Révision 251

BLAS/xTRSV/Makefile (revision 251)
1 
# Makefile designed for Debian Buster
2 

  		





  1
  3
  
    
SOURCE=xTRSV.c


  		





  2
  4
  
    

  		





  3
  5
  
    
CC=gcc


  		





  4
  6
  
    
FC=gfortran


  		





  5
  7
  
    
CFLAGS=-Wall -O3


  		





  6
  8
  
    
LDFLAGS=-lm


  		





  7
  
  
    
CUDASRC=/usr/share/doc/nvidia-cuda-doc/examples/


  		





  
  9
  
    

  		





  
  10
  
    
CUDAINC?=/usr/include


  		





  
  11
  
    
CUDALIB?=/usr/lib/x86_64-linux-gnu/


  		





  
  12
  
    
CUDASRC?=/usr/share/doc/nvidia-cuda-doc/examples/


  		





  8
  13
  
    
THUNKING=fortran_thunking.c


  		





  9
  14
  
    
CUDASRCINC=fortran_common.h


  		





  10
  
  
    
CUDAINC=/usr/include


  		





  11
  
  
    
CUDALIB=/usr/lib/x86_64-linux-gnu/


  		





  12
  15
  
    

  		





  13
  16
  
    
GSLINC=/usr/include/gsl


  		





  14
  17
  
    

  		





  15
  
  
    
GOTO2=/opt/GotoBLAS2


  		





  
  18
  
    
# OPENBLAS=/usr/lib/x86_64-linux-gnu


  		





  
  19
  
    
# ATLAS=/usr/lib/x86_64-linux-gnu


  		





  16
  20
  
    

  		





  17
  
  
    
OPENBLAS=/usr/lib


  		





  18
  
  
    

  		





  19
  
  
    
ATLAS=/usr/lib/atlas-base


  		





  20
  
  
    

  		





  21
  21
  
    
ACML=/opt/acml


  		





  22
  22
  
    
ACMLINC=$(ACML)/gfortran64_mp/include


  		





  23
  23
  
    
ACMLLIB=$(ACML)/gfortran64_mp/lib


  		





  24
  24
  
    

  		





  25
  
  
    
#EXECUTABLE=cblas fblas gsl cublas thunking gotoblas acml openblas


  		





  
  25
  
    
#EXECUTABLE=cblas fblas gsl cublas thunking acml openblas


  		





  26
  26
  
    
EXECUTABLE=cblas fblas gsl cublas thunking openblas


  		





  27
  27
  
    

  		





  28
  
  
    
#FORMAT=DOUBLE


  		





  29
  
  
    
FORMAT=FLOAT


  		





  
  28
  
    
# FORMAT define the precision of Floating Point numbers


  		





  
  29
  
    
# FP32 for simple of 32 bits size


  		





  
  30
  
    
# FP64 for double of 64 bits size


  		





  
  31
  
    
#FORMAT=FP64


  		





  
  32
  
    
FORMAT=FP32


  		





  30
  33
  
    

  		





  31
  34
  
    
#DIRECTIVES=-D$(FORMAT) -DPRINT -DUNIT


  		





  32
  35
  
    
#DIRECTIVES=-D$(FORMAT) -DUNIT -DRESULTS -DQUIET


  		





  ......




  37
  40
  
    

  		





  38
  41
  
    
cblas: $(SOURCE)


  		





  39
  42
  
    

  		





  40
  
  
    
	$(CC) $(CFLAGS) $(DIRECTIVES) -DFLOAT -DCBLAS $(LDFLAGS) \


  		





  41
  
  
    
		$(SOURCE) $(ATLAS)/libcblas.a $(ATLAS)/libatlas.a -o $(SOURCE:.c=)_SP_$@


  		





  
  43
  
    
	$(CC) $(CFLAGS) $(DIRECTIVES) -DFP32 -DCBLAS $(LDFLAGS) \


  		





  
  44
  
    
		$(SOURCE) -latlas -lcblas -o $(SOURCE:.c=)_SP_$@


  		





  42
  45
  
    

  		





  43
  
  
    
	$(CC) $(CFLAGS) $(DIRECTIVES) -DDOUBLE -DCBLAS $(LDFLAGS) \


  		





  44
  
  
    
		$(SOURCE) $(ATLAS)/libcblas.a $(ATLAS)/libatlas.a -o $(SOURCE:.c=)_DP_$@


  		





  
  46
  
    
	$(CC) $(CFLAGS) $(DIRECTIVES) -DFP64 -DCBLAS $(LDFLAGS) \


  		





  
  47
  
    
		$(SOURCE) -latlas -lcblas -o $(SOURCE:.c=)_DP_$@


  		





  45
  48
  
    

  		





  46
  49
  
    
openblas: $(SOURCE)


  		





  47
  50
  
    

  		





  48
  
  
    
	$(CC) $(CFLAGS) $(DIRECTIVES) -DFLOAT -DCBLAS $(LDFLAGS) \


  		





  49
  
  
    
		$(SOURCE) $(OPENBLAS)/libopenblas.a -lpthread -o $(SOURCE:.c=)_SP_$@


  		





  
  51
  
    
	$(CC) $(CFLAGS) $(DIRECTIVES) -DFP32 -DCBLAS $(LDFLAGS) \


  		





  
  52
  
    
		$(SOURCE) -lopenblas -lpthread -o $(SOURCE:.c=)_SP_$@


  		





  50
  53
  
    

  		





  51
  
  
    
	$(CC) $(CFLAGS) $(DIRECTIVES) -DDOUBLE -DCBLAS $(LDFLAGS) \


  		





  52
  
  
    
		$(SOURCE) $(OPENBLAS)/libopenblas.a -lpthread -o $(SOURCE:.c=)_DP_$@


  		





  
  54
  
    
	$(CC) $(CFLAGS) $(DIRECTIVES) -DFP64 -DCBLAS $(LDFLAGS) \


  		





  
  55
  
    
		$(SOURCE) -lopenblas -lpthread -o $(SOURCE:.c=)_DP_$@


  		





  53
  56
  
    

  		





  54
  57
  
    
acml: $(SOURCE)


  		





  55
  58
  
    

  		





  56
  
  
    
	$(CC) -I$(ACMLINC) $(CFLAGS) $(DIRECTIVES) -DFLOAT -DACML $(LDFLAGS) \


  		





  
  59
  
    
	$(CC) -I$(ACMLINC) $(CFLAGS) $(DIRECTIVES) -DFP32 -DACML $(LDFLAGS) \


  		





  57
  60
  
    
		$(SOURCE) -L$(ACMLLIB) -lacml_mp -lacml_mv \


  		





  58
  61
  
    
		-lgomp -lgfortran -lpthread -o $(SOURCE:.c=)_SP_$@


  		





  59
  62
  
    

  		





  60
  
  
    
	$(CC) -I$(ACMLINC) $(CFLAGS) $(DIRECTIVES) -DDOUBLE -DACML $(LDFLAGS) \


  		





  
  63
  
    
	$(CC) -I$(ACMLINC) $(CFLAGS) $(DIRECTIVES) -DFP64 -DACML $(LDFLAGS) \


  		





  61
  64
  
    
		$(SOURCE) -L$(ACMLLIB) -lacml_mp -lacml_mv \


  		





  62
  65
  
    
		-lgomp -lgfortran -lpthread -o $(SOURCE:.c=)_DP_$@


  		





  63
  66
  
    

  		





  64
  67
  
    
fblas: $(SOURCE)


  		





  65
  68
  
    

  		





  66
  
  
    
	$(CC) $(CFLAGS) $(DIRECTIVES) -DFLOAT -DFBLAS $(LDFLAGS) \


  		





  67
  
  
    
		$(SOURCE) $(ATLAS)/libf77blas.a $(ATLAS)/libatlas.a -lgfortran -o $(SOURCE:.c=)_SP_$@


  		





  
  69
  
    
	# ATLAS version


  		





  
  70
  
    
	# $(CC) -L$(ATLAS) $(CFLAGS) $(DIRECTIVES) -DFP32 -DFBLAS $(LDFLAGS) \


  		





  
  71
  
    
	#  	$(SOURCE) -latlas -lf77blas -o $(SOURCE:.c=)_SP_$@


  		





  68
  72
  
    

  		





  69
  
  
    
	$(CC) $(CFLAGS) $(DIRECTIVES) -DDOUBLE -DFBLAS $(LDFLAGS) \


  		





  70
  
  
    
		$(SOURCE) $(ATLAS)/libf77blas.a $(ATLAS)/libatlas.a -lgfortran -o $(SOURCE:.c=)_DP_$@


  		





  
  73
  
    
	# $(CC) -L$(ATLAS) $(CFLAGS) $(DIRECTIVES) -DFP64 -DFBLAS $(LDFLAGS) \


  		





  
  74
  
    
	#  	$(SOURCE) -latlas -lf77blas -o $(SOURCE:.c=)_DP_$@


  		





  71
  75
  
    

  		





  
  76
  
    
	# OpenBLAS version


  		





  
  77
  
    
	$(CC) $(CFLAGS) $(DIRECTIVES) -DFP32 -DFBLAS $(LDFLAGS) \


  		





  
  78
  
    
	  	$(SOURCE) -lopenblas -lpthread -o $(SOURCE:.c=)_SP_$@


  		





  72
  79
  
    

  		





  
  80
  
    
	$(CC) $(CFLAGS) $(DIRECTIVES) -DFP64 -DFBLAS $(LDFLAGS) \


  		





  
  81
  
    
	  	$(SOURCE) -lopenblas -lpthread -o $(SOURCE:.c=)_DP_$@


  		





  
  82
  
    

  		





  
  83
  
    

  		





  73
  84
  
    
gsl: $(SOURCE)


  		





  74
  85
  
    

  		





  75
  
  
    
	$(CC) -I$(GSLINC) $(CFLAGS) $(DIRECTIVES) -DFLOAT -DGSL $(LDFLAGS) \


  		





  
  86
  
    
	$(CC) -I$(GSLINC) $(CFLAGS) $(DIRECTIVES) -DFP32 -DGSL $(LDFLAGS) \


  		





  76
  87
  
    
		$(SOURCE) -lgslcblas -o $(SOURCE:.c=)_SP_$@


  		





  77
  88
  
    

  		





  78
  
  
    
	$(CC) -I$(GSLINC) $(CFLAGS) $(DIRECTIVES) -DDOUBLE -DGSL $(LDFLAGS) \


  		





  
  89
  
    
	$(CC) -I$(GSLINC) $(CFLAGS) $(DIRECTIVES) -DFP64 -DGSL $(LDFLAGS) \


  		





  79
  90
  
    
		$(SOURCE) -lgslcblas -o $(SOURCE:.c=)_DP_$@


  		





  80
  91
  
    

  		





  81
  92
  
    

  		





  82
  93
  
    
cublas: $(SOURCE)


  		





  83
  94
  
    

  		





  84
  
  
    
	$(CC) -I$(CUDAINC) -L$(CUDALIB) $(CFLAGS) -DCUBLAS -DFLOAT $(LDFLAGS) \


  		





  
  95
  
    
	$(CC) -I$(CUDAINC) -L$(CUDALIB) $(CFLAGS) -DCUBLAS -DFP32 $(LDFLAGS) \


  		





  85
  96
  
    
		$(DIRECTIVES) $(SOURCE) -lcublas -o $(SOURCE:.c=)_SP_$@


  		





  86
  97
  
    

  		





  87
  
  
    
	$(CC) -I$(CUDAINC) -L$(CUDALIB) $(CFLAGS) -DCUBLAS -DDOUBLE $(LDFLAGS) \


  		





  
  98
  
    
	$(CC) -I$(CUDAINC) -L$(CUDALIB) $(CFLAGS) -DCUBLAS -DFP64 $(LDFLAGS) \


  		





  88
  99
  
    
		$(DIRECTIVES) $(SOURCE) -lcublas -o $(SOURCE:.c=)_DP_$@


  		





  89
  100
  
    

  		





  90
  101
  
    
thunking: $(SOURCE)


  		





  ......




  96
  107
  
    
	$(CC) -DCUBLAS_GFORTRAN -I$(CUDAINC) $(CFLAGS) -c $(THUNKING)


  		





  97
  108
  
    

  		





  98
  109
  
    
	$(CC) -DCUBLAS_GFORTRAN -I$(CUDAINC) -L$(CUDALIB) $(CFLAGS) -DTHUNKING \


  		





  99
  
  
    
		$(LDFLAGS) $(DIRECTIVES) $(SOURCE) -DFLOAT \


  		





  
  110
  
    
		$(LDFLAGS) $(DIRECTIVES) $(SOURCE) -DFP32 \


  		





  100
  111
  
    
		$(THUNKING:.c=.o) -lcublas -o $(SOURCE:.c=)_SP_$@


  		





  101
  112
  
    

  		





  102
  113
  
    
	$(CC) -DCUBLAS_GFORTRAN -I$(CUDAINC) -L$(CUDALIB) $(CFLAGS) -DTHUNKING \


  		





  103
  
  
    
		$(LDFLAGS) $(DIRECTIVES) $(SOURCE) -DDOUBLE \


  		





  
  114
  
    
		$(LDFLAGS) $(DIRECTIVES) $(SOURCE) -DFP64 \


  		





  104
  115
  
    
		$(THUNKING:.c=.o) -lcublas -o $(SOURCE:.c=)_DP_$@


  		





  105
  116
  
    

  		





  106
  117
  
    
clean: $(SOURCE)


  		












  

    
      BLAS/xTRSV/xTRSV.c	(revision 251)
    
  






  25
  25
  
    
#include "fortran_common.h"


  		





  26
  26
  
    
#include "fortran_thunking.h"


  		





  27
  27
  
    
#elif FBLAS


  		





  28
  
  
    
#include <cblas.h>


  		





  29
  
  
    
#include <cblas_f77.h>


  		





  
  28
  
    
#include <f77blas.h>


  		





  30
  29
  
    
#elif GSL


  		





  31
  30
  
    
#include <gsl_cblas.h>


  		





  32
  31
  
    
#elif ACML


  		





  33
  32
  
    
#include <acml.h>


  		





  34
  33
  
    
#else


  		





  35
  34
  
    
#include <cblas.h>


  		





  36
  
  
    
#include <blaswrap.h>


  		





  
  35
  
    
// #include <blaswrap.h>


  		





  37
  36
  
    
#endif


  		





  38
  37
  
    

  		





  39
  
  
    
#ifdef DOUBLE


  		





  
  38
  
    
#ifdef FP64


  		





  40
  39
  
    
#define LENGTH double


  		





  41
  40
  
    
#else


  		





  42
  41
  
    
#define LENGTH float


  		





  ......




  264
  263
  
    

  		





  265
  264
  
    
  for (i=0;i<RUNS;i++)


  		





  266
  265
  
    
    {


  		





  267
  
  
    
#ifdef DOUBLE


  		





  
  266
  
    
#ifdef FP64


  		





  268
  267
  
    

  		





  269
  268
  
    
      printVectorGPU(dim,devPtrX,"X","Roots");


  		





  270
  269
  
    

  		





  ......




  346
  345
  
    

  		





  347
  346
  
    
  for (i=0;i<RUNS;i++)


  		





  348
  347
  
    
    {


  		





  349
  
  
    
#ifdef DOUBLE


  		





  
  348
  
    
#ifdef FP64


  		





  350
  349
  
    
      


  		





  351
  350
  
    
      printVector(dim,X,"X","Roots");


  		





  352
  351
  
    
      


  		





  ......




  412
  411
  
    
  


  		





  413
  412
  
    
  for (i=0;i<RUNS;i++)


  		





  414
  413
  
    
    {


  		





  415
  
  
    
#ifdef DOUBLE


  		





  
  414
  
    
#ifdef FP64


  		





  416
  415
  
    
      


  		





  417
  416
  
    
      printVector(dim,X,"X","Initial roots");


  		





  418
  417
  
    
      


  		





  
  418
  
    
      /* /\* Multiply A by X as Y <- A.X *\/ */


  		





  
  419
  
    
      /* dgemv_(&trans,&dim,&dim,&alpha,A,&dim,X,&incx,&beta,Y,&incx); */


  		





  
  420
  
    
      


  		





  
  421
  
    
      /* printVector(dim,Y,"Y<-A.X","Estimated results"); */


  		





  
  422
  
    
      


  		





  
  423
  
    
      /* /\* Solve linear system *\/ */


  		





  
  424
  
    
      /* dtrsv_(&uplo,&trans,&diag,&dim,A,&dim,Y,&incx); */


  		





  
  425
  
    
      


  		





  
  426
  
    
      /* printVector(dim,Y,"X","Solutions from A.X=Y"); */


  		





  
  427
  
    
      


  		





  
  428
  
    
      /* /\* Compare the roots X and Y *\/ */


  		





  
  429
  
    
      /* daxpy_(&dim,&beta2,Y,&incx,X,&incx); */


  		





  
  430
  
    
      


  		





  
  431
  
    
      /* printVector(dim,X,"X","Differences initial and estimated"); */


  		





  
  432
  
    
      


  		





  
  433
  
    
      /* /\* Store the checker of errors *\/ */


  		





  
  434
  
    
      /* dnrm2_(&dim,X,&incx,&checksA[i]); */


  		





  
  435
  
    
            


  		





  
  436
  
    
      /* /\* Swap vector X and Y *\/ */


  		





  
  437
  
    
      /* dswap_(&dim,X,&incx,Y,&incx); */


  		





  
  438
  
    

  		





  419
  439
  
    
      /* Multiply A by X as Y <- A.X */


  		





  420
  440
  
    
      dgemv_(&trans,&dim,&dim,&alpha,A,&dim,X,&incx,&beta,Y,&incx);


  		





  421
  441
  
    
      


  		





  ......




  432
  452
  
    
      printVector(dim,X,"X","Differences initial and estimated");


  		





  433
  453
  
    
      


  		





  434
  454
  
    
      /* Store the checker of errors */


  		





  435
  
  
    
      dnrm2_(&dim,X,&incx,&checksA[i]);


  		





  
  455
  
    
      checksA[i]=(double)dnrm2_(&dim,X,&incx);


  		





  436
  456
  
    
            


  		





  437
  457
  
    
      /* Swap vector X and Y */


  		





  438
  458
  
    
      dswap_(&dim,X,&incx,Y,&incx);


  		





  ......




  457
  477
  
    
      printVector(dim,X,"X","Errors");


  		





  458
  478
  
    

  		





  459
  479
  
    
      /* Store the checker of errors */


  		





  460
  
  
    
      snrm2_(&dim,X,&incx,&checksA[i]);


  		





  
  480
  
    
      checksA[i]=(LENGTH)snrm2_(&dim,X,&incx);


  		





  461
  481
  
    

  		





  462
  482
  
    
      /* Swap vector X and Y */


  		





  463
  483
  
    
      sswap_(&dim,X,&incx,Y,&incx);


  		





  ......




  475
  495
  
    
  


  		





  476
  496
  
    
  for (i=0;i<RUNS;i++)


  		





  477
  497
  
    
    {


  		





  478
  
  
    
#ifdef DOUBLE


  		





  
  498
  
    
#ifdef FP64


  		





  479
  499
  
    
      


  		





  480
  500
  
    
      printVector(dim,X,"X","Roots");


  		





  481
  501
  
    
      


  		





  ......




  542
  562
  
    
  for (i=0;i<RUNS;i++)


  		





  543
  563
  
    
    {  


  		





  544
  564
  
    

  		





  545
  
  
    
#ifdef DOUBLE


  		





  
  565
  
    
#ifdef FP64


  		





  546
  566
  
    

  		





  547
  567
  
    
      printVector(dim,X,"X","Roots");


  		





  548
  568
  
    

  		





  ......




  609
  629
  
    
  for (i=0;i<RUNS;i++)


  		





  610
  630
  
    
    {  


  		





  611
  631
  
    

  		





  612
  
  
    
#ifdef DOUBLE


  		





  
  632
  
    
#ifdef FP64


  		





  613
  633
  
    

  		





  614
  634
  
    
      printVector(dim,X,"X","Roots");


  		





  615
  635
  
    

  		












Formats disponibles :
  Unified diff