Commit 3359c1f1 authored by lisj's avatar lisj
Browse files

增加GKLib

parent f2c80b44
/*!
\file gk_macros.h
\brief This file contains various macros
\date Started 3/27/2007
\author George
\version\verbatim $Id: gk_macros.h 15048 2013-08-31 19:38:14Z karypis $ \endverbatim
*/
#ifndef _GK_MACROS_H_
#define _GK_MACROS_H_
/*-------------------------------------------------------------
* Usefull commands
*-------------------------------------------------------------*/
#define gk_max(a, b) ((a) >= (b) ? (a) : (b))
#define gk_min(a, b) ((a) >= (b) ? (b) : (a))
#define gk_max3(a, b, c) ((a) >= (b) && (a) >= (c) ? (a) : ((b) >= (a) && (b) >= (c) ? (b) : (c)))
#define gk_SWAP(a, b, tmp) do {(tmp) = (a); (a) = (b); (b) = (tmp);} while(0)
#define INC_DEC(a, b, val) do {(a) += (val); (b) -= (val);} while(0)
#define sign(a, b) ((a >= 0 ? b : -b))
#define ONEOVERRANDMAX (1.0/(RAND_MAX+1.0))
#define RandomInRange(u) ((int) (ONEOVERRANDMAX*(u)*rand()))
#define RandomInRange_r(s, u) ((int) (ONEOVERRANDMAX*(u)*rand_r(s)))
#define gk_abs(x) ((x) >= 0 ? (x) : -(x))
/*-------------------------------------------------------------
* Timing macros
*-------------------------------------------------------------*/
#define gk_clearcputimer(tmr) (tmr = 0.0)
#define gk_startcputimer(tmr) (tmr -= gk_CPUSeconds())
#define gk_stopcputimer(tmr) (tmr += gk_CPUSeconds())
#define gk_getcputimer(tmr) (tmr)
#define gk_clearwctimer(tmr) (tmr = 0.0)
#define gk_startwctimer(tmr) (tmr -= gk_WClockSeconds())
#define gk_stopwctimer(tmr) (tmr += gk_WClockSeconds())
#define gk_getwctimer(tmr) (tmr)
/*-------------------------------------------------------------
* dbglvl handling macros
*-------------------------------------------------------------*/
#define IFSET(a, flag, cmd) if ((a)&(flag)) (cmd);
/*-------------------------------------------------------------
* gracefull library exit macro
*-------------------------------------------------------------*/
#define GKSETJMP() (setjmp(gk_return_to_entry))
#define gk_sigcatch() (setjmp(gk_jbufs[gk_cur_jbufs]))
/*-------------------------------------------------------------
* Debuging memory leaks
*-------------------------------------------------------------*/
#ifdef DMALLOC
# define MALLOC_CHECK(ptr) \
if (malloc_verify((ptr)) == DMALLOC_VERIFY_ERROR) { \
printf("***MALLOC_CHECK failed on line %d of file %s: " #ptr "\n", \
__LINE__, __FILE__); \
abort(); \
}
#else
# define MALLOC_CHECK(ptr) ;
#endif
/*-------------------------------------------------------------
* CSR conversion macros
*-------------------------------------------------------------*/
#define MAKECSR(i, n, a) \
do { \
for (i=1; i<n; i++) a[i] += a[i-1]; \
for (i=n; i>0; i--) a[i] = a[i-1]; \
a[0] = 0; \
} while(0)
#define SHIFTCSR(i, n, a) \
do { \
for (i=n; i>0; i--) a[i] = a[i-1]; \
a[0] = 0; \
} while(0)
/*-------------------------------------------------------------
* ASSERTS that cannot be turned off!
*-------------------------------------------------------------*/
#define GKASSERT(expr) \
if (!(expr)) { \
printf("***ASSERTION failed on line %d of file %s: " #expr "\n", \
__LINE__, __FILE__); \
abort(); \
}
#define GKASSERTP(expr,msg) \
if (!(expr)) { \
printf("***ASSERTION failed on line %d of file %s: " #expr "\n", \
__LINE__, __FILE__); \
printf msg ; \
printf("\n"); \
abort(); \
}
#define GKCUASSERT(expr) \
if (!(expr)) { \
printf("***ASSERTION failed on line %d of file %s: " #expr "\n", \
__LINE__, __FILE__); \
}
#define GKWARN(expr) \
if (!(expr)) { \
printf("***ASSERTION failed on line %d of file %s: " #expr "\n", \
__LINE__, __FILE__); \
}
#define GKCUASSERTP(expr,msg) \
if (!(expr)) { \
printf("***ASSERTION failed on line %d of file %s: " #expr "\n", \
__LINE__, __FILE__); \
printf msg ; \
printf("\n"); \
}
#define GKWARNP(expr,msg) \
if (!(expr)) { \
printf("***ASSERTION failed on line %d of file %s: " #expr "\n", \
__LINE__, __FILE__); \
printf msg ; \
printf("\n"); \
}
/*-------------------------------------------------------------
* Program Assertions
*-------------------------------------------------------------*/
#ifndef NDEBUG
# define ASSERT(expr) \
if (!(expr)) { \
printf("***ASSERTION failed on line %d of file %s: " #expr "\n", \
__LINE__, __FILE__); \
assert(expr); \
}
# define ASSERTP(expr,msg) \
if (!(expr)) { \
printf("***ASSERTION failed on line %d of file %s: " #expr "\n", \
__LINE__, __FILE__); \
printf msg ; \
printf("\n"); \
assert(expr); \
}
#else
# define ASSERT(expr) ;
# define ASSERTP(expr,msg) ;
#endif
#ifndef NDEBUG2
# define ASSERT2 ASSERT
# define ASSERTP2 ASSERTP
#else
# define ASSERT2(expr) ;
# define ASSERTP2(expr,msg) ;
#endif
#endif
/*!
\file gk_mkblas.h
\brief Templates for BLAS-like routines
\date Started 3/28/07
\author George
\version\verbatim $Id: gk_mkblas.h 16304 2014-02-25 14:27:19Z karypis $ \endverbatim
*/
#ifndef _GK_MKBLAS_H_
#define _GK_MKBLAS_H_
#define GK_MKBLAS(PRFX, TYPE, OUTTYPE) \
/*************************************************************************/\
/*! The macro for gk_?incset()-class of routines */\
/*************************************************************************/\
TYPE *PRFX ## incset(size_t n, TYPE baseval, TYPE *x)\
{\
size_t i;\
\
for (i=0; i<n; i++)\
x[i] = baseval+i;\
\
return x;\
}\
\
/*************************************************************************/\
/*! The macro for gk_?max()-class of routines */\
/*************************************************************************/\
TYPE PRFX ## max(size_t n, TYPE *x, size_t incx)\
{\
size_t i;\
TYPE max;\
\
if (n <= 0) return (TYPE) 0;\
\
for (max=(*x), x+=incx, i=1; i<n; i++, x+=incx)\
max = ((*x) > max ? (*x) : max);\
\
return max;\
}\
\
\
/*************************************************************************/\
/*! The macro for gk_?min()-class of routines */\
/*************************************************************************/\
TYPE PRFX ## min(size_t n, TYPE *x, size_t incx)\
{\
size_t i;\
TYPE min;\
\
if (n <= 0) return (TYPE) 0;\
\
for (min=(*x), x+=incx, i=1; i<n; i++, x+=incx)\
min = ((*x) < min ? (*x) : min);\
\
return min;\
}\
\
\
/*************************************************************************/\
/*! The macro for gk_?argmax()-class of routines */\
/*************************************************************************/\
size_t PRFX ## argmax(size_t n, TYPE *x, size_t incx)\
{\
size_t i, j, max=0;\
\
for (i=1, j=incx; i<n; i++, j+=incx)\
max = (x[j] > x[max] ? j : max);\
\
return (size_t)(max/incx);\
}\
\
\
/*************************************************************************/\
/*! The macro for gk_?argmin()-class of routines */\
/*************************************************************************/\
size_t PRFX ## argmin(size_t n, TYPE *x, size_t incx)\
{\
size_t i, j, min=0;\
\
for (i=1, j=incx; i<n; i++, j+=incx)\
min = (x[j] < x[min] ? j : min);\
\
return (size_t)(min/incx);\
}\
\
\
/*************************************************************************/\
/*! The macro for gk_?argmax_n()-class of routines */\
/*************************************************************************/\
size_t PRFX ## argmax_n(size_t n, TYPE *x, size_t incx, size_t k)\
{\
size_t i, j, max_n;\
PRFX ## kv_t *cand;\
\
cand = PRFX ## kvmalloc(n, "GK_ARGMAX_N: cand");\
\
for (i=0, j=0; i<n; i++, j+=incx) {\
cand[i].val = i;\
cand[i].key = x[j];\
}\
PRFX ## kvsortd(n, cand);\
\
max_n = cand[k-1].val;\
\
gk_free((void *)&cand, LTERM);\
\
return max_n;\
}\
\
\
/*************************************************************************/\
/*! The macro for gk_?sum()-class of routines */\
/**************************************************************************/\
OUTTYPE PRFX ## sum(size_t n, TYPE *x, size_t incx)\
{\
size_t i;\
OUTTYPE sum = 0;\
\
for (i=0; i<n; i++, x+=incx)\
sum += (*x);\
\
return sum;\
}\
\
\
/*************************************************************************/\
/*! The macro for gk_?scale()-class of routines */\
/**************************************************************************/\
TYPE *PRFX ## scale(size_t n, TYPE alpha, TYPE *x, size_t incx)\
{\
size_t i;\
\
for (i=0; i<n; i++, x+=incx)\
(*x) *= alpha;\
\
return x;\
}\
\
\
/*************************************************************************/\
/*! The macro for gk_?norm2()-class of routines */\
/**************************************************************************/\
OUTTYPE PRFX ## norm2(size_t n, TYPE *x, size_t incx)\
{\
size_t i;\
OUTTYPE partial = 0;\
\
for (i=0; i<n; i++, x+=incx)\
partial += (*x) * (*x);\
\
return (partial > 0 ? (OUTTYPE)sqrt((double)partial) : (OUTTYPE)0);\
}\
\
\
/*************************************************************************/\
/*! The macro for gk_?dot()-class of routines */\
/**************************************************************************/\
OUTTYPE PRFX ## dot(size_t n, TYPE *x, size_t incx, TYPE *y, size_t incy)\
{\
size_t i;\
OUTTYPE partial = 0.0;\
\
for (i=0; i<n; i++, x+=incx, y+=incy)\
partial += (*x) * (*y);\
\
return partial;\
}\
\
\
/*************************************************************************/\
/*! The macro for gk_?axpy()-class of routines */\
/**************************************************************************/\
TYPE *PRFX ## axpy(size_t n, TYPE alpha, TYPE *x, size_t incx, TYPE *y, size_t incy)\
{\
size_t i;\
TYPE *y_in = y;\
\
for (i=0; i<n; i++, x+=incx, y+=incy)\
*y += alpha*(*x);\
\
return y_in;\
}\
#define GK_MKBLAS_PROTO(PRFX, TYPE, OUTTYPE) \
TYPE *PRFX ## incset(size_t n, TYPE baseval, TYPE *x);\
TYPE PRFX ## max(size_t n, TYPE *x, size_t incx);\
TYPE PRFX ## min(size_t n, TYPE *x, size_t incx);\
size_t PRFX ## argmax(size_t n, TYPE *x, size_t incx);\
size_t PRFX ## argmin(size_t n, TYPE *x, size_t incx);\
size_t PRFX ## argmax_n(size_t n, TYPE *x, size_t incx, size_t k);\
OUTTYPE PRFX ## sum(size_t n, TYPE *x, size_t incx);\
TYPE *PRFX ## scale(size_t n, TYPE alpha, TYPE *x, size_t incx);\
OUTTYPE PRFX ## norm2(size_t n, TYPE *x, size_t incx);\
OUTTYPE PRFX ## dot(size_t n, TYPE *x, size_t incx, TYPE *y, size_t incy);\
TYPE *PRFX ## axpy(size_t n, TYPE alpha, TYPE *x, size_t incx, TYPE *y, size_t incy);\
#endif
/*!
\file gk_mkmemory.h
\brief Templates for memory allocation routines
\date Started 3/29/07
\author George
\version\verbatim $Id: gk_mkmemory.h 10711 2011-08-31 22:23:04Z karypis $ \endverbatim
*/
#ifndef _GK_MKMEMORY_H_
#define _GK_MKMEMORY_H_
#define GK_MKALLOC(PRFX, TYPE)\
/*************************************************************************/\
/*! The macro for gk_?malloc()-class of routines */\
/**************************************************************************/\
TYPE *PRFX ## malloc(size_t n, char *msg)\
{\
return (TYPE *)gk_malloc(sizeof(TYPE)*n, msg);\
}\
\
\
/*************************************************************************/\
/*! The macro for gk_?realloc()-class of routines */\
/**************************************************************************/\
TYPE *PRFX ## realloc(TYPE *ptr, size_t n, char *msg)\
{\
return (TYPE *)gk_realloc((void *)ptr, sizeof(TYPE)*n, msg);\
}\
\
\
/*************************************************************************/\
/*! The macro for gk_?smalloc()-class of routines */\
/**************************************************************************/\
TYPE *PRFX ## smalloc(size_t n, TYPE ival, char *msg)\
{\
TYPE *ptr;\
\
ptr = (TYPE *)gk_malloc(sizeof(TYPE)*n, msg);\
if (ptr == NULL) \
return NULL; \
\
return PRFX ## set(n, ival, ptr); \
}\
\
\
/*************************************************************************/\
/*! The macro for gk_?set()-class of routines */\
/*************************************************************************/\
TYPE *PRFX ## set(size_t n, TYPE val, TYPE *x)\
{\
size_t i;\
\
for (i=0; i<n; i++)\
x[i] = val;\
\
return x;\
}\
\
\
/*************************************************************************/\
/*! The macro for gk_?set()-class of routines */\
/*************************************************************************/\
TYPE *PRFX ## copy(size_t n, TYPE *a, TYPE *b)\
{\
return (TYPE *)memmove((void *)b, (void *)a, sizeof(TYPE)*n);\
}\
\
\
/*************************************************************************/\
/*! The macro for gk_?AllocMatrix()-class of routines */\
/**************************************************************************/\
TYPE **PRFX ## AllocMatrix(size_t ndim1, size_t ndim2, TYPE value, char *errmsg)\
{\
gk_idx_t i, j;\
TYPE **matrix;\
\
matrix = (TYPE **)gk_malloc(ndim1*sizeof(TYPE *), errmsg);\
if (matrix == NULL) \
return NULL;\
\
for (i=0; i<ndim1; i++) { \
matrix[i] = PRFX ## smalloc(ndim2, value, errmsg);\
if (matrix[i] == NULL) { \
for (j=0; j<i; j++) \
gk_free((void **)&matrix[j], LTERM); \
return NULL; \
} \
}\
\
return matrix;\
}\
\
\
/*************************************************************************/\
/*! The macro for gk_?AllocMatrix()-class of routines */\
/**************************************************************************/\
void PRFX ## FreeMatrix(TYPE ***r_matrix, size_t ndim1, size_t ndim2)\
{\
gk_idx_t i;\
TYPE **matrix;\
\
if (*r_matrix == NULL) \
return; \
\
matrix = *r_matrix;\
\
for (i=0; i<ndim1; i++) \
gk_free((void **)&(matrix[i]), LTERM);\
\
gk_free((void **)r_matrix, LTERM);\
}\
\
\
/*************************************************************************/\
/*! The macro for gk_?SetMatrix()-class of routines */\
/**************************************************************************/\
void PRFX ## SetMatrix(TYPE **matrix, size_t ndim1, size_t ndim2, TYPE value)\
{\
gk_idx_t i, j;\
\
for (i=0; i<ndim1; i++) {\
for (j=0; j<ndim2; j++)\
matrix[i][j] = value;\
}\
}\
#define GK_MKALLOC_PROTO(PRFX, TYPE)\
TYPE *PRFX ## malloc(size_t n, char *msg);\
TYPE *PRFX ## realloc(TYPE *ptr, size_t n, char *msg);\
TYPE *PRFX ## smalloc(size_t n, TYPE ival, char *msg);\
TYPE *PRFX ## set(size_t n, TYPE val, TYPE *x);\
TYPE *PRFX ## copy(size_t n, TYPE *a, TYPE *b);\
TYPE **PRFX ## AllocMatrix(size_t ndim1, size_t ndim2, TYPE value, char *errmsg);\
void PRFX ## FreeMatrix(TYPE ***r_matrix, size_t ndim1, size_t ndim2);\
void PRFX ## SetMatrix(TYPE **matrix, size_t ndim1, size_t ndim2, TYPE value);\
#endif
/*!
\file gk_mkpqueue.h
\brief Templates for priority queues
\date Started 4/09/07
\author George
\version\verbatim $Id: gk_mkpqueue.h 21742 2018-01-26 16:59:15Z karypis $ \endverbatim
*/
#ifndef _GK_MKPQUEUE_H
#define _GK_MKPQUEUE_H
#define GK_MKPQUEUE(FPRFX, PQT, KVT, KT, VT, KVMALLOC, KMAX, KEY_LT)\
/*************************************************************************/\
/*! This function creates and initializes a priority queue */\
/**************************************************************************/\
PQT *FPRFX ## Create(size_t maxnodes)\
{\
PQT *queue; \
\
queue = (PQT *)gk_malloc(sizeof(PQT), "gk_pqCreate: queue");\
FPRFX ## Init(queue, maxnodes);\
\
return queue;\
}\
\
\
/*************************************************************************/\
/*! This function initializes the data structures of the priority queue */\
/**************************************************************************/\
void FPRFX ## Init(PQT *queue, size_t maxnodes)\
{\
queue->nnodes = 0;\
queue->maxnodes = maxnodes;\
\
queue->heap = KVMALLOC(maxnodes, "gk_PQInit: heap");\
queue->locator = gk_idxsmalloc(maxnodes, -1, "gk_PQInit: locator");\
}\
\
\
/*************************************************************************/\
/*! This function resets the priority queue */\
/**************************************************************************/\
void FPRFX ## Reset(PQT *queue)\
{\
ssize_t i;\
ssize_t *locator=queue->locator;\
KVT *heap=queue->heap;\
\
for (i=queue->nnodes-1; i>=0; i--)\
locator[heap[i].val] = -1;\
queue->nnodes = 0;\
}\
\
\
/*************************************************************************/\
/*! This function frees the internal datastructures of the priority queue */\
/**************************************************************************/\
void FPRFX ## Free(PQT *queue)\
{\
if (queue == NULL) return;\
gk_free((void **)&queue->heap, &queue->locator, LTERM);\
queue->maxnodes = 0;\
}\
\
\
/*************************************************************************/\
/*! This function frees the internal datastructures of the priority queue \
and the queue itself */\
/**************************************************************************/\
void FPRFX ## Destroy(PQT *queue)\
{\
if (queue == NULL) return;\
FPRFX ## Free(queue);\
gk_free((void **)&queue, LTERM);\
}\
\
\
/*************************************************************************/\
/*! This function returns the length of the queue */\
/**************************************************************************/\
size_t FPRFX ## Length(PQT *queue)\
{\
return queue->nnodes;\
}\
\
\
/*************************************************************************/\
/*! This function adds an item in the priority queue */\
/**************************************************************************/\
int FPRFX ## Insert(PQT *queue, VT node, KT key)\
{\
ssize_t i, j;\
ssize_t *locator=queue->locator;\
KVT *heap=queue->heap;\
\
ASSERT2(FPRFX ## CheckHeap(queue));\
\
ASSERT(locator[node] == -1);\
\
i = queue->nnodes++;\
while (i > 0) {\
j = (i-1)>>1;\
if (KEY_LT(key, heap[j].key)) {\
heap[i] = heap[j];\
locator[heap[i].val] = i;\
i = j;\
}\
else\
break;\
}\
ASSERT(i >= 0);\
heap[i].key = key;\
heap[i].val = node;\
locator[node] = i;\
\
ASSERT2(FPRFX ## CheckHeap(queue));\
\
return 0;\
}\
\
\
/*************************************************************************/\
/*! This function deletes an item from the priority queue */\
/**************************************************************************/\
int FPRFX ## Delete(PQT *queue, VT node)\
{\
ssize_t i, j;\
size_t nnodes;\
KT newkey, oldkey;\
ssize_t *locator=queue->locator;\
KVT *heap=queue->heap;\
\
ASSERT(locator[node] != -1);\
ASSERT(heap[locator[node]].val == node);\
\
ASSERT2(FPRFX ## CheckHeap(queue));\
\
i = locator[node];\
locator[node] = -1;\
\
if (--queue->nnodes > 0 && heap[queue->nnodes].val != node) {\
node = heap[queue->nnodes].val;\
newkey = heap[queue->nnodes].key;\
oldkey = heap[i].key;\
\
if (KEY_LT(newkey, oldkey)) { /* Filter-up */\
while (i > 0) {\
j = (i-1)>>1;\
if (KEY_LT(newkey, heap[j].key)) {\
heap[i] = heap[j];\
locator[heap[i].val] = i;\
i = j;\
}\
else\
break;\
}\
}\
else { /* Filter down */\
nnodes = queue->nnodes;\
while ((j=(i<<1)+1) < nnodes) {\
if (KEY_LT(heap[j].key, newkey)) {\
if (j+1 < nnodes && KEY_LT(heap[j+1].key, heap[j].key))\
j++;\
heap[i] = heap[j];\
locator[heap[i].val] = i;\
i = j;\
}\
else if (j+1 < nnodes && KEY_LT(heap[j+1].key, newkey)) {\
j++;\
heap[i] = heap[j];\
locator[heap[i].val] = i;\
i = j;\
}\
else\
break;\
}\
}\
\
heap[i].key = newkey;\
heap[i].val = node;\
locator[node] = i;\
}\
\
ASSERT2(FPRFX ## CheckHeap(queue));\
\
return 0;\
}\
\
\
/*************************************************************************/\
/*! This function updates the key values associated for a particular item */ \
/**************************************************************************/\
void FPRFX ## Update(PQT *queue, VT node, KT newkey)\
{\
ssize_t i, j;\
size_t nnodes;\
KT oldkey;\
ssize_t *locator=queue->locator;\
KVT *heap=queue->heap;\
\
oldkey = heap[locator[node]].key;\
if (!KEY_LT(newkey, oldkey) && !KEY_LT(oldkey, newkey)) return;\
\
ASSERT(locator[node] != -1);\
ASSERT(heap[locator[node]].val == node);\
ASSERT2(FPRFX ## CheckHeap(queue));\
\
i = locator[node];\
\
if (KEY_LT(newkey, oldkey)) { /* Filter-up */\
while (i > 0) {\
j = (i-1)>>1;\
if (KEY_LT(newkey, heap[j].key)) {\
heap[i] = heap[j];\
locator[heap[i].val] = i;\
i = j;\
}\
else\
break;\
}\
}\
else { /* Filter down */\
nnodes = queue->nnodes;\
while ((j=(i<<1)+1) < nnodes) {\
if (KEY_LT(heap[j].key, newkey)) {\
if (j+1 < nnodes && KEY_LT(heap[j+1].key, heap[j].key))\
j++;\
heap[i] = heap[j];\
locator[heap[i].val] = i;\
i = j;\
}\
else if (j+1 < nnodes && KEY_LT(heap[j+1].key, newkey)) {\
j++;\
heap[i] = heap[j];\
locator[heap[i].val] = i;\
i = j;\
}\
else\
break;\
}\
}\
\
heap[i].key = newkey;\
heap[i].val = node;\
locator[node] = i;\
\
ASSERT2(FPRFX ## CheckHeap(queue));\
\
return;\
}\
\
\
/*************************************************************************/\
/*! This function returns the item at the top of the queue and removes\
it from the priority queue */\
/**************************************************************************/\
VT FPRFX ## GetTop(PQT *queue)\
{\
ssize_t i, j;\
ssize_t *locator;\
KVT *heap;\
VT vtx, node;\
KT key;\
\
ASSERT2(FPRFX ## CheckHeap(queue));\
\
if (queue->nnodes == 0)\
return -1;\
\
queue->nnodes--;\
\
heap = queue->heap;\
locator = queue->locator;\
\
vtx = heap[0].val;\
locator[vtx] = -1;\
\
if ((i = queue->nnodes) > 0) {\
key = heap[i].key;\
node = heap[i].val;\
i = 0;\
while ((j=2*i+1) < queue->nnodes) {\
if (KEY_LT(heap[j].key, key)) {\
if (j+1 < queue->nnodes && KEY_LT(heap[j+1].key, heap[j].key))\
j = j+1;\
heap[i] = heap[j];\
locator[heap[i].val] = i;\
i = j;\
}\
else if (j+1 < queue->nnodes && KEY_LT(heap[j+1].key, key)) {\
j = j+1;\
heap[i] = heap[j];\
locator[heap[i].val] = i;\
i = j;\
}\
else\
break;\
}\
\
heap[i].key = key;\
heap[i].val = node;\
locator[node] = i;\
}\
\
ASSERT2(FPRFX ## CheckHeap(queue));\
return vtx;\
}\
\
\
/*************************************************************************/\
/*! This function returns the item at the top of the queue. The item is not\
deleted from the queue. */\
/**************************************************************************/\
VT FPRFX ## SeeTopVal(PQT *queue)\
{\
return (queue->nnodes == 0 ? -1 : queue->heap[0].val);\
}\
\
\
/*************************************************************************/\
/*! This function returns the key of the top item. The item is not\
deleted from the queue. */\
/**************************************************************************/\
KT FPRFX ## SeeTopKey(PQT *queue)\
{\
return (queue->nnodes == 0 ? KMAX : queue->heap[0].key);\
}\
\
\
/*************************************************************************/\
/*! This function returns the key of a specific item */\
/**************************************************************************/\
KT FPRFX ## SeeKey(PQT *queue, VT node)\
{\
ssize_t *locator;\
KVT *heap;\
\
heap = queue->heap;\
locator = queue->locator;\
\
return heap[locator[node]].key;\
}\
\
\
/*************************************************************************/\
/*! This function returns the first item in a breadth-first traversal of\
the heap whose key is less than maxwgt. This function is here due to\
hMETIS and is not general!*/\
/**************************************************************************/\
/*\
VT FPRFX ## SeeConstraintTop(PQT *queue, KT maxwgt, KT *wgts)\
{\
ssize_t i;\
\
if (queue->nnodes == 0)\
return -1;\
\
if (maxwgt <= 1000)\
return FPRFX ## SeeTopVal(queue);\
\
for (i=0; i<queue->nnodes; i++) {\
if (queue->heap[i].key > 0) {\
if (wgts[queue->heap[i].val] <= maxwgt)\
return queue->heap[i].val;\
}\
else {\
if (queue->heap[i/2].key <= 0)\
break;\
}\
}\
\
return queue->heap[0].val;\
\
}\
*/\
\
\
/*************************************************************************/\
/*! This functions checks the consistency of the heap */\
/**************************************************************************/\
int FPRFX ## CheckHeap(PQT *queue)\
{\
ssize_t i, j;\
size_t nnodes;\
ssize_t *locator;\
KVT *heap;\
\
heap = queue->heap;\
locator = queue->locator;\
nnodes = queue->nnodes;\
\
if (nnodes == 0)\
return 1;\
\
ASSERT(locator[heap[0].val] == 0);\
for (i=1; i<nnodes; i++) {\
ASSERT(locator[heap[i].val] == i);\
ASSERT(!KEY_LT(heap[i].key, heap[(i-1)/2].key));\
}\
for (i=1; i<nnodes; i++)\
ASSERT(!KEY_LT(heap[i].key, heap[0].key));\
\
for (j=i=0; i<queue->maxnodes; i++) {\
if (locator[i] != -1)\
j++;\
}\
ASSERTP(j == nnodes, ("%jd %jd\n", (intmax_t)j, (intmax_t)nnodes));\
\
return 1;\
}\
#define GK_MKPQUEUE_PROTO(FPRFX, PQT, KT, VT)\
PQT * FPRFX ## Create(size_t maxnodes);\
void FPRFX ## Init(PQT *queue, size_t maxnodes);\
void FPRFX ## Reset(PQT *queue);\
void FPRFX ## Free(PQT *queue);\
void FPRFX ## Destroy(PQT *queue);\
size_t FPRFX ## Length(PQT *queue);\
int FPRFX ## Insert(PQT *queue, VT node, KT key);\
int FPRFX ## Delete(PQT *queue, VT node);\
void FPRFX ## Update(PQT *queue, VT node, KT newkey);\
VT FPRFX ## GetTop(PQT *queue);\
VT FPRFX ## SeeTopVal(PQT *queue);\
KT FPRFX ## SeeTopKey(PQT *queue);\
KT FPRFX ## SeeKey(PQT *queue, VT node);\
VT FPRFX ## SeeConstraintTop(PQT *queue, KT maxwgt, KT *wgts);\
int FPRFX ## CheckHeap(PQT *queue);\
/* This is how these macros are used
GK_MKPQUEUE(gk_dkvPQ, gk_dkvPQ_t, double, gk_idx_t, gk_dkvmalloc, DBL_MAX)
GK_MKPQUEUE_PROTO(gk_dkvPQ, gk_dkvPQ_t, double, gk_idx_t)
*/
#endif
/*!
\file gk_mkpqueue2.h
\brief Templates for priority queues that do not utilize locators and as such
they can use different types of values.
\date Started 4/09/07
\author George
\version\verbatim $Id: gk_mkpqueue2.h 13005 2012-10-23 22:34:36Z karypis $ \endverbatim
*/
#ifndef _GK_MKPQUEUE2_H
#define _GK_MKPQUEUE2_H
#define GK_MKPQUEUE2(FPRFX, PQT, KT, VT, KMALLOC, VMALLOC, KMAX, KEY_LT)\
/*************************************************************************/\
/*! This function creates and initializes a priority queue */\
/**************************************************************************/\
PQT *FPRFX ## Create2(ssize_t maxnodes)\
{\
PQT *queue; \
\
if ((queue = (PQT *)gk_malloc(sizeof(PQT), "gk_pqCreate2: queue")) != NULL) {\
memset(queue, 0, sizeof(PQT));\
queue->nnodes = 0;\
queue->maxnodes = maxnodes;\
queue->keys = KMALLOC(maxnodes, "gk_pqCreate2: keys");\
queue->vals = VMALLOC(maxnodes, "gk_pqCreate2: vals");\
\
if (queue->keys == NULL || queue->vals == NULL)\
gk_free((void **)&queue->keys, &queue->vals, &queue, LTERM);\
}\
\
return queue;\
}\
\
\
/*************************************************************************/\
/*! This function resets the priority queue */\
/**************************************************************************/\
void FPRFX ## Reset2(PQT *queue)\
{\
queue->nnodes = 0;\
}\
\
\
/*************************************************************************/\
/*! This function frees the internal datastructures of the priority queue */\
/**************************************************************************/\
void FPRFX ## Destroy2(PQT **r_queue)\
{\
PQT *queue = *r_queue; \
if (queue == NULL) return;\
gk_free((void **)&queue->keys, &queue->vals, &queue, LTERM);\
*r_queue = NULL;\
}\
\
\
/*************************************************************************/\
/*! This function returns the length of the queue */\
/**************************************************************************/\
size_t FPRFX ## Length2(PQT *queue)\
{\
return queue->nnodes;\
}\
\
\
/*************************************************************************/\
/*! This function adds an item in the priority queue. */\
/**************************************************************************/\
int FPRFX ## Insert2(PQT *queue, VT val, KT key)\
{\
ssize_t i, j;\
KT *keys=queue->keys;\
VT *vals=queue->vals;\
\
ASSERT2(FPRFX ## CheckHeap2(queue));\
\
if (queue->nnodes == queue->maxnodes) \
return 0;\
\
ASSERT2(FPRFX ## CheckHeap2(queue));\
\
i = queue->nnodes++;\
while (i > 0) {\
j = (i-1)>>1;\
if (KEY_LT(key, keys[j])) {\
keys[i] = keys[j];\
vals[i] = vals[j];\
i = j;\
}\
else\
break;\
}\
ASSERT(i >= 0);\
keys[i] = key;\
vals[i] = val;\
\
ASSERT2(FPRFX ## CheckHeap2(queue));\
\
return 1;\
}\
\
\
/*************************************************************************/\
/*! This function returns the item at the top of the queue and removes\
it from the priority queue */\
/**************************************************************************/\
int FPRFX ## GetTop2(PQT *queue, VT *r_val)\
{\
ssize_t i, j;\
KT key, *keys=queue->keys;\
VT val, *vals=queue->vals;\
\
ASSERT2(FPRFX ## CheckHeap2(queue));\
\
if (queue->nnodes == 0)\
return 0;\
\
queue->nnodes--;\
\
*r_val = vals[0];\
\
if ((i = queue->nnodes) > 0) {\
key = keys[i];\
val = vals[i];\
i = 0;\
while ((j=2*i+1) < queue->nnodes) {\
if (KEY_LT(keys[j], key)) {\
if (j+1 < queue->nnodes && KEY_LT(keys[j+1], keys[j]))\
j = j+1;\
keys[i] = keys[j];\
vals[i] = vals[j];\
i = j;\
}\
else if (j+1 < queue->nnodes && KEY_LT(keys[j+1], key)) {\
j = j+1;\
keys[i] = keys[j];\
vals[i] = vals[j];\
i = j;\
}\
else\
break;\
}\
\
keys[i] = key;\
vals[i] = val;\
}\
\
ASSERT2(FPRFX ## CheckHeap2(queue));\
\
return 1;\
}\
\
\
/*************************************************************************/\
/*! This function returns the item at the top of the queue. The item is not\
deleted from the queue. */\
/**************************************************************************/\
int FPRFX ## SeeTopVal2(PQT *queue, VT *r_val)\
{\
if (queue->nnodes == 0) \
return 0;\
\
*r_val = queue->vals[0];\
\
return 1;\
}\
\
\
/*************************************************************************/\
/*! This function returns the key of the top item. The item is not\
deleted from the queue. */\
/**************************************************************************/\
KT FPRFX ## SeeTopKey2(PQT *queue)\
{\
return (queue->nnodes == 0 ? KMAX : queue->keys[0]);\
}\
\
\
/*************************************************************************/\
/*! This functions checks the consistency of the heap */\
/**************************************************************************/\
int FPRFX ## CheckHeap2(PQT *queue)\
{\
ssize_t i;\
KT *keys=queue->keys;\
\
if (queue->nnodes == 0)\
return 1;\
\
for (i=1; i<queue->nnodes; i++) {\
ASSERT(!KEY_LT(keys[i], keys[(i-1)/2]));\
}\
for (i=1; i<queue->nnodes; i++)\
ASSERT(!KEY_LT(keys[i], keys[0]));\
\
return 1;\
}\
#define GK_MKPQUEUE2_PROTO(FPRFX, PQT, KT, VT)\
PQT * FPRFX ## Create2(ssize_t maxnodes);\
void FPRFX ## Reset2(PQT *queue);\
void FPRFX ## Destroy2(PQT **r_queue);\
size_t FPRFX ## Length2(PQT *queue);\
int FPRFX ## Insert2(PQT *queue, VT node, KT key);\
int FPRFX ## GetTop2(PQT *queue, VT *r_val);\
int FPRFX ## SeeTopVal2(PQT *queue, VT *r_val);\
KT FPRFX ## SeeTopKey2(PQT *queue);\
int FPRFX ## CheckHeap2(PQT *queue);\
#endif
/*!
\file
\brief Templates for portable random number generation
\date Started 5/17/07
\author George
\version\verbatim $Id: gk_mkrandom.h 10711 2011-08-31 22:23:04Z karypis $ \endverbatim
*/
#ifndef _GK_MKRANDOM_H
#define _GK_MKRANDOM_H
/*************************************************************************/\
/*! The generator for the rand() related routines. \
\params RNGT the datatype that defines the range of values over which\
random numbers will be generated\
\params VALT the datatype that defines the contents of the array to \
be permuted by randArrayPermute() \
\params FPRFX the function prefix \
*/\
/**************************************************************************/\
#define GK_MKRANDOM(FPRFX, RNGT, VALT)\
/*************************************************************************/\
/*! Initializes the generator */ \
/**************************************************************************/\
void FPRFX ## srand(RNGT seed) \
{\
gk_randinit((uint64_t) seed);\
}\
\
\
/*************************************************************************/\
/*! Returns a random number */ \
/**************************************************************************/\
RNGT FPRFX ## rand() \
{\
if (sizeof(RNGT) <= sizeof(int32_t)) \
return (RNGT)gk_randint32(); \
else \
return (RNGT)gk_randint64(); \
}\
\
\
/*************************************************************************/\
/*! Returns a random number between [0, max) */ \
/**************************************************************************/\
RNGT FPRFX ## randInRange(RNGT max) \
{\
return (RNGT)((FPRFX ## rand())%max); \
}\
\
\
/*************************************************************************/\
/*! Randomly permutes the elements of an array p[]. \
flag == 1, p[i] = i prior to permutation, \
flag == 0, p[] is not initialized. */\
/**************************************************************************/\
void FPRFX ## randArrayPermute(RNGT n, VALT *p, RNGT nshuffles, int flag)\
{\
RNGT i, u, v;\
VALT tmp;\
\
if (flag == 1) {\
for (i=0; i<n; i++)\
p[i] = (VALT)i;\
}\
\
if (n < 10) {\
for (i=0; i<n; i++) {\
v = FPRFX ## randInRange(n);\
u = FPRFX ## randInRange(n);\
gk_SWAP(p[v], p[u], tmp);\
}\
}\
else {\
for (i=0; i<nshuffles; i++) {\
v = FPRFX ## randInRange(n-3);\
u = FPRFX ## randInRange(n-3);\
/*gk_SWAP(p[v+0], p[u+0], tmp);*/\
/*gk_SWAP(p[v+1], p[u+1], tmp);*/\
/*gk_SWAP(p[v+2], p[u+2], tmp);*/\
/*gk_SWAP(p[v+3], p[u+3], tmp);*/\
gk_SWAP(p[v+0], p[u+2], tmp);\
gk_SWAP(p[v+1], p[u+3], tmp);\
gk_SWAP(p[v+2], p[u+0], tmp);\
gk_SWAP(p[v+3], p[u+1], tmp);\
}\
}\
}\
\
\
/*************************************************************************/\
/*! Randomly permutes the elements of an array p[]. \
flag == 1, p[i] = i prior to permutation, \
flag == 0, p[] is not initialized. */\
/**************************************************************************/\
void FPRFX ## randArrayPermuteFine(RNGT n, VALT *p, int flag)\
{\
RNGT i, v;\
VALT tmp;\
\
if (flag == 1) {\
for (i=0; i<n; i++)\
p[i] = (VALT)i;\
}\
\
for (i=0; i<n; i++) {\
v = FPRFX ## randInRange(n);\
gk_SWAP(p[i], p[v], tmp);\
}\
}\
#define GK_MKRANDOM_PROTO(FPRFX, RNGT, VALT)\
void FPRFX ## srand(RNGT seed); \
RNGT FPRFX ## rand(); \
RNGT FPRFX ## randInRange(RNGT max); \
void FPRFX ## randArrayPermute(RNGT n, VALT *p, RNGT nshuffles, int flag);\
void FPRFX ## randArrayPermuteFine(RNGT n, VALT *p, int flag);\
#endif
/*!
\file gk_mksort.h
\brief Templates for the qsort routine
\date Started 3/28/07
\author George
\version\verbatim $Id: gk_mksort.h 21051 2017-05-25 04:36:14Z karypis $ \endverbatim
*/
#ifndef _GK_MKSORT_H_
#define _GK_MKSORT_H_
/* Adopted from GNU glibc by Mjt.
* See stdlib/qsort.c in glibc */
/* Copyright (C) 1991, 1992, 1996, 1997, 1999 Free Software Foundation, Inc.
This file is part of the GNU C Library.
Written by Douglas C. Schmidt (schmidt@ics.uci.edu).
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, write to the Free
Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
02111-1307 USA. */
/* in-line qsort implementation. Differs from traditional qsort() routine
* in that it is a macro, not a function, and instead of passing an address
* of a comparision routine to the function, it is possible to inline
* comparision routine, thus speed up sorting alot.
*
* Usage:
* #include "iqsort.h"
* #define islt(a,b) (strcmp((*a),(*b))<0)
* char *arr[];
* int n;
* GKQSORT(char*, arr, n, islt);
*
* The "prototype" and 4 arguments are:
* GKQSORT(TYPE,BASE,NELT,ISLT)
* 1) type of each element, TYPE,
* 2) address of the beginning of the array, of type TYPE*,
* 3) number of elements in the array, and
* 4) comparision routine.
* Array pointer and number of elements are referenced only once.
* This is similar to a call
* qsort(BASE,NELT,sizeof(TYPE),ISLT)
* with the difference in last parameter.
* Note the islt macro/routine (it receives pointers to two elements):
* the only condition of interest is whenever one element is less than
* another, no other conditions (greather than, equal to etc) are tested.
* So, for example, to define integer sort, use:
* #define islt(a,b) ((*a)<(*b))
* GKQSORT(int, arr, n, islt)
*
* The macro could be used to implement a sorting function (see examples
* below), or to implement the sorting algorithm inline. That is, either
* create a sorting function and use it whenever you want to sort something,
* or use GKQSORT() macro directly instead a call to such routine. Note that
* the macro expands to quite some code (compiled size of int qsort on x86
* is about 700..800 bytes).
*
* Using this macro directly it isn't possible to implement traditional
* qsort() routine, because the macro assumes sizeof(element) == sizeof(TYPE),
* while qsort() allows element size to be different.
*
* Several ready-to-use examples:
*
* Sorting array of integers:
* void int_qsort(int *arr, unsigned n) {
* #define int_lt(a,b) ((*a)<(*b))
* GKQSORT(int, arr, n, int_lt);
* }
*
* Sorting array of string pointers:
* void str_qsort(char *arr[], unsigned n) {
* #define str_lt(a,b) (strcmp((*a),(*b)) < 0)
* GKQSORT(char*, arr, n, str_lt);
* }
*
* Sorting array of structures:
*
* struct elt {
* int key;
* ...
* };
* void elt_qsort(struct elt *arr, unsigned n) {
* #define elt_lt(a,b) ((a)->key < (b)->key)
* GKQSORT(struct elt, arr, n, elt_lt);
* }
*
* And so on.
*/
/* Swap two items pointed to by A and B using temporary buffer t. */
#define _GKQSORT_SWAP(a, b, t) ((void)((t = *a), (*a = *b), (*b = t)))
/* Discontinue quicksort algorithm when partition gets below this size. */
#define _GKQSORT_MAX_THRESH 8
/* The next 4 #defines implement a very fast in-line stack abstraction. */
#define _GKQSORT_STACK_SIZE (8 * sizeof(size_t))
#define _GKQSORT_PUSH(top, low, high) (((top->_lo = (low)), (top->_hi = (high)), ++top))
#define _GKQSORT_POP(low, high, top) ((--top, (low = top->_lo), (high = top->_hi)))
#define _GKQSORT_STACK_NOT_EMPTY (_stack < _top)
/* The main code starts here... */
#define GK_MKQSORT(GKQSORT_TYPE,GKQSORT_BASE,GKQSORT_NELT,GKQSORT_LT) \
{ \
GKQSORT_TYPE *const _base = (GKQSORT_BASE); \
const size_t _elems = (GKQSORT_NELT); \
GKQSORT_TYPE _hold; \
\
if (_elems < 1) \
return; \
\
/* Don't declare two variables of type GKQSORT_TYPE in a single \
* statement: eg `TYPE a, b;', in case if TYPE is a pointer, \
* expands to `type* a, b;' wich isn't what we want. \
*/ \
\
if (_elems > _GKQSORT_MAX_THRESH) { \
GKQSORT_TYPE *_lo = _base; \
GKQSORT_TYPE *_hi = _lo + _elems - 1; \
struct { \
GKQSORT_TYPE *_hi; GKQSORT_TYPE *_lo; \
} _stack[_GKQSORT_STACK_SIZE], *_top = _stack + 1; \
\
while (_GKQSORT_STACK_NOT_EMPTY) { \
GKQSORT_TYPE *_left_ptr; GKQSORT_TYPE *_right_ptr; \
\
/* Select median value from among LO, MID, and HI. Rearrange \
LO and HI so the three values are sorted. This lowers the \
probability of picking a pathological pivot value and \
skips a comparison for both the LEFT_PTR and RIGHT_PTR in \
the while loops. */ \
\
GKQSORT_TYPE *_mid = _lo + ((_hi - _lo) >> 1); \
\
if (GKQSORT_LT (_mid, _lo)) \
_GKQSORT_SWAP (_mid, _lo, _hold); \
if (GKQSORT_LT (_hi, _mid)) \
_GKQSORT_SWAP (_mid, _hi, _hold); \
else \
goto _jump_over; \
if (GKQSORT_LT (_mid, _lo)) \
_GKQSORT_SWAP (_mid, _lo, _hold); \
_jump_over:; \
\
_left_ptr = _lo + 1; \
_right_ptr = _hi - 1; \
\
/* Here's the famous ``collapse the walls'' section of quicksort. \
Gotta like those tight inner loops! They are the main reason \
that this algorithm runs much faster than others. */ \
do { \
while (GKQSORT_LT (_left_ptr, _mid)) \
++_left_ptr; \
\
while (GKQSORT_LT (_mid, _right_ptr)) \
--_right_ptr; \
\
if (_left_ptr < _right_ptr) { \
_GKQSORT_SWAP (_left_ptr, _right_ptr, _hold); \
if (_mid == _left_ptr) \
_mid = _right_ptr; \
else if (_mid == _right_ptr) \
_mid = _left_ptr; \
++_left_ptr; \
--_right_ptr; \
} \
else if (_left_ptr == _right_ptr) { \
++_left_ptr; \
--_right_ptr; \
break; \
} \
} while (_left_ptr <= _right_ptr); \
\
/* Set up pointers for next iteration. First determine whether \
left and right partitions are below the threshold size. If so, \
ignore one or both. Otherwise, push the larger partition's \
bounds on the stack and continue sorting the smaller one. */ \
\
if (_right_ptr - _lo <= _GKQSORT_MAX_THRESH) { \
if (_hi - _left_ptr <= _GKQSORT_MAX_THRESH) \
/* Ignore both small partitions. */ \
_GKQSORT_POP (_lo, _hi, _top); \
else \
/* Ignore small left partition. */ \
_lo = _left_ptr; \
} \
else if (_hi - _left_ptr <= _GKQSORT_MAX_THRESH) \
/* Ignore small right partition. */ \
_hi = _right_ptr; \
else if (_right_ptr - _lo > _hi - _left_ptr) { \
/* Push larger left partition indices. */ \
_GKQSORT_PUSH (_top, _lo, _right_ptr); \
_lo = _left_ptr; \
} \
else { \
/* Push larger right partition indices. */ \
_GKQSORT_PUSH (_top, _left_ptr, _hi); \
_hi = _right_ptr; \
} \
} \
} \
\
/* Once the BASE array is partially sorted by quicksort the rest \
is completely sorted using insertion sort, since this is efficient \
for partitions below MAX_THRESH size. BASE points to the \
beginning of the array to sort, and END_PTR points at the very \
last element in the array (*not* one beyond it!). */ \
\
{ \
GKQSORT_TYPE *const _end_ptr = _base + _elems - 1; \
GKQSORT_TYPE *_tmp_ptr = _base; \
register GKQSORT_TYPE *_run_ptr; \
GKQSORT_TYPE *_thresh; \
\
_thresh = _base + _GKQSORT_MAX_THRESH; \
if (_thresh > _end_ptr) \
_thresh = _end_ptr; \
\
/* Find smallest element in first threshold and place it at the \
array's beginning. This is the smallest array element, \
and the operation speeds up insertion sort's inner loop. */ \
\
for (_run_ptr = _tmp_ptr + 1; _run_ptr <= _thresh; ++_run_ptr) \
if (GKQSORT_LT (_run_ptr, _tmp_ptr)) \
_tmp_ptr = _run_ptr; \
\
if (_tmp_ptr != _base) \
_GKQSORT_SWAP (_tmp_ptr, _base, _hold); \
\
/* Insertion sort, running from left-hand-side \
* up to right-hand-side. */ \
\
_run_ptr = _base + 1; \
while (++_run_ptr <= _end_ptr) { \
_tmp_ptr = _run_ptr - 1; \
while (GKQSORT_LT (_run_ptr, _tmp_ptr)) \
--_tmp_ptr; \
\
++_tmp_ptr; \
if (_tmp_ptr != _run_ptr) { \
GKQSORT_TYPE *_trav = _run_ptr + 1; \
while (--_trav >= _run_ptr) { \
GKQSORT_TYPE *_hi; GKQSORT_TYPE *_lo; \
_hold = *_trav; \
\
for (_hi = _lo = _trav; --_lo >= _tmp_ptr; _hi = _lo) \
*_hi = *_lo; \
*_hi = _hold; \
} \
} \
} \
} \
\
}
#endif
/*!
\file
\brief Templates for various utility routines
\date Started 5/28/07
\author George
\version\verbatim $Id: gk_mkutils.h 10711 2011-08-31 22:23:04Z karypis $ \endverbatim
*/
#ifndef _GK_MKUTILS_H_
#define _GK_MKUTILS_H_
#define GK_MKARRAY2CSR(PRFX, TYPE)\
/*************************************************************************/\
/*! The macro for gk_?array2csr() routine */\
/**************************************************************************/\
void PRFX ## array2csr(TYPE n, TYPE range, TYPE *array, TYPE *ptr, TYPE *ind)\
{\
TYPE i;\
\
for (i=0; i<=range; i++)\
ptr[i] = 0;\
\
for (i=0; i<n; i++)\
ptr[array[i]]++;\
\
/* Compute the ptr, ind structure */\
MAKECSR(i, range, ptr);\
for (i=0; i<n; i++)\
ind[ptr[array[i]]++] = i;\
SHIFTCSR(i, range, ptr);\
}
#define GK_MKARRAY2CSR_PROTO(PRFX, TYPE)\
void PRFX ## array2csr(TYPE n, TYPE range, TYPE *array, TYPE *ptr, TYPE *ind);\
#endif
/*!
\file gk_proto.h
\brief This file contains function prototypes
\date Started 3/27/2007
\author George
\version\verbatim $Id: gk_proto.h 21050 2017-05-25 03:53:58Z karypis $ \endverbatim
*/
#ifndef _GK_PROTO_H_
#define _GK_PROTO_H_
#ifdef __cplusplus
extern "C" {
#endif
/*-------------------------------------------------------------
* blas.c
*-------------------------------------------------------------*/
GK_MKBLAS_PROTO(gk_c, char, int)
GK_MKBLAS_PROTO(gk_i, int, int)
GK_MKBLAS_PROTO(gk_i8, int8_t, int8_t)
GK_MKBLAS_PROTO(gk_i16, int16_t, int16_t)
GK_MKBLAS_PROTO(gk_i32, int32_t, int32_t)
GK_MKBLAS_PROTO(gk_i64, int64_t, int64_t)
GK_MKBLAS_PROTO(gk_z, ssize_t, ssize_t)
GK_MKBLAS_PROTO(gk_zu, size_t, size_t)
GK_MKBLAS_PROTO(gk_f, float, float)
GK_MKBLAS_PROTO(gk_d, double, double)
GK_MKBLAS_PROTO(gk_idx, gk_idx_t, gk_idx_t)
/*-------------------------------------------------------------
* io.c
*-------------------------------------------------------------*/
FILE *gk_fopen(char *, char *, const char *);
void gk_fclose(FILE *);
ssize_t gk_read(int fd, void *vbuf, size_t count);
ssize_t gk_write(int fd, void *vbuf, size_t count);
gk_idx_t gk_getline(char **lineptr, size_t *n, FILE *stream);
char **gk_readfile(char *fname, size_t *r_nlines);
int32_t *gk_i32readfile(char *fname, size_t *r_nlines);
int64_t *gk_i64readfile(char *fname, size_t *r_nlines);
ssize_t *gk_zreadfile(char *fname, size_t *r_nlines);
int32_t *gk_i32readfilebin(char *fname, size_t *r_nelmnts);
size_t gk_i32writefilebin(char *fname, size_t n, int32_t *a);
int64_t *gk_i64readfilebin(char *fname, size_t *r_nelmnts);
size_t gk_i64writefilebin(char *fname, size_t n, int64_t *a);
ssize_t *gk_zreadfilebin(char *fname, size_t *r_nelmnts);
size_t gk_zwritefilebin(char *fname, size_t n, ssize_t *a);
float *gk_freadfilebin(char *fname, size_t *r_nelmnts);
size_t gk_fwritefilebin(char *fname, size_t n, float *a);
double *gk_dreadfilebin(char *fname, size_t *r_nelmnts);
size_t gk_dwritefilebin(char *fname, size_t n, double *a);
/*-------------------------------------------------------------
* fs.c
*-------------------------------------------------------------*/
int gk_fexists(char *);
int gk_dexists(char *);
ssize_t gk_getfsize(char *);
void gk_getfilestats(char *fname, size_t *r_nlines, size_t *r_ntokens,
size_t *r_max_nlntokens, size_t *r_nbytes);
char *gk_getbasename(char *path);
char *gk_getextname(char *path);
char *gk_getfilename(char *path);
char *gk_getpathname(char *path);
int gk_mkpath(char *);
int gk_rmpath(char *);
/*-------------------------------------------------------------
* memory.c
*-------------------------------------------------------------*/
GK_MKALLOC_PROTO(gk_c, char)
GK_MKALLOC_PROTO(gk_i, int)
GK_MKALLOC_PROTO(gk_i8, int8_t)
GK_MKALLOC_PROTO(gk_i16, int16_t)
GK_MKALLOC_PROTO(gk_i32, int32_t)
GK_MKALLOC_PROTO(gk_i64, int64_t)
GK_MKALLOC_PROTO(gk_ui8, uint8_t)
GK_MKALLOC_PROTO(gk_ui16, uint16_t)
GK_MKALLOC_PROTO(gk_ui32, uint32_t)
GK_MKALLOC_PROTO(gk_ui64, uint64_t)
GK_MKALLOC_PROTO(gk_z, ssize_t)
GK_MKALLOC_PROTO(gk_zu, size_t)
GK_MKALLOC_PROTO(gk_f, float)
GK_MKALLOC_PROTO(gk_d, double)
GK_MKALLOC_PROTO(gk_idx, gk_idx_t)
GK_MKALLOC_PROTO(gk_ckv, gk_ckv_t)
GK_MKALLOC_PROTO(gk_ikv, gk_ikv_t)
GK_MKALLOC_PROTO(gk_i8kv, gk_i8kv_t)
GK_MKALLOC_PROTO(gk_i16kv, gk_i16kv_t)
GK_MKALLOC_PROTO(gk_i32kv, gk_i32kv_t)
GK_MKALLOC_PROTO(gk_i64kv, gk_i64kv_t)
GK_MKALLOC_PROTO(gk_zkv, gk_zkv_t)
GK_MKALLOC_PROTO(gk_zukv, gk_zukv_t)
GK_MKALLOC_PROTO(gk_fkv, gk_fkv_t)
GK_MKALLOC_PROTO(gk_dkv, gk_dkv_t)
GK_MKALLOC_PROTO(gk_skv, gk_skv_t)
GK_MKALLOC_PROTO(gk_idxkv, gk_idxkv_t)
void gk_AllocMatrix(void ***, size_t, size_t , size_t);
void gk_FreeMatrix(void ***, size_t, size_t);
int gk_malloc_init();
void gk_malloc_cleanup(int showstats);
void *gk_malloc(size_t nbytes, char *msg);
void *gk_realloc(void *oldptr, size_t nbytes, char *msg);
void gk_free(void **ptr1,...);
size_t gk_GetCurMemoryUsed();
size_t gk_GetMaxMemoryUsed();
void gk_GetVMInfo(size_t *vmsize, size_t *vmrss);
/*-------------------------------------------------------------
* seq.c
*-------------------------------------------------------------*/
gk_seq_t *gk_seq_ReadGKMODPSSM(char *file_name);
gk_i2cc2i_t *gk_i2cc2i_create_common(char *alphabet);
void gk_seq_init(gk_seq_t *seq);
/*-------------------------------------------------------------
* error.c
*-------------------------------------------------------------*/
void gk_set_exit_on_error(int value);
void errexit(char *,...);
void gk_errexit(int signum, char *,...);
int gk_sigtrap();
int gk_siguntrap();
void gk_sigthrow(int signum);
void gk_SetSignalHandlers();
void gk_UnsetSignalHandlers();
void gk_NonLocalExit_Handler(int signum);
char *gk_strerror(int errnum);
void PrintBackTrace();
/*-------------------------------------------------------------
* util.c
*-------------------------------------------------------------*/
void gk_RandomPermute(size_t, int *, int);
void gk_array2csr(size_t n, size_t range, int *array, int *ptr, int *ind);
int gk_log2(int);
int gk_ispow2(int);
float gk_flog2(float);
/*-------------------------------------------------------------
* time.c
*-------------------------------------------------------------*/
gk_wclock_t gk_WClockSeconds(void);
double gk_CPUSeconds(void);
/*-------------------------------------------------------------
* string.c
*-------------------------------------------------------------*/
char *gk_strchr_replace(char *str, char *fromlist, char *tolist);
int gk_strstr_replace(char *str, char *pattern, char *replacement, char *options, char **new_str);
char *gk_strtprune(char *, char *);
char *gk_strhprune(char *, char *);
char *gk_strtoupper(char *);
char *gk_strtolower(char *);
char *gk_strdup(char *orgstr);
int gk_strcasecmp(char *s1, char *s2);
int gk_strrcmp(char *s1, char *s2);
char *gk_time2str(time_t time);
time_t gk_str2time(char *str);
int gk_GetStringID(gk_StringMap_t *strmap, char *key);
/*-------------------------------------------------------------
* sort.c
*-------------------------------------------------------------*/
void gk_csorti(size_t, char *);
void gk_csortd(size_t, char *);
void gk_isorti(size_t, int *);
void gk_isortd(size_t, int *);
void gk_i32sorti(size_t, int32_t *);
void gk_i32sortd(size_t, int32_t *);
void gk_i64sorti(size_t, int64_t *);
void gk_i64sortd(size_t, int64_t *);
void gk_ui32sorti(size_t, uint32_t *);
void gk_ui32sortd(size_t, uint32_t *);
void gk_ui64sorti(size_t, uint64_t *);
void gk_ui64sortd(size_t, uint64_t *);
void gk_fsorti(size_t, float *);
void gk_fsortd(size_t, float *);
void gk_dsorti(size_t, double *);
void gk_dsortd(size_t, double *);
void gk_idxsorti(size_t, gk_idx_t *);
void gk_idxsortd(size_t, gk_idx_t *);
void gk_ckvsorti(size_t, gk_ckv_t *);
void gk_ckvsortd(size_t, gk_ckv_t *);
void gk_ikvsorti(size_t, gk_ikv_t *);
void gk_ikvsortd(size_t, gk_ikv_t *);
void gk_i32kvsorti(size_t, gk_i32kv_t *);
void gk_i32kvsortd(size_t, gk_i32kv_t *);
void gk_i64kvsorti(size_t, gk_i64kv_t *);
void gk_i64kvsortd(size_t, gk_i64kv_t *);
void gk_zkvsorti(size_t, gk_zkv_t *);
void gk_zkvsortd(size_t, gk_zkv_t *);
void gk_zukvsorti(size_t, gk_zukv_t *);
void gk_zukvsortd(size_t, gk_zukv_t *);
void gk_fkvsorti(size_t, gk_fkv_t *);
void gk_fkvsortd(size_t, gk_fkv_t *);
void gk_dkvsorti(size_t, gk_dkv_t *);
void gk_dkvsortd(size_t, gk_dkv_t *);
void gk_skvsorti(size_t, gk_skv_t *);
void gk_skvsortd(size_t, gk_skv_t *);
void gk_idxkvsorti(size_t, gk_idxkv_t *);
void gk_idxkvsortd(size_t, gk_idxkv_t *);
/*-------------------------------------------------------------
* Selection routines
*-------------------------------------------------------------*/
int gk_dfkvkselect(size_t, int, gk_fkv_t *);
int gk_ifkvkselect(size_t, int, gk_fkv_t *);
/*-------------------------------------------------------------
* Priority queue
*-------------------------------------------------------------*/
GK_MKPQUEUE_PROTO(gk_ipq, gk_ipq_t, int, gk_idx_t)
GK_MKPQUEUE_PROTO(gk_i32pq, gk_i32pq_t, int32_t, gk_idx_t)
GK_MKPQUEUE_PROTO(gk_i64pq, gk_i64pq_t, int64_t, gk_idx_t)
GK_MKPQUEUE_PROTO(gk_fpq, gk_fpq_t, float, gk_idx_t)
GK_MKPQUEUE_PROTO(gk_dpq, gk_dpq_t, double, gk_idx_t)
GK_MKPQUEUE_PROTO(gk_idxpq, gk_idxpq_t, gk_idx_t, gk_idx_t)
/*-------------------------------------------------------------
* HTable routines
*-------------------------------------------------------------*/
gk_HTable_t *HTable_Create(int nelements);
void HTable_Reset(gk_HTable_t *htable);
void HTable_Resize(gk_HTable_t *htable, int nelements);
void HTable_Insert(gk_HTable_t *htable, int key, int val);
void HTable_Delete(gk_HTable_t *htable, int key);
int HTable_Search(gk_HTable_t *htable, int key);
int HTable_GetNext(gk_HTable_t *htable, int key, int *val, int type);
int HTable_SearchAndDelete(gk_HTable_t *htable, int key);
void HTable_Destroy(gk_HTable_t *htable);
int HTable_HFunction(int nelements, int key);
/*-------------------------------------------------------------
* Tokenizer routines
*-------------------------------------------------------------*/
void gk_strtokenize(char *line, char *delim, gk_Tokens_t *tokens);
void gk_freetokenslist(gk_Tokens_t *tokens);
/*-------------------------------------------------------------
* Encoder/Decoder
*-------------------------------------------------------------*/
void encodeblock(unsigned char *in, unsigned char *out);
void decodeblock(unsigned char *in, unsigned char *out);
void GKEncodeBase64(int nbytes, unsigned char *inbuffer, unsigned char *outbuffer);
void GKDecodeBase64(int nbytes, unsigned char *inbuffer, unsigned char *outbuffer);
/*-------------------------------------------------------------
* random.c
*-------------------------------------------------------------*/
GK_MKRANDOM_PROTO(gk_c, size_t, char)
GK_MKRANDOM_PROTO(gk_i, size_t, int)
GK_MKRANDOM_PROTO(gk_i32, size_t, int32_t)
GK_MKRANDOM_PROTO(gk_f, size_t, float)
GK_MKRANDOM_PROTO(gk_d, size_t, double)
GK_MKRANDOM_PROTO(gk_idx, size_t, gk_idx_t)
GK_MKRANDOM_PROTO(gk_z, size_t, ssize_t)
GK_MKRANDOM_PROTO(gk_zu, size_t, size_t)
void gk_randinit(uint64_t);
uint64_t gk_randint64(void);
uint32_t gk_randint32(void);
/*-------------------------------------------------------------
* OpenMP fake functions
*-------------------------------------------------------------*/
#if !defined(__OPENMP__)
void omp_set_num_threads(int num_threads);
int omp_get_num_threads(void);
int omp_get_max_threads(void);
int omp_get_thread_num(void);
int omp_get_num_procs(void);
int omp_in_parallel(void);
void omp_set_dynamic(int num_threads);
int omp_get_dynamic(void);
void omp_set_nested(int nested);
int omp_get_nested(void);
#endif /* __OPENMP__ */
/*-------------------------------------------------------------
* CSR-related functions
*-------------------------------------------------------------*/
gk_csr_t *gk_csr_Create();
void gk_csr_Init(gk_csr_t *mat);
void gk_csr_Free(gk_csr_t **mat);
void gk_csr_FreeContents(gk_csr_t *mat);
gk_csr_t *gk_csr_Dup(gk_csr_t *mat);
gk_csr_t *gk_csr_ExtractSubmatrix(gk_csr_t *mat, int rstart, int nrows);
gk_csr_t *gk_csr_ExtractRows(gk_csr_t *mat, int nrows, int *rind);
gk_csr_t *gk_csr_ExtractPartition(gk_csr_t *mat, int *part, int pid);
gk_csr_t **gk_csr_Split(gk_csr_t *mat, int *color);
int gk_csr_DetermineFormat(char *filename, int format);
gk_csr_t *gk_csr_Read(char *filename, int format, int readvals, int numbering);
void gk_csr_Write(gk_csr_t *mat, char *filename, int format, int writevals, int numbering);
gk_csr_t *gk_csr_Prune(gk_csr_t *mat, int what, int minf, int maxf);
gk_csr_t *gk_csr_LowFilter(gk_csr_t *mat, int what, int norm, float fraction);
gk_csr_t *gk_csr_TopKPlusFilter(gk_csr_t *mat, int what, int topk, float keepval);
gk_csr_t *gk_csr_ZScoreFilter(gk_csr_t *mat, int what, float zscore);
void gk_csr_CompactColumns(gk_csr_t *mat);
void gk_csr_SortIndices(gk_csr_t *mat, int what);
void gk_csr_CreateIndex(gk_csr_t *mat, int what);
void gk_csr_Normalize(gk_csr_t *mat, int what, int norm);
void gk_csr_Scale(gk_csr_t *mat, int type);
void gk_csr_ComputeSums(gk_csr_t *mat, int what);
void gk_csr_ComputeNorms(gk_csr_t *mat, int what);
void gk_csr_ComputeSquaredNorms(gk_csr_t *mat, int what);
gk_csr_t *gk_csr_Shuffle(gk_csr_t *mat, int what, int summetric);
gk_csr_t *gk_csr_Transpose(gk_csr_t *mat);
float gk_csr_ComputeSimilarity(gk_csr_t *mat, int i1, int i2, int what, int simtype);
float gk_csr_ComputePairSimilarity(gk_csr_t *mat_a, gk_csr_t *mat_b, int i1, int i2, int what, int simtype);
int gk_csr_GetSimilarRows(gk_csr_t *mat, int nqterms, int *qind, float *qval,
int simtype, int nsim, float minsim, gk_fkv_t *hits, int *_imarker,
gk_fkv_t *i_cand);
int gk_csr_FindConnectedComponents(gk_csr_t *mat, int32_t *cptr, int32_t *cind,
int32_t *cids);
gk_csr_t *gk_csr_MakeSymmetric(gk_csr_t *mat, int op);
gk_csr_t *gk_csr_ReorderSymmetric(gk_csr_t *mat, int32_t *perm, int32_t *iperm);
void gk_csr_ComputeBFSOrderingSymmetric(gk_csr_t *mat, int maxdegree, int v,
int32_t **r_perm, int32_t **r_iperm);
void gk_csr_ComputeBestFOrderingSymmetric(gk_csr_t *mat, int v, int type,
int32_t **r_perm, int32_t **r_iperm);
/* itemsets.c */
void gk_find_frequent_itemsets(int ntrans, ssize_t *tranptr, int *tranind,
int minfreq, int maxfreq, int minlen, int maxlen,
void (*process_itemset)(void *stateptr, int nitems, int *itemind,
int ntrans, int *tranind),
void *stateptr);
/* evaluate.c */
float ComputeAccuracy(int n, gk_fkv_t *list);
float ComputeROCn(int n, int maxN, gk_fkv_t *list);
float ComputeMedianRFP(int n, gk_fkv_t *list);
float ComputeMean (int n, float *values);
float ComputeStdDev(int n, float *values);
/* mcore.c */
gk_mcore_t *gk_mcoreCreate(size_t coresize);
gk_mcore_t *gk_gkmcoreCreate();
void gk_mcoreDestroy(gk_mcore_t **r_mcore, int showstats);
void gk_gkmcoreDestroy(gk_mcore_t **r_mcore, int showstats);
void *gk_mcoreMalloc(gk_mcore_t *mcore, size_t nbytes);
void gk_mcorePush(gk_mcore_t *mcore);
void gk_gkmcorePush(gk_mcore_t *mcore);
void gk_mcorePop(gk_mcore_t *mcore);
void gk_gkmcorePop(gk_mcore_t *mcore);
void gk_mcoreAdd(gk_mcore_t *mcore, int type, size_t nbytes, void *ptr);
void gk_gkmcoreAdd(gk_mcore_t *mcore, int type, size_t nbytes, void *ptr);
void gk_mcoreDel(gk_mcore_t *mcore, void *ptr);
void gk_gkmcoreDel(gk_mcore_t *mcore, void *ptr);
/* rw.c */
int gk_rw_PageRank(gk_csr_t *mat, float lamda, float eps, int max_niter, float *pr);
/* graph.c */
gk_graph_t *gk_graph_Create();
void gk_graph_Init(gk_graph_t *graph);
void gk_graph_Free(gk_graph_t **graph);
void gk_graph_FreeContents(gk_graph_t *graph);
gk_graph_t *gk_graph_Read(char *filename, int format, int hasvals,
int numbering, int isfewgts, int isfvwgts, int isfvsizes);
void gk_graph_Write(gk_graph_t *graph, char *filename, int format);
gk_graph_t *gk_graph_Dup(gk_graph_t *graph);
gk_graph_t *gk_graph_ExtractSubgraph(gk_graph_t *graph, int vstart, int nvtxs);
gk_graph_t *gk_graph_Reorder(gk_graph_t *graph, int32_t *perm, int32_t *iperm);
int gk_graph_FindComponents(gk_graph_t *graph, int32_t *cptr, int32_t *cind);
void gk_graph_ComputeBFSOrdering(gk_graph_t *graph, int v, int32_t **r_perm,
int32_t **r_iperm);
void gk_graph_ComputeBestFOrdering0(gk_graph_t *graph, int v, int type,
int32_t **r_perm, int32_t **r_iperm);
void gk_graph_ComputeBestFOrdering(gk_graph_t *graph, int v, int type,
int32_t **r_perm, int32_t **r_iperm);
void gk_graph_SingleSourceShortestPaths(gk_graph_t *graph, int v, void **r_sps);
#ifdef __cplusplus
}
#endif
#endif
/*!
\file gk_struct.h
\brief This file contains various datastructures used/provided by GKlib
\date Started 3/27/2007
\author George
\version\verbatim $Id: gk_struct.h 21032 2017-05-23 02:49:02Z karypis $ \endverbatim
*/
#ifndef _GK_STRUCT_H_
#define _GK_STRUCT_H_
/********************************************************************/
/*! Generator for gk_??KeyVal_t data structure */
/********************************************************************/
#define GK_MKKEYVALUE_T(NAME, KEYTYPE, VALTYPE) \
typedef struct {\
KEYTYPE key;\
VALTYPE val;\
} NAME;\
/* The actual KeyVal data structures */
GK_MKKEYVALUE_T(gk_ckv_t, char, ssize_t)
GK_MKKEYVALUE_T(gk_ikv_t, int, ssize_t)
GK_MKKEYVALUE_T(gk_i8kv_t, int8_t, ssize_t)
GK_MKKEYVALUE_T(gk_i16kv_t, int16_t, ssize_t)
GK_MKKEYVALUE_T(gk_i32kv_t, int32_t, ssize_t)
GK_MKKEYVALUE_T(gk_i64kv_t, int64_t, ssize_t)
GK_MKKEYVALUE_T(gk_zkv_t, ssize_t, ssize_t)
GK_MKKEYVALUE_T(gk_zukv_t, size_t, ssize_t)
GK_MKKEYVALUE_T(gk_fkv_t, float, ssize_t)
GK_MKKEYVALUE_T(gk_dkv_t, double, ssize_t)
GK_MKKEYVALUE_T(gk_skv_t, char *, ssize_t)
GK_MKKEYVALUE_T(gk_idxkv_t, gk_idx_t, gk_idx_t)
/********************************************************************/
/*! Generator for gk_?pq_t data structure */
/********************************************************************/
#define GK_MKPQUEUE_T(NAME, KVTYPE)\
typedef struct {\
size_t nnodes;\
size_t maxnodes;\
\
/* Heap version of the data structure */ \
KVTYPE *heap;\
ssize_t *locator;\
} NAME;\
GK_MKPQUEUE_T(gk_ipq_t, gk_ikv_t)
GK_MKPQUEUE_T(gk_i32pq_t, gk_i32kv_t)
GK_MKPQUEUE_T(gk_i64pq_t, gk_i64kv_t)
GK_MKPQUEUE_T(gk_fpq_t, gk_fkv_t)
GK_MKPQUEUE_T(gk_dpq_t, gk_dkv_t)
GK_MKPQUEUE_T(gk_idxpq_t, gk_idxkv_t)
#define GK_MKPQUEUE2_T(NAME, KTYPE, VTYPE)\
typedef struct {\
ssize_t nnodes;\
ssize_t maxnodes;\
\
/* Heap version of the data structure */ \
KTYPE *keys;\
VTYPE *vals;\
} NAME;\
/*-------------------------------------------------------------
* The following data structure stores a sparse CSR format
*-------------------------------------------------------------*/
typedef struct gk_csr_t {
int32_t nrows, ncols;
ssize_t *rowptr, *colptr;
int32_t *rowind, *colind;
int32_t *rowids, *colids;
int32_t *rlabels, *clabels;
int32_t *rmap, *cmap;
float *rowval, *colval;
float *rnorms, *cnorms;
float *rsums, *csums;
float *rsizes, *csizes;
float *rvols, *cvols;
float *rwgts, *cwgts;
} gk_csr_t;
/*-------------------------------------------------------------
* The following data structure stores a sparse graph
*-------------------------------------------------------------*/
typedef struct gk_graph_t {
int32_t nvtxs; /*!< The number of vertices in the graph */
ssize_t *xadj; /*!< The ptr-structure of the adjncy list */
int32_t *adjncy; /*!< The adjacency list of the graph */
int32_t *iadjwgt; /*!< The integer edge weights */
float *fadjwgt; /*!< The floating point edge weights */
int32_t *ivwgts; /*!< The integer vertex weights */
float *fvwgts; /*!< The floating point vertex weights */
int32_t *ivsizes; /*!< The integer vertex sizes */
float *fvsizes; /*!< The floating point vertex sizes */
int32_t *vlabels; /*!< The labels of the vertices */
} gk_graph_t;
/*-------------------------------------------------------------
* The following data structure stores stores a string as a
* pair of its allocated buffer and the buffer itself.
*-------------------------------------------------------------*/
typedef struct gk_str_t {
size_t len;
char *buf;
} gk_str_t;
/*-------------------------------------------------------------
* The following data structure implements a string-2-int mapping
* table used for parsing command-line options
*-------------------------------------------------------------*/
typedef struct gk_StringMap_t {
char *name;
int id;
} gk_StringMap_t;
/*------------------------------------------------------------
* This structure implements a simple hash table
*------------------------------------------------------------*/
typedef struct gk_HTable_t {
int nelements; /* The overall size of the hash-table */
int htsize; /* The current size of the hash-table */
gk_ikv_t *harray; /* The actual hash-table */
} gk_HTable_t;
/*------------------------------------------------------------
* This structure implements a gk_Tokens_t list returned by the
* string tokenizer
*------------------------------------------------------------*/
typedef struct gk_Tokens_t {
int ntoks; /* The number of tokens in the input string */
char *strbuf; /* The memory that stores all the entries */
char **list; /* Pointers to the strbuf for each element */
} gk_Tokens_t;
/*------------------------------------------------------------
* This structure implements storage for an atom in a pdb file
*------------------------------------------------------------*/
typedef struct atom {
int serial;
char *name;
char altLoc;
char *resname;
char chainid;
int rserial;
char icode;
char element;
double x;
double y;
double z;
double opcy;
double tmpt;
} atom;
/*------------------------------------------------------------
* This structure implements storage for a center of mass for
* a single residue.
*------------------------------------------------------------*/
typedef struct center_of_mass {
char name;
double x;
double y;
double z;
} center_of_mass;
/*------------------------------------------------------------
* This structure implements storage for a pdb protein
*------------------------------------------------------------*/
typedef struct pdbf {
int natoms; /* Number of atoms */
int nresidues; /* Number of residues based on coordinates */
int ncas;
int nbbs;
int corruption;
char *resSeq; /* Residue sequence based on coordinates */
char **threeresSeq; /* three-letter residue sequence */
atom *atoms;
atom **bbs;
atom **cas;
center_of_mass *cm;
} pdbf;
/*************************************************************
* Localization Structures for converting characters to integers
**************************************************************/
typedef struct gk_i2cc2i_t {
int n;
char *i2c;
int *c2i;
} gk_i2cc2i_t;
/*******************************************************************
*This structure implements storage of a protein sequence
* *****************************************************************/
typedef struct gk_seq_t {
int len; /*Number of Residues */
int *sequence; /* Stores the sequence*/
int **pssm; /* Stores the pssm matrix */
int **psfm; /* Stores the psfm matrix */
char *name; /* Stores the name of the sequence */
int nsymbols;
} gk_seq_t;
/*************************************************************************/
/*! The following data structure stores information about a memory
allocation operation that can either be served from gk_mcore_t or by
a gk_malloc if not sufficient workspace memory is available. */
/*************************************************************************/
typedef struct gk_mop_t {
int type;
ssize_t nbytes;
void *ptr;
} gk_mop_t;
/*************************************************************************/
/*! The following structure stores information used by Metis */
/*************************************************************************/
typedef struct gk_mcore_t {
/* Workspace information */
size_t coresize; /*!< The amount of core memory that has been allocated */
size_t corecpos; /*!< Index of the first free location in core */
void *core; /*!< Pointer to the core itself */
/* These are for implementing a stack-based allocation scheme using both
core and also dynamically allocated memory */
size_t nmops; /*!< The number of maop_t entries that have been allocated */
size_t cmop; /*!< Index of the first free location in maops */
gk_mop_t *mops; /*!< The array recording the maop_t operations */
/* These are for keeping various statistics for wspacemalloc */
size_t num_callocs; /*!< The number of core mallocs */
size_t num_hallocs; /*!< The number of heap mallocs */
size_t size_callocs; /*!< The total # of bytes in core mallocs */
size_t size_hallocs; /*!< The total # of bytes in heap mallocs */
size_t cur_callocs; /*!< The current # of bytes in core mallocs */
size_t cur_hallocs; /*!< The current # of bytes in heap mallocs */
size_t max_callocs; /*!< The maximum # of bytes in core mallocs at any given time */
size_t max_hallocs; /*!< The maximum # of bytes in heap mallocs at any given time */
} gk_mcore_t;
#endif
/*!
\file gk_types.h
\brief This file contains basic scalar datatype used in GKlib
\date Started 3/27/2007
\author George
\version\verbatim $Id: gk_types.h 10711 2011-08-31 22:23:04Z karypis $ \endverbatim
*/
#ifndef _GK_TYPES_H_
#define _GK_TYPES_H_
/*************************************************************************
* Basic data type definitions. These definitions allow GKlib to separate
* the following elemental types:
* - loop iterator variables, which are set to size_t
* - signed and unsigned int variables that can be set to any # of bits
* - signed and unsigned long variables that can be set to any # of bits
* - real variables, which can be set to single or double precision.
**************************************************************************/
/*typedef ptrdiff_t gk_idx_t; */ /* index variable */
typedef ssize_t gk_idx_t; /* index variable */
typedef int32_t gk_int_t; /* integer values */
typedef uint32_t gk_uint_t; /* unsigned integer values */
typedef int64_t gk_long_t; /* long integer values */
typedef uint64_t gk_ulong_t; /* unsigned long integer values */
typedef float gk_real_t; /* real type */
typedef double gk_dreal_t; /* double precission real type */
typedef double gk_wclock_t; /* wall-clock time */
/*#define GK_IDX_MAX PTRDIFF_MAX*/
#define GK_IDX_MAX ((SIZE_MAX>>1)-2)
#define PRIGKIDX "zd"
#define SCNGKIDX "zd"
#endif
/*!
\file util.c
\brief Various utility routines
\date Started 4/12/2007
\author George
\version\verbatim $Id: gk_util.c 16223 2014-02-15 21:34:09Z karypis $ \endverbatim
*/
#include <GKlib.h>
/*************************************************************************
* This file randomly permutes the contents of an array.
* flag == 0, don't initialize perm
* flag == 1, set p[i] = i
**************************************************************************/
void gk_RandomPermute(size_t n, int *p, int flag)
{
size_t i, u, v;
int tmp;
if (flag == 1) {
for (i=0; i<n; i++)
p[i] = i;
}
for (i=0; i<n/2; i++) {
v = RandomInRange(n);
u = RandomInRange(n);
gk_SWAP(p[v], p[u], tmp);
}
}
/************************************************************************/
/*!
\brief Converts an element-based set membership into a CSR-format set-based
membership.
For example, it takes an array such as part[] that stores where each
element belongs to and returns a pair of arrays (pptr[], pind[]) that
store in CSF format the list of elements belonging in each partition.
\param n
the number of elements in the array (e.g., # of vertices)
\param range
the cardinality of the set (e.g., # of partitions)
\param array
the array that stores the per-element set membership
\param ptr
the array that will store the starting indices in ind for
the elements of each set. This is filled by the routine and
its size should be at least range+1.
\param ind
the array that stores consecutively which elements belong to
each set. The size of this array should be n.
*/
/************************************************************************/
void gk_array2csr(size_t n, size_t range, int *array, int *ptr, int *ind)
{
size_t i;
gk_iset(range+1, 0, ptr);
for (i=0; i<n; i++)
ptr[array[i]]++;
/* Compute the ptr, ind structure */
MAKECSR(i, range, ptr);
for (i=0; i<n; i++)
ind[ptr[array[i]]++] = i;
SHIFTCSR(i, range, ptr);
}
/*************************************************************************
* This function returns the log2(x)
**************************************************************************/
int gk_log2(int a)
{
size_t i;
for (i=1; a > 1; i++, a = a>>1);
return i-1;
}
/*************************************************************************
* This function checks if the argument is a power of 2
**************************************************************************/
int gk_ispow2(int a)
{
return (a == (1<<gk_log2(a)));
}
/*************************************************************************
* This function returns the log2(x)
**************************************************************************/
float gk_flog2(float a)
{
return log(a)/log(2.0);
}
This source diff could not be displayed because it is too large. You can view the blob instead.
/* Definitions for data structures and routines for the regular
expression library.
Copyright (C) 1985,1989-93,1995-98,2000,2001,2002,2003,2005,2006
Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, write to the Free
Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
02111-1307 USA. */
#ifndef _REGEX_H
#define _REGEX_H 1
#include <sys/types.h>
/* Allow the use in C++ code. */
#ifdef __cplusplus
extern "C" {
#endif
/* The following two types have to be signed and unsigned integer type
wide enough to hold a value of a pointer. For most ANSI compilers
ptrdiff_t and size_t should be likely OK. Still size of these two
types is 2 for Microsoft C. Ugh... */
typedef long int s_reg_t;
typedef unsigned long int active_reg_t;
/* The following bits are used to determine the regexp syntax we
recognize. The set/not-set meanings are chosen so that Emacs syntax
remains the value 0. The bits are given in alphabetical order, and
the definitions shifted by one from the previous bit; thus, when we
add or remove a bit, only one other definition need change. */
typedef unsigned long int reg_syntax_t;
/* If this bit is not set, then \ inside a bracket expression is literal.
If set, then such a \ quotes the following character. */
#define RE_BACKSLASH_ESCAPE_IN_LISTS ((unsigned long int) 1)
/* If this bit is not set, then + and ? are operators, and \+ and \? are
literals.
If set, then \+ and \? are operators and + and ? are literals. */
#define RE_BK_PLUS_QM (RE_BACKSLASH_ESCAPE_IN_LISTS << 1)
/* If this bit is set, then character classes are supported. They are:
[:alpha:], [:upper:], [:lower:], [:digit:], [:alnum:], [:xdigit:],
[:space:], [:print:], [:punct:], [:graph:], and [:cntrl:].
If not set, then character classes are not supported. */
#define RE_CHAR_CLASSES (RE_BK_PLUS_QM << 1)
/* If this bit is set, then ^ and $ are always anchors (outside bracket
expressions, of course).
If this bit is not set, then it depends:
^ is an anchor if it is at the beginning of a regular
expression or after an open-group or an alternation operator;
$ is an anchor if it is at the end of a regular expression, or
before a close-group or an alternation operator.
This bit could be (re)combined with RE_CONTEXT_INDEP_OPS, because
POSIX draft 11.2 says that * etc. in leading positions is undefined.
We already implemented a previous draft which made those constructs
invalid, though, so we haven't changed the code back. */
#define RE_CONTEXT_INDEP_ANCHORS (RE_CHAR_CLASSES << 1)
/* If this bit is set, then special characters are always special
regardless of where they are in the pattern.
If this bit is not set, then special characters are special only in
some contexts; otherwise they are ordinary. Specifically,
* + ? and intervals are only special when not after the beginning,
open-group, or alternation operator. */
#define RE_CONTEXT_INDEP_OPS (RE_CONTEXT_INDEP_ANCHORS << 1)
/* If this bit is set, then *, +, ?, and { cannot be first in an re or
immediately after an alternation or begin-group operator. */
#define RE_CONTEXT_INVALID_OPS (RE_CONTEXT_INDEP_OPS << 1)
/* If this bit is set, then . matches newline.
If not set, then it doesn't. */
#define RE_DOT_NEWLINE (RE_CONTEXT_INVALID_OPS << 1)
/* If this bit is set, then . doesn't match NUL.
If not set, then it does. */
#define RE_DOT_NOT_NULL (RE_DOT_NEWLINE << 1)
/* If this bit is set, nonmatching lists [^...] do not match newline.
If not set, they do. */
#define RE_HAT_LISTS_NOT_NEWLINE (RE_DOT_NOT_NULL << 1)
/* If this bit is set, either \{...\} or {...} defines an
interval, depending on RE_NO_BK_BRACES.
If not set, \{, \}, {, and } are literals. */
#define RE_INTERVALS (RE_HAT_LISTS_NOT_NEWLINE << 1)
/* If this bit is set, +, ? and | aren't recognized as operators.
If not set, they are. */
#define RE_LIMITED_OPS (RE_INTERVALS << 1)
/* If this bit is set, newline is an alternation operator.
If not set, newline is literal. */
#define RE_NEWLINE_ALT (RE_LIMITED_OPS << 1)
/* If this bit is set, then `{...}' defines an interval, and \{ and \}
are literals.
If not set, then `\{...\}' defines an interval. */
#define RE_NO_BK_BRACES (RE_NEWLINE_ALT << 1)
/* If this bit is set, (...) defines a group, and \( and \) are literals.
If not set, \(...\) defines a group, and ( and ) are literals. */
#define RE_NO_BK_PARENS (RE_NO_BK_BRACES << 1)
/* If this bit is set, then \<digit> matches <digit>.
If not set, then \<digit> is a back-reference. */
#define RE_NO_BK_REFS (RE_NO_BK_PARENS << 1)
/* If this bit is set, then | is an alternation operator, and \| is literal.
If not set, then \| is an alternation operator, and | is literal. */
#define RE_NO_BK_VBAR (RE_NO_BK_REFS << 1)
/* If this bit is set, then an ending range point collating higher
than the starting range point, as in [z-a], is invalid.
If not set, then when ending range point collates higher than the
starting range point, the range is ignored. */
#define RE_NO_EMPTY_RANGES (RE_NO_BK_VBAR << 1)
/* If this bit is set, then an unmatched ) is ordinary.
If not set, then an unmatched ) is invalid. */
#define RE_UNMATCHED_RIGHT_PAREN_ORD (RE_NO_EMPTY_RANGES << 1)
/* If this bit is set, succeed as soon as we match the whole pattern,
without further backtracking. */
#define RE_NO_POSIX_BACKTRACKING (RE_UNMATCHED_RIGHT_PAREN_ORD << 1)
/* If this bit is set, do not process the GNU regex operators.
If not set, then the GNU regex operators are recognized. */
#define RE_NO_GNU_OPS (RE_NO_POSIX_BACKTRACKING << 1)
/* If this bit is set, turn on internal regex debugging.
If not set, and debugging was on, turn it off.
This only works if regex.c is compiled -DDEBUG.
We define this bit always, so that all that's needed to turn on
debugging is to recompile regex.c; the calling code can always have
this bit set, and it won't affect anything in the normal case. */
#define RE_DEBUG (RE_NO_GNU_OPS << 1)
/* If this bit is set, a syntactically invalid interval is treated as
a string of ordinary characters. For example, the ERE 'a{1' is
treated as 'a\{1'. */
#define RE_INVALID_INTERVAL_ORD (RE_DEBUG << 1)
/* If this bit is set, then ignore case when matching.
If not set, then case is significant. */
#define RE_ICASE (RE_INVALID_INTERVAL_ORD << 1)
/* This bit is used internally like RE_CONTEXT_INDEP_ANCHORS but only
for ^, because it is difficult to scan the regex backwards to find
whether ^ should be special. */
#define RE_CARET_ANCHORS_HERE (RE_ICASE << 1)
/* If this bit is set, then \{ cannot be first in an bre or
immediately after an alternation or begin-group operator. */
#define RE_CONTEXT_INVALID_DUP (RE_CARET_ANCHORS_HERE << 1)
/* If this bit is set, then no_sub will be set to 1 during
re_compile_pattern. */
#define RE_NO_SUB (RE_CONTEXT_INVALID_DUP << 1)
/* This global variable defines the particular regexp syntax to use (for
some interfaces). When a regexp is compiled, the syntax used is
stored in the pattern buffer, so changing this does not affect
already-compiled regexps. */
extern reg_syntax_t re_syntax_options;
/* Define combinations of the above bits for the standard possibilities.
(The [[[ comments delimit what gets put into the Texinfo file, so
don't delete them!) */
/* [[[begin syntaxes]]] */
#define RE_SYNTAX_EMACS 0
#define RE_SYNTAX_AWK \
(RE_BACKSLASH_ESCAPE_IN_LISTS | RE_DOT_NOT_NULL \
| RE_NO_BK_PARENS | RE_NO_BK_REFS \
| RE_NO_BK_VBAR | RE_NO_EMPTY_RANGES \
| RE_DOT_NEWLINE | RE_CONTEXT_INDEP_ANCHORS \
| RE_UNMATCHED_RIGHT_PAREN_ORD | RE_NO_GNU_OPS)
#define RE_SYNTAX_GNU_AWK \
((RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS | RE_DEBUG) \
& ~(RE_DOT_NOT_NULL | RE_INTERVALS | RE_CONTEXT_INDEP_OPS \
| RE_CONTEXT_INVALID_OPS ))
#define RE_SYNTAX_POSIX_AWK \
(RE_SYNTAX_POSIX_EXTENDED | RE_BACKSLASH_ESCAPE_IN_LISTS \
| RE_INTERVALS | RE_NO_GNU_OPS)
#define RE_SYNTAX_GREP \
(RE_BK_PLUS_QM | RE_CHAR_CLASSES \
| RE_HAT_LISTS_NOT_NEWLINE | RE_INTERVALS \
| RE_NEWLINE_ALT)
#define RE_SYNTAX_EGREP \
(RE_CHAR_CLASSES | RE_CONTEXT_INDEP_ANCHORS \
| RE_CONTEXT_INDEP_OPS | RE_HAT_LISTS_NOT_NEWLINE \
| RE_NEWLINE_ALT | RE_NO_BK_PARENS \
| RE_NO_BK_VBAR)
#define RE_SYNTAX_POSIX_EGREP \
(RE_SYNTAX_EGREP | RE_INTERVALS | RE_NO_BK_BRACES \
| RE_INVALID_INTERVAL_ORD)
/* P1003.2/D11.2, section 4.20.7.1, lines 5078ff. */
#define RE_SYNTAX_ED RE_SYNTAX_POSIX_BASIC
#define RE_SYNTAX_SED RE_SYNTAX_POSIX_BASIC
/* Syntax bits common to both basic and extended POSIX regex syntax. */
#define _RE_SYNTAX_POSIX_COMMON \
(RE_CHAR_CLASSES | RE_DOT_NEWLINE | RE_DOT_NOT_NULL \
| RE_INTERVALS | RE_NO_EMPTY_RANGES)
#define RE_SYNTAX_POSIX_BASIC \
(_RE_SYNTAX_POSIX_COMMON | RE_BK_PLUS_QM | RE_CONTEXT_INVALID_DUP)
/* Differs from ..._POSIX_BASIC only in that RE_BK_PLUS_QM becomes
RE_LIMITED_OPS, i.e., \? \+ \| are not recognized. Actually, this
isn't minimal, since other operators, such as \`, aren't disabled. */
#define RE_SYNTAX_POSIX_MINIMAL_BASIC \
(_RE_SYNTAX_POSIX_COMMON | RE_LIMITED_OPS)
#define RE_SYNTAX_POSIX_EXTENDED \
(_RE_SYNTAX_POSIX_COMMON | RE_CONTEXT_INDEP_ANCHORS \
| RE_CONTEXT_INDEP_OPS | RE_NO_BK_BRACES \
| RE_NO_BK_PARENS | RE_NO_BK_VBAR \
| RE_CONTEXT_INVALID_OPS | RE_UNMATCHED_RIGHT_PAREN_ORD)
/* Differs from ..._POSIX_EXTENDED in that RE_CONTEXT_INDEP_OPS is
removed and RE_NO_BK_REFS is added. */
#define RE_SYNTAX_POSIX_MINIMAL_EXTENDED \
(_RE_SYNTAX_POSIX_COMMON | RE_CONTEXT_INDEP_ANCHORS \
| RE_CONTEXT_INVALID_OPS | RE_NO_BK_BRACES \
| RE_NO_BK_PARENS | RE_NO_BK_REFS \
| RE_NO_BK_VBAR | RE_UNMATCHED_RIGHT_PAREN_ORD)
/* [[[end syntaxes]]] */
/* Maximum number of duplicates an interval can allow. Some systems
(erroneously) define this in other header files, but we want our
value, so remove any previous define. */
#ifdef RE_DUP_MAX
# undef RE_DUP_MAX
#endif
/* If sizeof(int) == 2, then ((1 << 15) - 1) overflows. */
#define RE_DUP_MAX (0x7fff)
/* POSIX `cflags' bits (i.e., information for `regcomp'). */
/* If this bit is set, then use extended regular expression syntax.
If not set, then use basic regular expression syntax. */
#define REG_EXTENDED 1
/* If this bit is set, then ignore case when matching.
If not set, then case is significant. */
#define REG_ICASE (REG_EXTENDED << 1)
/* If this bit is set, then anchors do not match at newline
characters in the string.
If not set, then anchors do match at newlines. */
#define REG_NEWLINE (REG_ICASE << 1)
/* If this bit is set, then report only success or fail in regexec.
If not set, then returns differ between not matching and errors. */
#define REG_NOSUB (REG_NEWLINE << 1)
/* POSIX `eflags' bits (i.e., information for regexec). */
/* If this bit is set, then the beginning-of-line operator doesn't match
the beginning of the string (presumably because it's not the
beginning of a line).
If not set, then the beginning-of-line operator does match the
beginning of the string. */
#define REG_NOTBOL 1
/* Like REG_NOTBOL, except for the end-of-line. */
#define REG_NOTEOL (1 << 1)
/* Use PMATCH[0] to delimit the start and end of the search in the
buffer. */
#define REG_STARTEND (1 << 2)
/* If any error codes are removed, changed, or added, update the
`re_error_msg' table in regex.c. */
typedef enum
{
#ifdef _XOPEN_SOURCE
REG_ENOSYS = -1, /* This will never happen for this implementation. */
#endif
REG_NOERROR = 0, /* Success. */
REG_NOMATCH, /* Didn't find a match (for regexec). */
/* POSIX regcomp return error codes. (In the order listed in the
standard.) */
REG_BADPAT, /* Invalid pattern. */
REG_ECOLLATE, /* Inalid collating element. */
REG_ECTYPE, /* Invalid character class name. */
REG_EESCAPE, /* Trailing backslash. */
REG_ESUBREG, /* Invalid back reference. */
REG_EBRACK, /* Unmatched left bracket. */
REG_EPAREN, /* Parenthesis imbalance. */
REG_EBRACE, /* Unmatched \{. */
REG_BADBR, /* Invalid contents of \{\}. */
REG_ERANGE, /* Invalid range end. */
REG_ESPACE, /* Ran out of memory. */
REG_BADRPT, /* No preceding re for repetition op. */
/* Error codes we've added. */
REG_EEND, /* Premature end. */
REG_ESIZE, /* Compiled pattern bigger than 2^16 bytes. */
REG_ERPAREN /* Unmatched ) or \); not returned from regcomp. */
} reg_errcode_t;
/* This data structure represents a compiled pattern. Before calling
the pattern compiler, the fields `buffer', `allocated', `fastmap',
`translate', and `no_sub' can be set. After the pattern has been
compiled, the `re_nsub' field is available. All other fields are
private to the regex routines. */
#ifndef RE_TRANSLATE_TYPE
# define RE_TRANSLATE_TYPE unsigned char *
#endif
struct re_pattern_buffer
{
/* Space that holds the compiled pattern. It is declared as
`unsigned char *' because its elements are sometimes used as
array indexes. */
unsigned char *buffer;
/* Number of bytes to which `buffer' points. */
unsigned long int allocated;
/* Number of bytes actually used in `buffer'. */
unsigned long int used;
/* Syntax setting with which the pattern was compiled. */
reg_syntax_t syntax;
/* Pointer to a fastmap, if any, otherwise zero. re_search uses the
fastmap, if there is one, to skip over impossible starting points
for matches. */
char *fastmap;
/* Either a translate table to apply to all characters before
comparing them, or zero for no translation. The translation is
applied to a pattern when it is compiled and to a string when it
is matched. */
RE_TRANSLATE_TYPE translate;
/* Number of subexpressions found by the compiler. */
size_t re_nsub;
/* Zero if this pattern cannot match the empty string, one else.
Well, in truth it's used only in `re_search_2', to see whether or
not we should use the fastmap, so we don't set this absolutely
perfectly; see `re_compile_fastmap' (the `duplicate' case). */
unsigned can_be_null : 1;
/* If REGS_UNALLOCATED, allocate space in the `regs' structure
for `max (RE_NREGS, re_nsub + 1)' groups.
If REGS_REALLOCATE, reallocate space if necessary.
If REGS_FIXED, use what's there. */
#define REGS_UNALLOCATED 0
#define REGS_REALLOCATE 1
#define REGS_FIXED 2
unsigned regs_allocated : 2;
/* Set to zero when `regex_compile' compiles a pattern; set to one
by `re_compile_fastmap' if it updates the fastmap. */
unsigned fastmap_accurate : 1;
/* If set, `re_match_2' does not return information about
subexpressions. */
unsigned no_sub : 1;
/* If set, a beginning-of-line anchor doesn't match at the beginning
of the string. */
unsigned not_bol : 1;
/* Similarly for an end-of-line anchor. */
unsigned not_eol : 1;
/* If true, an anchor at a newline matches. */
unsigned newline_anchor : 1;
};
typedef struct re_pattern_buffer regex_t;
/* Type for byte offsets within the string. POSIX mandates this. */
typedef int regoff_t;
/* This is the structure we store register match data in. See
regex.texinfo for a full description of what registers match. */
struct re_registers
{
unsigned num_regs;
regoff_t *start;
regoff_t *end;
};
/* If `regs_allocated' is REGS_UNALLOCATED in the pattern buffer,
`re_match_2' returns information about at least this many registers
the first time a `regs' structure is passed. */
#ifndef RE_NREGS
# define RE_NREGS 30
#endif
/* POSIX specification for registers. Aside from the different names than
`re_registers', POSIX uses an array of structures, instead of a
structure of arrays. */
typedef struct
{
regoff_t rm_so; /* Byte offset from string's start to substring's start. */
regoff_t rm_eo; /* Byte offset from string's start to substring's end. */
} regmatch_t;
/* Declarations for routines. */
/* Sets the current default syntax to SYNTAX, and return the old syntax.
You can also simply assign to the `re_syntax_options' variable. */
extern reg_syntax_t re_set_syntax (reg_syntax_t __syntax);
/* Compile the regular expression PATTERN, with length LENGTH
and syntax given by the global `re_syntax_options', into the buffer
BUFFER. Return NULL if successful, and an error string if not. */
extern const char *re_compile_pattern (const char *__pattern, size_t __length,
struct re_pattern_buffer *__buffer);
/* Compile a fastmap for the compiled pattern in BUFFER; used to
accelerate searches. Return 0 if successful and -2 if was an
internal error. */
extern int re_compile_fastmap (struct re_pattern_buffer *__buffer);
/* Search in the string STRING (with length LENGTH) for the pattern
compiled into BUFFER. Start searching at position START, for RANGE
characters. Return the starting position of the match, -1 for no
match, or -2 for an internal error. Also return register
information in REGS (if REGS and BUFFER->no_sub are nonzero). */
extern int re_search (struct re_pattern_buffer *__buffer, const char *__string,
int __length, int __start, int __range,
struct re_registers *__regs);
/* Like `re_search', but search in the concatenation of STRING1 and
STRING2. Also, stop searching at index START + STOP. */
extern int re_search_2 (struct re_pattern_buffer *__buffer,
const char *__string1, int __length1,
const char *__string2, int __length2, int __start,
int __range, struct re_registers *__regs, int __stop);
/* Like `re_search', but return how many characters in STRING the regexp
in BUFFER matched, starting at position START. */
extern int re_match (struct re_pattern_buffer *__buffer, const char *__string,
int __length, int __start, struct re_registers *__regs);
/* Relates to `re_match' as `re_search_2' relates to `re_search'. */
extern int re_match_2 (struct re_pattern_buffer *__buffer,
const char *__string1, int __length1,
const char *__string2, int __length2, int __start,
struct re_registers *__regs, int __stop);
/* Set REGS to hold NUM_REGS registers, storing them in STARTS and
ENDS. Subsequent matches using BUFFER and REGS will use this memory
for recording register information. STARTS and ENDS must be
allocated with malloc, and must each be at least `NUM_REGS * sizeof
(regoff_t)' bytes long.
If NUM_REGS == 0, then subsequent matches should allocate their own
register data.
Unless this function is called, the first search or match using
PATTERN_BUFFER will allocate its own register data, without
freeing the old data. */
extern void re_set_registers (struct re_pattern_buffer *__buffer,
struct re_registers *__regs,
unsigned int __num_regs,
regoff_t *__starts, regoff_t *__ends);
#if defined _REGEX_RE_COMP || defined _LIBC
# ifndef _CRAY
/* 4.2 bsd compatibility. */
extern char *re_comp (const char *);
extern int re_exec (const char *);
# endif
#endif
/* GCC 2.95 and later have "__restrict"; C99 compilers have
"restrict", and "configure" may have defined "restrict". */
#ifndef __restrict
# if ! (2 < __GNUC__ || (2 == __GNUC__ && 95 <= __GNUC_MINOR__))
# if defined restrict || 199901L <= __STDC_VERSION__
# define __restrict restrict
# else
# define __restrict
# endif
# endif
#endif
/* gcc 3.1 and up support the [restrict] syntax. */
#ifndef __restrict_arr
# if (__GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ >= 1)) \
&& !defined __GNUG__
# define __restrict_arr __restrict
# else
# define __restrict_arr
# endif
#endif
/* POSIX compatibility. */
extern int regcomp (regex_t *__restrict __preg,
const char *__restrict __pattern,
int __cflags);
extern int regexec (const regex_t *__restrict __preg,
const char *__restrict __string, size_t __nmatch,
regmatch_t __pmatch[__restrict_arr],
int __eflags);
extern size_t regerror (int __errcode, const regex_t *__restrict __preg,
char *__restrict __errbuf, size_t __errbuf_size);
extern void regfree (regex_t *__preg);
#ifdef __cplusplus
}
#endif /* C++ */
#endif /* regex.h */
/*!
* \file
*
* \brief Various routines with dealing with sparse graphs
*
* \author George Karypis
* \version\verbatim $Id: graph.c 21482 2017-09-18 05:53:12Z karypis $ \endverbatim
*/
#include <GKlib.h>
#define OMPMINOPS 50000
/*************************************************************************/
/*! Allocate memory for a graph and initializes it
\returns the allocated graph. The various fields are set to NULL.
*/
/**************************************************************************/
gk_graph_t *gk_graph_Create()
{
gk_graph_t *graph;
graph = (gk_graph_t *)gk_malloc(sizeof(gk_graph_t), "gk_graph_Create: graph");
gk_graph_Init(graph);
return graph;
}
/*************************************************************************/
/*! Initializes the graph.
\param graph is the graph to be initialized.
*/
/*************************************************************************/
void gk_graph_Init(gk_graph_t *graph)
{
memset(graph, 0, sizeof(gk_graph_t));
graph->nvtxs = -1;
}
/*************************************************************************/
/*! Frees all the memory allocated for a graph.
\param graph is the graph to be freed.
*/
/*************************************************************************/
void gk_graph_Free(gk_graph_t **graph)
{
if (*graph == NULL)
return;
gk_graph_FreeContents(*graph);
gk_free((void **)graph, LTERM);
}
/*************************************************************************/
/*! Frees only the memory allocated for the graph's different fields and
sets them to NULL.
\param graph is the graph whose contents will be freed.
*/
/*************************************************************************/
void gk_graph_FreeContents(gk_graph_t *graph)
{
gk_free((void *)&graph->xadj, &graph->adjncy,
&graph->iadjwgt, &graph->fadjwgt,
&graph->ivwgts, &graph->fvwgts,
&graph->ivsizes, &graph->fvsizes,
&graph->vlabels,
LTERM);
}
/**************************************************************************/
/*! Reads a sparse graph from the supplied file
\param filename is the file that stores the data.
\param format is the graph format. The supported values are:
GK_GRAPH_FMT_METIS, GK_GRAPH_FMT_IJV.
\param hasvals is 1 if the input file has values
\param numbering is 1 if the input file numbering starts from one
\param isfewgts is 1 if the edge-weights should be read as floats
\param isfvwgts is 1 if the vertex-weights should be read as floats
\param isfvsizes is 1 if the vertex-sizes should be read as floats
\returns the graph that was read.
*/
/**************************************************************************/
gk_graph_t *gk_graph_Read(char *filename, int format, int hasvals,
int numbering, int isfewgts, int isfvwgts, int isfvsizes)
{
ssize_t i, k, l;
size_t nfields, nvtxs, nedges, fmt, ncon, lnlen;
ssize_t *xadj;
int32_t ival, *iinds=NULL, *jinds=NULL, *ivals=NULL, *adjncy, *iadjwgt;
float fval, *fvals=NULL, *fadjwgt;
int readsizes=0, readwgts=0, readvals=0;
char *line=NULL, *head, *tail, fmtstr[256];
FILE *fpin=NULL;
gk_graph_t *graph=NULL;
if (!gk_fexists(filename))
gk_errexit(SIGERR, "File %s does not exist!\n", filename);
switch (format) {
case GK_GRAPH_FMT_METIS:
fpin = gk_fopen(filename, "r", "gk_graph_Read: fpin");
do {
if (gk_getline(&line, &lnlen, fpin) <= 0)
gk_errexit(SIGERR, "Premature end of input file: file:%s\n", filename);
} while (line[0] == '%');
fmt = ncon = 0;
nfields = sscanf(line, "%zu %zu %zu %zu", &nvtxs, &nedges, &fmt, &ncon);
if (nfields < 2)
gk_errexit(SIGERR, "Header line must contain at least 2 integers (#vtxs and #edges).\n");
nedges *= 2;
if (fmt > 111)
gk_errexit(SIGERR, "Cannot read this type of file format [fmt=%zu]!\n", fmt);
sprintf(fmtstr, "%03zu", fmt%1000);
readsizes = (fmtstr[0] == '1');
readwgts = (fmtstr[1] == '1');
readvals = (fmtstr[2] == '1');
numbering = 1;
ncon = (ncon == 0 ? 1 : ncon);
graph = gk_graph_Create();
graph->nvtxs = nvtxs;
graph->xadj = gk_zmalloc(nvtxs+1, "gk_graph_Read: xadj");
graph->adjncy = gk_i32malloc(nedges, "gk_graph_Read: adjncy");
if (readvals) {
if (isfewgts)
graph->fadjwgt = gk_fmalloc(nedges, "gk_graph_Read: fadjwgt");
else
graph->iadjwgt = gk_i32malloc(nedges, "gk_graph_Read: iadjwgt");
}
if (readsizes) {
if (isfvsizes)
graph->fvsizes = gk_fmalloc(nvtxs, "gk_graph_Read: fvsizes");
else
graph->ivsizes = gk_i32malloc(nvtxs, "gk_graph_Read: ivsizes");
}
if (readwgts) {
if (isfvwgts)
graph->fvwgts = gk_fmalloc(nvtxs*ncon, "gk_graph_Read: fvwgts");
else
graph->ivwgts = gk_i32malloc(nvtxs*ncon, "gk_graph_Read: ivwgts");
}
/*----------------------------------------------------------------------
* Read the sparse graph file
*---------------------------------------------------------------------*/
numbering = (numbering ? - 1 : 0);
for (graph->xadj[0]=0, k=0, i=0; i<nvtxs; i++) {
do {
if (gk_getline(&line, &lnlen, fpin) == -1)
gk_errexit(SIGERR, "Pregraphure end of input file: file while reading row %d\n", i);
} while (line[0] == '%');
head = line;
tail = NULL;
/* Read vertex sizes */
if (readsizes) {
if (isfvsizes) {
#ifdef __MSC__
graph->fvsizes[i] = (float)strtod(head, &tail);
#else
graph->fvsizes[i] = strtof(head, &tail);
#endif
if (tail == head)
gk_errexit(SIGERR, "The line for vertex %zd does not have size information\n", i+1);
if (graph->fvsizes[i] < 0)
gk_errexit(SIGERR, "The size for vertex %zd must be >= 0\n", i+1);
}
else {
graph->ivsizes[i] = strtol(head, &tail, 0);
if (tail == head)
gk_errexit(SIGERR, "The line for vertex %zd does not have size information\n", i+1);
if (graph->ivsizes[i] < 0)
gk_errexit(SIGERR, "The size for vertex %zd must be >= 0\n", i+1);
}
head = tail;
}
/* Read vertex weights */
if (readwgts) {
for (l=0; l<ncon; l++) {
if (isfvwgts) {
#ifdef __MSC__
graph->fvwgts[i*ncon+l] = (float)strtod(head, &tail);
#else
graph->fvwgts[i*ncon+l] = strtof(head, &tail);
#endif
if (tail == head)
gk_errexit(SIGERR, "The line for vertex %zd does not have enough weights "
"for the %d constraints.\n", i+1, ncon);
if (graph->fvwgts[i*ncon+l] < 0)
gk_errexit(SIGERR, "The weight vertex %zd and constraint %zd must be >= 0\n", i+1, l);
}
else {
graph->ivwgts[i*ncon+l] = strtol(head, &tail, 0);
if (tail == head)
gk_errexit(SIGERR, "The line for vertex %zd does not have enough weights "
"for the %d constraints.\n", i+1, ncon);
if (graph->ivwgts[i*ncon+l] < 0)
gk_errexit(SIGERR, "The weight vertex %zd and constraint %zd must be >= 0\n", i+1, l);
}
head = tail;
}
}
/* Read the rest of the row */
while (1) {
ival = (int)strtol(head, &tail, 0);
if (tail == head)
break;
head = tail;
if ((graph->adjncy[k] = ival + numbering) < 0)
gk_errexit(SIGERR, "Error: Invalid column number %d at row %zd.\n", ival, i);
if (readvals) {
if (isfewgts) {
#ifdef __MSC__
fval = (float)strtod(head, &tail);
#else
fval = strtof(head, &tail);
#endif
if (tail == head)
gk_errexit(SIGERR, "Value could not be found for edge! Vertex:%zd, NNZ:%zd\n", i, k);
graph->fadjwgt[k] = fval;
}
else {
ival = strtol(head, &tail, 0);
if (tail == head)
gk_errexit(SIGERR, "Value could not be found for edge! Vertex:%zd, NNZ:%zd\n", i, k);
graph->iadjwgt[k] = ival;
}
head = tail;
}
k++;
}
graph->xadj[i+1] = k;
}
if (k != nedges)
gk_errexit(SIGERR, "gk_graph_Read: Something wrong with the number of edges in "
"the input file. nedges=%zd, Actualnedges=%zd.\n", nedges, k);
gk_fclose(fpin);
gk_free((void **)&line, LTERM);
break;
case GK_GRAPH_FMT_IJV:
gk_getfilestats(filename, &nvtxs, &nedges, NULL, NULL);
if (hasvals == 1 && 3*nvtxs != nedges)
gk_errexit(SIGERR, "Error: The number of numbers (%zd %d) in the input file is not a multiple of 3.\n", nedges, hasvals);
if (hasvals == 0 && 2*nvtxs != nedges)
gk_errexit(SIGERR, "Error: The number of numbers (%zd %d) in the input file is not a multiple of 2.\n", nedges, hasvals);
nedges = nvtxs;
numbering = (numbering ? - 1 : 0);
/* read the data into three arrays */
iinds = gk_i32malloc(nedges, "iinds");
jinds = gk_i32malloc(nedges, "jinds");
if (hasvals) {
if (isfewgts)
fvals = gk_fmalloc(nedges, "fvals");
else
ivals = gk_i32malloc(nedges, "ivals");
}
fpin = gk_fopen(filename, "r", "gk_graph_Read: fpin");
for (nvtxs=0, i=0; i<nedges; i++) {
if (hasvals) {
if (isfewgts) {
if (fscanf(fpin, "%"PRId32" %"PRId32" %f", &iinds[i], &jinds[i], &fvals[i]) != 3)
gk_errexit(SIGERR, "Error: Failed to read (i, j, val) for nedge: %zd.\n", i);
}
else {
if (fscanf(fpin, "%"PRId32" %"PRId32" %"PRId32, &iinds[i], &jinds[i], &ivals[i]) != 3)
gk_errexit(SIGERR, "Error: Failed to read (i, j, val) for nedge: %zd.\n", i);
}
}
else {
if (fscanf(fpin, "%"PRId32" %"PRId32, &iinds[i], &jinds[i]) != 2)
gk_errexit(SIGERR, "Error: Failed to read (i, j) value for nedge: %zd.\n", i);
}
iinds[i] += numbering;
jinds[i] += numbering;
if (nvtxs < iinds[i])
nvtxs = iinds[i];
if (nvtxs < jinds[i])
nvtxs = jinds[i];
}
gk_fclose(fpin);
/* convert (i, j, v) into a graph format */
graph = gk_graph_Create();
graph->nvtxs = ++nvtxs;
xadj = graph->xadj = gk_zsmalloc(nvtxs+1, 0, "xadj");
adjncy = graph->adjncy = gk_i32malloc(nedges, "adjncy");
if (hasvals) {
if (isfewgts)
fadjwgt = graph->fadjwgt = gk_fmalloc(nedges, "fadjwgt");
else
iadjwgt = graph->iadjwgt = gk_i32malloc(nedges, "iadjwgt");
}
for (i=0; i<nedges; i++)
xadj[iinds[i]]++;
MAKECSR(i, nvtxs, xadj);
for (i=0; i<nedges; i++) {
adjncy[xadj[iinds[i]]] = jinds[i];
if (hasvals) {
if (isfewgts)
fadjwgt[xadj[iinds[i]]] = fvals[i];
else
iadjwgt[xadj[iinds[i]]] = ivals[i];
}
xadj[iinds[i]]++;
}
SHIFTCSR(i, nvtxs, xadj);
gk_free((void **)&iinds, &jinds, &fvals, &ivals, LTERM);
break;
default:
gk_errexit(SIGERR, "Unrecognized format: %d\n", format);
}
return graph;
}
/**************************************************************************/
/*! Writes a graph into a file.
\param graph is the graph to be written,
\param filename is the name of the output file.
\param format is one of GK_GRAPH_FMT_METIS specifying
the format of the output file.
*/
/**************************************************************************/
void gk_graph_Write(gk_graph_t *graph, char *filename, int format)
{
ssize_t i, j;
int hasvwgts, hasvsizes, hasewgts;
FILE *fpout;
if (format != GK_GRAPH_FMT_METIS)
gk_errexit(SIGERR, "Unknown file format. %d\n", format);
if (filename)
fpout = gk_fopen(filename, "w", "gk_graph_Write: fpout");
else
fpout = stdout;
hasewgts = (graph->iadjwgt || graph->fadjwgt);
hasvwgts = (graph->ivwgts || graph->fvwgts);
hasvsizes = (graph->ivsizes || graph->fvsizes);
/* write the header line */
fprintf(fpout, "%d %zd", graph->nvtxs, graph->xadj[graph->nvtxs]/2);
if (hasvwgts || hasvsizes || hasewgts)
fprintf(fpout, " %d%d%d", hasvsizes, hasvwgts, hasewgts);
fprintf(fpout, "\n");
for (i=0; i<graph->nvtxs; i++) {
if (hasvsizes) {
if (graph->ivsizes)
fprintf(fpout, " %d", graph->ivsizes[i]);
else
fprintf(fpout, " %f", graph->fvsizes[i]);
}
if (hasvwgts) {
if (graph->ivwgts)
fprintf(fpout, " %d", graph->ivwgts[i]);
else
fprintf(fpout, " %f", graph->fvwgts[i]);
}
for (j=graph->xadj[i]; j<graph->xadj[i+1]; j++) {
fprintf(fpout, " %d", graph->adjncy[j]+1);
if (hasewgts) {
if (graph->iadjwgt)
fprintf(fpout, " %d", graph->iadjwgt[j]);
else
fprintf(fpout, " %f", graph->fadjwgt[j]);
}
}
fprintf(fpout, "\n");
}
if (filename)
gk_fclose(fpout);
}
/*************************************************************************/
/*! Returns a copy of a graph.
\param graph is the graph to be duplicated.
\returns the newly created copy of the graph.
*/
/**************************************************************************/
gk_graph_t *gk_graph_Dup(gk_graph_t *graph)
{
gk_graph_t *ngraph;
ngraph = gk_graph_Create();
ngraph->nvtxs = graph->nvtxs;
/* copy the adjacency structure */
if (graph->xadj)
ngraph->xadj = gk_zcopy(graph->nvtxs+1, graph->xadj,
gk_zmalloc(graph->nvtxs+1, "gk_graph_Dup: xadj"));
if (graph->ivwgts)
ngraph->ivwgts = gk_i32copy(graph->nvtxs, graph->ivwgts,
gk_i32malloc(graph->nvtxs, "gk_graph_Dup: ivwgts"));
if (graph->ivsizes)
ngraph->ivsizes = gk_i32copy(graph->nvtxs, graph->ivsizes,
gk_i32malloc(graph->nvtxs, "gk_graph_Dup: ivsizes"));
if (graph->vlabels)
ngraph->vlabels = gk_i32copy(graph->nvtxs, graph->vlabels,
gk_i32malloc(graph->nvtxs, "gk_graph_Dup: ivlabels"));
if (graph->fvwgts)
ngraph->fvwgts = gk_fcopy(graph->nvtxs, graph->fvwgts,
gk_fmalloc(graph->nvtxs, "gk_graph_Dup: fvwgts"));
if (graph->fvsizes)
ngraph->fvsizes = gk_fcopy(graph->nvtxs, graph->fvsizes,
gk_fmalloc(graph->nvtxs, "gk_graph_Dup: fvsizes"));
if (graph->adjncy)
ngraph->adjncy = gk_i32copy(graph->xadj[graph->nvtxs], graph->adjncy,
gk_i32malloc(graph->xadj[graph->nvtxs], "gk_graph_Dup: adjncy"));
if (graph->iadjwgt)
ngraph->iadjwgt = gk_i32copy(graph->xadj[graph->nvtxs], graph->iadjwgt,
gk_i32malloc(graph->xadj[graph->nvtxs], "gk_graph_Dup: iadjwgt"));
if (graph->fadjwgt)
ngraph->fadjwgt = gk_fcopy(graph->xadj[graph->nvtxs], graph->fadjwgt,
gk_fmalloc(graph->xadj[graph->nvtxs], "gk_graph_Dup: fadjwgt"));
return ngraph;
}
/*************************************************************************/
/*! Returns a subgraph containing a set of consecutive vertices.
\param graph is the original graph.
\param vstart is the starting vertex.
\param nvtxs is the number of vertices from vstart to extract.
\returns the newly created subgraph.
*/
/**************************************************************************/
gk_graph_t *gk_graph_ExtractSubgraph(gk_graph_t *graph, int vstart, int nvtxs)
{
ssize_t i;
gk_graph_t *ngraph;
if (vstart+nvtxs > graph->nvtxs)
return NULL;
ngraph = gk_graph_Create();
ngraph->nvtxs = nvtxs;
/* copy the adjancy structure */
if (graph->xadj)
ngraph->xadj = gk_zcopy(nvtxs+1, graph->xadj+vstart,
gk_zmalloc(nvtxs+1, "gk_graph_ExtractSubgraph: xadj"));
for (i=nvtxs; i>=0; i--)
ngraph->xadj[i] -= ngraph->xadj[0];
ASSERT(ngraph->xadj[0] == 0);
if (graph->ivwgts)
ngraph->ivwgts = gk_i32copy(nvtxs, graph->ivwgts+vstart,
gk_i32malloc(nvtxs, "gk_graph_ExtractSubgraph: ivwgts"));
if (graph->ivsizes)
ngraph->ivsizes = gk_i32copy(nvtxs, graph->ivsizes+vstart,
gk_i32malloc(nvtxs, "gk_graph_ExtractSubgraph: ivsizes"));
if (graph->vlabels)
ngraph->vlabels = gk_i32copy(nvtxs, graph->vlabels+vstart,
gk_i32malloc(nvtxs, "gk_graph_ExtractSubgraph: vlabels"));
if (graph->fvwgts)
ngraph->fvwgts = gk_fcopy(nvtxs, graph->fvwgts+vstart,
gk_fmalloc(nvtxs, "gk_graph_ExtractSubgraph: fvwgts"));
if (graph->fvsizes)
ngraph->fvsizes = gk_fcopy(nvtxs, graph->fvsizes+vstart,
gk_fmalloc(nvtxs, "gk_graph_ExtractSubgraph: fvsizes"));
ASSERT(ngraph->xadj[nvtxs] == graph->xadj[vstart+nvtxs]-graph->xadj[vstart]);
if (graph->adjncy)
ngraph->adjncy = gk_i32copy(graph->xadj[vstart+nvtxs]-graph->xadj[vstart],
graph->adjncy+graph->xadj[vstart],
gk_i32malloc(graph->xadj[vstart+nvtxs]-graph->xadj[vstart],
"gk_graph_ExtractSubgraph: adjncy"));
if (graph->iadjwgt)
ngraph->iadjwgt = gk_i32copy(graph->xadj[vstart+nvtxs]-graph->xadj[vstart],
graph->iadjwgt+graph->xadj[vstart],
gk_i32malloc(graph->xadj[vstart+nvtxs]-graph->xadj[vstart],
"gk_graph_ExtractSubgraph: iadjwgt"));
if (graph->fadjwgt)
ngraph->fadjwgt = gk_fcopy(graph->xadj[vstart+nvtxs]-graph->xadj[vstart],
graph->fadjwgt+graph->xadj[vstart],
gk_fmalloc(graph->xadj[vstart+nvtxs]-graph->xadj[vstart],
"gk_graph_ExtractSubgraph: fadjwgt"));
return ngraph;
}
/*************************************************************************/
/*! Returns a graph that has been reordered according to the permutation.
\param[IN] graph is the graph to be re-ordered.
\param[IN] perm is the new ordering of the graph's vertices
\param[IN] iperm is the original ordering of the re-ordered graph's vertices
\returns the newly created copy of the graph.
\note Either perm or iperm can be NULL but not both.
*/
/**************************************************************************/
gk_graph_t *gk_graph_Reorder(gk_graph_t *graph, int32_t *perm, int32_t *iperm)
{
ssize_t j, jj, *xadj;
int i, k, u, v, nvtxs;
int freeperm=0, freeiperm=0;
int32_t *adjncy;
gk_graph_t *ngraph;
if (perm == NULL && iperm == NULL)
return NULL;
ngraph = gk_graph_Create();
ngraph->nvtxs = nvtxs = graph->nvtxs;
xadj = graph->xadj;
adjncy = graph->adjncy;
/* allocate memory for the different structures that are present in graph */
if (graph->xadj)
ngraph->xadj = gk_zmalloc(nvtxs+1, "gk_graph_Reorder: xadj");
if (graph->ivwgts)
ngraph->ivwgts = gk_i32malloc(nvtxs, "gk_graph_Reorder: ivwgts");
if (graph->ivsizes)
ngraph->ivsizes = gk_i32malloc(nvtxs, "gk_graph_Reorder: ivsizes");
if (graph->vlabels)
ngraph->vlabels = gk_i32malloc(nvtxs, "gk_graph_Reorder: ivlabels");
if (graph->fvwgts)
ngraph->fvwgts = gk_fmalloc(nvtxs, "gk_graph_Reorder: fvwgts");
if (graph->fvsizes)
ngraph->fvsizes = gk_fmalloc(nvtxs, "gk_graph_Reorder: fvsizes");
if (graph->adjncy)
ngraph->adjncy = gk_i32malloc(graph->xadj[nvtxs], "gk_graph_Reorder: adjncy");
if (graph->iadjwgt)
ngraph->iadjwgt = gk_i32malloc(graph->xadj[nvtxs], "gk_graph_Reorder: iadjwgt");
if (graph->fadjwgt)
ngraph->fadjwgt = gk_fmalloc(graph->xadj[nvtxs], "gk_graph_Reorder: fadjwgt");
/* create perm/iperm if not provided */
if (perm == NULL) {
freeperm = 1;
perm = gk_i32malloc(nvtxs, "gk_graph_Reorder: perm");
for (i=0; i<nvtxs; i++)
perm[iperm[i]] = i;
}
if (iperm == NULL) {
freeiperm = 1;
iperm = gk_i32malloc(nvtxs, "gk_graph_Reorder: iperm");
for (i=0; i<nvtxs; i++)
iperm[perm[i]] = i;
}
/* fill-in the information of the re-ordered graph */
ngraph->xadj[0] = jj = 0;
for (v=0; v<nvtxs; v++) {
u = iperm[v];
for (j=xadj[u]; j<xadj[u+1]; j++, jj++) {
ngraph->adjncy[jj] = perm[adjncy[j]];
if (graph->iadjwgt)
ngraph->iadjwgt[jj] = graph->iadjwgt[j];
if (graph->fadjwgt)
ngraph->fadjwgt[jj] = graph->fadjwgt[j];
}
if (graph->ivwgts)
ngraph->ivwgts[v] = graph->ivwgts[u];
if (graph->fvwgts)
ngraph->fvwgts[v] = graph->fvwgts[u];
if (graph->ivsizes)
ngraph->ivsizes[v] = graph->ivsizes[u];
if (graph->fvsizes)
ngraph->fvsizes[v] = graph->fvsizes[u];
if (graph->vlabels)
ngraph->vlabels[v] = graph->vlabels[u];
ngraph->xadj[v+1] = jj;
}
/* free memory */
if (freeperm)
gk_free((void **)&perm, LTERM);
if (freeiperm)
gk_free((void **)&iperm, LTERM);
return ngraph;
}
/*************************************************************************/
/*! This function finds the connected components in a graph.
\param graph is the graph structure
\param cptr is the ptr structure of the CSR representation of the
components. The length of this vector must be graph->nvtxs+1.
\param cind is the indices structure of the CSR representation of
the components. The length of this vector must be graph->nvtxs.
\returns the number of components that it found.
\note The cptr and cind parameters can be NULL, in which case only the
number of connected components is returned.
*/
/*************************************************************************/
int gk_graph_FindComponents(gk_graph_t *graph, int32_t *cptr, int32_t *cind)
{
ssize_t i, ii, j, jj, k, nvtxs, first, last, ntodo, ncmps;
ssize_t *xadj;
int32_t *adjncy, *pos, *todo;
int32_t mustfree_ccsr=0;
nvtxs = graph->nvtxs;
xadj = graph->xadj;
adjncy = graph->adjncy;
/* Deal with NULL supplied cptr/cind vectors */
if (cptr == NULL) {
cptr = gk_i32malloc(nvtxs+1, "gk_graph_FindComponents: cptr");
cind = gk_i32malloc(nvtxs, "gk_graph_FindComponents: cind");
mustfree_ccsr = 1;
}
/* The list of vertices that have not been touched yet.
The valid entries are from [0..ntodo). */
todo = gk_i32incset(nvtxs, 0, gk_i32malloc(nvtxs, "gk_graph_FindComponents: todo"));
/* For a vertex that has not been visited, pos[i] is the position in the
todo list that this vertex is stored.
If a vertex has been visited, pos[i] = -1. */
pos = gk_i32incset(nvtxs, 0, gk_i32malloc(nvtxs, "gk_graph_FindComponents: pos"));
/* Find the connected componends */
ncmps = -1;
ntodo = nvtxs; /* All vertices have not been visited */
first = last = 0; /* Point to the first and last vertices that have been touched
but not explored.
These vertices are stored in cind[first]...cind[last-1]. */
while (1) {
if (first == last) { /* Find another starting vertex */
cptr[++ncmps] = first; /* Mark the end of the current CC */
if (ntodo > 0) {
/* put the first vertex in the todo list as the start of the new CC */
GKASSERT(pos[todo[0]] != -1);
cind[last++] = todo[0];
pos[todo[0]] = -1;
todo[0] = todo[--ntodo];
pos[todo[0]] = 0;
}
else {
break;
}
}
i = cind[first++]; /* Get the first visited but unexplored vertex */
for (j=xadj[i]; j<xadj[i+1]; j++) {
k = adjncy[j];
if (pos[k] != -1) {
cind[last++] = k;
/* Remove k from the todo list and put the last item in the todo
list at the position that k was so that the todo list will be
consequtive. The pos[] array is updated accordingly to keep track
the location of the vertices in the todo[] list. */
todo[pos[k]] = todo[--ntodo];
pos[todo[pos[k]]] = pos[k];
pos[k] = -1;
}
}
}
GKASSERT(first == nvtxs);
if (mustfree_ccsr)
gk_free((void **)&cptr, &cind, LTERM);
gk_free((void **)&pos, &todo, LTERM);
return (int) ncmps;
}
/*************************************************************************/
/*! This function computes a permutation of the vertices based on a
breadth-first-traversal. It can be used for re-ordering the graph
to reduce its bandwidth for better cache locality.
The algorithm used is a simplified version of the method used to find
the connected components.
\param[IN] graph is the graph structure
\param[IN] v is the starting vertex of the BFS
\param[OUT] perm[i] stores the ID of vertex i in the re-ordered graph.
\param[OUT] iperm[i] stores the ID of the vertex that corresponds to
the ith vertex in the re-ordered graph.
\note The perm or iperm (but not both) can be NULL, at which point,
the corresponding arrays are not returned. Though the program
works fine when both are NULL, doing that is not smart.
The returned arrays should be freed with gk_free().
*/
/*************************************************************************/
void gk_graph_ComputeBFSOrdering(gk_graph_t *graph, int v, int32_t **r_perm,
int32_t **r_iperm)
{
ssize_t j, *xadj;
int i, k, nvtxs, first, last;
int32_t *adjncy, *cot, *pos;
if (graph->nvtxs <= 0)
return;
nvtxs = graph->nvtxs;
xadj = graph->xadj;
adjncy = graph->adjncy;
/* This array will function like pos + touched of the CC method */
pos = gk_i32incset(nvtxs, 0, gk_i32malloc(nvtxs, "gk_graph_ComputeBFSOrdering: pos"));
/* This array ([C]losed[O]pen[T]odo => cot) serves three purposes.
Positions from [0...first) is the current iperm[] vector of the explored vertices;
Positions from [first...last) is the OPEN list (i.e., visited vertices);
Positions from [last...nvtxs) is the todo list. */
cot = gk_i32incset(nvtxs, 0, gk_i32malloc(nvtxs, "gk_graph_ComputeBFSOrdering: cot"));
/* put v at the front of the todo list */
pos[0] = cot[0] = v;
pos[v] = cot[v] = 0;
/* compute a BFS ordering from the seed vertex */
first = last = 0;
while (first < nvtxs) {
if (first == last) { /* Find another starting vertex */
k = cot[last];
ASSERT(pos[k] != -1);
pos[k] = -1; /* mark node as being visited */
last++;
}
i = cot[first++]; /* the ++ advances the explored vertices */
for (j=xadj[i]; j<xadj[i+1]; j++) {
k = adjncy[j];
/* if a node has already been visited, its pos[] will be -1 */
if (pos[k] != -1) {
/* pos[k] is the location within cot[] where k resides (it is in the 'todo' part);
It is placed in that location cot[last] (end of OPEN list) that we
are about to overwrite and update pos[cot[last]] to reflect that. */
cot[pos[k]] = cot[last]; /* put the head of the todo list to
where k was in the todo list */
pos[cot[last]] = pos[k]; /* update perm to reflect the move */
cot[last++] = k; /* put node at the end of the OPEN list */
pos[k] = -1; /* mark node as being visited */
}
}
}
/* time to decide what to return */
if (r_perm != NULL) {
/* use the 'pos' array to build the perm array */
for (i=0; i<nvtxs; i++)
pos[cot[i]] = i;
*r_perm = pos;
pos = NULL;
}
if (r_iperm != NULL) {
*r_iperm = cot;
cot = NULL;
}
/* cleanup memory */
gk_free((void **)&pos, &cot, LTERM);
}
/*************************************************************************/
/*! This function computes a permutation of the vertices based on a
best-first-traversal. It can be used for re-ordering the graph
to reduce its bandwidth for better cache locality.
\param[IN] graph is the graph structure.
\param[IN] v is the starting vertex of the best-first traversal.
\param[IN] type indicates the criteria to use to measure the 'bestness'
of a vertex.
\param[OUT] perm[i] stores the ID of vertex i in the re-ordered graph.
\param[OUT] iperm[i] stores the ID of the vertex that corresponds to
the ith vertex in the re-ordered graph.
\note The perm or iperm (but not both) can be NULL, at which point,
the corresponding arrays are not returned. Though the program
works fine when both are NULL, doing that is not smart.
The returned arrays should be freed with gk_free().
*/
/*************************************************************************/
void gk_graph_ComputeBestFOrdering0(gk_graph_t *graph, int v, int type,
int32_t **r_perm, int32_t **r_iperm)
{
ssize_t j, jj, *xadj;
int i, k, u, nvtxs;
int32_t *adjncy, *perm, *degrees, *minIDs, *open;
gk_i32pq_t *queue;
if (graph->nvtxs <= 0)
return;
nvtxs = graph->nvtxs;
xadj = graph->xadj;
adjncy = graph->adjncy;
/* the degree of the vertices in the closed list */
degrees = gk_i32smalloc(nvtxs, 0, "gk_graph_ComputeBestFOrdering: degrees");
/* the minimum vertex ID of an open vertex to the closed list */
minIDs = gk_i32smalloc(nvtxs, nvtxs+1, "gk_graph_ComputeBestFOrdering: minIDs");
/* the open list */
open = gk_i32malloc(nvtxs, "gk_graph_ComputeBestFOrdering: open");
/* if perm[i] >= 0, then perm[i] is the order of vertex i;
otherwise perm[i] == -1.
*/
perm = gk_i32smalloc(nvtxs, -1, "gk_graph_ComputeBestFOrdering: perm");
/* create the queue and put everything in it */
queue = gk_i32pqCreate(nvtxs);
for (i=0; i<nvtxs; i++)
gk_i32pqInsert(queue, i, 0);
gk_i32pqUpdate(queue, v, 1);
open[0] = v;
/* start processing the nodes */
for (i=0; i<nvtxs; i++) {
if ((v = gk_i32pqGetTop(queue)) == -1)
gk_errexit(SIGERR, "The priority queue got empty ahead of time [i=%d].\n", i);
if (perm[v] != -1)
gk_errexit(SIGERR, "The perm[%d] has already been set.\n", v);
perm[v] = i;
for (j=xadj[v]; j<xadj[v+1]; j++) {
u = adjncy[j];
if (perm[u] == -1) {
degrees[u]++;
minIDs[u] = (i < minIDs[u] ? i : minIDs[u]);
switch (type) {
case 1: /* DFS */
gk_i32pqUpdate(queue, u, 1);
break;
case 2: /* Max in closed degree */
gk_i32pqUpdate(queue, u, degrees[u]);
break;
case 3: /* Sum of orders in closed list */
for (k=0, jj=xadj[u]; jj<xadj[u+1]; jj++) {
if (perm[adjncy[jj]] != -1)
k += perm[adjncy[jj]];
}
gk_i32pqUpdate(queue, u, k);
break;
case 4: /* Sum of order-differences (w.r.t. current number) in closed
list (updated once in a while) */
for (k=0, jj=xadj[u]; jj<xadj[u+1]; jj++) {
if (perm[adjncy[jj]] != -1)
k += (i-perm[adjncy[jj]]);
}
gk_i32pqUpdate(queue, u, k);
break;
default:
;
}
}
}
}
/* time to decide what to return */
if (r_perm != NULL) {
*r_perm = perm;
perm = NULL;
}
if (r_iperm != NULL) {
/* use the 'degrees' array to build the iperm array */
for (i=0; i<nvtxs; i++)
degrees[perm[i]] = i;
*r_iperm = degrees;
degrees = NULL;
}
/* cleanup memory */
gk_i32pqDestroy(queue);
gk_free((void **)&perm, &degrees, &minIDs, &open, LTERM);
}
/*************************************************************************/
/*! This function computes a permutation of the vertices based on a
best-first-traversal. It can be used for re-ordering the graph
to reduce its bandwidth for better cache locality.
\param[IN] graph is the graph structure.
\param[IN] v is the starting vertex of the best-first traversal.
\param[IN] type indicates the criteria to use to measure the 'bestness'
of a vertex.
\param[OUT] perm[i] stores the ID of vertex i in the re-ordered graph.
\param[OUT] iperm[i] stores the ID of the vertex that corresponds to
the ith vertex in the re-ordered graph.
\note The perm or iperm (but not both) can be NULL, at which point,
the corresponding arrays are not returned. Though the program
works fine when both are NULL, doing that is not smart.
The returned arrays should be freed with gk_free().
*/
/*************************************************************************/
void gk_graph_ComputeBestFOrdering(gk_graph_t *graph, int v, int type,
int32_t **r_perm, int32_t **r_iperm)
{
ssize_t j, jj, *xadj;
int i, k, u, nvtxs, nopen, ntodo;
int32_t *adjncy, *perm, *degrees, *sod, *level, *ot, *pos;
int64_t *wdegrees;
gk_i32pq_t *queue;
if (graph->nvtxs <= 0)
return;
nvtxs = graph->nvtxs;
xadj = graph->xadj;
adjncy = graph->adjncy;
/* the degree of the vertices in the closed list */
degrees = gk_i32smalloc(nvtxs, 0, "gk_graph_ComputeBestFOrdering: degrees");
/* the weighted degree of the vertices in the closed list for type==3 */
wdegrees = gk_i64smalloc(nvtxs, 0, "gk_graph_ComputeBestFOrdering: wdegrees");
/* the sum of differences for type==4 */
sod = gk_i32smalloc(nvtxs, 0, "gk_graph_ComputeBestFOrdering: sod");
/* the encountering level of a vertex type==5 */
level = gk_i32smalloc(nvtxs, 0, "gk_graph_ComputeBestFOrdering: level");
/* The open+todo list of vertices.
The vertices from [0..nopen] are the open vertices.
The vertices from [nopen..ntodo) are the todo vertices.
*/
ot = gk_i32incset(nvtxs, 0, gk_i32malloc(nvtxs, "gk_graph_FindComponents: ot"));
/* For a vertex that has not been explored, pos[i] is the position in the ot list. */
pos = gk_i32incset(nvtxs, 0, gk_i32malloc(nvtxs, "gk_graph_FindComponents: pos"));
/* if perm[i] >= 0, then perm[i] is the order of vertex i; otherwise perm[i] == -1. */
perm = gk_i32smalloc(nvtxs, -1, "gk_graph_ComputeBestFOrdering: perm");
/* create the queue and put the starting vertex in it */
queue = gk_i32pqCreate(nvtxs);
gk_i32pqInsert(queue, v, 1);
/* put v at the front of the open list */
pos[0] = ot[0] = v;
pos[v] = ot[v] = 0;
nopen = 1;
ntodo = nvtxs;
/* start processing the nodes */
for (i=0; i<nvtxs; i++) {
if (nopen == 0) { /* deal with non-connected graphs */
gk_i32pqInsert(queue, ot[0], 1);
nopen++;
}
if ((v = gk_i32pqGetTop(queue)) == -1)
gk_errexit(SIGERR, "The priority queue got empty ahead of time [i=%d].\n", i);
if (perm[v] != -1)
gk_errexit(SIGERR, "The perm[%d] has already been set.\n", v);
perm[v] = i;
if (ot[pos[v]] != v)
gk_errexit(SIGERR, "Something went wrong [ot[pos[%d]]!=%d.\n", v, v);
if (pos[v] >= nopen)
gk_errexit(SIGERR, "The position of v is not in open list. pos[%d]=%d is >=%d.\n", v, pos[v], nopen);
/* remove v from the open list and re-arrange the todo part of the list */
ot[pos[v]] = ot[nopen-1];
pos[ot[nopen-1]] = pos[v];
if (ntodo > nopen) {
ot[nopen-1] = ot[ntodo-1];
pos[ot[ntodo-1]] = nopen-1;
}
nopen--;
ntodo--;
for (j=xadj[v]; j<xadj[v+1]; j++) {
u = adjncy[j];
if (perm[u] == -1) {
/* update ot list, if u is not in the open list by putting it at the end
of the open list. */
if (degrees[u] == 0) {
ot[pos[u]] = ot[nopen];
pos[ot[nopen]] = pos[u];
ot[nopen] = u;
pos[u] = nopen;
nopen++;
level[u] = level[v]+1;
gk_i32pqInsert(queue, u, 0);
}
/* update the in-closed degree */
degrees[u]++;
/* update the queues based on the type */
switch (type) {
case 1: /* DFS */
gk_i32pqUpdate(queue, u, 1000*(i+1)+degrees[u]);
break;
case 2: /* Max in closed degree */
gk_i32pqUpdate(queue, u, degrees[u]);
break;
case 3: /* Sum of orders in closed list */
wdegrees[u] += i;
gk_i32pqUpdate(queue, u, (int32_t)sqrt(wdegrees[u]));
break;
case 4: /* Sum of order-differences */
/* this is handled at the end of the loop */
;
break;
case 5: /* BFS with in degree priority */
gk_i32pqUpdate(queue, u, -(1000*level[u] - degrees[u]));
break;
case 6: /* Hybrid of 1+2 */
gk_i32pqUpdate(queue, u, (i+1)*degrees[u]);
break;
default:
;
}
}
}
if (type == 4) { /* update all the vertices in the open list */
for (j=0; j<nopen; j++) {
u = ot[j];
if (perm[u] != -1)
gk_errexit(SIGERR, "For i=%d, the open list contains a closed vertex: ot[%zd]=%d, perm[%d]=%d.\n", i, j, u, u, perm[u]);
sod[u] += degrees[u];
if (i<1000 || i%25==0)
gk_i32pqUpdate(queue, u, sod[u]);
}
}
/*
for (j=0; j<ntodo; j++) {
if (pos[ot[j]] != j)
gk_errexit(SIGERR, "pos[ot[%zd]] != %zd.\n", j, j);
}
*/
}
/* time to decide what to return */
if (r_perm != NULL) {
*r_perm = perm;
perm = NULL;
}
if (r_iperm != NULL) {
/* use the 'degrees' array to build the iperm array */
for (i=0; i<nvtxs; i++)
degrees[perm[i]] = i;
*r_iperm = degrees;
degrees = NULL;
}
/* cleanup memory */
gk_i32pqDestroy(queue);
gk_free((void **)&perm, &degrees, &wdegrees, &sod, &ot, &pos, &level, LTERM);
}
/*************************************************************************/
/*! This function computes the single-source shortest path lengths from the
root node to all the other nodes in the graph. If the graph is not
connected then, the sortest part to the vertices in the other components
is -1.
\param[IN] graph is the graph structure.
\param[IN] v is the root of the single-source shortest path computations.
\param[IN] type indicates the criteria to use to measure the 'bestness'
of a vertex.
\param[OUT] sps[i] stores the length of the shortest path from v to vertex i.
If no such path exists, then it is -1. Note that the returned
array will be either an array of int32_t or an array of floats.
The specific type is determined by the existance of non NULL
iadjwgt and fadjwgt arrays. If both of these arrays exist, then
priority is given to iadjwgt.
\note The returned array should be freed with gk_free().
*/
/*************************************************************************/
void gk_graph_SingleSourceShortestPaths(gk_graph_t *graph, int v, void **r_sps)
{
ssize_t *xadj;
int i, u, nvtxs;
int32_t *adjncy, *inqueue;
if (graph->nvtxs <= 0)
return;
nvtxs = graph->nvtxs;
xadj = graph->xadj;
adjncy = graph->adjncy;
inqueue = gk_i32smalloc(nvtxs, 0, "gk_graph_SingleSourceShortestPaths: inqueue");
/* determine if you will be computing using int32_t or float and proceed from there */
if (graph->iadjwgt != NULL) {
gk_i32pq_t *queue;
int32_t *adjwgt;
int32_t *sps;
adjwgt = graph->iadjwgt;
queue = gk_i32pqCreate(nvtxs);
gk_i32pqInsert(queue, v, 0);
inqueue[v] = 1;
sps = gk_i32smalloc(nvtxs, -1, "gk_graph_SingleSourceShortestPaths: sps");
sps[v] = 0;
/* start processing the nodes */
while ((v = gk_i32pqGetTop(queue)) != -1) {
inqueue[v] = 2;
/* relax the adjacent edges */
for (i=xadj[v]; i<xadj[v+1]; i++) {
u = adjncy[i];
if (inqueue[u] == 2)
continue;
if (sps[u] < 0 || sps[v]+adjwgt[i] < sps[u]) {
sps[u] = sps[v]+adjwgt[i];
if (inqueue[u])
gk_i32pqUpdate(queue, u, -sps[u]);
else {
gk_i32pqInsert(queue, u, -sps[u]);
inqueue[u] = 1;
}
}
}
}
*r_sps = (void *)sps;
gk_i32pqDestroy(queue);
}
else {
gk_fpq_t *queue;
float *adjwgt;
float *sps;
adjwgt = graph->fadjwgt;
queue = gk_fpqCreate(nvtxs);
gk_fpqInsert(queue, v, 0);
inqueue[v] = 1;
sps = gk_fsmalloc(nvtxs, -1, "gk_graph_SingleSourceShortestPaths: sps");
sps[v] = 0;
/* start processing the nodes */
while ((v = gk_fpqGetTop(queue)) != -1) {
inqueue[v] = 2;
/* relax the adjacent edges */
for (i=xadj[v]; i<xadj[v+1]; i++) {
u = adjncy[i];
if (inqueue[u] == 2)
continue;
if (sps[u] < 0 || sps[v]+adjwgt[i] < sps[u]) {
sps[u] = sps[v]+adjwgt[i];
if (inqueue[u])
gk_fpqUpdate(queue, u, -sps[u]);
else {
gk_fpqInsert(queue, u, -sps[u]);
inqueue[u] = 1;
}
}
}
}
*r_sps = (void *)sps;
gk_fpqDestroy(queue);
}
gk_free((void **)&inqueue, LTERM);
}
#ifdef XXX
/*************************************************************************/
/*! Sorts the adjacency lists in increasing vertex order
\param graph the graph itself,
*/
/**************************************************************************/
void gk_graph_SortAdjacencies(gk_graph_t *graph)
{
int n, nn=0;
ssize_t *ptr;
int *ind;
float *val;
switch (what) {
case GK_CSR_ROW:
if (!graph->rowptr)
gk_errexit(SIGERR, "Row-based view of the graphrix does not exists.\n");
n = graph->nrows;
ptr = graph->rowptr;
ind = graph->rowind;
val = graph->rowval;
break;
case GK_CSR_COL:
if (!graph->colptr)
gk_errexit(SIGERR, "Column-based view of the graphrix does not exists.\n");
n = graph->ncols;
ptr = graph->colptr;
ind = graph->colind;
val = graph->colval;
break;
default:
gk_errexit(SIGERR, "Invalid index type of %d.\n", what);
return;
}
#pragma omp parallel if (n > 100)
{
ssize_t i, j, k;
gk_ikv_t *cand;
float *tval;
#pragma omp single
for (i=0; i<n; i++)
nn = gk_max(nn, ptr[i+1]-ptr[i]);
cand = gk_ikvmalloc(nn, "gk_graph_SortIndices: cand");
tval = gk_fmalloc(nn, "gk_graph_SortIndices: tval");
#pragma omp for schedule(static)
for (i=0; i<n; i++) {
for (k=0, j=ptr[i]; j<ptr[i+1]; j++) {
if (j > ptr[i] && ind[j] < ind[j-1])
k = 1; /* an inversion */
cand[j-ptr[i]].val = j-ptr[i];
cand[j-ptr[i]].key = ind[j];
tval[j-ptr[i]] = val[j];
}
if (k) {
gk_ikvsorti(ptr[i+1]-ptr[i], cand);
for (j=ptr[i]; j<ptr[i+1]; j++) {
ind[j] = cand[j-ptr[i]].key;
val[j] = tval[cand[j-ptr[i]].val];
}
}
}
gk_free((void **)&cand, &tval, LTERM);
}
}
/*************************************************************************/
/*! Returns a subgraphrix containing a certain set of rows.
\param graph is the original graphrix.
\param nrows is the number of rows to extract.
\param rind is the set of row numbers to extract.
\returns the row structure of the newly created subgraphrix.
*/
/**************************************************************************/
gk_graph_t *gk_graph_ExtractRows(gk_graph_t *graph, int nrows, int *rind)
{
ssize_t i, ii, j, nnz;
gk_graph_t *ngraph;
ngraph = gk_graph_Create();
ngraph->nrows = nrows;
ngraph->ncols = graph->ncols;
for (nnz=0, i=0; i<nrows; i++)
nnz += graph->rowptr[rind[i]+1]-graph->rowptr[rind[i]];
ngraph->rowptr = gk_zmalloc(ngraph->nrows+1, "gk_graph_ExtractPartition: rowptr");
ngraph->rowind = gk_imalloc(nnz, "gk_graph_ExtractPartition: rowind");
ngraph->rowval = gk_fmalloc(nnz, "gk_graph_ExtractPartition: rowval");
ngraph->rowptr[0] = 0;
for (nnz=0, j=0, ii=0; ii<nrows; ii++) {
i = rind[ii];
gk_icopy(graph->rowptr[i+1]-graph->rowptr[i], graph->rowind+graph->rowptr[i], ngraph->rowind+nnz);
gk_fcopy(graph->rowptr[i+1]-graph->rowptr[i], graph->rowval+graph->rowptr[i], ngraph->rowval+nnz);
nnz += graph->rowptr[i+1]-graph->rowptr[i];
ngraph->rowptr[++j] = nnz;
}
ASSERT(j == ngraph->nrows);
return ngraph;
}
/*************************************************************************/
/*! Returns a subgraphrix corresponding to a specified partitioning of rows.
\param graph is the original graphrix.
\param part is the partitioning vector of the rows.
\param pid is the partition ID that will be extracted.
\returns the row structure of the newly created subgraphrix.
*/
/**************************************************************************/
gk_graph_t *gk_graph_ExtractPartition(gk_graph_t *graph, int *part, int pid)
{
ssize_t i, j, nnz;
gk_graph_t *ngraph;
ngraph = gk_graph_Create();
ngraph->nrows = 0;
ngraph->ncols = graph->ncols;
for (nnz=0, i=0; i<graph->nrows; i++) {
if (part[i] == pid) {
ngraph->nrows++;
nnz += graph->rowptr[i+1]-graph->rowptr[i];
}
}
ngraph->rowptr = gk_zmalloc(ngraph->nrows+1, "gk_graph_ExtractPartition: rowptr");
ngraph->rowind = gk_imalloc(nnz, "gk_graph_ExtractPartition: rowind");
ngraph->rowval = gk_fmalloc(nnz, "gk_graph_ExtractPartition: rowval");
ngraph->rowptr[0] = 0;
for (nnz=0, j=0, i=0; i<graph->nrows; i++) {
if (part[i] == pid) {
gk_icopy(graph->rowptr[i+1]-graph->rowptr[i], graph->rowind+graph->rowptr[i], ngraph->rowind+nnz);
gk_fcopy(graph->rowptr[i+1]-graph->rowptr[i], graph->rowval+graph->rowptr[i], ngraph->rowval+nnz);
nnz += graph->rowptr[i+1]-graph->rowptr[i];
ngraph->rowptr[++j] = nnz;
}
}
ASSERT(j == ngraph->nrows);
return ngraph;
}
/*************************************************************************/
/*! Splits the graphrix into multiple sub-graphrices based on the provided
color array.
\param graph is the original graphrix.
\param color is an array of size equal to the number of non-zeros
in the graphrix (row-wise structure). The graphrix is split into
as many parts as the number of colors. For meaningfull results,
the colors should be numbered consecutively starting from 0.
\returns an array of graphrices for each supplied color number.
*/
/**************************************************************************/
gk_graph_t **gk_graph_Split(gk_graph_t *graph, int *color)
{
ssize_t i, j;
int nrows, ncolors;
ssize_t *rowptr;
int *rowind;
float *rowval;
gk_graph_t **sgraphs;
nrows = graph->nrows;
rowptr = graph->rowptr;
rowind = graph->rowind;
rowval = graph->rowval;
ncolors = gk_imax(rowptr[nrows], color)+1;
sgraphs = (gk_graph_t **)gk_malloc(sizeof(gk_graph_t *)*ncolors, "gk_graph_Split: sgraphs");
for (i=0; i<ncolors; i++) {
sgraphs[i] = gk_graph_Create();
sgraphs[i]->nrows = graph->nrows;
sgraphs[i]->ncols = graph->ncols;
sgraphs[i]->rowptr = gk_zsmalloc(nrows+1, 0, "gk_graph_Split: sgraphs[i]->rowptr");
}
for (i=0; i<nrows; i++) {
for (j=rowptr[i]; j<rowptr[i+1]; j++)
sgraphs[color[j]]->rowptr[i]++;
}
for (i=0; i<ncolors; i++)
MAKECSR(j, nrows, sgraphs[i]->rowptr);
for (i=0; i<ncolors; i++) {
sgraphs[i]->rowind = gk_imalloc(sgraphs[i]->rowptr[nrows], "gk_graph_Split: sgraphs[i]->rowind");
sgraphs[i]->rowval = gk_fmalloc(sgraphs[i]->rowptr[nrows], "gk_graph_Split: sgraphs[i]->rowval");
}
for (i=0; i<nrows; i++) {
for (j=rowptr[i]; j<rowptr[i+1]; j++) {
sgraphs[color[j]]->rowind[sgraphs[color[j]]->rowptr[i]] = rowind[j];
sgraphs[color[j]]->rowval[sgraphs[color[j]]->rowptr[i]] = rowval[j];
sgraphs[color[j]]->rowptr[i]++;
}
}
for (i=0; i<ncolors; i++)
SHIFTCSR(j, nrows, sgraphs[i]->rowptr);
return sgraphs;
}
/*************************************************************************/
/*! Prunes certain rows/columns of the graphrix. The prunning takes place
by analyzing the row structure of the graphrix. The prunning takes place
by removing rows/columns but it does not affect the numbering of the
remaining rows/columns.
\param graph the graphrix to be prunned,
\param what indicates if the rows (GK_CSR_ROW) or the columns (GK_CSR_COL)
of the graphrix will be prunned,
\param minf is the minimum number of rows (columns) that a column (row) must
be present in order to be kept,
\param maxf is the maximum number of rows (columns) that a column (row) must
be present at in order to be kept.
\returns the prunned graphrix consisting only of its row-based structure.
The input graphrix is not modified.
*/
/**************************************************************************/
gk_graph_t *gk_graph_Prune(gk_graph_t *graph, int what, int minf, int maxf)
{
ssize_t i, j, nnz;
int nrows, ncols;
ssize_t *rowptr, *nrowptr;
int *rowind, *nrowind, *collen;
float *rowval, *nrowval;
gk_graph_t *ngraph;
ngraph = gk_graph_Create();
nrows = ngraph->nrows = graph->nrows;
ncols = ngraph->ncols = graph->ncols;
rowptr = graph->rowptr;
rowind = graph->rowind;
rowval = graph->rowval;
nrowptr = ngraph->rowptr = gk_zmalloc(nrows+1, "gk_graph_Prune: nrowptr");
nrowind = ngraph->rowind = gk_imalloc(rowptr[nrows], "gk_graph_Prune: nrowind");
nrowval = ngraph->rowval = gk_fmalloc(rowptr[nrows], "gk_graph_Prune: nrowval");
switch (what) {
case GK_CSR_COL:
collen = gk_ismalloc(ncols, 0, "gk_graph_Prune: collen");
for (i=0; i<nrows; i++) {
for (j=rowptr[i]; j<rowptr[i+1]; j++) {
ASSERT(rowind[j] < ncols);
collen[rowind[j]]++;
}
}
for (i=0; i<ncols; i++)
collen[i] = (collen[i] >= minf && collen[i] <= maxf ? 1 : 0);
nrowptr[0] = 0;
for (nnz=0, i=0; i<nrows; i++) {
for (j=rowptr[i]; j<rowptr[i+1]; j++) {
if (collen[rowind[j]]) {
nrowind[nnz] = rowind[j];
nrowval[nnz] = rowval[j];
nnz++;
}
}
nrowptr[i+1] = nnz;
}
gk_free((void **)&collen, LTERM);
break;
case GK_CSR_ROW:
nrowptr[0] = 0;
for (nnz=0, i=0; i<nrows; i++) {
if (rowptr[i+1]-rowptr[i] >= minf && rowptr[i+1]-rowptr[i] <= maxf) {
for (j=rowptr[i]; j<rowptr[i+1]; j++, nnz++) {
nrowind[nnz] = rowind[j];
nrowval[nnz] = rowval[j];
}
}
nrowptr[i+1] = nnz;
}
break;
default:
gk_graph_Free(&ngraph);
gk_errexit(SIGERR, "Unknown prunning type of %d\n", what);
return NULL;
}
return ngraph;
}
/*************************************************************************/
/*! Normalizes the rows/columns of the graphrix to be unit
length.
\param graph the graphrix itself,
\param what indicates what will be normalized and is obtained by
specifying GK_CSR_ROW, GK_CSR_COL, GK_CSR_ROW|GK_CSR_COL.
\param norm indicates what norm is to normalize to, 1: 1-norm, 2: 2-norm
*/
/**************************************************************************/
void gk_graph_Normalize(gk_graph_t *graph, int what, int norm)
{
ssize_t i, j;
int n;
ssize_t *ptr;
float *val, sum;
if (what&GK_CSR_ROW && graph->rowval) {
n = graph->nrows;
ptr = graph->rowptr;
val = graph->rowval;
#pragma omp parallel if (ptr[n] > OMPMINOPS)
{
#pragma omp for private(j,sum) schedule(static)
for (i=0; i<n; i++) {
for (sum=0.0, j=ptr[i]; j<ptr[i+1]; j++){
if (norm == 2)
sum += val[j]*val[j];
else if (norm == 1)
sum += val[j]; /* assume val[j] > 0 */
}
if (sum > 0) {
if (norm == 2)
sum=1.0/sqrt(sum);
else if (norm == 1)
sum=1.0/sum;
for (j=ptr[i]; j<ptr[i+1]; j++)
val[j] *= sum;
}
}
}
}
if (what&GK_CSR_COL && graph->colval) {
n = graph->ncols;
ptr = graph->colptr;
val = graph->colval;
#pragma omp parallel if (ptr[n] > OMPMINOPS)
{
#pragma omp for private(j,sum) schedule(static)
for (i=0; i<n; i++) {
for (sum=0.0, j=ptr[i]; j<ptr[i+1]; j++)
if (norm == 2)
sum += val[j]*val[j];
else if (norm == 1)
sum += val[j];
if (sum > 0) {
if (norm == 2)
sum=1.0/sqrt(sum);
else if (norm == 1)
sum=1.0/sum;
for (j=ptr[i]; j<ptr[i+1]; j++)
val[j] *= sum;
}
}
}
}
}
#endif
/*
* Copyright 2004, Regents of the University of Minnesota
*
* This file contains routines for manipulating a direct-access hash table
*
* Started 3/22/04
* George
*
*/
#include <GKlib.h>
/******************************************************************************
* This function creates the hash-table
*******************************************************************************/
gk_HTable_t *HTable_Create(int nelements)
{
gk_HTable_t *htable;
htable = gk_malloc(sizeof(gk_HTable_t), "HTable_Create: htable");
htable->harray = gk_ikvmalloc(nelements, "HTable_Create: harray");
htable->nelements = nelements;
HTable_Reset(htable);
return htable;
}
/******************************************************************************
* This function resets the data-structures associated with the hash-table
*******************************************************************************/
void HTable_Reset(gk_HTable_t *htable)
{
int i;
for (i=0; i<htable->nelements; i++)
htable->harray[i].key = HTABLE_EMPTY;
htable->htsize = 0;
}
/******************************************************************************
* This function resizes the hash-table
*******************************************************************************/
void HTable_Resize(gk_HTable_t *htable, int nelements)
{
int i, old_nelements;
gk_ikv_t *old_harray;
old_nelements = htable->nelements;
old_harray = htable->harray;
/* prepare larger hash */
htable->nelements = nelements;
htable->htsize = 0;
htable->harray = gk_ikvmalloc(nelements, "HTable_Resize: harray");
for (i=0; i<nelements; i++)
htable->harray[i].key = HTABLE_EMPTY;
/* reassign the values */
for (i=0; i<old_nelements; i++)
if (old_harray[i].key != HTABLE_EMPTY)
HTable_Insert(htable, old_harray[i].key, old_harray[i].val);
/* remove old harray */
gk_free((void **)&old_harray, LTERM);
}
/******************************************************************************
* This function inserts a key-value pair in the array
*******************************************************************************/
void HTable_Insert(gk_HTable_t *htable, int key, int val)
{
int i, first;
if (htable->htsize > htable->nelements/2)
HTable_Resize(htable, 2*htable->nelements);
first = HTable_HFunction(htable->nelements, key);
for (i=first; i<htable->nelements; i++) {
if (htable->harray[i].key == HTABLE_EMPTY || htable->harray[i].key == HTABLE_DELETED) {
htable->harray[i].key = key;
htable->harray[i].val = val;
htable->htsize++;
return;
}
}
for (i=0; i<first; i++) {
if (htable->harray[i].key == HTABLE_EMPTY || htable->harray[i].key == HTABLE_DELETED) {
htable->harray[i].key = key;
htable->harray[i].val = val;
htable->htsize++;
return;
}
}
}
/******************************************************************************
* This function deletes key from the htable
*******************************************************************************/
void HTable_Delete(gk_HTable_t *htable, int key)
{
int i, first;
first = HTable_HFunction(htable->nelements, key);
for (i=first; i<htable->nelements; i++) {
if (htable->harray[i].key == key) {
htable->harray[i].key = HTABLE_DELETED;
htable->htsize--;
return;
}
}
for (i=0; i<first; i++) {
if (htable->harray[i].key == key) {
htable->harray[i].key = HTABLE_DELETED;
htable->htsize--;
return;
}
}
}
/******************************************************************************
* This function returns the data associated with the key in the hastable
*******************************************************************************/
int HTable_Search(gk_HTable_t *htable, int key)
{
int i, first;
first = HTable_HFunction(htable->nelements, key);
for (i=first; i<htable->nelements; i++) {
if (htable->harray[i].key == key)
return htable->harray[i].val;
else if (htable->harray[i].key == HTABLE_EMPTY)
return -1;
}
for (i=0; i<first; i++) {
if (htable->harray[i].key == key)
return htable->harray[i].val;
else if (htable->harray[i].key == HTABLE_EMPTY)
return -1;
}
return -1;
}
/******************************************************************************
* This function returns the next key/val
*******************************************************************************/
int HTable_GetNext(gk_HTable_t *htable, int key, int *r_val, int type)
{
int i;
static int first, last;
if (type == HTABLE_FIRST)
first = last = HTable_HFunction(htable->nelements, key);
if (first > last) {
for (i=first; i<htable->nelements; i++) {
if (htable->harray[i].key == key) {
*r_val = htable->harray[i].val;
first = i+1;
return 1;
}
else if (htable->harray[i].key == HTABLE_EMPTY)
return -1;
}
first = 0;
}
for (i=first; i<last; i++) {
if (htable->harray[i].key == key) {
*r_val = htable->harray[i].val;
first = i+1;
return 1;
}
else if (htable->harray[i].key == HTABLE_EMPTY)
return -1;
}
return -1;
}
/******************************************************************************
* This function returns the data associated with the key in the hastable
*******************************************************************************/
int HTable_SearchAndDelete(gk_HTable_t *htable, int key)
{
int i, first;
first = HTable_HFunction(htable->nelements, key);
for (i=first; i<htable->nelements; i++) {
if (htable->harray[i].key == key) {
htable->harray[i].key = HTABLE_DELETED;
htable->htsize--;
return htable->harray[i].val;
}
else if (htable->harray[i].key == HTABLE_EMPTY)
gk_errexit(SIGERR, "HTable_SearchAndDelete: Failed to find the key!\n");
}
for (i=0; i<first; i++) {
if (htable->harray[i].key == key) {
htable->harray[i].key = HTABLE_DELETED;
htable->htsize--;
return htable->harray[i].val;
}
else if (htable->harray[i].key == HTABLE_EMPTY)
gk_errexit(SIGERR, "HTable_SearchAndDelete: Failed to find the key!\n");
}
return -1;
}
/******************************************************************************
* This function destroys the data structures associated with the hash-table
*******************************************************************************/
void HTable_Destroy(gk_HTable_t *htable)
{
gk_free((void **)&htable->harray, &htable, LTERM);
}
/******************************************************************************
* This is the hash-function. Based on multiplication
*******************************************************************************/
int HTable_HFunction(int nelements, int key)
{
return (int)(key%nelements);
}
/*!
\file io.c
\brief Various file I/O functions.
This file contains various functions that perform I/O.
\date Started 4/10/95
\author George
\version\verbatim $Id: io.c 18951 2015-08-08 20:10:46Z karypis $ \endverbatim
*/
#ifdef HAVE_GETLINE
/* Get getline to be defined. */
#define _GNU_SOURCE
#include <stdio.h>
#undef _GNU_SOURCE
#endif
#include <GKlib.h>
/*************************************************************************
* This function opens a file
**************************************************************************/
FILE *gk_fopen(char *fname, char *mode, const char *msg)
{
FILE *fp;
char errmsg[8192];
fp = fopen(fname, mode);
if (fp != NULL)
return fp;
sprintf(errmsg,"file: %s, mode: %s, [%s]", fname, mode, msg);
perror(errmsg);
errexit("Failed on gk_fopen()\n");
return NULL;
}
/*************************************************************************
* This function closes a file
**************************************************************************/
void gk_fclose(FILE *fp)
{
fclose(fp);
}
/*************************************************************************/
/*! This function is a wrapper around the read() function that ensures
that all data is been read, by issuing multiple read requests.
The only time when not 'count' items are read is when the EOF has been
reached.
*/
/*************************************************************************/
ssize_t gk_read(int fd, void *vbuf, size_t count)
{
char *buf = (char *)vbuf;
ssize_t rsize, tsize=count;
do {
if ((rsize = read(fd, buf, tsize)) == -1)
return -1;
buf += rsize;
tsize -= rsize;
} while (tsize > 0 && rsize > 0);
return count-tsize;
}
/*************************************************************************/
/*! This function is a wrapper around the write() function that ensures
that all data is been written, by issueing multiple write requests.
*/
/*************************************************************************/
ssize_t gk_write(int fd, void *vbuf, size_t count)
{
char *buf = (char *)vbuf;
ssize_t size, tsize=count;
do {
if ((size = write(fd, buf, tsize)) == -1)
return -1;
buf += size;
tsize -= size;
} while (tsize > 0);
return count;
}
/*************************************************************************/
/*! This function is the GKlib implementation of glibc's getline()
function.
\returns -1 if the EOF has been reached, otherwise it returns the
number of bytes read.
*/
/*************************************************************************/
gk_idx_t gk_getline(char **lineptr, size_t *n, FILE *stream)
{
#ifdef HAVE_GETLINE
return getline(lineptr, n, stream);
#else
size_t i;
int ch;
if (feof(stream))
return -1;
/* Initial memory allocation if *lineptr is NULL */
if (*lineptr == NULL || *n == 0) {
*n = 1024;
*lineptr = gk_malloc((*n)*sizeof(char), "gk_getline: lineptr");
}
/* get into the main loop */
i = 0;
while ((ch = getc(stream)) != EOF) {
(*lineptr)[i++] = (char)ch;
/* reallocate memory if reached at the end of the buffer. The +1 is for '\0' */
if (i+1 == *n) {
*n = 2*(*n);
*lineptr = gk_realloc(*lineptr, (*n)*sizeof(char), "gk_getline: lineptr");
}
if (ch == '\n')
break;
}
(*lineptr)[i] = '\0';
return (i == 0 ? -1 : i);
#endif
}
/*************************************************************************/
/*! This function reads the contents of a text file and returns it in the
form of an array of strings.
\param fname is the name of the file
\param r_nlines is the number of lines in the file. If it is NULL,
this information is not returned.
*/
/*************************************************************************/
char **gk_readfile(char *fname, size_t *r_nlines)
{
size_t lnlen, nlines=0;
char *line=NULL, **lines=NULL;
FILE *fpin;
gk_getfilestats(fname, &nlines, NULL, NULL, NULL);
if (nlines > 0) {
lines = (char **)gk_malloc(nlines*sizeof(char *), "gk_readfile: lines");
fpin = gk_fopen(fname, "r", "gk_readfile");
nlines = 0;
while (gk_getline(&line, &lnlen, fpin) != -1) {
gk_strtprune(line, "\n\r");
lines[nlines++] = gk_strdup(line);
}
gk_fclose(fpin);
}
gk_free((void **)&line, LTERM);
if (r_nlines != NULL)
*r_nlines = nlines;
return lines;
}
/*************************************************************************/
/*! This function reads the contents of a file and returns it in the
form of an array of int32_t.
\param fname is the name of the file
\param r_nlines is the number of lines in the file. If it is NULL,
this information is not returned.
*/
/*************************************************************************/
int32_t *gk_i32readfile(char *fname, size_t *r_nlines)
{
size_t lnlen, nlines=0;
char *line=NULL;
int32_t *array=NULL;
FILE *fpin;
gk_getfilestats(fname, &nlines, NULL, NULL, NULL);
if (nlines > 0) {
array = gk_i32malloc(nlines, "gk_i32readfile: array");
fpin = gk_fopen(fname, "r", "gk_readfile");
nlines = 0;
while (gk_getline(&line, &lnlen, fpin) != -1) {
sscanf(line, "%"SCNd32, &array[nlines++]);
}
gk_fclose(fpin);
}
gk_free((void **)&line, LTERM);
if (r_nlines != NULL)
*r_nlines = nlines;
return array;
}
/*************************************************************************/
/*! This function reads the contents of a file and returns it in the
form of an array of int64_t.
\param fname is the name of the file
\param r_nlines is the number of lines in the file. If it is NULL,
this information is not returned.
*/
/*************************************************************************/
int64_t *gk_i64readfile(char *fname, size_t *r_nlines)
{
size_t lnlen, nlines=0;
char *line=NULL;
int64_t *array=NULL;
FILE *fpin;
gk_getfilestats(fname, &nlines, NULL, NULL, NULL);
if (nlines > 0) {
array = gk_i64malloc(nlines, "gk_i64readfile: array");
fpin = gk_fopen(fname, "r", "gk_readfile");
nlines = 0;
while (gk_getline(&line, &lnlen, fpin) != -1) {
sscanf(line, "%"SCNd64, &array[nlines++]);
}
gk_fclose(fpin);
}
gk_free((void **)&line, LTERM);
if (r_nlines != NULL)
*r_nlines = nlines;
return array;
}
/*************************************************************************/
/*! This function reads the contents of a file and returns it in the
form of an array of ssize_t.
\param fname is the name of the file
\param r_nlines is the number of lines in the file. If it is NULL,
this information is not returned.
*/
/*************************************************************************/
ssize_t *gk_zreadfile(char *fname, size_t *r_nlines)
{
size_t lnlen, nlines=0;
char *line=NULL;
ssize_t *array=NULL;
FILE *fpin;
gk_getfilestats(fname, &nlines, NULL, NULL, NULL);
if (nlines > 0) {
array = gk_zmalloc(nlines, "gk_zreadfile: array");
fpin = gk_fopen(fname, "r", "gk_readfile");
nlines = 0;
while (gk_getline(&line, &lnlen, fpin) != -1) {
sscanf(line, "%zd", &array[nlines++]);
}
gk_fclose(fpin);
}
gk_free((void **)&line, LTERM);
if (r_nlines != NULL)
*r_nlines = nlines;
return array;
}
/*************************************************************************/
/*! This function reads the contents of a binary file and returns it in the
form of an array of int32_t.
\param fname is the name of the file
\param r_nlines is the number of lines in the file. If it is NULL,
this information is not returned.
*/
/*************************************************************************/
int32_t *gk_i32readfilebin(char *fname, size_t *r_nelmnts)
{
size_t nelmnts;
ssize_t fsize;
int32_t *array=NULL;
FILE *fpin;
*r_nelmnts = 0;
fsize = gk_getfsize(fname);
if (fsize == -1) {
gk_errexit(SIGERR, "Failed to fstat(%s).\n", fname);
return NULL;
}
if (fsize%sizeof(int32_t) != 0) {
gk_errexit(SIGERR, "The size [%zd] of the file [%s] is not in multiples of sizeof(int32_t).\n", fsize, fname);
return NULL;
}
nelmnts = fsize/sizeof(int32_t);
array = gk_i32malloc(nelmnts, "gk_i32readfilebin: array");
fpin = gk_fopen(fname, "rb", "gk_i32readfilebin");
if (fread(array, sizeof(int32_t), nelmnts, fpin) != nelmnts) {
gk_errexit(SIGERR, "Failed to read the number of words requested. %zd\n", nelmnts);
gk_free((void **)&array, LTERM);
return NULL;
}
gk_fclose(fpin);
*r_nelmnts = nelmnts;
return array;
}
/*************************************************************************/
/*! This function writes the contents of an array into a binary file.
\param fname is the name of the file
\param n the number of elements in the array.
\param a the array to be written out.
*/
/*************************************************************************/
size_t gk_i32writefilebin(char *fname, size_t n, int32_t *a)
{
size_t fsize;
FILE *fp;
fp = gk_fopen(fname, "wb", "gk_writefilebin");
fsize = fwrite(a, sizeof(int32_t), n, fp);
gk_fclose(fp);
return fsize;
}
/*************************************************************************/
/*! This function reads the contents of a binary file and returns it in the
form of an array of int64_t.
\param fname is the name of the file
\param r_nlines is the number of lines in the file. If it is NULL,
this information is not returned.
*/
/*************************************************************************/
int64_t *gk_i64readfilebin(char *fname, size_t *r_nelmnts)
{
size_t nelmnts;
ssize_t fsize;
int64_t *array=NULL;
FILE *fpin;
*r_nelmnts = 0;
fsize = gk_getfsize(fname);
if (fsize == -1) {
gk_errexit(SIGERR, "Failed to fstat(%s).\n", fname);
return NULL;
}
if (fsize%sizeof(int64_t) != 0) {
gk_errexit(SIGERR, "The size of the file is not in multiples of sizeof(int64_t).\n");
return NULL;
}
nelmnts = fsize/sizeof(int64_t);
array = gk_i64malloc(nelmnts, "gk_i64readfilebin: array");
fpin = gk_fopen(fname, "rb", "gk_i64readfilebin");
if (fread(array, sizeof(int64_t), nelmnts, fpin) != nelmnts) {
gk_errexit(SIGERR, "Failed to read the number of words requested. %zd\n", nelmnts);
gk_free((void **)&array, LTERM);
return NULL;
}
gk_fclose(fpin);
*r_nelmnts = nelmnts;
return array;
}
/*************************************************************************/
/*! This function writes the contents of an array into a binary file.
\param fname is the name of the file
\param n the number of elements in the array.
\param a the array to be written out.
*/
/*************************************************************************/
size_t gk_i64writefilebin(char *fname, size_t n, int64_t *a)
{
size_t fsize;
FILE *fp;
fp = gk_fopen(fname, "wb", "gk_writefilebin");
fsize = fwrite(a, sizeof(int64_t), n, fp);
gk_fclose(fp);
return fsize;
}
/*************************************************************************/
/*! This function reads the contents of a binary file and returns it in the
form of an array of ssize_t.
\param fname is the name of the file
\param r_nlines is the number of lines in the file. If it is NULL,
this information is not returned.
*/
/*************************************************************************/
ssize_t *gk_zreadfilebin(char *fname, size_t *r_nelmnts)
{
size_t nelmnts;
ssize_t fsize;
ssize_t *array=NULL;
FILE *fpin;
*r_nelmnts = 0;
fsize = gk_getfsize(fname);
if (fsize == -1) {
gk_errexit(SIGERR, "Failed to fstat(%s).\n", fname);
return NULL;
}
if (fsize%sizeof(ssize_t) != 0) {
gk_errexit(SIGERR, "The size of the file is not in multiples of sizeof(ssize_t).\n");
return NULL;
}
nelmnts = fsize/sizeof(ssize_t);
array = gk_zmalloc(nelmnts, "gk_zreadfilebin: array");
fpin = gk_fopen(fname, "rb", "gk_zreadfilebin");
if (fread(array, sizeof(ssize_t), nelmnts, fpin) != nelmnts) {
gk_errexit(SIGERR, "Failed to read the number of words requested. %zd\n", nelmnts);
gk_free((void **)&array, LTERM);
return NULL;
}
gk_fclose(fpin);
*r_nelmnts = nelmnts;
return array;
}
/*************************************************************************/
/*! This function writes the contents of an array into a binary file.
\param fname is the name of the file
\param n the number of elements in the array.
\param a the array to be written out.
*/
/*************************************************************************/
size_t gk_zwritefilebin(char *fname, size_t n, ssize_t *a)
{
size_t fsize;
FILE *fp;
fp = gk_fopen(fname, "wb", "gk_writefilebin");
fsize = fwrite(a, sizeof(ssize_t), n, fp);
gk_fclose(fp);
return fsize;
}
/*************************************************************************/
/*! This function reads the contents of a binary file and returns it in the
form of an array of float.
\param fname is the name of the file
\param r_nlines is the number of lines in the file. If it is NULL,
this information is not returned.
*/
/*************************************************************************/
float *gk_freadfilebin(char *fname, size_t *r_nelmnts)
{
size_t nelmnts;
ssize_t fsize;
float *array=NULL;
FILE *fpin;
*r_nelmnts = 0;
fsize = gk_getfsize(fname);
if (fsize == -1) {
gk_errexit(SIGERR, "Failed to fstat(%s).\n", fname);
return NULL;
}
if (fsize%sizeof(float) != 0) {
gk_errexit(SIGERR, "The size of the file is not in multiples of sizeof(float).\n");
return NULL;
}
nelmnts = fsize/sizeof(float);
array = gk_fmalloc(nelmnts, "gk_freadfilebin: array");
fpin = gk_fopen(fname, "rb", "gk_freadfilebin");
if (fread(array, sizeof(float), nelmnts, fpin) != nelmnts) {
gk_errexit(SIGERR, "Failed to read the number of words requested. %zd\n", nelmnts);
gk_free((void **)&array, LTERM);
return NULL;
}
gk_fclose(fpin);
*r_nelmnts = nelmnts;
return array;
}
/*************************************************************************/
/*! This function writes the contents of an array into a binary file.
\param fname is the name of the file
\param n the number of elements in the array.
\param a the array to be written out.
*/
/*************************************************************************/
size_t gk_fwritefilebin(char *fname, size_t n, float *a)
{
size_t fsize;
FILE *fp;
fp = gk_fopen(fname, "wb", "gk_fwritefilebin");
fsize = fwrite(a, sizeof(float), n, fp);
gk_fclose(fp);
return fsize;
}
/*************************************************************************/
/*! This function reads the contents of a binary file and returns it in the
form of an array of double.
\param fname is the name of the file
\param r_nlines is the number of lines in the file. If it is NULL,
this information is not returned.
*/
/*************************************************************************/
double *gk_dreadfilebin(char *fname, size_t *r_nelmnts)
{
size_t nelmnts;
ssize_t fsize;
double *array=NULL;
FILE *fpin;
*r_nelmnts = 0;
fsize = gk_getfsize(fname);
if (fsize == -1) {
gk_errexit(SIGERR, "Failed to fstat(%s).\n", fname);
return NULL;
}
if (fsize%sizeof(double) != 0) {
gk_errexit(SIGERR, "The size of the file is not in multiples of sizeof(double).\n");
return NULL;
}
nelmnts = fsize/sizeof(double);
array = gk_dmalloc(nelmnts, "gk_dreadfilebin: array");
fpin = gk_fopen(fname, "rb", "gk_dreadfilebin");
if (fread(array, sizeof(double), nelmnts, fpin) != nelmnts) {
gk_errexit(SIGERR, "Failed to read the number of words requested. %zd\n", nelmnts);
gk_free((void **)&array, LTERM);
return NULL;
}
gk_fclose(fpin);
*r_nelmnts = nelmnts;
return array;
}
/*************************************************************************/
/*! This function writes the contents of an array into a binary file.
\param fname is the name of the file
\param n the number of elements in the array.
\param a the array to be written out.
*/
/*************************************************************************/
size_t gk_dwritefilebin(char *fname, size_t n, double *a)
{
size_t fsize;
FILE *fp;
fp = gk_fopen(fname, "wb", "gk_writefilebin");
fsize = fwrite(a, sizeof(double), n, fp);
gk_fclose(fp);
return fsize;
}
/*!
* \file
* \brief Frequent/Closed itemset discovery routines
*
* This file contains the code for finding frequent/closed itemests. These routines
* are implemented using a call-back mechanism to deal with the discovered itemsets.
*
* \date 6/13/2008
* \author George Karypis
* \version\verbatim $Id: itemsets.c 19240 2015-10-22 12:41:19Z karypis $ \endverbatim
*/
#include <GKlib.h>
/*-------------------------------------------------------------*/
/*! Data structures for use within this module */
/*-------------------------------------------------------------*/
typedef struct {
int minfreq; /* the minimum frequency of a pattern */
int maxfreq; /* the maximum frequency of a pattern */
int minlen; /* the minimum length of the requested pattern */
int maxlen; /* the maximum length of the requested pattern */
int tnitems; /* the initial range of the item space */
/* the call-back function */
void (*callback)(void *stateptr, int nitems, int *itemids, int ntrans, int *transids);
void *stateptr; /* the user-supplied pointer to pass to the callback */
/* workspace variables */
int *rmarker;
gk_ikv_t *cand;
} isparams_t;
/*-------------------------------------------------------------*/
/*! Prototypes for this module */
/*-------------------------------------------------------------*/
void itemsets_find_frequent_itemsets(isparams_t *params, gk_csr_t *mat,
int preflen, int *prefix);
gk_csr_t *itemsets_project_matrix(isparams_t *param, gk_csr_t *mat, int cid);
/*************************************************************************/
/*! The entry point of the frequent itemset discovery code */
/*************************************************************************/
void gk_find_frequent_itemsets(int ntrans, ssize_t *tranptr, int *tranind,
int minfreq, int maxfreq, int minlen, int maxlen,
void (*process_itemset)(void *stateptr, int nitems, int *itemids,
int ntrans, int *transids),
void *stateptr)
{
ssize_t i;
gk_csr_t *mat, *pmat;
isparams_t params;
int *pattern;
/* Create the matrix */
mat = gk_csr_Create();
mat->nrows = ntrans;
mat->ncols = tranind[gk_iargmax(tranptr[ntrans], tranind, 1)]+1;
mat->rowptr = gk_zcopy(ntrans+1, tranptr, gk_zmalloc(ntrans+1, "gk_find_frequent_itemsets: mat.rowptr"));
mat->rowind = gk_icopy(tranptr[ntrans], tranind, gk_imalloc(tranptr[ntrans], "gk_find_frequent_itemsets: mat.rowind"));
mat->colids = gk_iincset(mat->ncols, 0, gk_imalloc(mat->ncols, "gk_find_frequent_itemsets: mat.colids"));
/* Setup the parameters */
params.minfreq = minfreq;
params.maxfreq = (maxfreq == -1 ? mat->nrows : maxfreq);
params.minlen = minlen;
params.maxlen = (maxlen == -1 ? mat->ncols : maxlen);
params.tnitems = mat->ncols;
params.callback = process_itemset;
params.stateptr = stateptr;
params.rmarker = gk_ismalloc(mat->nrows, 0, "gk_find_frequent_itemsets: rmarker");
params.cand = gk_ikvmalloc(mat->ncols, "gk_find_frequent_itemsets: cand");
/* Perform the initial projection */
gk_csr_CreateIndex(mat, GK_CSR_COL);
pmat = itemsets_project_matrix(&params, mat, -1);
gk_csr_Free(&mat);
pattern = gk_imalloc(pmat->ncols, "gk_find_frequent_itemsets: pattern");
itemsets_find_frequent_itemsets(&params, pmat, 0, pattern);
gk_csr_Free(&pmat);
gk_free((void **)&pattern, &params.rmarker, &params.cand, LTERM);
}
/*************************************************************************/
/*! The recursive routine for DFS-based frequent pattern discovery */
/*************************************************************************/
void itemsets_find_frequent_itemsets(isparams_t *params, gk_csr_t *mat,
int preflen, int *prefix)
{
ssize_t i;
gk_csr_t *cmat;
/* Project each frequent column */
for (i=0; i<mat->ncols; i++) {
prefix[preflen] = mat->colids[i];
if (preflen+1 >= params->minlen)
(*params->callback)(params->stateptr, preflen+1, prefix,
mat->colptr[i+1]-mat->colptr[i], mat->colind+mat->colptr[i]);
if (preflen+1 < params->maxlen) {
cmat = itemsets_project_matrix(params, mat, i);
itemsets_find_frequent_itemsets(params, cmat, preflen+1, prefix);
gk_csr_Free(&cmat);
}
}
}
/******************************************************************************/
/*! This function projects a matrix w.r.t. to a particular column.
It performs the following steps:
- Determines the length of each column that is remaining.
- Sorts the columns in increasing length.
- Creates a column-based version of the matrix with the proper
column ordering.
*/
/*******************************************************************************/
gk_csr_t *itemsets_project_matrix(isparams_t *params, gk_csr_t *mat, int cid)
{
ssize_t i, j, k, ii, pnnz;
int nrows, ncols, pnrows, pncols;
ssize_t *colptr, *pcolptr;
int *colind, *colids, *pcolind, *pcolids, *rmarker;
gk_csr_t *pmat;
gk_ikv_t *cand;
nrows = mat->nrows;
ncols = mat->ncols;
colptr = mat->colptr;
colind = mat->colind;
colids = mat->colids;
rmarker = params->rmarker;
cand = params->cand;
/* Allocate space for the projected matrix based on what you know thus far */
pmat = gk_csr_Create();
pmat->nrows = pnrows = (cid == -1 ? nrows : colptr[cid+1]-colptr[cid]);
/* Mark the rows that will be kept and determine the prowids */
if (cid == -1) { /* Initial projection */
gk_iset(nrows, 1, rmarker);
}
else { /* The other projections */
for (i=colptr[cid]; i<colptr[cid+1]; i++)
rmarker[colind[i]] = 1;
}
/* Determine the length of each column that will be left in the projected matrix */
for (pncols=0, pnnz=0, i=cid+1; i<ncols; i++) {
for (k=0, j=colptr[i]; j<colptr[i+1]; j++) {
k += rmarker[colind[j]];
}
if (k >= params->minfreq && k <= params->maxfreq) {
cand[pncols].val = i;
cand[pncols++].key = k;
pnnz += k;
}
}
/* Sort the columns in increasing order */
gk_ikvsorti(pncols, cand);
/* Allocate space for the remaining fields of the projected matrix */
pmat->ncols = pncols;
pmat->colids = pcolids = gk_imalloc(pncols, "itemsets_project_matrix: pcolids");
pmat->colptr = pcolptr = gk_zmalloc(pncols+1, "itemsets_project_matrix: pcolptr");
pmat->colind = pcolind = gk_imalloc(pnnz, "itemsets_project_matrix: pcolind");
/* Populate the projected matrix */
pcolptr[0] = 0;
for (pnnz=0, ii=0; ii<pncols; ii++) {
i = cand[ii].val;
for (j=colptr[i]; j<colptr[i+1]; j++) {
if (rmarker[colind[j]])
pcolind[pnnz++] = colind[j];
}
pcolids[ii] = colids[i];
pcolptr[ii+1] = pnnz;
}
/* Reset the rmarker array */
if (cid == -1) { /* Initial projection */
gk_iset(nrows, 0, rmarker);
}
else { /* The other projections */
for (i=colptr[cid]; i<colptr[cid+1]; i++)
rmarker[colind[i]] = 0;
}
return pmat;
}
/*!
\file
\brief Functions dealing with creating and allocating mcores
\date Started 5/30/11
\author George
\author Copyright 1997-2011, Regents of the University of Minnesota
\version $Id: mcore.c 13953 2013-03-30 16:20:07Z karypis $
*/
#include <GKlib.h>
/*************************************************************************/
/*! This function creates an mcore
*/
/*************************************************************************/
gk_mcore_t *gk_mcoreCreate(size_t coresize)
{
gk_mcore_t *mcore;
mcore = (gk_mcore_t *)gk_malloc(sizeof(gk_mcore_t), "gk_mcoreCreate: mcore");
memset(mcore, 0, sizeof(gk_mcore_t));
mcore->coresize = coresize;
mcore->corecpos = 0;
mcore->core = (coresize == 0 ? NULL : gk_malloc(mcore->coresize, "gk_mcoreCreate: core"));
/* allocate the memory for keeping track of malloc ops */
mcore->nmops = 2048;
mcore->cmop = 0;
mcore->mops = (gk_mop_t *)gk_malloc(mcore->nmops*sizeof(gk_mop_t), "gk_mcoreCreate: mcore->mops");
return mcore;
}
/*************************************************************************/
/*! This function creates an mcore. This version is used for gkmcore.
*/
/*************************************************************************/
gk_mcore_t *gk_gkmcoreCreate()
{
gk_mcore_t *mcore;
if ((mcore = (gk_mcore_t *)malloc(sizeof(gk_mcore_t))) == NULL)
return NULL;
memset(mcore, 0, sizeof(gk_mcore_t));
/* allocate the memory for keeping track of malloc ops */
mcore->nmops = 2048;
mcore->cmop = 0;
if ((mcore->mops = (gk_mop_t *)malloc(mcore->nmops*sizeof(gk_mop_t))) == NULL) {
free(mcore);
return NULL;
}
return mcore;
}
/*************************************************************************/
/*! This function destroys an mcore.
*/
/*************************************************************************/
void gk_mcoreDestroy(gk_mcore_t **r_mcore, int showstats)
{
gk_mcore_t *mcore = *r_mcore;
if (mcore == NULL)
return;
if (showstats)
printf("\n gk_mcore statistics\n"
" coresize: %12zu nmops: %12zu cmop: %6zu\n"
" num_callocs: %12zu num_hallocs: %12zu\n"
" size_callocs: %12zu size_hallocs: %12zu\n"
" cur_callocs: %12zu cur_hallocs: %12zu\n"
" max_callocs: %12zu max_hallocs: %12zu\n",
mcore->coresize, mcore->nmops, mcore->cmop,
mcore->num_callocs, mcore->num_hallocs,
mcore->size_callocs, mcore->size_hallocs,
mcore->cur_callocs, mcore->cur_hallocs,
mcore->max_callocs, mcore->max_hallocs);
if (mcore->cur_callocs != 0 || mcore->cur_hallocs != 0 || mcore->cmop != 0) {
printf("***Warning: mcore memory was not fully freed when destroyed.\n"
" cur_callocs: %6zu cur_hallocs: %6zu cmop: %6zu\n",
mcore->cur_callocs, mcore->cur_hallocs, mcore->cmop);
}
gk_free((void **)&mcore->core, &mcore->mops, &mcore, LTERM);
*r_mcore = NULL;
}
/*************************************************************************/
/*! This function destroys an mcore. This version is for gkmcore.
*/
/*************************************************************************/
void gk_gkmcoreDestroy(gk_mcore_t **r_mcore, int showstats)
{
gk_mcore_t *mcore = *r_mcore;
if (mcore == NULL)
return;
if (showstats)
printf("\n gk_mcore statistics\n"
" nmops: %12zu cmop: %6zu\n"
" num_hallocs: %12zu\n"
" size_hallocs: %12zu\n"
" cur_hallocs: %12zu\n"
" max_hallocs: %12zu\n",
mcore->nmops, mcore->cmop,
mcore->num_hallocs,
mcore->size_hallocs,
mcore->cur_hallocs,
mcore->max_hallocs);
if (mcore->cur_hallocs != 0 || mcore->cmop != 0) {
printf("***Warning: mcore memory was not fully freed when destroyed.\n"
" cur_hallocs: %6zu cmop: %6zu\n",
mcore->cur_hallocs, mcore->cmop);
}
free(mcore->mops);
free(mcore);
*r_mcore = NULL;
}
/*************************************************************************/
/*! This function allocate space from the core/heap
*/
/*************************************************************************/
void *gk_mcoreMalloc(gk_mcore_t *mcore, size_t nbytes)
{
void *ptr;
/* pad to make pointers 8-byte aligned */
nbytes += (nbytes%8 == 0 ? 0 : 8 - nbytes%8);
if (mcore->corecpos + nbytes < mcore->coresize) {
/* service this request from the core */
ptr = ((char *)mcore->core)+mcore->corecpos;
mcore->corecpos += nbytes;
gk_mcoreAdd(mcore, GK_MOPT_CORE, nbytes, ptr);
}
else {
/* service this request from the heap */
ptr = gk_malloc(nbytes, "gk_mcoremalloc: ptr");
gk_mcoreAdd(mcore, GK_MOPT_HEAP, nbytes, ptr);
}
/*
printf("MCMALLOC: %zu %d %8zu\n", mcore->cmop-1,
mcore->mops[mcore->cmop-1].type, mcore->mops[mcore->cmop-1].nbytes);
*/
return ptr;
}
/*************************************************************************/
/*! This function sets a marker in the stack of malloc ops to be used
subsequently for freeing purposes
*/
/*************************************************************************/
void gk_mcorePush(gk_mcore_t *mcore)
{
gk_mcoreAdd(mcore, GK_MOPT_MARK, 0, NULL);
/* printf("MCPPUSH: %zu\n", mcore->cmop-1); */
}
/*************************************************************************/
/*! This function sets a marker in the stack of malloc ops to be used
subsequently for freeing purposes. This is the gkmcore version.
*/
/*************************************************************************/
void gk_gkmcorePush(gk_mcore_t *mcore)
{
gk_gkmcoreAdd(mcore, GK_MOPT_MARK, 0, NULL);
/* printf("MCPPUSH: %zu\n", mcore->cmop-1); */
}
/*************************************************************************/
/*! This function frees all mops since the last push
*/
/*************************************************************************/
void gk_mcorePop(gk_mcore_t *mcore)
{
while (mcore->cmop > 0) {
mcore->cmop--;
switch (mcore->mops[mcore->cmop].type) {
case GK_MOPT_MARK: /* push marker */
goto DONE;
break;
case GK_MOPT_CORE: /* core free */
if (mcore->corecpos < mcore->mops[mcore->cmop].nbytes)
errexit("Internal Error: wspace's core is about to be over-freed [%zu, %zu, %zd]\n",
mcore->coresize, mcore->corecpos, mcore->mops[mcore->cmop].nbytes);
mcore->corecpos -= mcore->mops[mcore->cmop].nbytes;
mcore->cur_callocs -= mcore->mops[mcore->cmop].nbytes;
break;
case GK_MOPT_HEAP: /* heap free */
gk_free((void **)&mcore->mops[mcore->cmop].ptr, LTERM);
mcore->cur_hallocs -= mcore->mops[mcore->cmop].nbytes;
break;
default:
gk_errexit(SIGMEM, "Unknown mop type of %d\n", mcore->mops[mcore->cmop].type);
}
}
DONE:
;
/*printf("MCPPOP: %zu\n", mcore->cmop); */
}
/*************************************************************************/
/*! This function frees all mops since the last push. This version is
for poping the gkmcore and it uses free instead of gk_free.
*/
/*************************************************************************/
void gk_gkmcorePop(gk_mcore_t *mcore)
{
while (mcore->cmop > 0) {
mcore->cmop--;
switch (mcore->mops[mcore->cmop].type) {
case GK_MOPT_MARK: /* push marker */
goto DONE;
break;
case GK_MOPT_HEAP: /* heap free */
free(mcore->mops[mcore->cmop].ptr);
mcore->cur_hallocs -= mcore->mops[mcore->cmop].nbytes;
break;
default:
gk_errexit(SIGMEM, "Unknown mop type of %d\n", mcore->mops[mcore->cmop].type);
}
}
DONE:
;
}
/*************************************************************************/
/*! Adds a memory allocation at the end of the list.
*/
/*************************************************************************/
void gk_mcoreAdd(gk_mcore_t *mcore, int type, size_t nbytes, void *ptr)
{
if (mcore->cmop == mcore->nmops) {
mcore->nmops *= 2;
mcore->mops = realloc(mcore->mops, mcore->nmops*sizeof(gk_mop_t));
if (mcore->mops == NULL)
gk_errexit(SIGMEM, "***Memory allocation for gkmcore failed.\n");
}
mcore->mops[mcore->cmop].type = type;
mcore->mops[mcore->cmop].nbytes = nbytes;
mcore->mops[mcore->cmop].ptr = ptr;
mcore->cmop++;
switch (type) {
case GK_MOPT_MARK:
break;
case GK_MOPT_CORE:
mcore->num_callocs++;
mcore->size_callocs += nbytes;
mcore->cur_callocs += nbytes;
if (mcore->max_callocs < mcore->cur_callocs)
mcore->max_callocs = mcore->cur_callocs;
break;
case GK_MOPT_HEAP:
mcore->num_hallocs++;
mcore->size_hallocs += nbytes;
mcore->cur_hallocs += nbytes;
if (mcore->max_hallocs < mcore->cur_hallocs)
mcore->max_hallocs = mcore->cur_hallocs;
break;
default:
gk_errexit(SIGMEM, "Incorrect mcore type operation.\n");
}
}
/*************************************************************************/
/*! Adds a memory allocation at the end of the list. This is the gkmcore
version.
*/
/*************************************************************************/
void gk_gkmcoreAdd(gk_mcore_t *mcore, int type, size_t nbytes, void *ptr)
{
if (mcore->cmop == mcore->nmops) {
mcore->nmops *= 2;
mcore->mops = realloc(mcore->mops, mcore->nmops*sizeof(gk_mop_t));
if (mcore->mops == NULL)
gk_errexit(SIGMEM, "***Memory allocation for gkmcore failed.\n");
}
mcore->mops[mcore->cmop].type = type;
mcore->mops[mcore->cmop].nbytes = nbytes;
mcore->mops[mcore->cmop].ptr = ptr;
mcore->cmop++;
switch (type) {
case GK_MOPT_MARK:
break;
case GK_MOPT_HEAP:
mcore->num_hallocs++;
mcore->size_hallocs += nbytes;
mcore->cur_hallocs += nbytes;
if (mcore->max_hallocs < mcore->cur_hallocs)
mcore->max_hallocs = mcore->cur_hallocs;
break;
default:
gk_errexit(SIGMEM, "Incorrect mcore type operation.\n");
}
}
/*************************************************************************/
/*! This function deletes the mop associated with the supplied pointer.
The mop has to be a heap allocation, otherwise it fails violently.
*/
/*************************************************************************/
void gk_mcoreDel(gk_mcore_t *mcore, void *ptr)
{
int i;
for (i=mcore->cmop-1; i>=0; i--) {
if (mcore->mops[i].type == GK_MOPT_MARK)
gk_errexit(SIGMEM, "Could not find pointer %p in mcore\n", ptr);
if (mcore->mops[i].ptr == ptr) {
if (mcore->mops[i].type != GK_MOPT_HEAP)
gk_errexit(SIGMEM, "Trying to delete a non-HEAP mop.\n");
mcore->cur_hallocs -= mcore->mops[i].nbytes;
mcore->mops[i] = mcore->mops[--mcore->cmop];
return;
}
}
gk_errexit(SIGMEM, "mcoreDel should never have been here!\n");
}
/*************************************************************************/
/*! This function deletes the mop associated with the supplied pointer.
The mop has to be a heap allocation, otherwise it fails violently.
This is the gkmcore version.
*/
/*************************************************************************/
void gk_gkmcoreDel(gk_mcore_t *mcore, void *ptr)
{
int i;
for (i=mcore->cmop-1; i>=0; i--) {
if (mcore->mops[i].type == GK_MOPT_MARK)
gk_errexit(SIGMEM, "Could not find pointer %p in mcore\n", ptr);
if (mcore->mops[i].ptr == ptr) {
if (mcore->mops[i].type != GK_MOPT_HEAP)
gk_errexit(SIGMEM, "Trying to delete a non-HEAP mop.\n");
mcore->cur_hallocs -= mcore->mops[i].nbytes;
mcore->mops[i] = mcore->mops[--mcore->cmop];
return;
}
}
gk_errexit(SIGMEM, "gkmcoreDel should never have been here!\n");
}
/*!
\file memory.c
\brief This file contains various allocation routines
The allocation routines included are for 1D and 2D arrays of the
most datatypes that GKlib support. Many of these routines are
defined with the help of the macros in gk_memory.h. These macros
can be used to define other memory allocation routines.
\date Started 4/3/2007
\author George
\version\verbatim $Id: memory.c 21050 2017-05-25 03:53:58Z karypis $ \endverbatim
*/
#include <GKlib.h>
/* This is for the global mcore that tracks all heap allocations */
static __thread gk_mcore_t *gkmcore = NULL;
/*************************************************************************/
/*! Define the set of memory allocation routines for each data type */
/**************************************************************************/
GK_MKALLOC(gk_c, char)
GK_MKALLOC(gk_i, int)
GK_MKALLOC(gk_i8, int8_t)
GK_MKALLOC(gk_i16, int16_t)
GK_MKALLOC(gk_i32, int32_t)
GK_MKALLOC(gk_i64, int64_t)
GK_MKALLOC(gk_ui8, uint8_t)
GK_MKALLOC(gk_ui16, uint16_t)
GK_MKALLOC(gk_ui32, uint32_t)
GK_MKALLOC(gk_ui64, uint64_t)
GK_MKALLOC(gk_z, ssize_t)
GK_MKALLOC(gk_zu, size_t)
GK_MKALLOC(gk_f, float)
GK_MKALLOC(gk_d, double)
GK_MKALLOC(gk_idx, gk_idx_t)
GK_MKALLOC(gk_ckv, gk_ckv_t)
GK_MKALLOC(gk_ikv, gk_ikv_t)
GK_MKALLOC(gk_i8kv, gk_i8kv_t)
GK_MKALLOC(gk_i16kv, gk_i16kv_t)
GK_MKALLOC(gk_i32kv, gk_i32kv_t)
GK_MKALLOC(gk_i64kv, gk_i64kv_t)
GK_MKALLOC(gk_zkv, gk_zkv_t)
GK_MKALLOC(gk_zukv, gk_zukv_t)
GK_MKALLOC(gk_fkv, gk_fkv_t)
GK_MKALLOC(gk_dkv, gk_dkv_t)
GK_MKALLOC(gk_skv, gk_skv_t)
GK_MKALLOC(gk_idxkv, gk_idxkv_t)
/*************************************************************************/
/*! This function allocates a two-dimensional matrix.
*/
/*************************************************************************/
void gk_AllocMatrix(void ***r_matrix, size_t elmlen, size_t ndim1, size_t ndim2)
{
size_t i, j;
void **matrix;
*r_matrix = NULL;
if ((matrix = (void **)gk_malloc(ndim1*sizeof(void *), "gk_AllocMatrix: matrix")) == NULL)
return;
for (i=0; i<ndim1; i++) {
if ((matrix[i] = (void *)gk_malloc(ndim2*elmlen, "gk_AllocMatrix: matrix[i]")) == NULL) {
for (j=0; j<i; j++)
gk_free((void **)&matrix[j], LTERM);
return;
}
}
*r_matrix = matrix;
}
/*************************************************************************/
/*! This function frees a two-dimensional matrix.
*/
/*************************************************************************/
void gk_FreeMatrix(void ***r_matrix, size_t ndim1, size_t ndim2)
{
size_t i;
void **matrix;
if ((matrix = *r_matrix) == NULL)
return;
for (i=0; i<ndim1; i++)
gk_free((void **)&matrix[i], LTERM);
gk_free((void **)r_matrix, LTERM);
}
/*************************************************************************/
/*! This function initializes tracking of heap allocations.
*/
/*************************************************************************/
int gk_malloc_init()
{
if (gkmcore == NULL)
gkmcore = gk_gkmcoreCreate();
if (gkmcore == NULL)
return 0;
gk_gkmcorePush(gkmcore);
return 1;
}
/*************************************************************************/
/*! This function frees the memory that has been allocated since the
last call to gk_malloc_init().
*/
/*************************************************************************/
void gk_malloc_cleanup(int showstats)
{
if (gkmcore != NULL) {
gk_gkmcorePop(gkmcore);
if (gkmcore->cmop == 0) {
gk_gkmcoreDestroy(&gkmcore, showstats);
gkmcore = NULL;
}
}
}
/*************************************************************************/
/*! This function is my wrapper around malloc that provides the following
enhancements over malloc:
* It always allocates one byte of memory, even if 0 bytes are requested.
This is to ensure that checks of returned values do not lead to NULL
due to 0 bytes requested.
* It zeros-out the memory that is allocated. This is for a quick init
of the underlying datastructures.
*/
/**************************************************************************/
void *gk_malloc(size_t nbytes, char *msg)
{
void *ptr=NULL;
if (nbytes == 0)
nbytes++; /* Force mallocs to actually allocate some memory */
ptr = (void *)malloc(nbytes);
if (ptr == NULL) {
fprintf(stderr, " Current memory used: %10zu bytes\n", gk_GetCurMemoryUsed());
fprintf(stderr, " Maximum memory used: %10zu bytes\n", gk_GetMaxMemoryUsed());
gk_errexit(SIGMEM, "***Memory allocation failed for %s. Requested size: %zu bytes",
msg, nbytes);
return NULL;
}
/* add this memory allocation */
if (gkmcore != NULL) gk_gkmcoreAdd(gkmcore, GK_MOPT_HEAP, nbytes, ptr);
return ptr;
}
/*************************************************************************
* This function is my wrapper around realloc
**************************************************************************/
void *gk_realloc(void *oldptr, size_t nbytes, char *msg)
{
void *ptr=NULL;
if (nbytes == 0)
nbytes++; /* Force mallocs to actually allocate some memory */
/* remove this memory de-allocation */
if (gkmcore != NULL && oldptr != NULL) gk_gkmcoreDel(gkmcore, oldptr);
ptr = (void *)realloc(oldptr, nbytes);
if (ptr == NULL) {
fprintf(stderr, " Maximum memory used: %10zu bytes\n", gk_GetMaxMemoryUsed());
fprintf(stderr, " Current memory used: %10zu bytes\n", gk_GetCurMemoryUsed());
gk_errexit(SIGMEM, "***Memory realloc failed for %s. " "Requested size: %zu bytes",
msg, nbytes);
return NULL;
}
/* add this memory allocation */
if (gkmcore != NULL) gk_gkmcoreAdd(gkmcore, GK_MOPT_HEAP, nbytes, ptr);
return ptr;
}
/*************************************************************************
* This function is my wrapper around free, allows multiple pointers
**************************************************************************/
void gk_free(void **ptr1,...)
{
va_list plist;
void **ptr;
if (*ptr1 != NULL) {
free(*ptr1);
/* remove this memory de-allocation */
if (gkmcore != NULL)
gk_gkmcoreDel(gkmcore, *ptr1);
}
*ptr1 = NULL;
va_start(plist, ptr1);
while ((ptr = va_arg(plist, void **)) != LTERM) {
if (*ptr != NULL) {
free(*ptr);
/* remove this memory de-allocation */
if (gkmcore != NULL)
gk_gkmcoreDel(gkmcore, *ptr);
}
*ptr = NULL;
}
va_end(plist);
}
/*************************************************************************
* This function returns the current ammount of dynamically allocated
* memory that is used by the system
**************************************************************************/
size_t gk_GetCurMemoryUsed()
{
if (gkmcore == NULL)
return 0;
else
return gkmcore->cur_hallocs;
}
/*************************************************************************
* This function returns the maximum ammount of dynamically allocated
* memory that was used by the system
**************************************************************************/
size_t gk_GetMaxMemoryUsed()
{
if (gkmcore == NULL)
return 0;
else
return gkmcore->max_hallocs;
}
/*************************************************************************/
/*! This function returns the VmSize and VmRSS of the calling process. */
/*************************************************************************/
void gk_GetVMInfo(size_t *vmsize, size_t *vmrss)
{
FILE *fp;
char fname[1024];
sprintf(fname, "/proc/%d/statm", getpid());
fp = gk_fopen(fname, "r", "proc/pid/statm");
if (fscanf(fp, "%zu %zu", vmsize, vmrss) != 2)
errexit("Failed to read to values from %s\n", fname);
gk_fclose(fp);
/*
*vmsize *= sysconf(_SC_PAGESIZE);
*vmrss *= sysconf(_SC_PAGESIZE);
*/
return;
}
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment