1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213
|
//------------------------------------------------------------------------------
// GB_unjumble: unjumble the vectors of a matrix
//------------------------------------------------------------------------------
// SuiteSparse:GraphBLAS, Timothy A. Davis, (c) 2017-2025, All Rights Reserved.
// SPDX-License-Identifier: Apache-2.0
//------------------------------------------------------------------------------
#include "sort/GB_sort.h"
#include "unaryop/GB_unop.h"
#include "jitifyer/GB_stringify.h"
#define GB_FREE_ALL GB_WERK_POP (A_slice, int64_t) ;
GrB_Info GB_unjumble // unjumble a matrix
(
GrB_Matrix A, // matrix to unjumble
GB_Werk Werk
)
{
//--------------------------------------------------------------------------
// check inputs
//--------------------------------------------------------------------------
ASSERT_MATRIX_OK (A, "A to unjumble", GB0) ;
ASSERT (!GB_ZOMBIES (A)) ; // zombies must be killed first
ASSERT (GB_PENDING_OK (A)) ; // pending tuples are not modified
GB_nvec_nonempty_update (A) ;
if (!A->jumbled)
{
// nothing to do
return (GrB_SUCCESS) ;
}
// full and bitmap matrices are never jumbled
ASSERT (!GB_IS_FULL (A)) ;
ASSERT (!GB_IS_BITMAP (A)) ;
ASSERT (GB_IS_SPARSE (A) || GB_IS_HYPERSPARSE (A)) ;
//--------------------------------------------------------------------------
// get A
//--------------------------------------------------------------------------
const int64_t anvec = A->nvec ;
const int64_t anz = GB_nnz (A) ;
GB_Ap_DECLARE (Ap, const) ; GB_Ap_PTR (Ap, A) ;
GB_Ai_DECLARE_U (Ai, ) ; GB_Ai_PTR (Ai, A) ;
bool Ai_is_32 = A->i_is_32 ;
const size_t asize = (A->iso) ? 0 : A->type->size ;
//--------------------------------------------------------------------------
// determine the number of threads to use
//--------------------------------------------------------------------------
int nthreads_max = GB_Context_nthreads_max ( ) ;
double chunk = GB_Context_chunk ( ) ;
int nthreads = GB_nthreads (anz + anvec, chunk, nthreads_max) ;
int ntasks = (nthreads == 1) ? 1 : (32 * nthreads) ;
ntasks = GB_IMIN (ntasks, anvec) ;
ntasks = GB_IMAX (ntasks, 1) ;
//--------------------------------------------------------------------------
// slice the work
//--------------------------------------------------------------------------
GB_WERK_DECLARE (A_slice, int64_t) ;
GB_WERK_PUSH (A_slice, ntasks + 1, int64_t) ;
if (A_slice == NULL)
{
// out of memory
return (GrB_OUT_OF_MEMORY) ;
}
GB_p_slice (A_slice, Ap, A->p_is_32, anvec, ntasks, false) ;
//--------------------------------------------------------------------------
// sort the vectors
//--------------------------------------------------------------------------
GrB_Info info = GrB_NO_VALUE ;
if (asize == 0)
{
//----------------------------------------------------------------------
// iso matrices of any type; only sort the pattern
//----------------------------------------------------------------------
#define GB_QSORT \
if (Ai_is_32) GB_qsort_1_32 (Ai32+p, n) ; \
else GB_qsort_1_64 (Ai64+p, n) ;
#include "wait/template/GB_unjumbled_template.c"
info = GrB_SUCCESS ;
}
else
{
//----------------------------------------------------------------------
// factory kernels for non-iso matrices
//----------------------------------------------------------------------
#ifndef GBCOMPACT
GB_IF_FACTORY_KERNELS_ENABLED
{
switch (asize)
{
case GB_1BYTE : // bool, uint8, int8, and user types of size 1
{
uint8_t *Ax = (uint8_t *) A->x ;
#define GB_QSORT \
if (Ai_is_32) GB_qsort_1b_32_size1 (Ai32+p, Ax+p, n) ; \
else GB_qsort_1b_64_size1 (Ai64+p, Ax+p, n) ;
#include "wait/template/GB_unjumbled_template.c"
info = GrB_SUCCESS ;
}
break ;
case GB_2BYTE : // uint16, int16, and user types of size 2
{
uint16_t *Ax = (uint16_t *) A->x ;
#define GB_QSORT \
if (Ai_is_32) GB_qsort_1b_32_size2 (Ai32+p, Ax+p, n) ; \
else GB_qsort_1b_64_size2 (Ai64+p, Ax+p, n) ;
#include "wait/template/GB_unjumbled_template.c"
info = GrB_SUCCESS ;
}
break ;
case GB_4BYTE : // uint32, int32, float, and 4-byte user
{
uint32_t *Ax = (uint32_t *) A->x ;
#define GB_QSORT \
if (Ai_is_32) GB_qsort_1b_32_size4 (Ai32+p, Ax+p, n) ; \
else GB_qsort_1b_64_size4 (Ai64+p, Ax+p, n) ;
#include "wait/template/GB_unjumbled_template.c"
info = GrB_SUCCESS ;
}
break ;
case GB_8BYTE : // uint64, int64, double, float complex,
// and 8-byte user-defined types
{
uint64_t *Ax = (uint64_t *) A->x ;
#define GB_QSORT \
if (Ai_is_32) GB_qsort_1b_32_size8 (Ai32+p, Ax+p, n) ; \
else GB_qsort_1b_64_size8 (Ai64+p, Ax+p, n) ;
#include "wait/template/GB_unjumbled_template.c"
info = GrB_SUCCESS ;
}
break ;
case GB_16BYTE : // double complex, and user types of size 16
{
GB_blob16 *Ax = (GB_blob16 *) A->x ;
#define GB_QSORT \
if (Ai_is_32) GB_qsort_1b_32_size16 (Ai32+p, Ax+p, n) ; \
else GB_qsort_1b_64_size16 (Ai64+p, Ax+p, n) ;
#include "wait/template/GB_unjumbled_template.c"
info = GrB_SUCCESS ;
}
break ;
default:;
}
}
#endif
}
//--------------------------------------------------------------------------
// via the JIT kernel
//--------------------------------------------------------------------------
if (info == GrB_NO_VALUE)
{
GBURBLE ("(unjumble: jit kernel) ") ;
struct GB_UnaryOp_opaque op_header ;
GB_Operator op = GB_unop_identity (A->type, &op_header) ;
info = GB_unjumble_jit (A, op, A_slice, ntasks, nthreads) ;
}
//--------------------------------------------------------------------------
// via the generic kernel
//--------------------------------------------------------------------------
if (info == GrB_NO_VALUE)
{
GBURBLE ("(unjumble: generic kernel) ") ;
GB_void *Ax = (GB_void *) A->x ;
#define GB_QSORT \
if (Ai_is_32) GB_qsort_1b_32_generic (Ai32+p, Ax+p*asize, asize, n) ; \
else GB_qsort_1b_64_generic (Ai64+p, Ax+p*asize, asize, n) ;
#include "wait/template/GB_unjumbled_template.c"
info = GrB_SUCCESS ;
}
//--------------------------------------------------------------------------
// free workspace and return result
//--------------------------------------------------------------------------
GB_FREE_ALL ;
if (info == GrB_SUCCESS)
{
A->jumbled = false ; // A has been unjumbled
ASSERT_MATRIX_OK (A, "A unjumbled", GB0) ;
ASSERT (GB_nvec_nonempty_get (A) >= 0)
}
return (info) ;
}
|