Spaces:
Sleeping
Sleeping
File size: 12,954 Bytes
66c9c8a | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 | #include "warp.h"
#include <algorithm>
#include <numeric>
#include <vector>
namespace
{
// Specialized is_zero and accumulation function for common block sizes
// Rely on compiler to unroll loops when block size is known
template <int N, typename T> bool bsr_fixed_block_is_zero(const T *val, int value_size)
{
return std::all_of(val, val + N, [](float v) { return v == T(0); });
}
template <typename T> bool bsr_dyn_block_is_zero(const T *val, int value_size)
{
return std::all_of(val, val + value_size, [](float v) { return v == T(0); });
}
template <int N, typename T> void bsr_fixed_block_accumulate(const T *val, T *sum, int value_size)
{
for (int i = 0; i < N; ++i, ++val, ++sum)
{
*sum += *val;
}
}
template <typename T> void bsr_dyn_block_accumulate(const T *val, T *sum, int value_size)
{
for (int i = 0; i < value_size; ++i, ++val, ++sum)
{
*sum += *val;
}
}
template <int Rows, int Cols, typename T>
void bsr_fixed_block_transpose(const T *src, T *dest, int row_count, int col_count)
{
for (int r = 0; r < Rows; ++r)
{
for (int c = 0; c < Cols; ++c)
{
dest[c * Rows + r] = src[r * Cols + c];
}
}
}
template <typename T> void bsr_dyn_block_transpose(const T *src, T *dest, int row_count, int col_count)
{
for (int r = 0; r < row_count; ++r)
{
for (int c = 0; c < col_count; ++c)
{
dest[c * row_count + r] = src[r * col_count + c];
}
}
}
} // namespace
template <typename T>
int bsr_matrix_from_triplets_host(const int rows_per_block, const int cols_per_block, const int row_count,
const int nnz, const int *tpl_rows, const int *tpl_columns, const T *tpl_values,
int *bsr_offsets, int *bsr_columns, T *bsr_values)
{
// get specialized accumulator for common block sizes (1,1), (1,2), (1,3),
// (2,2), (2,3), (3,3)
const int block_size = rows_per_block * cols_per_block;
void (*block_accumulate_func)(const T *, T *, int);
bool (*block_is_zero_func)(const T *, int);
switch (block_size)
{
case 1:
block_accumulate_func = bsr_fixed_block_accumulate<1, T>;
block_is_zero_func = bsr_fixed_block_is_zero<1, T>;
break;
case 2:
block_accumulate_func = bsr_fixed_block_accumulate<2, T>;
block_is_zero_func = bsr_fixed_block_is_zero<2, T>;
break;
case 3:
block_accumulate_func = bsr_fixed_block_accumulate<3, T>;
block_is_zero_func = bsr_fixed_block_is_zero<3, T>;
break;
case 4:
block_accumulate_func = bsr_fixed_block_accumulate<4, T>;
block_is_zero_func = bsr_fixed_block_is_zero<4, T>;
break;
case 6:
block_accumulate_func = bsr_fixed_block_accumulate<6, T>;
block_is_zero_func = bsr_fixed_block_is_zero<6, T>;
break;
case 9:
block_accumulate_func = bsr_fixed_block_accumulate<9, T>;
block_is_zero_func = bsr_fixed_block_is_zero<9, T>;
break;
default:
block_accumulate_func = bsr_dyn_block_accumulate<T>;
block_is_zero_func = bsr_dyn_block_is_zero<T>;
}
std::vector<int> block_indices(nnz);
std::iota(block_indices.begin(), block_indices.end(), 0);
// remove zero block indices
if (tpl_values)
{
block_indices.erase(std::remove_if(block_indices.begin(), block_indices.end(),
[block_is_zero_func, tpl_values, block_size](int i) {
return block_is_zero_func(tpl_values + i * block_size, block_size);
}),
block_indices.end());
}
// sort block indices according to lexico order
std::sort(block_indices.begin(), block_indices.end(), [tpl_rows, tpl_columns](int i, int j) -> bool {
return tpl_rows[i] < tpl_rows[j] || (tpl_rows[i] == tpl_rows[j] && tpl_columns[i] < tpl_columns[j]);
});
// accumulate blocks at same locations, count blocks per row
std::fill_n(bsr_offsets, row_count + 1, 0);
int current_row = -1;
int current_col = -1;
// so that we get back to the start for the first block
if (bsr_values)
{
bsr_values -= block_size;
}
for (int i = 0; i < block_indices.size(); ++i)
{
int idx = block_indices[i];
int row = tpl_rows[idx];
int col = tpl_columns[idx];
const T *val = tpl_values + idx * block_size;
if (row == current_row && col == current_col)
{
if (bsr_values)
{
block_accumulate_func(val, bsr_values, block_size);
}
}
else
{
*(bsr_columns++) = col;
if (bsr_values)
{
bsr_values += block_size;
std::copy_n(val, block_size, bsr_values);
}
bsr_offsets[row + 1]++;
current_row = row;
current_col = col;
}
}
// build postfix sum of row counts
std::partial_sum(bsr_offsets, bsr_offsets + row_count + 1, bsr_offsets);
return bsr_offsets[row_count];
}
template <typename T>
void bsr_transpose_host(int rows_per_block, int cols_per_block, int row_count, int col_count, int nnz,
const int *bsr_offsets, const int *bsr_columns, const T *bsr_values,
int *transposed_bsr_offsets, int *transposed_bsr_columns, T *transposed_bsr_values)
{
const int block_size = rows_per_block * cols_per_block;
void (*block_transpose_func)(const T *, T *, int, int) = bsr_dyn_block_transpose<T>;
switch (rows_per_block)
{
case 1:
switch (cols_per_block)
{
case 1:
block_transpose_func = bsr_fixed_block_transpose<1, 1, T>;
break;
case 2:
block_transpose_func = bsr_fixed_block_transpose<1, 2, T>;
break;
case 3:
block_transpose_func = bsr_fixed_block_transpose<1, 3, T>;
break;
}
break;
case 2:
switch (cols_per_block)
{
case 1:
block_transpose_func = bsr_fixed_block_transpose<2, 1, T>;
break;
case 2:
block_transpose_func = bsr_fixed_block_transpose<2, 2, T>;
break;
case 3:
block_transpose_func = bsr_fixed_block_transpose<2, 3, T>;
break;
}
break;
case 3:
switch (cols_per_block)
{
case 1:
block_transpose_func = bsr_fixed_block_transpose<3, 1, T>;
break;
case 2:
block_transpose_func = bsr_fixed_block_transpose<3, 2, T>;
break;
case 3:
block_transpose_func = bsr_fixed_block_transpose<3, 3, T>;
break;
}
break;
}
std::vector<int> block_indices(nnz), bsr_rows(nnz);
std::iota(block_indices.begin(), block_indices.end(), 0);
// Fill row indices from offsets
for (int row = 0; row < row_count; ++row)
{
std::fill(bsr_rows.begin() + bsr_offsets[row], bsr_rows.begin() + bsr_offsets[row + 1], row);
}
// sort block indices according to (transposed) lexico order
std::sort(block_indices.begin(), block_indices.end(), [&bsr_rows, bsr_columns](int i, int j) -> bool {
return bsr_columns[i] < bsr_columns[j] || (bsr_columns[i] == bsr_columns[j] && bsr_rows[i] < bsr_rows[j]);
});
// Count blocks per column and transpose blocks
std::fill_n(transposed_bsr_offsets, col_count + 1, 0);
for (int i = 0; i < nnz; ++i)
{
int idx = block_indices[i];
int row = bsr_rows[idx];
int col = bsr_columns[idx];
++transposed_bsr_offsets[col + 1];
transposed_bsr_columns[i] = row;
const T *src_block = bsr_values + idx * block_size;
T *dst_block = transposed_bsr_values + i * block_size;
block_transpose_func(src_block, dst_block, rows_per_block, cols_per_block);
}
// build postfix sum of column counts
std::partial_sum(transposed_bsr_offsets, transposed_bsr_offsets + col_count + 1, transposed_bsr_offsets);
}
WP_API int bsr_matrix_from_triplets_float_host(int rows_per_block, int cols_per_block, int row_count, int nnz,
uint64_t tpl_rows, uint64_t tpl_columns, uint64_t tpl_values,
uint64_t bsr_offsets, uint64_t bsr_columns, uint64_t bsr_values)
{
return bsr_matrix_from_triplets_host(
rows_per_block, cols_per_block, row_count, nnz, reinterpret_cast<const int *>(tpl_rows),
reinterpret_cast<const int *>(tpl_columns), reinterpret_cast<const float *>(tpl_values),
reinterpret_cast<int *>(bsr_offsets), reinterpret_cast<int *>(bsr_columns),
reinterpret_cast<float *>(bsr_values));
}
WP_API int bsr_matrix_from_triplets_double_host(int rows_per_block, int cols_per_block, int row_count, int nnz,
uint64_t tpl_rows, uint64_t tpl_columns, uint64_t tpl_values,
uint64_t bsr_offsets, uint64_t bsr_columns, uint64_t bsr_values)
{
return bsr_matrix_from_triplets_host(
rows_per_block, cols_per_block, row_count, nnz, reinterpret_cast<const int *>(tpl_rows),
reinterpret_cast<const int *>(tpl_columns), reinterpret_cast<const double *>(tpl_values),
reinterpret_cast<int *>(bsr_offsets), reinterpret_cast<int *>(bsr_columns),
reinterpret_cast<double *>(bsr_values));
}
WP_API void bsr_transpose_float_host(int rows_per_block, int cols_per_block, int row_count, int col_count, int nnz,
uint64_t bsr_offsets, uint64_t bsr_columns, uint64_t bsr_values,
uint64_t transposed_bsr_offsets, uint64_t transposed_bsr_columns,
uint64_t transposed_bsr_values)
{
bsr_transpose_host(rows_per_block, cols_per_block, row_count, col_count, nnz,
reinterpret_cast<const int *>(bsr_offsets), reinterpret_cast<const int *>(bsr_columns),
reinterpret_cast<const float *>(bsr_values), reinterpret_cast<int *>(transposed_bsr_offsets),
reinterpret_cast<int *>(transposed_bsr_columns),
reinterpret_cast<float *>(transposed_bsr_values));
}
WP_API void bsr_transpose_double_host(int rows_per_block, int cols_per_block, int row_count, int col_count, int nnz,
uint64_t bsr_offsets, uint64_t bsr_columns, uint64_t bsr_values,
uint64_t transposed_bsr_offsets, uint64_t transposed_bsr_columns,
uint64_t transposed_bsr_values)
{
bsr_transpose_host(rows_per_block, cols_per_block, row_count, col_count, nnz,
reinterpret_cast<const int *>(bsr_offsets), reinterpret_cast<const int *>(bsr_columns),
reinterpret_cast<const double *>(bsr_values), reinterpret_cast<int *>(transposed_bsr_offsets),
reinterpret_cast<int *>(transposed_bsr_columns),
reinterpret_cast<double *>(transposed_bsr_values));
}
#if !WP_ENABLE_CUDA
WP_API int bsr_matrix_from_triplets_float_device(int rows_per_block, int cols_per_block, int row_count, int nnz,
uint64_t tpl_rows, uint64_t tpl_columns, uint64_t tpl_values,
uint64_t bsr_offsets, uint64_t bsr_columns, uint64_t bsr_values)
{
return 0;
}
WP_API int bsr_matrix_from_triplets_double_device(int rows_per_block, int cols_per_block, int row_count, int nnz,
uint64_t tpl_rows, uint64_t tpl_columns, uint64_t tpl_values,
uint64_t bsr_offsets, uint64_t bsr_columns, uint64_t bsr_values)
{
return 0;
}
WP_API void bsr_transpose_float_device(int rows_per_block, int cols_per_block, int row_count, int col_count, int nnz,
uint64_t bsr_offsets, uint64_t bsr_columns, uint64_t bsr_values,
uint64_t transposed_bsr_offsets, uint64_t transposed_bsr_columns,
uint64_t transposed_bsr_values)
{
}
WP_API void bsr_transpose_double_device(int rows_per_block, int cols_per_block, int row_count, int col_count, int nnz,
uint64_t bsr_offsets, uint64_t bsr_columns, uint64_t bsr_values,
uint64_t transposed_bsr_offsets, uint64_t transposed_bsr_columns,
uint64_t transposed_bsr_values)
{
}
#endif |