Timing results for ring reduce

# perform one-side communication (suggested) instead of reduce (only if MPI is active)

OPT += -DONE_SIDE

# write the full 3D cube of gridded visibilities and its FFT transform

OPT += -DWRITE_DATA

#OPT += -DWRITE_DATA

# write the final image

OPT += -DWRITE_IMAGE

#OPT += -DWRITE_IMAGE

# perform w-stacking phase correction

OPT += -DPHASE_ON

# perform ring reduce

OPT += -DRING

#perform binomial reduce

#OPT += -DBINOMIAL

#perform debuging

#OPT += -DDEBUG

DEPS = w-stacking.h main.c w-stacking.cu phase_correction.cu allvars.h init.c gridding.c fourier_transform.c result.c reduce.c numa.h

int max_level = 0;

double *end_4, *end_reduce;

int dsize_4, iter=0;  

struct timing_r timing;

struct timingmpi_r timingmpi;

extern int max_level;

extern double *end_4, *end_reduce;

extern int dsize_4, iter;

extern struct timing_r { double rtime, ttotal, treduce, tspin, tspin_in, tmovmemory, tsum;} timing ;

extern struct timingmpi_r{ double tmpi, tmpi_reduce, tmpi_reduce_wait, tmpi_setup;} timingmpi ; 

double check_global_value ;

#endif

struct { double rtime, ttotal, treduce, tspin, tspin_in, tmovmemory, tsum;} timing = {0};

struct { double tmpi, tmpi_reduce, tmpi_reduce_wait, tmpi_setup;} timingmpi = {0};

//struct { double rtime, ttotal, treduce, tspin, tspin_in, tmovmemory, tsum;} timing = {0};

//struct { double tmpi, tmpi_reduce, tmpi_reduce_wait, tmpi_setup;} timingmpi = {0};

int_t summations = 0;

     #if defined(ONE_SIDE)

      //printf("%14s time : %f sec\n", "Reduce sh", wt_timing.reduce_sh);

      printf("%14s time : %f sec\n", "Reduce ring", wt_timing.reduce_ring);

      printf("%14s time : %f sec\n", "Shared mem reduce ring", timing.treduce);

      printf("%14s time : %f sec\n", "Shmem reduce multi host", timing.treduce+timingmpi.tmpi_reduce);

      //printf("%14s time : %f sec\n", "Mmove", wt_timing.mmove);

      //printf("%14s time : %f sec\n", "ReduceMPI", wt_timing.reduce_mpi);

     #endif