git.sesse.net Git - remoteglot-book/blob - binloader.cpp

   1 //#define _GLIBCXX_PARALLEL
   2
   3 // Usage: ./binloader IN1 IN2 IN3 ... OUT NUM_BUCKETS NUM_POS_PER_SUBSHARD
   4
   5 #include <stdio.h>
   6 #include <vector>
   7 #include <mtbl.h>
   8 #include <algorithm>
   9 #include <utility>
  10 #include <memory>
  11 #include <string>
  12 #include <unordered_set>
  13 #include <string.h>
  14 #include "count.pb.h"
  15 #include "hash.h"
  16
  17 #define DUMMY_TIMESTAMP 32503680000
  18
  19 using namespace std;
  20
  21 static inline int memcmp_different_len(const void *s1, size_t n1, const void *s2, size_t n2)
  22 {
  23         size_t shared_len = min(n1, n2);
  24         if (shared_len >= 8) {
  25                 uint64_t a1 = *(const uint64_t *)s1;
  26                 uint64_t a2 = *(const uint64_t *)s2;
  27                 if (a1 != a2) {
  28                         a1 = __builtin_bswap64(a1);
  29                         a2 = __builtin_bswap64(a2);
  30                         return (a1 < a2) ? -1 : 1;
  31                 }
  32         }
  33
  34         int s = memcmp(s1, s2, shared_len);
  35         if (s != 0) {
  36                 return s;
  37         }
  38
  39         return n2 - n1;
  40 }
  41
  42 enum Result { WHITE = 0, DRAW, BLACK };
  43 struct Element {
  44         char bpfen[32];  // includes prev_board_hash
  45         int bpfen_len;
  46         char move[8];   // Na1xc3+
  47         Result result;
  48         int opening_num, white_elo, black_elo;
  49         int file_num;
  50         time_t timestamp;
  51         long start_position;
  52         int computer;
  53
  54         bool operator< (const Element& other) const {
  55                 int s = memcmp_different_len(bpfen, bpfen_len, other.bpfen, other.bpfen_len);
  56                 return s < 0;
  57         }
  58 };
  59
  60 struct ShardData {
  61         vector<Element> elems;
  62         int num_written_subshards = 0;
  63 };
  64
  65 void write_subshard(const char *basename, ShardData* shard, int bucket)
  66 {
  67         string buf;  // Keep allocated.
  68         char filename[256];
  69         snprintf(filename, sizeof(filename), "%s.part%04d.subshard%04d",
  70                 basename, bucket, shard->num_written_subshards++);
  71         printf("Writing SSTable %s...\n", filename);
  72
  73         sort(shard->elems.begin(), shard->elems.end());
  74
  75         mtbl_writer_options* wopt = mtbl_writer_options_init();
  76         mtbl_writer_options_set_compression(wopt, MTBL_COMPRESSION_SNAPPY);
  77         mtbl_writer* mtbl = mtbl_writer_init(filename, wopt);
  78         Count c;
  79         unordered_set<string> moves;
  80         for (size_t i = 0; i < shard->elems.size(); ++i) {
  81                 const Element &e = shard->elems[i];
  82                 if (e.result == WHITE) {
  83                         c.set_white(c.white() + 1);
  84                 } else if (e.result == DRAW) {
  85                         c.set_draw(c.draw() + 1);
  86                 } else if (e.result == BLACK) {
  87                         c.set_black(c.black() + 1);
  88                 }
  89                 c.set_computer(c.computer() + e.computer);
  90                 if (e.white_elo >= 100 && e.black_elo >= 100) {
  91                         c.set_sum_white_elo(c.sum_white_elo() + e.white_elo);
  92                         c.set_sum_black_elo(c.sum_black_elo() + e.black_elo);
  93                         c.set_num_elo(c.num_elo() + 1);
  94                 }
  95                 if (!c.has_first_timestamp() || e.timestamp < c.first_timestamp()) {
  96                         if (e.timestamp != DUMMY_TIMESTAMP) {
  97                                 c.set_first_timestamp(e.timestamp);
  98                         }
  99                         c.set_opening_num(e.opening_num);
 100                         c.set_pgn_file_num(e.file_num);
 101                         c.set_pgn_start_position(e.start_position);
 102                 }
 103                 if (strlen(e.move) > 0 && !moves.count(e.move)) {
 104                         moves.insert(e.move);
 105                         c.add_move(e.move);
 106                 }
 107                 if (i == shard->elems.size() - 1 ||
 108                     e.bpfen_len != shard->elems[i + 1].bpfen_len ||
 109                     memcmp(e.bpfen, shard->elems[i + 1].bpfen, e.bpfen_len) != 0) {
 110                         c.SerializeToString(&buf);
 111                         mtbl_res res = mtbl_writer_add(mtbl,
 112                                 (const uint8_t *)e.bpfen, e.bpfen_len,
 113                                 (const uint8_t *)buf.data(), buf.size());
 114                         assert(res == mtbl_res_success);
 115                         c = Count();
 116                         moves.clear();
 117                 }
 118         }
 119         mtbl_writer_destroy(&mtbl);
 120
 121         shard->elems.clear();
 122 }
 123
 124 int main(int argc, char **argv)
 125 {
 126         int num_buckets = atoi(argv[argc - 2]);
 127         size_t num_pos_per_subshard = atoi(argv[argc - 1]);  // 500000 is a reasonable value.
 128
 129         vector<ShardData> shards;
 130         shards.resize(num_buckets);
 131
 132         for (int i = 0; i < num_buckets; ++i) {
 133                 shards[i].elems.reserve(num_pos_per_subshard);
 134         }
 135
 136         size_t num_elems = 0;
 137         for (int i = 1; i < argc - 3; ++i) {
 138                 FILE *fp;
 139                 if (strcmp(argv[i], "-") == 0) {
 140                         fp = stdin;
 141                 } else {
 142                         fp = fopen(argv[i], "rb");
 143                         if (fp == NULL) {
 144                                 perror(argv[i]);
 145                                 exit(1);
 146                         }
 147                 }
 148                 for ( ;; ) {
 149                         char bpfen[256];
 150
 151                         int bpfen_len = getc(fp);
 152                         if (bpfen_len == -1) {
 153                                 break;
 154                         }
 155                         if (bpfen_len <= 0) {
 156                                 fprintf(stderr, "Underlong BPFEN (%d bytes)\n", bpfen_len);
 157                                 exit(1);
 158                         }
 159                         if (bpfen_len >= 32) {
 160                                 fprintf(stderr, "Overlong BPFEN (%d bytes)\n", bpfen_len);
 161                                 exit(1);
 162                         }
 163                         if (fread(bpfen, bpfen_len, 1, fp) != 1) {
 164                                 perror("fread(bpfen)");
 165                                 exit(1);
 166                         }
 167
 168                         int r = getc(fp);
 169                         if (r == -1) {
 170                                 perror("getc()");
 171                                 exit(1);
 172                         }
 173
 174                         int opening_num, white_elo, black_elo, file_num;
 175                         time_t timestamp;
 176                         long start_position;
 177                         if (fread(&white_elo, sizeof(white_elo), 1, fp) != 1) {
 178                                 perror("fread(white_elo)");
 179                                 exit(1);
 180                         }
 181                         if (fread(&black_elo, sizeof(black_elo), 1, fp) != 1) {
 182                                 perror("fread(black_elo)");
 183                                 exit(1);
 184                         }
 185                         if (fread(&opening_num, sizeof(opening_num), 1, fp) != 1) {
 186                                 perror("fread(opening_num)");
 187                                 exit(1);
 188                         }
 189                         if (fread(&timestamp, sizeof(timestamp), 1, fp) != 1) {
 190                                 perror("fread(timestamp)");
 191                                 exit(1);
 192                         }
 193                         if (fread(&file_num, sizeof(file_num), 1, fp) != 1) {
 194                                 perror("fread(file_num)");
 195                                 exit(1);
 196                         }
 197                         if (fread(&start_position, sizeof(start_position), 1, fp) != 1) {
 198                                 perror("fread(start_position)");
 199                                 exit(1);
 200                         }
 201
 202                         int computer = getc(fp);
 203                         if (computer == -1) {
 204                                 perror("getc(computer)");
 205                                 exit(1);
 206                         }
 207
 208                         char move[8];
 209                         int l = getc(fp);
 210                         if (l == -1) {
 211                                 perror("getc(move_length)");
 212                                 break;
 213                         }
 214                         if (l >= int(sizeof(move))) {
 215                                 fprintf(stderr, "Overlong move (%d bytes)\n", l);
 216                                 exit(1);
 217                         }
 218                         if (l == 0) {
 219                                 move[0] = 0;
 220                         } else if (fread(&move[0], l, 1, fp) != 1) {
 221                                 perror("fread(move)");
 222                                 exit(1);
 223                         }
 224                         move[l] = 0;
 225
 226                         int bucket = hash_key_to_bucket(bpfen, bpfen_len, num_buckets);
 227                         Element e;
 228                         memcpy(e.bpfen, bpfen, bpfen_len);
 229                         e.bpfen_len = bpfen_len;
 230                         strcpy(e.move, move);
 231                         e.result = Result(r);
 232                         e.computer = computer;
 233                         e.opening_num = opening_num;
 234                         e.white_elo = white_elo;
 235                         e.black_elo = black_elo;
 236                         e.file_num = file_num;
 237                         e.timestamp = timestamp;
 238                         e.start_position = start_position;
 239                         shards[bucket].elems.push_back(e);
 240                         ++num_elems;
 241
 242                         if (shards[bucket].elems.size() >= num_pos_per_subshard) {
 243                                 write_subshard(argv[argc - 3], &shards[bucket], bucket);
 244                                 shards[bucket].elems.reserve(num_pos_per_subshard);
 245                         }
 246                 }
 247                 fclose(fp);
 248
 249                 printf("Read %ld elems\n", num_elems);
 250         }
 251
 252         for (int i = 0; i < num_buckets; ++i) {
 253                 write_subshard(argv[argc - 3], &shards[i], i);
 254         }
 255 }
 256