = 0;
for(FFT->n_field = 1, FFT->n_field_R_local = 1, FFT->n_field_k_local = 1; i_d < FFT->n_d; i_d++) {
FFT->n_field *= (size_t)FFT->n[i_d];
FFT->n_field_R_local *= (size_t)FFT->n_R_local[i_d];
FFT->n_field_k_local *= (size_t)FFT->n_k_local[i_d];
}
// Clear the field
clear_field(FFT);
// Initialize the FFT's real-space grid
FFT->R_field = (double **)SID_malloc(sizeof(double *) * FFT->n_d);
FFT->dR = (double *)SID_malloc(sizeof(double *) * FFT->n_d);
for(ptrdiff_t i_d = 0; i_d < FFT->n_d; i_d++) {
FFT->R_field[i_d] = (double *)SID_malloc(sizeof(double) * (FFT->n[i_d] + 1));
FFT->dR[i_d] = FFT->L[i_d] / (double)(FFT->n[i_d]);
for(ptrdiff_t i_i = 0; i_i < FFT->n[i_d]; i_i++)
FFT->R_field[i_d][i_i] = FFT->L[i_d] * ((double)i_i / (double)(FFT->n[i_d]));
FFT->R_field[i_d][FFT->n[i_d]] = FFT->L[i_d];
}
// Initialize the FFT's k-space grid
FFT->k_field = (double **)SID_malloc(sizeof(double *) * FFT->n_d);
FFT->dk = (double *)SID_malloc(sizeof(double *) * FFT->n_d);
FFT->k_Nyquist = (double *)SID_malloc(sizeof(double *) * FFT->n_d);
for(ptrdiff_t i_d = 0; i_d < FFT->n_d; i_d++) {
FFT->k_field[i_d] = (double *)SID_malloc(sizeof(double) * FFT->n[i_d]);
FFT->dk[i_d] = TWO_PI / FFT->L[i_d];
FFT->k_Nyquist[i_d] = TWO_PI * (double)(FFT->n[i_d]) / FFT->L[i_d] / 2.;
for(ptrdiff_t i_i = 0; i_i < FFT->n[i_d]; i_i++) {
if(i_i >= FFT->n[i_d] / 2)
FFT->k_field[i_d][i_i] = TWO_PI * (double)(i_i - FFT->n[i_d]) / FFT->L[i_d];
else
FFT->k_field[i_d][i_i] = TWO_PI * (double)(i_i) / FFT->L[i_d];
}
}
// Flags
FFT->flag_padded = GBP_FALSE;
// Slab info
FFT->slab.n_x_local = FFT->n_R_local[0];
FFT->slab.i_x_start_local = FFT->i_R_start_local[0];
FFT->slab.i_x_stop_local = FFT->i_R_stop_local[0];
FFT->slab.x_min_local = FFT->R_field[0][FFT->i_R_start_local[0]];
if(FFT->slab.n_x_local > 0)
FFT->slab.x_max_local = FFT->R_field[0][FFT->i_R_stop_local[0] + 1];
else
FFT->slab.x_max_local = FFT->slab.x_min_local;
SID_Allreduce(&(FFT->slab.x_max_local), &(FFT->slab.x_max), 1, SID_DOUBLE, SID_MAX, SID_COMM_WORLD);
#if USE_MPI
// All ranks are not necessarily assigned any slices, so
// we need to figure out what ranks are to the right and the left for
// buffer exchanges
n_x_rank = (ptrdiff_t *)SID_malloc(sizeof(ptrdiff_t) * SID.n_proc);
n_x_rank[SID.My_rank] = (ptrdiff_t)FFT->slab.n_x_local;
if(n_x_rank[SID.My_rank] > 0)
flag_active = GBP_TRUE;
else
flag_active = GBP_FALSE;
SID_Allreduce(&flag_active, &n_active, 1, SID_INT, SID_SUM, SID_COMM_WORLD);
SID_Allreduce(&n_x_rank[SID.My_rank], &min_size, 1, SID_INT, SID_MIN, SID_COMM_WORLD);
SID_Allreduce(&n_x_rank[SID.My_rank], &max_size, 1, SID_INT, SID_MAX, SID_COMM_WORLD);
for(int i_rank = 0; i_rank < SID.n_proc; i_rank++)
SID_Bcast(&(n_x_rank[i_rank]), 1, SID_INT, i_rank, SID_COMM_WORLD);
FFT->slab.rank_to_right = -1;
for(int i_rank = SID.My_rank + 1; i_rank < SID.My_rank + SID.n_proc && FFT->slab.rank_to_right < 0; i_rank++) {
int j_rank = i_rank % SID.n_proc;
if(n_x_rank[j_rank] > 0)
FFT->slab.rank_to_right = j_rank;
}
if(FFT->slab.rank_to_right < 0)
FFT->slab.rank_to_right = SID.My_rank;
FFT->slab.rank_to_left = -1;
for(int i_rank = SID.My_rank - 1; i_rank > SID.My_rank - SID.n_proc && FFT->slab.rank_to_left < 0; i_rank--) {
int j_rank = i_rank;
if(i_rank < 0)
j_rank = i_rank + SID.n_proc;
if(n_x_rank[j_rank] > 0)
FFT->slab.rank_to_left = j_rank;
}
if(FFT->slab.rank_to_left < 0)
FFT->slab.rank_to_left = SID.My_rank;
free(n_x_rank);
SID_log("(%d cores unused, min/max slab size=%d/%d)...", SID_LOG_CONTINUE, SID.n_proc - n_active, min_size, max_size);
#else
FFT->slab.rank_to_right = SID.My_rank;
FFT->slab.rank_to_left = SID.My_rank;
if(FFT->slab.n_x_local > 0) {
flag_active = GBP_TRUE;
n_active = 1;
min_size = FFT->slab.n_x_local;
max_size = FFT->slab.n_x_local;
} else {
flag_active = GBP_FALSE;
n_active = 0;
min_size = 0;
max_size = 0;
}
#endif
SID_log("Done.", SID_LOG_CLOSE);
}
int fopen_multifile(const char *filename_root,
size_t data_size,
fp_multifile_info *fp_out,
...){
va_list vargs;
va_start(vargs,fp_out);
int r_val =TRUE;
int flag_filefound=FALSE;
// Sort out what file format we're working with
fp_out->fp_multifile=NULL;
if(SID.I_am_Master){
int i_file;
char filename_multifile[MAX_FILENAME_LENGTH];
// Set some filename information
vsprintf(fp_out->filename_root,filename_root,vargs);
strcpy(fp_out->filename_base,fp_out->filename_root);
strip_path(fp_out->filename_base);
// Try reading a multifile first ...
sprintf(filename_multifile,"%s/%s.%d",fp_out->filename_root,fp_out->filename_base,0);
fp_out->fp_multifile=fopen(filename_multifile,"r");
if(fp_out->fp_multifile==NULL){
sprintf(filename_multifile,"%s",fp_out->filename_root);
// ... if we didn't find a multi-file, try reading a single file ...
fp_out->fp_multifile=fopen(filename_multifile,"r");
if(fp_out->fp_multifile==NULL){
r_val=FALSE;
// Don't report error here so that we can use this routine to check if datasets exist without termination
//SID_trap_error("Could not open multifile {%s}",ERROR_LOGIC,filename_root);
}
// ... we found a single file. Set flags.
else
fp_out->flag_multifile=FALSE;
}
// ... we found a multi-file. Set flags.
else
fp_out->flag_multifile=TRUE;
// Load/set header information
if(fp_out->fp_multifile!=NULL){
fread_verify(&(fp_out->i_file), sizeof(int),1,fp_out->fp_multifile);
fread_verify(&(fp_out->n_files), sizeof(int),1,fp_out->fp_multifile);
fread_verify(&(fp_out->n_items_file), sizeof(int),1,fp_out->fp_multifile);
fread_verify(&(fp_out->n_items_total),sizeof(int),1,fp_out->fp_multifile);
fclose(fp_out->fp_multifile);
fp_out->fp_multifile=NULL;
}
}
SID_Bcast(fp_out,sizeof(fp_multifile_info),MASTER_RANK,SID.COMM_WORLD);
// Set the data size
fp_out->data_size=data_size;
if(r_val){
// Initialize things by opening the first file. Even if we want a item
// that's deep in the list, we have to scan all the headers (starting
// with the first) to find where it is.
fp_out->fp_multifile=NULL;
fp_out->i_file =0;
fp_out->i_item =0;
if(!(r_val=fopen_multifile_nth_file(fp_out,0)))
SID_trap_error("Error opening multifile file.",ERROR_IO_OPEN);
}
va_end(vargs);
return(r_val);
}
int read_matches_header(char *filename_root_in,
int i_read_start,
int i_read_stop,
int i_read_step,
int * n_files_return,
int **n_subgroups_return,
int **n_groups_return,
int * n_subgroups_max,
int * n_groups_max,
int * n_halos_max) {
char filename_out[256];
char filename_out_dir[256];
char filename_out_name[256];
FILE * fp_test;
FILE * fp_out;
SID_fp fp_in;
int i_read, k_read, l_read;
int i_read_start_file;
int i_read_stop_file;
int i_read_step_file;
int k_match;
int n_k_match;
char group_text_prefix[5];
int n_matches;
int j_read;
char filename_cat1[256];
char filename_cat2[256];
char filename_cat1_order[256];
char filename_cat2_order[256];
char filename_out_dir_snap[256];
int n_groups_1;
int n_groups_1_local;
int n_groups_2;
int n_groups_2_local;
int i_group;
int buffered_count;
int buffered_count_local;
int j_group;
int index_test;
int i_rank;
int * n_particles;
int * n_sub_group;
int * match_id = NULL;
float * match_score = NULL;
char cat_name_1[20];
char cat_name_2[20];
size_t * match_rank = NULL;
size_t * match_index = NULL;
size_t offset;
plist_info plist1;
plist_info plist2;
void * buffer;
int * buffer_int;
size_t * buffer_size_t;
float * buffer_float;
int n_buffer_max = 1000;
int n_buffer;
int i_buffer;
int j_buffer;
int flag_sucessful_completion = GBP_TRUE;
SID_log("Reading header information...", SID_LOG_OPEN);
// Count the number of snapshots we are going to use and
// initialize the arrays that are to be returned
for(i_read = i_read_stop, (*n_files_return) = 0; i_read >= i_read_start; i_read -= i_read_step)
(*n_files_return)++;
(*n_subgroups_return) = (int *)SID_malloc(sizeof(int) * (*n_files_return));
(*n_groups_return) = (int *)SID_malloc(sizeof(int) * (*n_files_return));
if(SID.I_am_Master) {
FILE *fp_read_header;
// Loop for subgroups and then groups
for(k_match = 0; k_match < 2; k_match++) {
switch(k_match) {
case 0:
sprintf(group_text_prefix, "sub");
break;
case 1:
sprintf(group_text_prefix, "");
break;
}
// Open file and read header
int i_read_start_in;
int i_read_stop_in;
int n_search_total_in;
int n_files_in;
int i_read_in;
sprintf(filename_out, "%s/%sgroup_matches_header.dat", filename_root_in, group_text_prefix);
if((fp_read_header = fopen(filename_out, "r")) == NULL)
SID_exit_error("Could not open file {%s} when reading header information.", SID_ERROR_IO_OPEN,
filename_out);
SID_fread_verify(&i_read_start_in, sizeof(int), 1, fp_read_header);
SID_fread_verify(&i_read_stop_in, sizeof(int), 1, fp_read_header);
SID_fread_verify(&n_search_total_in, sizeof(int), 1, fp_read_header);
SID_fread_verify(&n_files_in, sizeof(int), 1, fp_read_header);
i_read_in = i_read_stop_in + 1;
// Loop for each snapshot we want to keep
int i_read_next;
for(j_read = 0, i_read_next = i_read_stop; j_read < (*n_files_return); j_read++, i_read_next -= i_read_step) {
// Read-forward to the desired snapshot
int flag_continue = GBP_TRUE;
while(flag_continue && i_read_in > i_read_start_in) {
SID_fread_verify(&i_read_in, sizeof(int), 1, fp_read_header);
SID_fread_verify(&n_groups_1, sizeof(int), 1, fp_read_header);
fseek(fp_read_header, n_groups_1 * sizeof(int), SEEK_CUR); // Skip halo sizes
if(k_match == 1)
fseek(fp_read_header, n_groups_1 * sizeof(int), SEEK_CUR); // Skip n_sub_per_group
if(i_read_in == i_read_next) {
switch(k_match) {
case 0:
(*n_subgroups_return)[j_read] = n_groups_1;
break;
case 1:
(*n_groups_return)[j_read] = n_groups_1;
break;
}
flag_continue = GBP_FALSE;
}
}
}
fclose(fp_read_header);
if(j_read != (*n_files_return))
SID_exit_error("Was not able to read the appriate number of group/subgroup sizes (i.e. %d!=%d)",
SID_ERROR_LOGIC, j_read, (*n_files_return));
}
}
SID_Bcast((*n_subgroups_return), (*n_files_return), SID_INT, SID_MASTER_RANK, SID_COMM_WORLD);
SID_Bcast((*n_groups_return), (*n_files_return), SID_INT, SID_MASTER_RANK, SID_COMM_WORLD);
// Compute some maxs (useful for array allocation)
calc_max((*n_subgroups_return), n_subgroups_max, (*n_files_return), SID_INT, CALC_MODE_DEFAULT);
calc_max((*n_groups_return), n_groups_max, (*n_files_return), SID_INT, CALC_MODE_DEFAULT);
(*n_halos_max) = GBP_MAX((*n_subgroups_max), (*n_groups_max));
SID_log("Done.", SID_LOG_CLOSE);
return (flag_sucessful_completion);
}
void SID_exit(int status){
int i_args=0;
int n_days;
int n_hrs;
int n_mins;
int n_secs;
size_t max_RAM;
size_t SID_max_RAM_local;
int i_rank;
char spacer[10];
char time_string[48];
// Deal with I/O channels
fflush(SID.fp_log);
SID_Barrier(SID.COMM_WORLD);
if(SID.fp_in!=stdin && SID.fp_in!=NULL)
fclose(SID.fp_in);
// Clean-up argument stuff
if(SID.args!=NULL){
while((SID_arg *)(SID.args[i_args])!=NULL){
if(SID.arg_alloc[i_args])
free(((SID_arg *)(SID.args[i_args]))->val);
i_args++;
}
if(i_args>0){
free(SID.arg_alloc);
free(SID.arg_set);
}
}
// Clean-up any memory still allocated
// Report execution statistics
if(status!=ERROR_SYNTAX){
if(SID.I_am_Master){
fprintf(SID.fp_log,"\n");
fprintf(SID.fp_log,"Run statistics:\n");
fprintf(SID.fp_log,"--------------\n");
(void)time(&(SID.time_stop));
n_secs =(int)(SID.time_stop-SID.time_start);
seconds2ascii(n_secs,time_string);
fprintf(SID.fp_log,"Time elapsed=%s.\n",time_string);
}
// Report memory usage
if(SID.I_am_Master){
fprintf(SID.fp_log,"\nMemory usage:\n");
fprintf(SID.fp_log,"------------\n");
}
#if USE_MPI
SID_Barrier(SID.COMM_WORLD);
SID_Allreduce(&(SID.max_RAM_local),&max_RAM,1,SID_SIZE_T,SID_SUM,SID.COMM_WORLD);
if(SID.n_proc>1){
for(i_rank=0;i_rank<SID.n_proc;i_rank++){
SID_max_RAM_local=SID.max_RAM_local;
if(i_rank!=MASTER_RANK)
SID_Bcast(&SID_max_RAM_local,sizeof(size_t),i_rank,SID.COMM_WORLD);
if(SID.I_am_Master){
if(SID.n_proc>1000){
if((float)SID.max_RAM_local/(float)SIZE_OF_MEGABYTE>1.)
fprintf(SID.fp_log,"Peak for rank %4d=%4.2lf Gb\n",i_rank,(float)SID_max_RAM_local/(float)SIZE_OF_GIGIBYTE);
else if((float)SID.max_RAM_local/(float)SIZE_OF_KILOBYTE>1.)
fprintf(SID.fp_log,"Peak for rank %4d=%4.2lf Mb\n",i_rank,(float)SID_max_RAM_local/(float)SIZE_OF_MEGABYTE);
else
fprintf(SID.fp_log,"Peak for rank %4d=%4.2lf kb\n",i_rank,(float)SID_max_RAM_local/(float)SIZE_OF_KILOBYTE);
strcpy(spacer," ");
}
else if(SID.n_proc>100){
if((float)SID.max_RAM_local/(float)SIZE_OF_MEGABYTE>1.)
fprintf(SID.fp_log,"Peak for rank %3d=%4.2lf Gb\n",i_rank,(float)SID_max_RAM_local/(float)SIZE_OF_GIGIBYTE);
else if((float)SID.max_RAM_local/(float)SIZE_OF_KILOBYTE>1.)
fprintf(SID.fp_log,"Peak for rank %3d=%4.2lf Mb\n",i_rank,(float)SID_max_RAM_local/(float)SIZE_OF_MEGABYTE);
else
fprintf(SID.fp_log,"Peak for rank %3d=%4.2lf kb\n",i_rank,(float)SID_max_RAM_local/(float)SIZE_OF_KILOBYTE);
strcpy(spacer," ");
}
else if(SID.n_proc>10){
if((float)SID.max_RAM_local/(float)SIZE_OF_MEGABYTE>1.)
fprintf(SID.fp_log,"Peak for rank %2d=%4.2lf Gb\n",i_rank,(float)SID_max_RAM_local/(float)SIZE_OF_GIGIBYTE);
else if((float)SID.max_RAM_local/(float)SIZE_OF_KILOBYTE>1.)
fprintf(SID.fp_log,"Peak for rank %2d=%4.2lf Mb\n",i_rank,(float)SID_max_RAM_local/(float)SIZE_OF_MEGABYTE);
else
fprintf(SID.fp_log,"Peak for rank %2d=%4.2lf kb\n",i_rank,(float)SID_max_RAM_local/(float)SIZE_OF_KILOBYTE);
strcpy(spacer," ");
}
else{
if((float)SID.max_RAM_local/(float)SIZE_OF_MEGABYTE>1.)
fprintf(SID.fp_log,"Peak for rank %d=%4.2lf Gb\n",i_rank,(float)SID_max_RAM_local/(float)SIZE_OF_GIGIBYTE);
else if((float)SID.max_RAM_local/(float)SIZE_OF_KILOBYTE>1.)
fprintf(SID.fp_log,"Peak for rank %d=%4.2lf Mb\n",i_rank,(float)SID_max_RAM_local/(float)SIZE_OF_MEGABYTE);
else
fprintf(SID.fp_log,"Peak for rank %d=%4.2lf kb\n",i_rank,(float)SID_max_RAM_local/(float)SIZE_OF_KILOBYTE);
strcpy(spacer," ");
}
}
SID_Barrier(SID.COMM_WORLD);
}
}
#else
strcpy(spacer,"\0");
max_RAM=SID.max_RAM_local;
#endif
if(SID.I_am_Master){
if((float)max_RAM/(float)SIZE_OF_MEGABYTE>1.)
fprintf(SID.fp_log,"Peak total %s=%4.2lf Gb\n",spacer,(float)max_RAM/(float)SIZE_OF_GIGIBYTE);
else if((float)max_RAM/(float)SIZE_OF_KILOBYTE>1.)
fprintf(SID.fp_log,"Peak total %s=%4.2lf Mb\n",spacer,(float)max_RAM/(float)SIZE_OF_MEGABYTE);
else
fprintf(SID.fp_log,"Peak total %s=%4.2lf kb\n",spacer,(float)max_RAM/(float)SIZE_OF_KILOBYTE);
}
}
// Free some arrays
SID_free(SID_FARG SID.time_start_level);
SID_free(SID_FARG SID.time_stop_level);
SID_free(SID_FARG SID.time_total_level);
SID_free(SID_FARG SID.flag_use_timer);
SID_free(SID_FARG SID.IO_size);
SID_free(SID_FARG SID.My_node);
// Finalize MPI
SID_Comm_free(&(SID.COMM_WORLD));
#if USE_MPI_IO
MPI_Info_free(&(SID.file_info));
#endif
#if USE_MPI
MPI_Finalize();
#endif
exit(status);
}
size_t SID_fread_chunked_ordered(void *buffer,
size_t n_x_read_local,
SID_fp *fp){
int i_chunk;
int i_group;
int i_rank;
size_t i_x_offset_local;
size_t n_x_read_local_bcast;
size_t i_x_read_chunk;
size_t i_x_offset_global;
size_t i_x_chunk;
size_t n_skip;
size_t n_x_chunk;
size_t n_x_chunk_max;
size_t header_size;
size_t r_val=0;
char filename_chunk[256];
// Operates the same way as SID_fread_chunked() except i_x_offset_local is
// computed here under the assumption that reads are done in order by rank
i_x_offset_local=0;
#if USE_MPI
for(i_rank=0;i_rank<SID.n_proc;i_rank++){
n_x_read_local_bcast=n_x_read_local;
SID_Bcast(&n_x_read_local_bcast,sizeof(size_t),i_rank,SID.COMM_WORLD);
if(i_rank<SID.My_rank)
i_x_offset_local+=n_x_read_local_bcast;
}
#endif
i_x_offset_global=fp->last_item;
for(i_chunk=0,i_x_read_chunk=0,i_x_chunk=i_x_offset_local+i_x_offset_global;
i_chunk<fp->chunked_header.n_chunk;
i_chunk++){
if(fp->i_x_start_chunk[i_chunk]<=i_x_chunk && fp->i_x_last_chunk[i_chunk]>=i_x_chunk){
n_skip =i_x_chunk-fp->i_x_start_chunk[i_chunk];
n_x_chunk=MIN(n_x_read_local-i_x_read_chunk,fp->i_x_step_chunk[i_chunk]-n_skip);
}
else{
n_x_chunk=0;
n_skip =0;
}
SID_Allreduce(&n_x_chunk,&n_x_chunk_max,1,SID_SIZE_T,SID_MAX,SID.COMM_WORLD);
if(n_x_chunk_max>0){
sprintf(filename_chunk,"%s.%d",fp->filename_root,i_chunk);
#if !USE_MPI_IO
for(i_group=0;i_group<SID.n_proc;i_group++){
if(SID.My_group==i_group && n_x_chunk>0){
#endif
SID_fopen(filename_chunk,"r",fp);
if(n_x_chunk>0){
SID_fseek(fp,
1,
fp->header_offset[i_chunk]+n_skip*fp->chunked_header.item_size,
SID_SEEK_SET);
SID_fread((char *)buffer+i_x_read_chunk*fp->chunked_header.item_size,
fp->chunked_header.item_size,
n_x_chunk,
fp);
i_x_chunk +=n_x_chunk;
i_x_read_chunk+=n_x_chunk;
}
SID_fclose(fp);
#if !USE_MPI_IO
}
SID_Barrier(SID.COMM_WORLD);
}
#endif
}
}
SID_Allreduce(&i_x_chunk,&(fp->last_item),1,SID_SIZE_T,SID_MAX,SID.COMM_WORLD);
//fp->last_item--;
return(r_val);
}
int main(int argc, char *argv[]){
int n_species;
int n_load;
int n_used;
int flag_used[N_GADGET_TYPE];
char species_name[256];
double h_Hubble;
double n_spec;
double redshift;
int i_species;
char n_string[64];
int n[3];
double L[3];
FILE *fp_1D;
FILE *fp_2D;
cosmo_info *cosmo;
field_info *field[N_GADGET_TYPE];
field_info *field_norm[N_GADGET_TYPE];
plist_info plist_header;
plist_info plist;
FILE *fp;
int i_temp;
int n_temp;
double *k_temp;
double *kmin_temp;
double *kmax_temp;
double *P_temp;
size_t *n_mode_temp;
double *sigma_P_temp;
double *shot_noise_temp;
double *dP_temp;
int snapshot_number;
int i_compute;
int distribution_scheme;
double k_min_1D;
double k_max_1D;
double k_min_2D;
double k_max_2D;
int n_k_1D;
int n_k_2D;
double *k_1D;
double *P_k_1D;
double *dP_k_1D;
int *n_modes_1D;
double *P_k_2D;
double *dP_k_2D;
int *n_modes_2D;
int n_groups=1;
double dk_1D;
double dk_2D;
char *grid_identifier;
// Initialization -- MPI etc.
SID_init(&argc,&argv,NULL,NULL);
// Parse arguments
int grid_size;
char filename_in_root[MAX_FILENAME_LENGTH];
char filename_out_root[MAX_FILENAME_LENGTH];
strcpy(filename_in_root, argv[1]);
snapshot_number=(int)atoi(argv[2]);
strcpy(filename_out_root, argv[3]);
grid_size =(int)atoi(argv[4]);
if(!strcmp(argv[5],"ngp") || !strcmp(argv[5],"NGP"))
distribution_scheme=MAP2GRID_DIST_NGP;
else if(!strcmp(argv[5],"cic") || !strcmp(argv[5],"CIC"))
distribution_scheme=MAP2GRID_DIST_CIC;
else if(!strcmp(argv[5],"tsc") || !strcmp(argv[5],"TSC"))
distribution_scheme=MAP2GRID_DIST_TSC;
else if(!strcmp(argv[5],"d12") || !strcmp(argv[5],"D12"))
distribution_scheme=MAP2GRID_DIST_DWT12;
else if(!strcmp(argv[5],"d20") || !strcmp(argv[5],"D20"))
distribution_scheme=MAP2GRID_DIST_DWT20;
else
SID_trap_error("Invalid distribution scheme {%s} specified.",ERROR_SYNTAX,argv[5]);
SID_log("Smoothing Gadget file {%s;snapshot=#%d} to a %dx%dx%d grid with %s kernel...",SID_LOG_OPEN|SID_LOG_TIMER,
filename_in_root,snapshot_number,grid_size,grid_size,grid_size,argv[5]);
// Initialization -- fetch header info
SID_log("Reading Gadget header...",SID_LOG_OPEN);
gadget_read_info fp_gadget;
int flag_filefound=init_gadget_read(filename_in_root,snapshot_number,&fp_gadget);
int flag_multifile=fp_gadget.flag_multifile;
int flag_file_type=fp_gadget.flag_file_type;
gadget_header_info header =fp_gadget.header;
double box_size =(double)(header.box_size);
size_t *n_all =(size_t *)SID_calloc(sizeof(size_t)*N_GADGET_TYPE);
size_t n_total;
if(flag_filefound){
if(SID.I_am_Master){
FILE *fp_in;
char filename[MAX_FILENAME_LENGTH];
int block_length_open;
int block_length_close;
set_gadget_filename(&fp_gadget,0,filename);
fp_in=fopen(filename,"r");
fread_verify(&block_length_open, sizeof(int),1,fp_in);
fread_verify(&header, sizeof(gadget_header_info),1,fp_in);
fread_verify(&block_length_close,sizeof(int),1,fp_in);
fclose(fp_in);
if(block_length_open!=block_length_close)
SID_trap_error("Block lengths don't match (ie. %d!=%d).",ERROR_LOGIC,block_length_open,block_length_close);
}
SID_Bcast(&header,sizeof(gadget_header_info),MASTER_RANK,SID.COMM_WORLD);
redshift=header.redshift;
h_Hubble=header.h_Hubble;
box_size=header.box_size;
if(SID.n_proc>1)
n_load=1;
else
n_load=header.n_files;
for(i_species=0,n_total=0,n_used=0;i_species<N_GADGET_TYPE;i_species++){
n_all[i_species]=(size_t)header.n_all_lo_word[i_species]+((size_t)header.n_all_hi_word[i_species])<<32;
n_total+=n_all[i_species];
if(n_all[i_species]>0){
n_used++;
flag_used[i_species]=TRUE;
}
else
flag_used[i_species]=FALSE;
}
// Initialize cosmology
double box_size =((double *)ADaPS_fetch(plist.data,"box_size"))[0];
double h_Hubble =((double *)ADaPS_fetch(plist.data,"h_Hubble"))[0];
double redshift =((double *)ADaPS_fetch(plist.data,"redshift"))[0];
double expansion_factor=((double *)ADaPS_fetch(plist.data,"expansion_factor"))[0];
double Omega_M =((double *)ADaPS_fetch(plist.data,"Omega_M"))[0];
double Omega_Lambda =((double *)ADaPS_fetch(plist.data,"Omega_Lambda"))[0];
double Omega_k =1.-Omega_Lambda-Omega_M;
double Omega_b=0.; // not needed, so doesn't matter
double f_gas =Omega_b/Omega_M;
double sigma_8=0.; // not needed, so doesn't matter
double n_spec =0.; // not needed, so doesn't matter
char cosmo_name[16];
sprintf(cosmo_name,"Gadget file's");
init_cosmo(&cosmo,
cosmo_name,
Omega_Lambda,
Omega_M,
Omega_k,
Omega_b,
f_gas,
h_Hubble,
sigma_8,
n_spec);
}
SID_log("Done.",SID_LOG_CLOSE);
grid_identifier=(char *)SID_calloc(GRID_IDENTIFIER_SIZE*sizeof(char));
// Only process if there are >0 particles present
if(n_used>0){
// Loop over ithe real-space and 3 redshift-space frames
int i_write;
int i_run;
int n_run;
int n_grids_total;
n_grids_total=4; // For now, hard-wire real-space density and velocity grids only
n_run=1; // For now, hard-wire real-space calculation only
for(i_run=0,i_write=0;i_run<n_run;i_run++){
// Read catalog
int n_grid;
char i_run_identifier[8];
switch(i_run){
case 0:
SID_log("Processing real-space ...",SID_LOG_OPEN|SID_LOG_TIMER);
sprintf(i_run_identifier,"r");
n_grid=4;
break;
case 1:
SID_log("Processing v_x redshift space...",SID_LOG_OPEN|SID_LOG_TIMER);
sprintf(i_run_identifier,"x");
n_grid=1;
break;
case 2:
SID_log("Processing v_y redshift space...",SID_LOG_OPEN|SID_LOG_TIMER);
sprintf(i_run_identifier,"y");
n_grid=1;
break;
case 3:
SID_log("Processing v_z redsift space...",SID_LOG_OPEN|SID_LOG_TIMER);
sprintf(i_run_identifier,"z");
n_grid=1;
break;
}
// For each i_run case, loop over the fields we want to produce
int i_grid;
for(i_grid=0;i_grid<n_grid;i_grid++){
char i_grid_identifier[8];
switch(i_grid){
case 0:
SID_log("Processing density grid ...",SID_LOG_OPEN|SID_LOG_TIMER);
sprintf(i_grid_identifier,"rho");
break;
case 1:
SID_log("Processing v_x velocity grid...",SID_LOG_OPEN|SID_LOG_TIMER);
sprintf(i_grid_identifier,"v_x");
break;
case 2:
SID_log("Processing v_y velocity grid...",SID_LOG_OPEN|SID_LOG_TIMER);
sprintf(i_grid_identifier,"v_y");
break;
case 3:
SID_log("Processing v_z velocity grid...",SID_LOG_OPEN|SID_LOG_TIMER);
sprintf(i_grid_identifier,"v_z");
break;
}
// Initialize the field that will hold the grid
int n[]={grid_size,grid_size,grid_size};
double L[]={box_size, box_size, box_size};
int i_init;
for(i_species=0;i_species<N_GADGET_TYPE;i_species++){
if(flag_used[i_species]){
field[i_species] =(field_info *)SID_malloc(sizeof(field_info));
field_norm[i_species]=(field_info *)SID_malloc(sizeof(field_info));
init_field(3,n,L,field[i_species]);
init_field(3,n,L,field_norm[i_species]);
i_init=i_species;
}
else{
field[i_species] =NULL;
field_norm[i_species]=NULL;
}
}
// Loop over all the files that this rank will read
int i_load;
for(i_load=0;i_load<n_load;i_load++){
if(n_load>1)
SID_log("Processing file No. %d of %d...",SID_LOG_OPEN|SID_LOG_TIMER,i_load+1,n_load);
// Initialization -- read gadget file
GBPREAL mass_array[N_GADGET_TYPE];
init_plist(&plist,&((field[i_init])->slab),GADGET_LENGTH,GADGET_MASS,GADGET_VELOCITY);
char filename_root[MAX_FILENAME_LENGTH];
read_gadget_binary_local(filename_in_root,
snapshot_number,
i_run,
i_load,
n_load,
mass_array,
&(field[i_init]->slab),
cosmo,
&plist);
// Generate power spectra
for(i_species=0;i_species<plist.n_species;i_species++){
// Determine how many particles of species i_species there are
if(n_all[i_species]>0){
// Fetch the needed information
size_t n_particles;
size_t n_particles_local;
int flag_alloc_m;
GBPREAL *x_particles_local;
GBPREAL *y_particles_local;
GBPREAL *z_particles_local;
GBPREAL *vx_particles_local;
GBPREAL *vy_particles_local;
GBPREAL *vz_particles_local;
GBPREAL *m_particles_local;
GBPREAL *v_particles_local;
GBPREAL *w_particles_local;
n_particles =((size_t *)ADaPS_fetch(plist.data,"n_all_%s",plist.species[i_species]))[0];
n_particles_local=((size_t *)ADaPS_fetch(plist.data,"n_%s", plist.species[i_species]))[0];
x_particles_local= (GBPREAL *)ADaPS_fetch(plist.data,"x_%s", plist.species[i_species]);
y_particles_local= (GBPREAL *)ADaPS_fetch(plist.data,"y_%s", plist.species[i_species]);
z_particles_local= (GBPREAL *)ADaPS_fetch(plist.data,"z_%s", plist.species[i_species]);
vx_particles_local=(GBPREAL *)ADaPS_fetch(plist.data,"vx_%s", plist.species[i_species]);
vy_particles_local=(GBPREAL *)ADaPS_fetch(plist.data,"vy_%s", plist.species[i_species]);
vz_particles_local=(GBPREAL *)ADaPS_fetch(plist.data,"vz_%s", plist.species[i_species]);
if(ADaPS_exist(plist.data,"M_%s",plist.species[i_species])){
flag_alloc_m=FALSE;
m_particles_local=(GBPREAL *)ADaPS_fetch(plist.data,"M_%s",plist.species[i_species]);
}
else{
flag_alloc_m=TRUE;
m_particles_local=(GBPREAL *)SID_malloc(n_particles_local*sizeof(GBPREAL));
int i_particle;
for(i_particle=0;i_particle<n_particles_local;i_particle++)
m_particles_local[i_particle]=mass_array[i_species];
}
// Decide the map_to_grid() mode
int mode;
if(n_load==1)
mode=MAP2GRID_MODE_DEFAULT;
else if(i_load==0 || n_load==1)
mode=MAP2GRID_MODE_DEFAULT|MAP2GRID_MODE_NONORM;
else if(i_load==(n_load-1))
mode=MAP2GRID_MODE_NOCLEAN;
else
mode=MAP2GRID_MODE_NOCLEAN|MAP2GRID_MODE_NONORM;
// Set the array that will weight the grid
field_info *field_i;
field_info *field_norm_i;
double factor;
switch(i_grid){
case 0:
v_particles_local=m_particles_local;
w_particles_local=NULL;
field_i =field[i_species];
field_norm_i =NULL;
mode|=MAP2GRID_MODE_APPLYFACTOR;
factor=pow((double)grid_size/box_size,3.);
break;
case 1:
v_particles_local=vx_particles_local;
w_particles_local=m_particles_local;
field_i =field[i_species];
field_norm_i =field_norm[i_species];
factor=1.;
break;
case 2:
v_particles_local=vy_particles_local;
w_particles_local=m_particles_local;
field_i =field[i_species];
field_norm_i =field_norm[i_species];
factor=1.;
break;
case 3:
v_particles_local=vz_particles_local;
w_particles_local=m_particles_local;
field_i =field[i_species];
field_norm_i =field_norm[i_species];
factor=1.;
break;
}
// Generate grid
map_to_grid(n_particles_local,
x_particles_local,
y_particles_local,
z_particles_local,
v_particles_local,
w_particles_local,
cosmo,
redshift,
distribution_scheme,
factor,
field_i,
field_norm_i,
mode);
if(flag_alloc_m)
SID_free(SID_FARG m_particles_local);
}
}
// Clean-up
free_plist(&plist);
if(n_load>1)
SID_log("Done.",SID_LOG_CLOSE);
} // loop over i_load
// Write results to disk
char filename_out_species[MAX_FILENAME_LENGTH];
init_plist(&plist,NULL,GADGET_LENGTH,GADGET_MASS,GADGET_VELOCITY);
for(i_species=0;i_species<plist.n_species;i_species++){
if(flag_used[i_species]){
sprintf(grid_identifier,"%s_%s_%s",i_grid_identifier,i_run_identifier,plist.species[i_species]);
sprintf(filename_out_species,"%s_%s",filename_out_root,plist.species[i_species]);
write_grid(field[i_species],
filename_out_species,
i_write,
n_grids_total,
distribution_scheme,
grid_identifier,
header.box_size);
free_field(field[i_species]);
free_field(field_norm[i_species]);
SID_free(SID_FARG field[i_species]);
SID_free(SID_FARG field_norm[i_species]);
i_write++;
}
}
// Clean-up
free_plist(&plist);
SID_log("Done.",SID_LOG_CLOSE);
} // loop over i_grid
SID_log("Done.",SID_LOG_CLOSE);
} // loop over i_run
} // if n_used>0
// Clean-up
free_cosmo(&cosmo);
SID_free(SID_FARG grid_identifier);
SID_free(SID_FARG n_all);
SID_log("Done.",SID_LOG_CLOSE);
SID_exit(ERROR_NONE);
}
void read_gadget_binary_local(char *filename_root_in,
int snapshot_number,
int i_coord,
int i_load,
int n_load,
GBPREAL mass_array[N_GADGET_TYPE],
slab_info *slab,
cosmo_info *cosmo,
plist_info *plist){
size_t n_of_type_local[N_GADGET_TYPE];
size_t n_of_type[N_GADGET_TYPE];
size_t type_counter[N_GADGET_TYPE];
GBPREAL *x_array[N_GADGET_TYPE];
GBPREAL *y_array[N_GADGET_TYPE];
GBPREAL *z_array[N_GADGET_TYPE];
GBPREAL *vx_array[N_GADGET_TYPE];
GBPREAL *vy_array[N_GADGET_TYPE];
GBPREAL *vz_array[N_GADGET_TYPE];
int i_type;
// Determine file format and read the header
gadget_read_info fp_gadget;
int flag_filefound=init_gadget_read(filename_root_in,snapshot_number,&fp_gadget);
int flag_multifile=fp_gadget.flag_multifile;
int flag_file_type=fp_gadget.flag_file_type;
gadget_header_info header =fp_gadget.header;
// A file was found ...
if(flag_filefound){
char **pname;
SID_log("Reading GADGET binary file...",SID_LOG_OPEN|SID_LOG_TIMER);
pname=plist->species;
// Expansion factor (or time)
ADaPS_store(&(plist->data),(void *)(&(header.time)),"expansion_factor",ADaPS_SCALAR_DOUBLE);
ADaPS_store(&(plist->data),(void *)(&(header.time)),"time", ADaPS_SCALAR_DOUBLE);
// Redshift
double d_value;
d_value=(double)header.redshift;
ADaPS_store(&(plist->data),(void *)(&d_value),"redshift",ADaPS_SCALAR_DOUBLE);
// Number of particles and masses for each species in all files
size_t n_all[N_GADGET_TYPE];
for(i_type=0;i_type<N_GADGET_TYPE;i_type++){
n_all[i_type] =(size_t)header.n_all_lo_word[i_type]+((size_t)header.n_all_hi_word[i_type])<<32;
mass_array[i_type]=(GBPREAL)header.mass_array[i_type];
}
// Number of files in this snapshot
int n_files;
ADaPS_store(&(plist->data),(void *)(&(header.n_files)),"n_files",ADaPS_SCALAR_INT);
n_files=header.n_files;
// Cosmology
// Omega_o
d_value=(double)header.Omega_M;
ADaPS_store(&(plist->data),(void *)(&d_value),"Omega_M",ADaPS_SCALAR_DOUBLE);
// Omega_Lambda
d_value=(double)header.Omega_Lambda;
ADaPS_store(&(plist->data),(void *)(&d_value),"Omega_Lambda",ADaPS_SCALAR_DOUBLE);
// Hubble parameter
double h_Hubble;
double redshift;
h_Hubble=(double)header.h_Hubble;
if(h_Hubble<1e-10) h_Hubble=1.;
ADaPS_store(&(plist->data),(void *)(&h_Hubble),"h_Hubble",ADaPS_SCALAR_DOUBLE);
redshift=header.redshift;
ADaPS_store(&(plist->data),(void *)(&redshift),"redshift",ADaPS_SCALAR_DOUBLE);
// Count and report the total number of particles
size_t n_particles_all;
int n_non_zero;
n_particles_all=0;
for(i_type=0,n_non_zero=0;i_type<N_GADGET_TYPE;i_type++){
if(n_all[i_type]>0){
n_particles_all+=n_all[i_type];
n_non_zero++;
}
}
SID_log("%zd",SID_LOG_CONTINUE,n_particles_all);
if(n_non_zero>0)
SID_log(" (",SID_LOG_CONTINUE,n_particles_all);
for(i_type=0;i_type<N_GADGET_TYPE;i_type++){
if(n_all[i_type]>0){
if(i_type==n_non_zero-1){
if(n_non_zero>1)
SID_log("and %lld %s",SID_LOG_CONTINUE,n_all[i_type],pname[i_type]);
else
SID_log("%lld %s",SID_LOG_CONTINUE,n_all[i_type],pname[i_type]);
}
else{
if(n_non_zero>1)
SID_log("%lld %s, ",SID_LOG_CONTINUE,n_all[i_type],pname[i_type]);
else
SID_log("%lld %s",SID_LOG_CONTINUE,n_all[i_type],pname[i_type]);
}
}
}
if(n_non_zero>0)
SID_log(") particles...",SID_LOG_CONTINUE);
else
SID_log(" particles...",SID_LOG_CONTINUE);
// Count the number of particles that will be scattered to each rank
char filename[MAX_FILENAME_LENGTH];
size_t k_particle;
int i_file;
int record_length_open;
int record_length_close;
size_t i_particle;
size_t i_buffer;
size_t i_step;
int i_type;
size_t index;
GBPREAL *pos_buffer;
GBPREAL *vel_buffer;
double pos_test;
// Initialize some arrays
pos_buffer=(GBPREAL *)SID_malloc(sizeof(GBPREAL)*READ_BUFFER_ALLOC_LOCAL);
vel_buffer=(GBPREAL *)SID_malloc(sizeof(GBPREAL)*READ_BUFFER_ALLOC_LOCAL);
for(i_type=0;i_type<N_GADGET_TYPE;i_type++){
n_of_type_local[i_type]=0;
n_of_type[i_type] =0;
type_counter[i_type] =0;
}
// Determine how many particles of each type will end-up on each core
SID_log("Performing domain decomposition...",SID_LOG_OPEN|SID_LOG_TIMER);
int n_read;
if(n_load<n_files)
n_read=n_files;
else
n_read=1;
for(i_file=i_load;i_file<(i_load+n_read);i_file++){
set_gadget_filename(&fp_gadget,i_file,filename);
// Read header and move to the positions
FILE *fp_pos;
FILE *fp_vel;
fp_pos=fopen(filename,"r");
fread_verify(&record_length_open,4,1,fp_pos);
fread_verify(&header,sizeof(gadget_header_info),1,fp_pos);
fread_verify(&record_length_close,4,1,fp_pos);
if(record_length_open!=record_length_close)
SID_log_warning("Problem with GADGET record size (close of header)",ERROR_LOGIC);
fread_verify(&record_length_open,4,1,fp_pos);
// Create a file pointer to the velocities
fp_vel=fopen(filename,"r");
fread_verify(&record_length_open,4,1,fp_vel);
fseeko(fp_vel,(off_t)(record_length_open),SEEK_CUR);
fread_verify(&record_length_close,4,1,fp_vel);
fread_verify(&record_length_open,4,1,fp_vel);
fseeko(fp_vel,(off_t)(record_length_open),SEEK_CUR);
fread_verify(&record_length_close,4,1,fp_vel);
if(record_length_open!=record_length_close)
SID_log_warning("Problem with GADGET record size (close of positons)",ERROR_LOGIC);
fread_verify(&record_length_open,4,1,fp_vel);
// We only have to worry about z-space effects for domain decomposition in this one case.
if(i_coord==1){
for(i_type=0;i_type<N_GADGET_TYPE;i_type++){
for(i_particle=0;i_particle<header.n_file[i_type];i_particle+=i_step){
i_step=MIN(READ_BUFFER_SIZE_LOCAL,header.n_file[i_type]-i_particle);
if(SID.I_am_Master){
fread_verify(pos_buffer,sizeof(GBPREAL),3*i_step,fp_pos);
fread_verify(vel_buffer,sizeof(GBPREAL),3*i_step,fp_vel);
}
SID_Bcast(pos_buffer,sizeof(GBPREAL)*3*i_step,MASTER_RANK,SID.COMM_WORLD);
SID_Bcast(vel_buffer,sizeof(GBPREAL)*3*i_step,MASTER_RANK,SID.COMM_WORLD);
for(i_buffer=0;i_buffer<i_step;i_buffer++){
index=3*i_buffer;
pos_test =(double)(pos_buffer[index]);
pos_test+=(double)(1e3*h_Hubble*((double)vel_buffer[index])/(a_of_z(redshift)*M_PER_MPC*H_convert(H_z(redshift,cosmo))));
if(pos_test<0) pos_test+=header.box_size;
if(pos_test>=header.box_size) pos_test-=header.box_size;
if(pos_test>=slab->x_min_local && pos_test<slab->x_max_local)
n_of_type_local[i_type]++;
}
}
}
}
else{
for(i_type=0;i_type<N_GADGET_TYPE;i_type++){
for(i_particle=0;i_particle<header.n_file[i_type];i_particle+=i_step){
i_step=MIN(READ_BUFFER_SIZE_LOCAL,header.n_file[i_type]-i_particle);
if(SID.I_am_Master){
fread_verify(pos_buffer,sizeof(GBPREAL),3*i_step,fp_pos);
fread_verify(vel_buffer,sizeof(GBPREAL),3*i_step,fp_vel);
}
SID_Bcast(pos_buffer,sizeof(GBPREAL)*3*i_step,MASTER_RANK,SID.COMM_WORLD);
SID_Bcast(vel_buffer,sizeof(GBPREAL)*3*i_step,MASTER_RANK,SID.COMM_WORLD);
for(i_buffer=0;i_buffer<i_step;i_buffer++){
pos_test=pos_buffer[3*i_buffer];
if(pos_test>=slab->x_min_local && pos_test<slab->x_max_local)
n_of_type_local[i_type]++;
}
}
i_step=MIN(READ_BUFFER_SIZE_LOCAL,header.n_file[i_type]-i_particle);
}
}
fclose(fp_pos);
fclose(fp_vel);
}
SID_log("Done.",SID_LOG_CLOSE);
// Allocate arrays
for(i_type=0;i_type<N_GADGET_TYPE;i_type++){
if(n_all[i_type]>0){
x_array[i_type] =(GBPREAL *)SID_malloc(sizeof(GBPREAL)*n_of_type_local[i_type]);
y_array[i_type] =(GBPREAL *)SID_malloc(sizeof(GBPREAL)*n_of_type_local[i_type]);
z_array[i_type] =(GBPREAL *)SID_malloc(sizeof(GBPREAL)*n_of_type_local[i_type]);
vx_array[i_type]=(GBPREAL *)SID_malloc(sizeof(GBPREAL)*n_of_type_local[i_type]);
vy_array[i_type]=(GBPREAL *)SID_malloc(sizeof(GBPREAL)*n_of_type_local[i_type]);
vz_array[i_type]=(GBPREAL *)SID_malloc(sizeof(GBPREAL)*n_of_type_local[i_type]);
}
}
// Perform read
SID_log("Performing read...",SID_LOG_OPEN|SID_LOG_TIMER);
for(i_file=i_load;i_file<(i_load+n_read);i_file++){
set_gadget_filename(&fp_gadget,i_file,filename);
// Read header and move to the positions
FILE *fp_pos;
FILE *fp_vel;
fp_pos=fopen(filename,"r");
fread_verify(&record_length_open,4,1,fp_pos);
fread_verify(&header,sizeof(gadget_header_info),1,fp_pos);
fread_verify(&record_length_close,4,1,fp_pos);
if(record_length_open!=record_length_close)
SID_log_warning("Problem with GADGET record size (close of header)",ERROR_LOGIC);
fread_verify(&record_length_open,4,1,fp_pos);
// Create a file pointer to the velocities
fp_vel=fopen(filename,"r");
fread_verify(&record_length_open,4,1,fp_vel);
fseeko(fp_vel,(off_t)(record_length_open),SEEK_CUR);
fread_verify(&record_length_close,4,1,fp_vel);
fread_verify(&record_length_open,4,1,fp_vel);
fseeko(fp_vel,(off_t)(record_length_open),SEEK_CUR);
fread_verify(&record_length_close,4,1,fp_vel);
if(record_length_open!=record_length_close)
SID_log_warning("Problem with GADGET record size (close of positions)",ERROR_LOGIC);
fread_verify(&record_length_open,4,1,fp_vel);
// Perform the read and populate the local position arrays
size_t i_particle;
size_t i_step;
int i_type;
for(i_type=0;i_type<N_GADGET_TYPE;i_type++){
for(i_particle=0;i_particle<header.n_file[i_type];i_particle+=i_step){
i_step=MIN(READ_BUFFER_SIZE_LOCAL,header.n_file[i_type]-i_particle);
if(SID.I_am_Master){
fread_verify(pos_buffer,sizeof(GBPREAL),3*i_step,fp_pos);
fread_verify(vel_buffer,sizeof(GBPREAL),3*i_step,fp_vel);
}
SID_Bcast(pos_buffer,sizeof(GBPREAL)*3*i_step,MASTER_RANK,SID.COMM_WORLD);
SID_Bcast(vel_buffer,sizeof(GBPREAL)*3*i_step,MASTER_RANK,SID.COMM_WORLD);
for(i_buffer=0;i_buffer<i_step;i_buffer++){
double x_test;
double y_test;
double z_test;
double vx_test;
double vy_test;
double vz_test;
index=3*i_buffer;
x_test =(double)pos_buffer[index+0];
y_test =(double)pos_buffer[index+1];
z_test =(double)pos_buffer[index+2];
vx_test=(double)vel_buffer[index+0];
vy_test=(double)vel_buffer[index+1];
vz_test=(double)vel_buffer[index+2];
switch(i_coord){
case 1:
x_test+=(1e3*h_Hubble*vx_test/(a_of_z(redshift)*M_PER_MPC*H_convert(H_z(redshift,cosmo))));
if(x_test<0) x_test+=header.box_size;
if(x_test>=header.box_size) x_test-=header.box_size;
break;
case 2:
y_test+=(1e3*h_Hubble*vy_test/(a_of_z(redshift)*M_PER_MPC*H_convert(H_z(redshift,cosmo))));
if(y_test<0) y_test+=header.box_size;
if(y_test>=header.box_size) y_test-=header.box_size;
break;
case 3:
z_test+=(1e3*h_Hubble*vz_test/(a_of_z(redshift)*M_PER_MPC*H_convert(H_z(redshift,cosmo))));
if(z_test<0) z_test+=header.box_size;
if(z_test>=header.box_size) z_test-=header.box_size;
break;
}
if(x_test>=slab->x_min_local && x_test<slab->x_max_local){
x_array[i_type][type_counter[i_type]] =x_test;
y_array[i_type][type_counter[i_type]] =y_test;
z_array[i_type][type_counter[i_type]] =z_test;
vx_array[i_type][type_counter[i_type]]=vx_test;
vy_array[i_type][type_counter[i_type]]=vy_test;
vz_array[i_type][type_counter[i_type]]=vz_test;
type_counter[i_type]++;
}
}
}
}
// Close file pointers
fclose(fp_pos);
fclose(fp_vel);
}
SID_free(SID_FARG pos_buffer);
SID_free(SID_FARG vel_buffer);
SID_log("Done.",SID_LOG_CLOSE);
// Sanity checks
size_t n_particles_local;
size_t n_particles_read;
size_t n_particles_test;
for(i_type=0,n_particles_local=0,n_particles_test=0;i_type<N_GADGET_TYPE;i_type++){
n_particles_local+=n_of_type_local[i_type];
n_particles_test +=n_all[i_type];
}
SID_Allreduce(&n_particles_local,&n_particles_read,1,SID_SIZE_T,SID_SUM,SID.COMM_WORLD);
if(n_particles_read!=n_particles_test && n_load==1)
SID_trap_error("Total particle counts don't make sense after read_gadget (ie. %zd!=%zd).",ERROR_LOGIC,n_particles_read,n_particles_test);
for(i_type=0;i_type<N_GADGET_TYPE;i_type++){
SID_Allreduce(&(n_of_type_local[i_type]),&(n_of_type[i_type]),1,SID_SIZE_T,SID_SUM,SID.COMM_WORLD);
if(n_of_type[i_type]!=n_all[i_type] && n_load==1)
SID_trap_error("Particle counts don't make sense after read_gadget (ie. %zd!=%zd).",ERROR_LOGIC,n_of_type[i_type],n_all[i_type]);
}
// Store results
for(i_type=0;i_type<N_GADGET_TYPE;i_type++){
if(n_of_type[i_type]>0){
ADaPS_store(&(plist->data),(void *)(&(n_of_type_local[i_type])),"n_%s", ADaPS_SCALAR_SIZE_T,pname[i_type]);
ADaPS_store(&(plist->data),(void *)(&(n_of_type[i_type])), "n_all_%s", ADaPS_SCALAR_SIZE_T,pname[i_type]);
}
}
ADaPS_store(&(plist->data),(void *)(&n_particles_all),"n_particles_all",ADaPS_SCALAR_SIZE_T);
for(i_type=0;i_type<N_GADGET_TYPE;i_type++){
if(n_of_type_local[i_type]>0){
ADaPS_store(&(plist->data),(void *)x_array[i_type], "x_%s", ADaPS_DEFAULT, pname[i_type]);
ADaPS_store(&(plist->data),(void *)y_array[i_type], "y_%s", ADaPS_DEFAULT, pname[i_type]);
ADaPS_store(&(plist->data),(void *)z_array[i_type], "z_%s", ADaPS_DEFAULT, pname[i_type]);
ADaPS_store(&(plist->data),(void *)vx_array[i_type],"vx_%s",ADaPS_DEFAULT,pname[i_type]);
ADaPS_store(&(plist->data),(void *)vy_array[i_type],"vy_%s",ADaPS_DEFAULT,pname[i_type]);
ADaPS_store(&(plist->data),(void *)vz_array[i_type],"vz_%s",ADaPS_DEFAULT,pname[i_type]);
}
}
SID_log("Done.",SID_LOG_CLOSE);
}
}
void read_smooth(plist_info *plist,
char *filename_root_in,
int snapshot_number,
int mode){
char filename[256];
size_t n_particles_all_mem;
size_t n_particles_local;
size_t n_particles_total;
int i_quantity;
int n_quantities=3;
int *used;
char *species_name;
char var_name[256];
char unit_name[256];
double unit_factor;
int i_file;
size_t *ids;
size_t *ids_index;
int n_particles_file;
int offset;
int n_files;
void *id_buf;
size_t *id_buf_index=NULL;
int *id_buf_i;
long long *id_buf_L;
long long *mark;
size_t i_particle;
size_t j_particle;
void *buffer;
float *local_array;
int n_mark;
float *r_smooth_array;
float *rho_array;
float *sigma_v_array;
char *read_array;
double expansion_factor;
double h_Hubble;
int flag_filefound=FALSE;
int flag_multifile=FALSE;
int flag_file_type;
int flag_LONGIDs;
int read_rank=MASTER_RANK;
int flag_log_sigma=FALSE;
int flag_log_rho=FALSE;
FILE *fp;
SID_log("Reading smooth file {%s}...",SID_LOG_OPEN|SID_LOG_TIMER,filename_root_in);
// Read header info
SID_log("Reading header information...",SID_LOG_OPEN);
smooth_header_info header;
flag_filefound =init_smooth_read(filename_root_in,snapshot_number,&flag_multifile,&flag_file_type,&header);
SID_log("n_files =%d", SID_LOG_COMMENT,header.n_files);
SID_log("n_particles=%lld",SID_LOG_COMMENT,header.n_particles_total);
SID_log("Done.",SID_LOG_CLOSE);
// Interpret the mode passed to this function
int flag_logs_used =FALSE;
int flag_log_quantity=FALSE;
if(check_mode_for_flag(mode,READ_SMOOTH_LOG_SIGMA)){
flag_log_sigma=TRUE;
flag_logs_used=TRUE;
}
else
flag_log_sigma=FALSE;
if(check_mode_for_flag(mode,READ_SMOOTH_LOG_RHO)){
flag_log_rho =TRUE;
flag_logs_used=TRUE;
}
else
flag_log_rho=FALSE;
// A file was found ...
if(flag_filefound){
// Communicate the header
n_particles_file =header.n_particles_file;
offset =header.offset;
n_particles_total=header.n_particles_total;
n_files =MAX(1,header.n_files);
SID_Bcast(&n_particles_file, (int)sizeof(int), read_rank,SID.COMM_WORLD);
SID_Bcast(&offset, (int)sizeof(int), read_rank,SID.COMM_WORLD);
SID_Bcast(&n_particles_total,(int)sizeof(long long),read_rank,SID.COMM_WORLD);
SID_Bcast(&n_files, (int)sizeof(int), read_rank,SID.COMM_WORLD);
// Fetch the number of particles and their ids
species_name =plist->species[GADGET_TYPE_DARK];
n_particles_local=((size_t *)ADaPS_fetch(plist->data,"n_%s", species_name))[0];
n_particles_all_mem =((size_t *)ADaPS_fetch(plist->data,"n_all_%s",species_name))[0];
ids = (size_t *)ADaPS_fetch(plist->data,"id_%s", species_name);
expansion_factor =((double *)ADaPS_fetch(plist->data,"expansion_factor"))[0];
h_Hubble =((double *)ADaPS_fetch(plist->data,"h_Hubble"))[0];
// Check if we are dealing with LONG IDs or not
if(ADaPS_exist(plist->data,"flag_LONGIDs"))
flag_LONGIDs=TRUE;
else
flag_LONGIDs=FALSE;
// Sort particle IDs
SID_log("Sorting particle IDs...",SID_LOG_OPEN|SID_LOG_TIMER);
merge_sort(ids,(size_t)n_particles_local,&ids_index,SID_SIZE_T,SORT_COMPUTE_INDEX,FALSE);
SID_log("Done.",SID_LOG_CLOSE);
// Allocate arrays
SID_log("Allocating arrays...",SID_LOG_OPEN);
read_array=(char *)SID_calloc(sizeof(char)*n_particles_local);
for(i_quantity=0;i_quantity<n_quantities;i_quantity++){
switch(i_quantity){
case 0:
r_smooth_array=(float *)SID_malloc(sizeof(float)*n_particles_local);
ADaPS_store(&(plist->data),r_smooth_array,"r_smooth_%s",ADaPS_DEFAULT,species_name);
break;
case 1:
rho_array=(float *)SID_malloc(sizeof(float)*n_particles_local);
ADaPS_store(&(plist->data),rho_array,"rho_%s",ADaPS_DEFAULT,species_name);
break;
case 2:
sigma_v_array=(float *)SID_malloc(sizeof(float)*n_particles_local);
ADaPS_store(&(plist->data),sigma_v_array,"sigma_v_%s",ADaPS_DEFAULT,species_name);
break;
}
}
SID_log("Done.",SID_LOG_CLOSE);
// Read each file in turn
pcounter_info pcounter;
size_t n_particles_read=0;
SID_log("Performing read...",SID_LOG_OPEN|SID_LOG_TIMER);
SID_init_pcounter(&pcounter,n_particles_total,10);
for(i_file=0;i_file<n_files;i_file++){
set_smooth_filename(filename_root_in,snapshot_number,i_file,flag_multifile,flag_file_type,filename);
if((fp=fopen(filename,"r"))!=NULL){
fread_verify(&(header.n_particles_file), sizeof(int), 1,fp);
fread_verify(&(header.offset), sizeof(int), 1,fp);
fread_verify(&(header.n_particles_total),sizeof(long long),1,fp);
fread_verify(&(header.n_files), sizeof(int), 1,fp);
}
else
SID_trap_error("Could not open smooth file {%s}",ERROR_IO_OPEN,filename);
n_particles_file =header.n_particles_file;
offset =header.offset;
n_particles_total=header.n_particles_total;
n_files =MAX(1,header.n_files);
SID_Bcast(&n_particles_file, (int)sizeof(int), read_rank,SID.COMM_WORLD);
SID_Bcast(&offset, (int)sizeof(int), read_rank,SID.COMM_WORLD);
SID_Bcast(&n_particles_total,(int)sizeof(long long),read_rank,SID.COMM_WORLD);
SID_Bcast(&n_files, (int)sizeof(int), read_rank,SID.COMM_WORLD);
// Read IDs
if(flag_LONGIDs){
if(i_file==0) SID_log("(long long) IDs...",SID_LOG_CONTINUE);
id_buf =SID_malloc(sizeof(long long)*n_particles_file);
id_buf_L=(long long *)id_buf;
if(SID.My_rank==read_rank){
fseeko(fp,(size_t)(3*n_particles_file*sizeof(float)),SEEK_CUR);
fread_verify(id_buf,sizeof(long long),n_particles_file,fp);
}
SID_Barrier(SID.COMM_WORLD);
SID_Bcast(id_buf_L,(int)(n_particles_file*sizeof(long long)),read_rank,SID.COMM_WORLD);
merge_sort(id_buf_L,(size_t)n_particles_file,&id_buf_index,SID_SIZE_T,SORT_COMPUTE_INDEX,FALSE);
}
else{
if(i_file==0) SID_log("(int) IDs...",SID_LOG_CONTINUE);
id_buf =SID_malloc(sizeof(int)*n_particles_file);
id_buf_i=(int *)id_buf;
if(SID.My_rank==read_rank){
fseeko(fp,(size_t)(3*n_particles_file*sizeof(float)),SEEK_CUR);
fread_verify(id_buf,sizeof(int),n_particles_file,fp);
}
SID_Barrier(SID.COMM_WORLD);
SID_Bcast(id_buf_i,(int)(n_particles_file*sizeof(int)),read_rank,SID.COMM_WORLD);
merge_sort(id_buf_i,(size_t)n_particles_file,&id_buf_index,SID_INT,SORT_COMPUTE_INDEX,FALSE);
}
// Create local particle mapping
int n_mark_i=0;
mark=(long long *)SID_malloc(sizeof(long long)*n_particles_file);
for(i_particle=0;i_particle<n_particles_file;i_particle++)
mark[i_particle]=-1;
if(flag_LONGIDs){
for(i_particle=0,j_particle=0,n_mark=0;i_particle<n_particles_file && j_particle<n_particles_local;i_particle++){
while(j_particle<n_particles_local-1 && ids[ids_index[j_particle]]<id_buf_L[id_buf_index[i_particle]]) j_particle++;
if(ids[ids_index[j_particle]]==id_buf_L[id_buf_index[i_particle]]){
mark[id_buf_index[i_particle]]=(long long)ids_index[j_particle];
n_mark++;
n_mark_i++;
}
}
}
else{
for(i_particle=0,j_particle=0,n_mark=0;i_particle<n_particles_file && j_particle<n_particles_local;i_particle++){
while(j_particle<n_particles_local-1 && ids[ids_index[j_particle]]<id_buf_i[id_buf_index[i_particle]]) j_particle++;
if(ids[ids_index[j_particle]]==id_buf_i[id_buf_index[i_particle]]){
mark[id_buf_index[i_particle]]=(long long)ids_index[j_particle];
n_mark++;
n_mark_i++;
}
}
}
SID_free(SID_FARG id_buf);
SID_free(SID_FARG id_buf_index);
// Move to the start of the particle quantities
if(SID.My_rank==read_rank){
rewind(fp);
fread_verify(&n_particles_file, sizeof(int), 1,fp);
fread_verify(&offset, sizeof(int), 1,fp);
fread_verify(&n_particles_total,sizeof(long long),1,fp);
fread_verify(&n_files, sizeof(int), 1,fp);
n_files=MAX(1,n_files);
}
SID_Bcast(&n_particles_file, (int)sizeof(int), read_rank,SID.COMM_WORLD);
SID_Bcast(&offset, (int)sizeof(int), read_rank,SID.COMM_WORLD);
SID_Bcast(&n_particles_total,(int)sizeof(long long),read_rank,SID.COMM_WORLD);
SID_Bcast(&n_files, (int)sizeof(int), read_rank,SID.COMM_WORLD);
buffer=SID_malloc(sizeof(float)*n_particles_file);
for(i_quantity=0;i_quantity<n_quantities;i_quantity++){
int flag_log_quantity;
switch(i_quantity){
case 0:
sprintf(var_name,"r_smooth_%s",species_name);
sprintf(unit_name,"Mpc");
unit_factor=plist->length_unit/h_Hubble;
local_array=r_smooth_array;
break;
case 1:
sprintf(var_name,"rho_%s",species_name);
sprintf(unit_name,"Msol/Mpc^3");
unit_factor=h_Hubble*h_Hubble*plist->mass_unit/pow(plist->length_unit,3.);
local_array=rho_array;
break;
case 2:
sprintf(var_name,"sigma_v_%s",species_name);
sprintf(unit_name,"km/s");
unit_factor=sqrt(expansion_factor)*plist->velocity_unit;
local_array=sigma_v_array;
break;
}
// Read next quantity
if(SID.My_rank==read_rank)
fread_verify(buffer,sizeof(float),n_particles_file,fp);
SID_Barrier(SID.COMM_WORLD);
SID_Bcast(buffer,(int)(n_particles_file*sizeof(float)),read_rank,SID.COMM_WORLD);
// Place in final array
if(n_mark_i>0){
if(i_quantity==0){
for(i_particle=0;i_particle<n_particles_file;i_particle++)
if(mark[i_particle]>=0) read_array[mark[i_particle]]=TRUE;
}
for(i_particle=0;i_particle<n_particles_file;i_particle++)
if(mark[i_particle]>=0) local_array[mark[i_particle]]=((float *)buffer)[i_particle]*unit_factor;
}
}
SID_free(SID_FARG mark);
SID_free(SID_FARG buffer);
if(SID.My_rank==read_rank)
fclose(fp);
n_particles_read+=n_particles_file;
if(n_files>1)
SID_check_pcounter(&pcounter,n_particles_read);
} // i_file
SID_free(SID_FARG ids_index);
SID_Barrier(SID.COMM_WORLD);
SID_log("Done.",SID_LOG_CLOSE);
// Check that all particles have been treated
size_t sum_check=0;
for(i_particle=0;i_particle<n_particles_local;i_particle++)
sum_check+=(size_t)read_array[i_particle];
SID_Allreduce(SID_IN_PLACE,&sum_check,1,SID_SIZE_T,SID_SUM,SID.COMM_WORLD);
if(sum_check!=n_particles_all_mem)
SID_trap_error("Only %lld of %lld particles were set.",ERROR_LOGIC,sum_check,n_particles_all_mem);
SID_free(SID_FARG read_array);
SID_log("Summary...",SID_LOG_OPEN);
for(i_quantity=0;i_quantity<n_quantities;i_quantity++){
char var_name[32];
switch(i_quantity){
case 0:
sprintf(var_name,"Lengths: ");
sprintf(unit_name,"Mpc");
unit_factor=1./M_PER_MPC;
local_array=r_smooth_array;
break;
case 1:
sprintf(var_name,"Densities: ");
sprintf(unit_name,"Msol/Mpc^3");
unit_factor=M_PER_MPC*M_PER_MPC*M_PER_MPC/M_SOL;
local_array=rho_array;
break;
case 2:
sprintf(var_name,"sigmas_v's:");
sprintf(unit_name,"km/s");
unit_factor=1e-3;
local_array=sigma_v_array;
break;
}
// Report some stats
float var_min,var_max,var_mean;
calc_min_global(local_array,
&var_min,
n_particles_local,
SID_FLOAT,
CALC_MODE_DEFAULT,
SID.COMM_WORLD);
calc_max_global(local_array,
&var_max,
n_particles_local,
SID_FLOAT,
CALC_MODE_DEFAULT,
SID.COMM_WORLD);
calc_mean_global(local_array,
&var_mean,
n_particles_local,
SID_FLOAT,
CALC_MODE_DEFAULT,
SID.COMM_WORLD);
// Remove NaN's from sigma_v's if needed
size_t n_NaN=0;
for(i_particle=0;i_particle<n_particles_local;i_particle++){
if(isnan(local_array[i_particle])){
local_array[i_particle]=1000.*0.5*(var_min+var_max);
n_NaN++;
}
}
if(n_NaN>0)
SID_log("%s min=%e max=%e mean=%e [%s] (%lld are NaN)",
SID_LOG_COMMENT,
var_name,
var_min*unit_factor,
var_max*unit_factor,
var_mean*unit_factor,
unit_name,
n_NaN);
else
SID_log("%s min=%e max=%e mean=%e [%s]",
SID_LOG_COMMENT,
var_name,
var_min*unit_factor,
var_max*unit_factor,
var_mean*unit_factor,
unit_name);
}
SID_log("",SID_LOG_CLOSE|SID_LOG_NOPRINT);
if(flag_logs_used){
SID_log("Taking needed logs of quantities...",SID_LOG_OPEN|SID_LOG_TIMER);
for(i_quantity=0;i_quantity<n_quantities;i_quantity++){
switch(i_quantity){
case 0:
unit_factor=1./M_PER_MPC;
local_array=r_smooth_array;
flag_log_quantity=FALSE;
break;
case 1:
unit_factor=M_PER_MPC*M_PER_MPC*M_PER_MPC/M_SOL;
local_array=rho_array;
flag_log_quantity=flag_log_rho;
break;
case 2:
unit_factor=1e-3;
local_array=sigma_v_array;
flag_log_quantity=flag_log_sigma;
break;
}
if(flag_log_quantity){
for(i_particle=0;i_particle<n_particles_local;i_particle++)
local_array[i_particle]=take_log10(local_array[i_particle]);
}
}
SID_log("Done.",SID_LOG_CLOSE);
}
SID_Barrier(SID.COMM_WORLD);
SID_log("Done.",SID_LOG_CLOSE);
}
else
SID_trap_error("Could not find file with root {%s}",ERROR_IO_OPEN,filename_root_in);
}
