Commit 59bb195b authored by Luca Tornatore's avatar Luca Tornatore
Browse files

debug fixes

parent 23fe68d4

Build/Makefile.systype

+1 −24
Original line number Diff line number Diff line 
CC       =  gcc-10

CXX      =  g++-10



MPICC    =  mpicc

MPIC++   =  mpiCC



OPTIMIZE = -ggdb3 -O4 -fopenmp -march=native -mavx -mavx2





GSL_INCL =

GSL_LIB =



FFTW_INCL=

FFTW_LIB=



NVCC =

NVFLAGS =

NVLIB =



CFLAGS +=



MPICHLIB =

HDF5INCL =

HDF5LIB  =
 
SYSTYPE=local

Makefile

+1 −0
Original line number Diff line number Diff line
@@ -15,6 +15,7 @@ SYSTYPE := $(SYSTYPE) 
include Build/Makefile.$(SYSTYPE)

else

include Build/Makefile.systype

include Build/Makefile.$(SYSTYPE)

endif



LINKER=$(MPICC)

allvars.h

+0 −1
Original line number Diff line number Diff line
@@ -40,7 +40,6 @@ 
#include "errcodes.h"



#define PI 3.14159265359

#define NUM_OF_SECTORS -1

#define MIN(X, Y) (((X) < (Y)) ? (X) : (Y))

#define MAX(X, Y) (((X) > (Y)) ? (X) : (Y))

#define NOVERBOSE

gridding_cpu.c

+5 −7
Original line number Diff line number Diff line
@@ -100,7 +100,7 @@ void gridding_data() 
					    (Nvissec*2+Nweightss)*sizeof(float_t) );



  if ( memory == NULL )

    shutdown(NOT_ENOUGH_MEM_STACKING, "Not enough memory for stacking", __FILE__, __LINE__);

    shutdown_wstacking(NOT_ENOUGH_MEM_STACKING, "Not enough memory for stacking", __FILE__, __LINE__);

  

  double_t *uus        = (double*) memory;

  double_t *vvs        = (double*) uus+Nsec;
@@ -121,7 +121,8 @@ void gridding_data() 
      uint inu = 0;



      #warning "this loop should be threaded"

      for(uint iphi = histo_send[isector]-1; iphi>=0; iphi--)

      #warning "the counter of this loop should not be int"

      for( int iphi = histo_send[isector]-1; iphi >=0 ; iphi--)

        {

	  uint ilocal = sectorarray[isector][iphi];
@@ -233,7 +234,7 @@ void gridding_data() 
		  sprintf( message, "Some problem occurred in the ring reduce "

			   "while processing sector %d", isector);

		  free( memory );

		  shutdown( ERR_REDUCE, message, __FILE__, __LINE__);

		  shutdown_wstacking( ERR_REDUCE, message, __FILE__, __LINE__);

		}

	      

	    }
@@ -248,9 +249,6 @@ void gridding_data() 


  free( memory );

  

  fclose(file.pFile1);



  

  if ( size > 1 )

    {

      double start = CPU_TIME_wt;

gridding_nccl.cpp

+20 −48
Original line number Diff line number Diff line
@@ -44,10 +44,6 @@ void gridding_data(){ 


  double shift = (double)(dx*yaxis);

    

 #ifndef USE_MPI

  file.pFile1 = fopen (out.outfile1,"w");

 #endif



  timing.kernel_time = 0.0;

  timing.kernel_time1 = 0.0;

  timing.reduce_time = 0.0;
@@ -149,23 +145,6 @@ void gridding_data(){ 
      timing.compose_time1 += (finishk.tv_sec - begink.tv_sec);

      timing.compose_time1 += (finishk.tv_nsec - begink.tv_nsec) / 1000000000.0;

      

     #ifndef USE_MPI

      double vvmin = 1e20;

      double uumax = -1e20;

      double vvmax = -1e20;

	 

      for (long ipart=0; ipart<Nsec; ipart++)

	{

	  uumin = MIN(uumin,uus[ipart]);

	  uumax = MAX(uumax,uus[ipart]);

	  vvmin = MIN(vvmin,vvs[ipart]);

	  vvmax = MAX(vvmax,vvs[ipart]);

	     

	  if(ipart%10 == 0)fprintf (file.pFile, "%ld %f %f %f\n",isector,uus[ipart],vvs[ipart]+isector*shift,wws[ipart]);

	}

	 

      printf("UU, VV, min, max = %f %f %f %f\n", uumin, uumax, vvmin, vvmax);

     #endif



      // Make convolution on the grid
@@ -270,17 +249,10 @@ void gridding_data(){ 


  //Copy data back from device to host (to be deleted in next steps)



 #ifdef NCCL_REDUCE

  cudaMemcpyAsync(grid, grid_gpu, 2*param.num_w_planes*xaxis*yaxis*sizeof(double), cudaMemcpyDeviceToHost, stream_reduce);

 #endif

    #ifndef USE_MPI

        fclose(file.pFile1);

    #endif



  

    #ifdef USE_MPI

  MPI_Barrier(MPI_COMM_WORLD);

    #endif





  end = clock();

  clock_gettime(CLOCK_MONOTONIC, &finish);
@@ -289,15 +261,15 @@ void gridding_data(){ 
  timing.process_time1 += (finish.tv_nsec - begin.tv_nsec) / 1000000000.0;

  clock_gettime(CLOCK_MONOTONIC, &begin);



   #ifdef NCCL_REDUCE

    cudaStreamSynchronize(stream_reduce);





  cudaFree(gridss_gpu);

  cudaFree(grid_gpu);

  

  cudaStreamDestroy(stream_reduce);

  

  ncclCommDestroy(comm);

   #endif



}



#endif