fanasina/y_project
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

add functions calculate parallel updates neurons, and debug some leak functions

Browse Source
This commit is contained in:
fanasina
2024-02-28 09:57:21 +01:00
parent 2fc2fe477f
commit 924ff3a7dd
11 changed files with 707 additions and 127 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tensor_t/src/tensor_t/cl_tensor_t.c
+4 -4
View File
@@ -181,7 +181,7 @@
void cl_tensorProd_##type(tensor_##type **MM, tensor_##type *M0, tensor_##type *M1) { \
dimension *dd; \
add_dimension(&dd, M0->dim, M1->dim); \
(*MM)=CREATE_TENSOR_##type(dd); \
_RECREATE_TENSOR_IF_NOT_THE_SAME_DIM_OR_NULL_##type(MM,dd); \
tensor_##type *M = *MM; \
char *file_cl_src = "../src/kernel_ProdTensor.cl"; \
char *func_cl_nameEndian = "prodTensorLin_" #type; \
@@ -235,7 +235,7 @@ void cl_tensorContractnProd_##type(tensor_##type** MM, tensor_##type *M0, tensor
dimension *dd;\
add_dimension(&dd, dSub0, dSub1);\
updateRankDim(dd);\
*MM = CREATE_TENSOR_##type(dd);\
_RECREATE_TENSOR_IF_NOT_THE_SAME_DIM_OR_NULL_##type(MM,dd);\
tensor_##type *M= *MM;\
char *file_cl_src = "../src/kernel_ProdContractnTensor.cl"; \
/*char *func_cl_name = "prodContractnTensorLin_" #type;*/ \
@@ -272,7 +272,7 @@ void cl_tensorContractnProd_##type(tensor_##type** MM, tensor_##type *M0, tensor
void cl2d_tensorProd_##type(tensor_##type **MM, tensor_##type *M0, tensor_##type *M1, size_t div0Wsz, size_t div1Wsz) { \
dimension *dd; \
add_dimension(&dd, M0->dim, M1->dim); \
(*MM)=CREATE_TENSOR_##type(dd); \
_RECREATE_TENSOR_IF_NOT_THE_SAME_DIM_OR_NULL_##type(MM,dd); \
tensor_##type *M = *MM; \
char *file_cl_src = "../src/kernel_2d_ProdTensor.cl"; \
/*char *func_cl_name = "prodTensor2dLin_" #type;*/ \
@@ -324,7 +324,7 @@ void cl2d_tensorContractnProd_##type(tensor_##type **MM, tensor_##type *M0, tens
dimension *dd;\
add_dimension(&dd, dSub0, dSub1);\
updateRankDim(dd);\
*MM = CREATE_TENSOR_##type(dd);\
_RECREATE_TENSOR_IF_NOT_THE_SAME_DIM_OR_NULL_##type(MM,dd);\
tensor_##type *M= *MM;\
char *file_cl_src = "../src/kernel_2d_ProdContractnTensor.cl"; \
char *func_cl_nameEndian = "prodContractnTensor2dLin_" #type; \
tensor_t/src/tensor_t/tensor_t.c
+256 -10
View File
@@ -61,6 +61,16 @@ long int decr(long int i) { return i - 1; }
return r_tens;\
}\
\
void _RECREATE_TENSOR_IF_NOT_THE_SAME_DIM_OR_NULL_##type(tensor_##type **M, dimension *dd){\
if(*M){ \
if(!is_equal_dim((*M)->dim, dd)){\
free_tensor_##type(*M);\
(*M)=CREATE_TENSOR_##type(dd);\
}else free_dimension(dd); /* because it is not used */\
}else{\
(*M)=CREATE_TENSOR_##type(dd);\
}\
}\
tensor_##type* init_tensor_head_##type(tensor_##type *troot ,dimension *dim){\
tensor_##type *r_tens=malloc(sizeof(tensor_##type));\
updateRankDim(dim);\
@@ -107,6 +117,18 @@ tensor_##type* init_copy_tensor_head_##type(tensor_##type *troot ,dimension *dim
r_tens->x = malloc(sizeof(type)*dim->rank);\
return r_tens;\
}\
\
tensor_##type* CLONE_TENSOR_##type(tensor_##type *tens){\
if(tens){\
tensor_##type *r_tens=malloc(sizeof(tensor_##type));\
r_tens->dim = clone_dim(tens->dim);\
r_tens->x = malloc(sizeof(type) * (tens->dim)->rank);\
for(size_t i=0; i<(tens->dim)->rank;++i)\
r_tens->x[i]=tens->x[i];\
return r_tens;\
}\
return NULL;\
}\
\
void free_tensor_##type(tensor_##type * tens){\
if(tens){\
@@ -486,7 +508,7 @@ void split_copy_tensor_##type(tensor_##type *Troot, tensor_##type **Tpart1, tens
void tensorProdNotOpt_##type(tensor_##type **MM, tensor_##type *M0, tensor_##type *M1) { \
dimension *dd; \
add_dimension(&dd, M0->dim, M1->dim); \
(*MM)=CREATE_TENSOR_##type(dd); \
_RECREATE_TENSOR_IF_NOT_THE_SAME_DIM_OR_NULL_##type(MM,dd); \
tensor_##type *M = *MM; \
size_t* coord; \
coord = malloc(sizeof(size_t)*(dd->size)); \
@@ -512,7 +534,7 @@ void tensorProdNotOpt_##type(tensor_##type **MM, tensor_##type *M0, tensor_##typ
void tensorProd_##type(tensor_##type **MM, tensor_##type *M0, tensor_##type *M1) { \
dimension *dd; \
add_dimension(&dd, M0->dim, M1->dim); \
(*MM)=CREATE_TENSOR_##type(dd); \
_RECREATE_TENSOR_IF_NOT_THE_SAME_DIM_OR_NULL_##type(MM,dd); \
tensor_##type *M = *MM; \
size_t m_idx;\
for(size_t i=0; i<M0->dim->rank; ++i){\
@@ -567,7 +589,7 @@ void tensorContractnProd_##type(tensor_##type** MM, tensor_##type *M0, tensor_##
add_dimension(&dd, dSub0, dSub1);\
/*printDebug_dimension(dd,"dd");*/\
updateRankDim(dd);\
*MM = CREATE_TENSOR_##type(dd);\
_RECREATE_TENSOR_IF_NOT_THE_SAME_DIM_OR_NULL_##type(MM,dd);\
tensor_##type *M= *MM;\
\
\
@@ -625,7 +647,7 @@ void* runProd_thread_##type(void *arg){\
void tensorProdThread_##type(tensor_##type **MM, tensor_##type *M0, tensor_##type *M1, size_t nbthread) { \
dimension *dd; \
add_dimension(&dd, M0->dim, M1->dim); \
(*MM)=CREATE_TENSOR_##type(dd); \
_RECREATE_TENSOR_IF_NOT_THE_SAME_DIM_OR_NULL_##type(MM,dd); \
tensor_##type *M = *MM; \
\
\
@@ -687,7 +709,7 @@ void* runProd_thread2d_##type(void *arg){\
void tensorProdThrea2d_##type(tensor_##type **MM, tensor_##type *M0, tensor_##type *M1, size_t nbthread) { \
dimension *dd; \
add_dimension(&dd, M0->dim, M1->dim); \
(*MM)=CREATE_TENSOR_##type(dd); \
_RECREATE_TENSOR_IF_NOT_THE_SAME_DIM_OR_NULL_##type(MM,dd); \
tensor_##type *M = *MM; \
\
\
@@ -779,7 +801,7 @@ void tensorContractnProdThread_##type(tensor_##type** MM, tensor_##type *M0, ten
dimension *dd;\
add_dimension(&dd, dSub0, dSub1);\
updateRankDim(dd);\
*MM = CREATE_TENSOR_##type(dd);\
_RECREATE_TENSOR_IF_NOT_THE_SAME_DIM_OR_NULL_##type(MM,dd);\
tensor_##type *M= *MM;\
\
\
@@ -876,7 +898,7 @@ void tensorContractnPro2dThread_##type(tensor_##type** MM, tensor_##type *M0, te
dimension *dd;\
add_dimension(&dd, dSub0, dSub1);\
updateRankDim(dd);\
*MM = CREATE_TENSOR_##type(dd);\
_RECREATE_TENSOR_IF_NOT_THE_SAME_DIM_OR_NULL_##type(MM,dd);\
tensor_##type *M= *MM;\
\
\
@@ -942,7 +964,7 @@ void tensorContractnProdNotOpt_##type(tensor_##type** MM, tensor_##type *M0, ten
add_dimension(&dd, dSub0, dSub1);\
/*printDebug_dimension(dd,"dd");*/\
updateRankDim(dd);\
*MM = CREATE_TENSOR_##type(dd);\
_RECREATE_TENSOR_IF_NOT_THE_SAME_DIM_OR_NULL_##type(MM,dd);\
tensor_##type *M= *MM;\
\
size_t* coord;\
@@ -1202,7 +1224,6 @@ void parseInputOutput_withDim_to_tensors_##type(tensor_##type **Tpart1, tensor_#
ttmp=ppEnd;\
}\
\
/*tens = create_tensor_from_list_array_##type(l_a,dim);*/\
*Tpart1 = create_tensor_from_list_array_##type(l_a1,ddim1);\
*Tpart2 = create_tensor_from_list_array_##type(l_a2,ddim2);\
free_array_chainlist_##type(l_a1);\
@@ -1396,7 +1417,7 @@ void parse_file_InputOutput_withDim_to_tensors_##type(tensor_##type **Tpart1, te
fclose(f_input);\
}\
\
tensor_##type ** formInput_to_array_tensor_##type(tensor_##type *tens){\
tensor_##type ** fromInput_to_array_tensor_##type(tensor_##type *tens){\
tensor_##type **re_tens=malloc((tens->dim)->perm[0]*sizeof(tensor_##type *));\
dimension *dim=create_dim((tens->dim)->size - 1);\
for(size_t i=0; i<dim->size; ++i) dim->perm[i]=(tens->dim)->perm[i+1];\
@@ -1498,6 +1519,231 @@ tensor_##type * permute_notOpt_tensor_##type(tensor_##type *org, dimension *dper
free(coord_tr);\
return tens_tr;\
}\
struct arg_1Update_##type{\
type *M0x;\
size_t beginRange;\
size_t endRange;\
type (*func)(type);\
};\
void* run1UpdatCalcfunc_thread_##type(void *arg){\
struct arg_1Update_##type *arg_t = arg;\
for (size_t i = arg_t->beginRange; i < arg_t->endRange; i++) {\
arg_t->M0x[i] = arg_t->func(arg_t->M0x[i]);\
}\
}\
\
void update_1tensor_func_##type(tensor_##type *M0, type (*func)(type), size_t nbthread){\
\
pthread_t *thrd = malloc(nbthread * sizeof(pthread_t));\
struct arg_1Update_##type **arg_th = malloc( nbthread * sizeof(struct arg_1Update_##type *));\
\
for(size_t i = 0; i < nbthread; ++i){\
arg_th[i]=malloc(sizeof(struct arg_1Update_##type));\
arg_th[i]->M0x=M0->x;\
arg_th[i]->func=func;\
arg_th[i]->beginRange = i*(M0->dim->rank)/nbthread ;\
arg_th[i]->endRange = (i+1)*(M0->dim->rank)/nbthread ;\
\
pthread_create(&thrd[i], NULL, run1UpdatCalcfunc_thread_##type, (void*)arg_th[i]);\
}\
\
for(size_t i=0; i< nbthread; ++i){\
pthread_join(thrd[i], NULL);\
free(arg_th[i]);\
}\
\
free(thrd);\
free(arg_th);\
} \
\
struct arg_2Update_##type{\
type *M0x;\
type *M1x;\
size_t beginRange;\
size_t endRange;\
type (*func)(type);\
};\
void* run2UpdatCalcfunc_thread_##type(void *arg){\
struct arg_2Update_##type *arg_t = arg;\
for (size_t i = arg_t->beginRange; i < arg_t->endRange; i++) {\
arg_t->M0x[i] = arg_t->func(arg_t->M1x[i]);\
}\
}\
\
void update_2tensor_func_##type(tensor_##type *M0, tensor_##type *M1, type (*func)(type), size_t nbthread){\
if ( is_equal_dim(M0->dim,M1->dim)){ \
pthread_t *thrd = malloc(nbthread * sizeof(pthread_t));\
struct arg_2Update_##type **arg_th = malloc( nbthread * sizeof(struct arg_2Update_##type *));\
\
for(size_t i = 0; i < nbthread; ++i){\
arg_th[i]=malloc(sizeof(struct arg_2Update_##type));\
arg_th[i]->M0x=M0->x;\
arg_th[i]->M1x=M1->x;\
arg_th[i]->func=func;\
arg_th[i]->beginRange = i*(M0->dim->rank)/nbthread ;\
arg_th[i]->endRange = (i+1)*(M0->dim->rank)/nbthread ;\
\
pthread_create(&thrd[i], NULL, run2UpdatCalcfunc_thread_##type, (void*)arg_th[i]);\
}\
\
for(size_t i=0; i< nbthread; ++i){\
pthread_join(thrd[i], NULL);\
free(arg_th[i]);\
}\
\
free(thrd);\
free(arg_th);\
}\
} \
\
struct arg_3Update_##type{\
type *M0x;\
type *M1x;\
type *M2x;\
size_t beginRange;\
size_t endRange;\
type (*func)(type, type);\
};\
void* run3UpdatCalcfunc_thread_##type(void *arg){\
struct arg_3Update_##type *arg_t = arg;\
for (size_t i = arg_t->beginRange; i < arg_t->endRange; i++) {\
arg_t->M0x[i] = arg_t->func(arg_t->M1x[i], arg_t->M2x[i]);\
}\
}\
\
void update_3tensor_func_##type(tensor_##type *M0, tensor_##type *M1, tensor_##type *M2, type (*func)(type,type), size_t nbthread){\
if ( is_equal_dim(M0->dim,M1->dim) && (is_equal_dim(M0->dim, M2->dim))){ \
pthread_t *thrd = malloc(nbthread * sizeof(pthread_t));\
struct arg_3Update_##type **arg_th = malloc( nbthread * sizeof(struct arg_3Update_##type *));\
\
for(size_t i = 0; i < nbthread; ++i){\
arg_th[i]=malloc(sizeof(struct arg_3Update_##type));\
arg_th[i]->M0x=M0->x;\
arg_th[i]->M1x=M1->x;\
arg_th[i]->M2x=M2->x;\
arg_th[i]->func=func;\
arg_th[i]->beginRange = i*(M0->dim->rank)/nbthread ;\
arg_th[i]->endRange = (i+1)*(M0->dim->rank)/nbthread ;\
\
pthread_create(&thrd[i], NULL, run3UpdatCalcfunc_thread_##type, (void*)arg_th[i]);\
}\
\
for(size_t i=0; i< nbthread; ++i){\
pthread_join(thrd[i], NULL);\
free(arg_th[i]);\
}\
\
free(thrd);\
free(arg_th);\
}\
} \
\
\
struct arg_4Update_##type{\
type *M0x;\
type *M1x;\
type *M2x;\
size_t beginRange;\
size_t endRange;\
type (*func)(type, type, type(*f1)(type));\
type(*f1)(type);\
};\
void* run4UpdatCalcfunc_thread_##type(void *arg){\
struct arg_4Update_##type *arg_t = arg;\
for (size_t i = arg_t->beginRange; i < arg_t->endRange; i++) {\
arg_t->M0x[i] = arg_t->func(arg_t->M1x[i], arg_t->M2x[i], arg_t->f1);\
}\
}\
\
void update_4tensor_func_##type(tensor_##type *M0, tensor_##type *M1, tensor_##type *M2, \
type (*func)(type, type, type(*f1)(type)),\
type(*f1)(type),\
size_t nbthread){\
/*printf(" rankM0=%ld , rank M2:%ld ; iseq :%d \n",(M0->dim)->rank,(M2->dim)->rank,is_equal_dim(M0->dim,M2->dim) );\
*/\
/* printDebug_dimension(M0->dim," dim M0 in update4 "); \
printDebug_dimension(M2->dim," dim M2 in update4 "); \
*/if ( is_equal_dim(M0->dim, M1->dim) /*&& (is_equal_dim(M0->dim, M2->dim))*/){ \
/*printDebug_dimension(M0->dim," dim M0 in update4 "); \
*/pthread_t *thrd = malloc(nbthread * sizeof(pthread_t));\
struct arg_4Update_##type **arg_th = malloc( nbthread * sizeof(struct arg_4Update_##type *));\
\
for(size_t i = 0; i < nbthread; ++i){\
arg_th[i]=malloc(sizeof(struct arg_4Update_##type));\
arg_th[i]->M0x=M0->x;\
arg_th[i]->M1x=M1->x;\
arg_th[i]->M2x=M2->x;\
arg_th[i]->func=func;\
arg_th[i]->f1=f1;\
arg_th[i]->beginRange = i*(M0->dim->rank)/nbthread ;\
arg_th[i]->endRange = (i+1)*(M0->dim->rank)/nbthread ;\
\
pthread_create(&thrd[i], NULL, run4UpdatCalcfunc_thread_##type, (void*)arg_th[i]);\
}\
\
for(size_t i=0; i< nbthread; ++i){\
pthread_join(thrd[i], NULL);\
free(arg_th[i]);\
}\
\
free(thrd);\
free(arg_th);\
}\
} \
\
struct arg_5Update_##type{\
type *M0x;\
type *M1x;\
type *M2x;\
type *M3x;\
size_t beginRange;\
size_t endRange;\
type (*func)(type, type, type, type(*f1)(type), type (*f2)(type,type) );\
type(*f1)(type);\
type (*f2)(type,type);\
};\
void* run5UpdatCalcfunc_thread_##type(void *arg){\
struct arg_5Update_##type *arg_t = arg;\
for (size_t i = arg_t->beginRange; i < arg_t->endRange; i++) {\
arg_t->M0x[i] = arg_t->func(arg_t->M1x[i], arg_t->M2x[i], arg_t->M3x[i], arg_t->f1, arg_t->f2);\
}\
}\
\
void update_5tensor_func_##type(tensor_##type *M0, tensor_##type *M1, tensor_##type *M2, tensor_##type *M3 , \
type (*func) (type, type, type, type(*f1)(type), type (*f2)(type,type)), \
type(*f1)(type), \
type (*f2)(type,type), \
size_t nbthread){\
if ( is_equal_dim(M0->dim,M1->dim) && (is_equal_dim(M0->dim, M2->dim))&& (is_equal_dim(M0->dim, M3->dim))){ \
pthread_t *thrd = malloc(nbthread * sizeof(pthread_t));\
struct arg_5Update_##type **arg_th = malloc( nbthread * sizeof(struct arg_5Update_##type *));\
/*printDebug_dimension(M0->dim," dim M0 in update5 "); */ \
for(size_t i = 0; i < nbthread; ++i){\
arg_th[i]=malloc(sizeof(struct arg_5Update_##type));\
arg_th[i]->M0x=M0->x;\
arg_th[i]->M1x=M1->x;\
arg_th[i]->M2x=M2->x;\
arg_th[i]->M3x=M3->x;\
arg_th[i]->func=func;\
arg_th[i]->f1=f1;\
arg_th[i]->f2=f2;\
arg_th[i]->beginRange = i*(M0->dim->rank)/nbthread ;\
arg_th[i]->endRange = (i+1)*(M0->dim->rank)/nbthread ;\
\
pthread_create(&thrd[i], NULL, run5UpdatCalcfunc_thread_##type, (void*)arg_th[i]);\
}\
\
for(size_t i=0; i< nbthread; ++i){\
pthread_join(thrd[i], NULL);\
free(arg_th[i]);\
}\
\
free(thrd);\
free(arg_th);\
}\
} \
\
GEN_FUNC_TENSOR(TYPE_FLOAT);
tensor_t/src/tensor_t/tensor_t.h
+19 -1
View File
@@ -17,6 +17,8 @@ struct tensor_##type{\
typedef struct tensor_##type tensor_##type;\
tensor_##type * CREATE_TENSOR_##type(dimension *dim); \
tensor_##type* CREATE_TENSOR_FROM_CPY_DIM_##type(dimension *dim);\
void _RECREATE_TENSOR_IF_NOT_THE_SAME_DIM_OR_NULL_##type(tensor_##type **M, dimension *dd);\
tensor_##type* CLONE_TENSOR_##type(tensor_##type *tens);\
void free_tensor_##type(tensor_##type * tens); \
tensor_##type * sub_minus_tensor_head_##type(tensor_##type *rootens, size_t minuSubdim, size_t rankInDim); \
tensor_##type * sub_minus_tensor_tail_##type(tensor_##type *rootens, size_t minuSubdim, size_t rankInDim); \
@@ -42,7 +44,7 @@ void init_random_x_##type(tensor_##type *M, type minR, type maxR, int randomRan
tensor_##type * parseInput_withDim_to_tensor_##type(char *input);\
void parseInputOutput_withDim_to_tensors_##type(tensor_##type **Tpart1, tensor_##type **Tpart2, char *input, size_t pivotSplit);\
void parse_file_InputOutput_withDim_to_tensors_##type(tensor_##type **Tpart1, tensor_##type **Tpart2, char *file_name_input, size_t pivotSplit);\
tensor_##type ** formInput_to_array_tensor_##type(tensor_##type *tens);\
tensor_##type ** fromInput_to_array_tensor_##type(tensor_##type *tens);\
struct array_chainlist_##type{\
size_t index;\
type x;\
@@ -54,6 +56,22 @@ tensor_##type * create_tensor_from_list_array_##type( array_chainlist_##type *l_
void free_array_chainlist_##type(array_chainlist_##type *l_a);\
tensor_##type * transpose_notOpt_tensor_##type(tensor_##type *org);\
tensor_##type * permute_notOpt_tensor_##type(tensor_##type *org, dimension *dperm);\
void update_1tensor_func_##type(tensor_##type *M0, \
type (*func)(type), size_t nbthread);\
void update_2tensor_func_##type(tensor_##type *M0, tensor_##type *M1, \
type (*func)(type), size_t nbthread);\
void update_3tensor_func_##type(tensor_##type *M0, tensor_##type *M1, tensor_##type *M2, \
type (*func)(type, type), size_t nbthread);\
void update_4tensor_func_##type(tensor_##type *M0, tensor_##type *M1, tensor_##type *M2, \
type (*func)(type, type, type(*f1)(type)),\
type(*f1)(type),\
size_t nbthread);\
void update_5tensor_func_##type(tensor_##type *M0, tensor_##type *M1, tensor_##type *M2, tensor_##type *M3 , \
type (*func) (type, type, type, type(*f1)(type), type (*f2)(type,type)), \
type(*f1)(type), \
type (*f2)(type,type), \
size_t nbthread);\
GENERATE_TENSOR_TYPE(TYPE_FLOAT);
tensor_t/test/is_good.c
+64 -39
View File
@@ -544,8 +544,8 @@ TEST(tensorProd ){
print_tensor_float(M1,"M1");
tensor_TYPE_FLOAT *M;
tensor_TYPE_FLOAT *Mn;
tensor_TYPE_FLOAT *M=NULL;
tensor_TYPE_FLOAT *Mn=NULL;
tensorProd_TYPE_FLOAT(&M,M0,M1);
tensorProdNotOpt_TYPE_FLOAT(&Mn,M0,M1);
@@ -603,8 +603,8 @@ TEST(tensorContractnProd_TYPE_FLOAT ){
print_tensor_float(M0,"M0");
print_tensor_float(M1,"M1");
tensor_TYPE_FLOAT *M;
tensor_TYPE_FLOAT *MnO;
tensor_TYPE_FLOAT *M=NULL;
tensor_TYPE_FLOAT *MnO=NULL;
tensorContractnProd_TYPE_FLOAT(&M, M0,M1,2);
tensorContractnProdNotOpt_TYPE_FLOAT(&MnO, M0,M1,2);
@@ -664,8 +664,8 @@ TEST(tensorContractnProd_TYPE_FLOAT2 ){
// print_tensor_float(M0,"M0");
// print_tensor_float(M1,"M1");
tensor_TYPE_FLOAT *M;
tensor_TYPE_FLOAT *MnO;
tensor_TYPE_FLOAT *M=NULL;
tensor_TYPE_FLOAT *MnO=NULL;
tensorContractnProd_TYPE_FLOAT(&M, M0,M1,2);
// print_tensor_float(M,"M");
@@ -717,8 +717,8 @@ TEST(tensorContractnProd_TYPE_DOUBLE_2_1 ){
print_tensor_double(M0,"M0");
print_tensor_double(M1,"M1");
tensor_TYPE_DOUBLE *M;
tensor_TYPE_DOUBLE *MnO;
tensor_TYPE_DOUBLE *M=NULL;
tensor_TYPE_DOUBLE *MnO=NULL;
tensorContractnProd_TYPE_DOUBLE(&M, M0,M1,1);
//print_tensor_double(M,"M");
@@ -774,8 +774,8 @@ TEST(tensorContractnProd_TYPE_DOUBLE_2_2 ){
print_tensor_double(M0,"M0");
print_tensor_double(M1,"M1");
tensor_TYPE_DOUBLE *M;
tensor_TYPE_DOUBLE *MnO;
tensor_TYPE_DOUBLE *M=NULL;
tensor_TYPE_DOUBLE *MnO=NULL;
tensorContractnProd_TYPE_DOUBLE(&M, M0,M1,1);
//print_tensor_double(M,"M");
@@ -836,8 +836,8 @@ TEST(tensorContractnProd_TYPE_DOUBLE2 ){
//print_tensor_double(M0,"M0");
//print_tensor_double(M1,"M1");
tensor_TYPE_DOUBLE *M;
tensor_TYPE_DOUBLE *MnO;
tensor_TYPE_DOUBLE *M=NULL;
tensor_TYPE_DOUBLE *MnO=NULL;
tensorContractnProd_TYPE_DOUBLE(&M, M0,M1,2);
//print_tensor_double(M,"M");
@@ -897,8 +897,8 @@ TEST(VStensorContractnProd_TYPE_DOUBLE2 ){
//print_tensor_double(M0,"M0");
//print_tensor_double(M1,"M1");
tensor_TYPE_DOUBLE *M;
tensor_TYPE_DOUBLE *MnO;
tensor_TYPE_DOUBLE *M=NULL;
tensor_TYPE_DOUBLE *MnO=NULL;
tensorContractnProd_TYPE_DOUBLE(&M, M0,M1,2);
//print_tensor_double(M,"M");
@@ -959,8 +959,8 @@ TEST(Pthread_tensorContractnPro2d_TYPE_DOUBLE2 ){
//print_tensor_double(M0,"M0");
//print_tensor_double(M1,"M1");
tensor_TYPE_DOUBLE *M;
tensor_TYPE_DOUBLE *MnO;
tensor_TYPE_DOUBLE *M=NULL;
tensor_TYPE_DOUBLE *MnO=NULL;
size_t nbthread = 5;
@@ -1017,8 +1017,8 @@ TEST(contract_dim1){
print_tensor_double(M0,"M0");
print_tensor_double(M1,"M1");
tensor_TYPE_DOUBLE *M;
tensor_TYPE_DOUBLE *MnO;
tensor_TYPE_DOUBLE *M=NULL;
tensor_TYPE_DOUBLE *MnO=NULL;
size_t nbthread = 5;
@@ -1085,8 +1085,8 @@ TEST(Pthread_tensorContractnProd_TYPE_DOUBLE2 ){
//print_tensor_double(M0,"M0");
//print_tensor_double(M1,"M1");
tensor_TYPE_DOUBLE *M;
tensor_TYPE_DOUBLE *MnO;
tensor_TYPE_DOUBLE *M=NULL;
tensor_TYPE_DOUBLE *MnO=NULL;
size_t nbthread = 5;
@@ -1146,8 +1146,8 @@ TEST(tensorProd_vs ){
for(size_t i=0; i<M1->dim->rank;++i) M1->x[i]=i*0.003 + 2;
tensor_TYPE_FLOAT *M;
tensor_TYPE_FLOAT *Mn;
tensor_TYPE_FLOAT *M=NULL;
tensor_TYPE_FLOAT *Mn=NULL;
tensorProd_TYPE_FLOAT(&M,M0,M1);
//tensorProdNotOpt_TYPE_FLOAT(&Mn,M0,M1);
@@ -1201,8 +1201,8 @@ TEST(tensorProd_vsThread ){
tensor_TYPE_FLOAT *M;
tensor_TYPE_FLOAT *Mn;
tensor_TYPE_FLOAT *M=NULL;
tensor_TYPE_FLOAT *Mn=NULL;
size_t nbthread = 5;
@@ -1258,8 +1258,8 @@ TEST(tensorProd_vsThread2d ){
for(size_t i=0; i<M1->dim->rank;++i) M1->x[i]=i*0.003 + 2;
tensor_TYPE_FLOAT *M;
tensor_TYPE_FLOAT *Mn;
tensor_TYPE_FLOAT *M=NULL;
tensor_TYPE_FLOAT *Mn=NULL;
size_t nbthread = 5;
@@ -1379,7 +1379,7 @@ TEST(parseInputOutput_unknownpart_to_tensor){
"((0,0,0,1,2)"\
"(0,0,0,2,4)) ";
tensor_TYPE_FLOAT *t0,*t1;
tensor_TYPE_FLOAT *t0=NULL,*t1=NULL;
parseInputOutput_withDim_to_tensors_TYPE_FLOAT(&t0,&t1 , input, 1);
print_tensor_msg_TYPE_FLOAT(t0," t0 from input" );
@@ -1401,7 +1401,7 @@ TEST(parseInputOutput_knownpart_to_tensor){
"((0,0,0,1,2)"\
"(0,0,0,2,4)) ";
tensor_TYPE_FLOAT *t0,*t1;
tensor_TYPE_FLOAT *t0=NULL,*t1=NULL;
parseInputOutput_withDim_to_tensors_TYPE_FLOAT(&t0,&t1 , input, 1);
print_tensor_msg_TYPE_FLOAT(t0," t0 from input" );
@@ -1423,7 +1423,7 @@ TEST(parseInputOutput_unknownpart2dimInput_to_tensor){
"((0,0,0),(8,8,8),1,2)"\
"(0,0,0),(8,8,8),2,4)) ";
tensor_TYPE_FLOAT *t0,*t1;
tensor_TYPE_FLOAT *t0=NULL,*t1=NULL;
parseInputOutput_withDim_to_tensors_TYPE_FLOAT(&t0,&t1 , input, 1);
print_tensor_msg_TYPE_FLOAT(t0," t0 from input" );
@@ -1445,7 +1445,7 @@ TEST(parseInputOutput_knownpart2dimInput_to_tensor){
"((0,0,0),(8,8,8),1,2)"\
"(0,0,0),(8,8,8),2,4)) ";
tensor_TYPE_FLOAT *t0,*t1;
tensor_TYPE_FLOAT *t0=NULL,*t1=NULL;
parseInputOutput_withDim_to_tensors_TYPE_FLOAT(&t0,&t1 , input, 1);
print_tensor_msg_TYPE_FLOAT(t0," t0 from input" );
@@ -1468,7 +1468,7 @@ TEST(parseInputOutput_unknownpart1dimInput_2output_to_tensor){
"((0,0,0),(8,8,8)8,1,2)"\
"(0,0,0),(8,8,8)8,2,4)) ";
tensor_TYPE_FLOAT *t0,*t1;
tensor_TYPE_FLOAT *t0=NULL,*t1=NULL;
parseInputOutput_withDim_to_tensors_TYPE_FLOAT(&t0,&t1 , input, 2);
print_tensor_msg_TYPE_FLOAT(t0," t0 from input" );
@@ -1491,7 +1491,7 @@ TEST(parseInputOutput_knownpart1dimInput_2output_to_tensor){
"((0,0,0),(8,8,8)8,1,2)"\
"(0,0,0),(8,8,8)8,2,4)) ";
tensor_TYPE_FLOAT *t0,*t1;
tensor_TYPE_FLOAT *t0=NULL,*t1=NULL;
parseInputOutput_withDim_to_tensors_TYPE_FLOAT(&t0,&t1 , input, 2);
print_tensor_msg_TYPE_FLOAT(t0," t0 from input" );
@@ -1514,7 +1514,7 @@ TEST(parseInputOutput_unknownpart1dimInput_1output_to_tensor){
"((0,0,0),(8,8,8)8,1,2)"\
"(0,0,0),(8,8,8)8,2,4)) ";
tensor_TYPE_FLOAT *t0,*t1;
tensor_TYPE_FLOAT *t0=NULL,*t1=NULL;
parseInputOutput_withDim_to_tensors_TYPE_FLOAT(&t0,&t1 , input, 1);
print_tensor_msg_TYPE_FLOAT(t0," t0 from input" );
@@ -1537,7 +1537,7 @@ TEST(parseInputOutput_knownpart1dimInput_1output_to_tensor){
"((0,0,0),(8,8,8)8,1,5)"\
"(0,0,0),(8,8,8)8,2,4)) ";
tensor_TYPE_FLOAT *t0,*t1;
tensor_TYPE_FLOAT *t0=NULL,*t1=NULL;
parseInputOutput_withDim_to_tensors_TYPE_FLOAT(&t0,&t1 , input, 1);
print_tensor_msg_TYPE_FLOAT(t0," t0 from input" );
@@ -1553,7 +1553,7 @@ TEST(parseInputOutput_file_knownpart1dimInput_1output_to_tensor){
endian=true;
char *inputfile="input.txt";
tensor_TYPE_FLOAT *t0,*t1;
tensor_TYPE_FLOAT *t0=NULL,*t1=NULL;
parse_file_InputOutput_withDim_to_tensors_TYPE_FLOAT(&t0,&t1 , inputfile, 1);
print_tensor_msg_TYPE_FLOAT(t0," t0 from inputfile" );
@@ -1571,7 +1571,7 @@ TEST(parseInputOutput_file_knownpart1dimInput_1output_to_tensor){
endian=true;
char *inputfile="unkinput.txt";
tensor_TYPE_FLOAT *t0,*t1;
tensor_TYPE_FLOAT *t0=NULL,*t1=NULL;
parse_file_InputOutput_withDim_to_tensors_TYPE_FLOAT(&t0,&t1 , inputfile, 1);
print_tensor_msg_TYPE_FLOAT(t0," t0 from inputfile" );
@@ -1589,14 +1589,14 @@ TEST(array_from_parseInputOutput_file_knownpart1dimInput_1output_to_tensor){
endian=true;
char *inputfile="unkinput.txt";
tensor_TYPE_FLOAT *t0,*t1;
tensor_TYPE_FLOAT *t0=NULL,*t1=NULL;
parse_file_InputOutput_withDim_to_tensors_TYPE_FLOAT(&t0,&t1 , inputfile, 1);
print_tensor_msg_TYPE_FLOAT(t0," t0 from inputfile" );
print_tensor_msg_TYPE_FLOAT(t1," t1 from inputfile" );
tensor_TYPE_FLOAT **arrt0 = formInput_to_array_tensor_TYPE_FLOAT(t0);
tensor_TYPE_FLOAT **arrt1 = formInput_to_array_tensor_TYPE_FLOAT(t1);
tensor_TYPE_FLOAT **arrt0 = fromInput_to_array_tensor_TYPE_FLOAT(t0);
tensor_TYPE_FLOAT **arrt1 = fromInput_to_array_tensor_TYPE_FLOAT(t1);
size_t sz0=(t0->dim)->perm[0];
@@ -1622,8 +1622,33 @@ TEST(array_from_parseInputOutput_file_knownpart1dimInput_1output_to_tensor){
free_tensor_TYPE_FLOAT(t1);
}
float func2(float x){
return x*x+1;
}
TEST(update_func_){
dimension *d0=create_dim(3);
d0->perm[0]=2;
d0->perm[1]=3;
d0->perm[2]=4;
updateRankDim(d0);
tensor_TYPE_FLOAT *M0 = CREATE_TENSOR_TYPE_FLOAT(d0);
LOG("M0->dim->rank = %ld\n",M0->dim->rank);
init_random_x_TYPE_FLOAT(M0,2.7,5.4,50001);
print_tensor_float(M0, "init M0 random");
update_1tensor_func_TYPE_FLOAT(M0, func2, 5);
print_tensor_float(M0, "x*x+1 M0 random");
}
tensor_t/test_cl/is_good.c
+32 -32
View File
@@ -86,8 +86,8 @@ TEST(tensorProd ){
print_tensor_float(M1,"M1");
tensor_TYPE_FLOAT *M;
tensor_TYPE_FLOAT *Mn;
tensor_TYPE_FLOAT *M=NULL;
tensor_TYPE_FLOAT *Mn=NULL;
tensorProd_TYPE_FLOAT(&M,M0,M1);
tensorProdNotOpt_TYPE_FLOAT(&Mn,M0,M1);
@@ -131,8 +131,8 @@ TEST(tensorContractnProd_TYPE_FLOAT ){
print_tensor_float(M0,"M0");
print_tensor_float(M1,"M1");
tensor_TYPE_FLOAT *M;
tensor_TYPE_FLOAT *MnO;
tensor_TYPE_FLOAT *M=NULL;
tensor_TYPE_FLOAT *MnO=NULL;
tensorContractnProd_TYPE_FLOAT(&M, M0,M1,1);
tensorContractnProdNotOpt_TYPE_FLOAT(&MnO, M0,M1,1);
@@ -180,8 +180,8 @@ TEST(tensorContractnProd_TYPE_FLOAT2 ){
// print_tensor_float(M0,"M0");
// print_tensor_float(M1,"M1");
tensor_TYPE_FLOAT *M;
tensor_TYPE_FLOAT *MnO;
tensor_TYPE_FLOAT *M=NULL;
tensor_TYPE_FLOAT *MnO=NULL;
tensorContractnProd_TYPE_FLOAT(&M, M0,M1,2);
// print_tensor_float(M,"M");
@@ -231,8 +231,8 @@ TEST(cl_tensorContractnProd_TYPE_FLOAT2 ){
// print_tensor_float(M0,"M0");
// print_tensor_float(M1,"M1");
tensor_TYPE_FLOAT *M;
tensor_TYPE_FLOAT *MnO;
tensor_TYPE_FLOAT *M=NULL;
tensor_TYPE_FLOAT *MnO=NULL;
tensorContractnProdNotOpt_TYPE_FLOAT(&M, M0,M1,2);
// print_tensor_float(M,"M");
@@ -280,8 +280,8 @@ TEST(cl_tensorContractnProd_TYPE_DOUBLE2 ){
//print_tensor_double(M0,"M0");
//print_tensor_double(M1,"M1");
tensor_TYPE_DOUBLE *M;
tensor_TYPE_DOUBLE *MnO;
tensor_TYPE_DOUBLE *M=NULL;
tensor_TYPE_DOUBLE *MnO=NULL;
tensorContractnProdNotOpt_TYPE_DOUBLE(&M, M0,M1,2);
//tensorContractnProd_TYPE_DOUBLE(&M, M0,M1,2);
@@ -341,8 +341,8 @@ TEST(tensorContractnProd_TYPE_DOUBLE2 ){
//print_tensor_double(M0,"M0");
//print_tensor_double(M1,"M1");
tensor_TYPE_DOUBLE *M;
tensor_TYPE_DOUBLE *MnO;
tensor_TYPE_DOUBLE *M=NULL;
tensor_TYPE_DOUBLE *MnO=NULL;
tensorContractnProd_TYPE_DOUBLE(&M, M0,M1,2);
//print_tensor_double(M,"M");
@@ -392,8 +392,8 @@ TEST(TensorProdCL){
tensor_TYPE_FLOAT *M;
tensor_TYPE_FLOAT *Mn;
tensor_TYPE_FLOAT *M=NULL;
tensor_TYPE_FLOAT *Mn=NULL;
tensorProd_TYPE_FLOAT(&M,M0,M1);
cl_tensorProd_TYPE_FLOAT(&Mn,M0,M1);
@@ -452,8 +452,8 @@ TEST(VS_thrd_tensorContractnProd_TYPE_DOUBLE2 ){
//print_tensor_double(M0,"M0");
//print_tensor_double(M1,"M1");
tensor_TYPE_DOUBLE *M;
tensor_TYPE_DOUBLE *MnO;
tensor_TYPE_DOUBLE *M=NULL;
tensor_TYPE_DOUBLE *MnO=NULL;
size_t nbth=10;
@@ -523,8 +523,8 @@ TEST(VS_thrd_tensorContractnProd_TYPE_DOUBLE2 ){
//print_tensor_double(M0,"M0");
//print_tensor_double(M1,"M1");
tensor_TYPE_DOUBLE *M;
tensor_TYPE_DOUBLE *MnO;
tensor_TYPE_DOUBLE *M=NULL;
tensor_TYPE_DOUBLE *MnO=NULL;
size_t nbth=15;
@@ -598,8 +598,8 @@ TEST(VS_thrd_tensorContractnProd_TYPE_DOUBLE2 ){
//print_tensor_double(M0,"M0");
//print_tensor_double(M1,"M1");
tensor_TYPE_DOUBLE *M;
tensor_TYPE_DOUBLE *MnO;
tensor_TYPE_DOUBLE *M=NULL;
tensor_TYPE_DOUBLE *MnO=NULL;
size_t nbth=10;
@@ -675,8 +675,8 @@ TEST(VScltensorContractnProd_TYPE_DOUBLE2 ){
//print_tensor_double(M0,"M0");
//print_tensor_double(M1,"M1");
tensor_TYPE_DOUBLE *M;
tensor_TYPE_DOUBLE *MnO;
tensor_TYPE_DOUBLE *M=NULL;
tensor_TYPE_DOUBLE *MnO=NULL;
size_t nbth = 10;
@@ -741,8 +741,8 @@ TEST(VScl2dtensorContractnProd_TYPE_DOUBLE2 ){
//print_tensor_double(M0,"M0");
//print_tensor_double(M1,"M1");
tensor_TYPE_DOUBLE *M;
tensor_TYPE_DOUBLE *MnO;
tensor_TYPE_DOUBLE *M=NULL;
tensor_TYPE_DOUBLE *MnO=NULL;
size_t nbth = 10;
@@ -799,8 +799,8 @@ TEST(tensorProd_vs2d ){
print_tensor_float(M1,"M1");*/
tensor_TYPE_FLOAT *M;
tensor_TYPE_FLOAT *Mn;
tensor_TYPE_FLOAT *M=NULL;
tensor_TYPE_FLOAT *Mn=NULL;
cl_tensorProd_TYPE_FLOAT(&M,M0,M1);
tensorProd_TYPE_FLOAT(&Mn,M0,M1);
@@ -847,8 +847,8 @@ TEST(tensorProd_vs2d ){
print_tensor_float(M1,"M1");*/
tensor_TYPE_FLOAT *M;
tensor_TYPE_FLOAT *Mn;
tensor_TYPE_FLOAT *M=NULL;
tensor_TYPE_FLOAT *Mn=NULL;
tensorProd_TYPE_FLOAT(&M,M0,M1);
//cl2d_tensorProd_TYPE_FLOAT(&Mn,M0,M1,24,24);
@@ -896,8 +896,8 @@ TEST(tensorProd_vs2d_Endian ){
print_tensor_float(M1,"M1");*/
tensor_TYPE_FLOAT *M;
tensor_TYPE_FLOAT *Mn;
tensor_TYPE_FLOAT *M=NULL;
tensor_TYPE_FLOAT *Mn=NULL;
cl_tensorProd_TYPE_FLOAT(&M,M0,M1);
tensorProd_TYPE_FLOAT(&Mn,M0,M1);
@@ -942,8 +942,8 @@ TEST(tensorProd_vs2d_Endian ){
tensor_TYPE_FLOAT *M;
tensor_TYPE_FLOAT *Mn;
tensor_TYPE_FLOAT *M=NULL;
tensor_TYPE_FLOAT *Mn=NULL;
tensorProd_TYPE_FLOAT(&M,M0,M1);
//cl2d_tensorProd_TYPE_FLOAT(&Mn,M0,M1,24,24);