| 1 | RCU Judy Array Design |
| 2 | Mathieu Desnoyers <mathieu.desnoyers@efficios.com> |
| 3 | March 8, 2012 |
| 4 | |
| 5 | Initial ideas based on the released Judy Shop Manual |
| 6 | (http://judy.sourceforge.net/). Judy was invented by Doug Baskins and |
| 7 | implemented by Hewlett-Packard. |
| 8 | |
| 9 | Thresholds and RCU-specific analysis is introduced in this document. |
| 10 | |
| 11 | Advantages of using Judy Array (compressed nodes) for RCU tree: |
| 12 | - no rebalancing |
| 13 | - no transplant |
| 14 | - RCU-friendly! |
| 15 | - favor cache-line alignment of structures |
| 16 | |
| 17 | Disadvantage: |
| 18 | - updates that need to reallocate nodes are slower than, e.g. non-rcu |
| 19 | red-black trees. |
| 20 | |
| 21 | Choice: Using 256 entries intermediate nodes (index can be represented |
| 22 | on 8 bits): 4 levels on 32-bit, 8 levels on 64-bit |
| 23 | |
| 24 | |
| 25 | * Node types (from less dense node to most dense) |
| 26 | |
| 27 | |
| 28 | - empty node: |
| 29 | |
| 30 | Parent pointer is NULL. |
| 31 | |
| 32 | |
| 33 | - Type A: sequential search in value and pointer arrays |
| 34 | |
| 35 | + Add/removal just needs to update value and pointer array, single-entry |
| 36 | (non-RCU...). For RCU, we might need to update the entire node anyway. |
| 37 | - Requires sequential search through all value array for lookup fail |
| 38 | test. |
| 39 | |
| 40 | Filled at 3 entries max 64-bit |
| 41 | 8 bits indicating number of children |
| 42 | Array of 8-bit values followed by array of associated pointers. |
| 43 | 64-bit: 1 byte + 3 bytes + 4 bytes pad + 3*8 = 32 bytes |
| 44 | |
| 45 | -> up to this point on 64-bit, sequential lookup and pointer read fit in |
| 46 | a 32-byte cache line. |
| 47 | - lookup fail&success: 1 cache-line. |
| 48 | |
| 49 | Filled at 6 entries max 32-bit, 7 entries max 64-bit |
| 50 | 8 bits indicating number of children |
| 51 | Array of 8-bit values followed by array of associated pointers. |
| 52 | 32-bit: 1 byte + 6 bytes + 1 byte pad + 6*4bytes = 32 bytes |
| 53 | 64-bit: 1 byte + 7 bytes + 7*8 = 64 bytes |
| 54 | |
| 55 | -> up to this point on 32-bit, sequential lookup and pointer read fit in |
| 56 | a 32-byte cache line. |
| 57 | - lookup fail&success: 1 cache-line. |
| 58 | |
| 59 | Filled at 12 entries max 32-bit, 14 entries max 64-bit |
| 60 | 8 bits indicating number of children |
| 61 | Array of 8-bit values followed by array of associated pointers. |
| 62 | 32-bit: 1 byte + 12 bytes + 3 bytes pad + 12*4bytes = 64 bytes |
| 63 | 64-bit: 1 byte + 14 bytes + 1 byte pad + 14*8 = 128 bytes |
| 64 | |
| 65 | Filled at 25 entries max 32-bit, 28 entries max 64-bit |
| 66 | 8 bits indicating number of children |
| 67 | Array of 8-bit values followed by array of associated pointers. |
| 68 | 32-bit: 1 byte + 25 bytes + 2 bytes pad + 25*4bytes = 128 bytes |
| 69 | 64-bit: 1 byte + 28 bytes + 3 bytes pad + 28*8 = 256 bytes |
| 70 | |
| 71 | ---> up to this point, on both 32-bit and 64-bit, the sequential lookup |
| 72 | in values array fits in a 32-byte cache line. |
| 73 | - lookup failure: 1 cache line. |
| 74 | - lookup success: 2 cache lines. |
| 75 | |
| 76 | The two below are listed for completeness sake, but because they require |
| 77 | 2 32-byte cache lines for lookup, these are deemed inappropriate. |
| 78 | |
| 79 | Filled at 51 entries max 32-bit, 56 entries max 64-bit |
| 80 | 8 bits indicating number of children |
| 81 | Array of 8-bit values followed by array of associated pointers. |
| 82 | 32-bit: 1 byte + 51 bytes + 51*4bytes = 256 bytes |
| 83 | 64-bit: 1 byte + 56 bytes + 7 bytes pad + 56*8 = 512 bytes |
| 84 | |
| 85 | Filled at 102 entries max 32-bit, 113 entries max 64-bit |
| 86 | 8 bits indicating number of children |
| 87 | Array of 8-bit values followed by array of associated pointers. |
| 88 | 32-bit: 1 byte + 102 bytes + 1 byte pad + 102*4bytes = 512 bytes |
| 89 | 64-bit: 1 byte + 113 bytes + 6 bytes pad + 113*8 = 1024 bytes |
| 90 | |
| 91 | |
| 92 | - Type B: pools of values and pointers arrays |
| 93 | |
| 94 | Pools of values and pointers arrays. Each pool values array is 32-bytes |
| 95 | in size (so it fits in a L1 cacheline). Each pool begins with an 8-bit |
| 96 | integer, which is the number of children in this pool, followed by an |
| 97 | array of 8-bit values, padding, and an array of pointers. Values and |
| 98 | pointer arrays are associated as in Type A. |
| 99 | |
| 100 | The entries of a node are associated to their respective pool based |
| 101 | on their index position. |
| 102 | |
| 103 | + Allows lookup failure to use 1 32-byte cache-line only. (1 cacheline) |
| 104 | lookup success: 2 cache lines. |
| 105 | |
| 106 | + Allows in-place updates without reallocation, except when a pool is |
| 107 | full. (this was not possible with bitmap-based nodes) |
| 108 | - If one pool exhausts its space, we need to increase the node size. |
| 109 | Therefore, for very dense populations, we will end up using the |
| 110 | pigeon-hole node type sooner, thus consuming more space. |
| 111 | |
| 112 | Pool configuration: |
| 113 | |
| 114 | Per pool, filled at 25 entries (32-bit), 28 entries (64-bit) |
| 115 | 32-bit: 1 byte + 25 bytes + 2 bytes pad + 25*4bytes = 128 bytes |
| 116 | 64-bit: 1 byte + 28 bytes + 3 bytes pad + 28*8 = 256 bytes |
| 117 | |
| 118 | Total up to 50 entries (32-bit), 56 entries (64-bit) |
| 119 | 2 pools: 32-bit = 256 bytes |
| 120 | 2 pools: 64-bit = 512 bytes |
| 121 | |
| 122 | Total up to 100 entries (32-bit), 112 entries (64-bit) |
| 123 | 4 pools: 32-bit = 512 bytes |
| 124 | 4 pools: 32-bit = 1024 bytes |
| 125 | |
| 126 | |
| 127 | * Choice of pool configuration distribution: |
| 128 | |
| 129 | We have pools of either 2 or 4 linear arrays. Their total size is |
| 130 | between 256 bytes (32-bit 2 arrays) and 1024 bytes (64-bit 4 arrays). |
| 131 | |
| 132 | Alignment on 256 bytes means that we can spare the 8 least significant |
| 133 | bits of the pointers. Given that the type selection already uses 3 bits, |
| 134 | we have 7 bits left. |
| 135 | |
| 136 | Alignment on 512 bytes -> 8 bits left. |
| 137 | |
| 138 | We can therefore encode which bit, or which two bits, are used as |
| 139 | distribution selection. We can use this technique to reequilibrate pools |
| 140 | if they become unbalanced (e.g. all children are within one of the two |
| 141 | linear arrays). |
| 142 | |
| 143 | Assuming that finding the exact sub-pool usage maximum for any given |
| 144 | distribution is NP complete (not proven). |
| 145 | |
| 146 | Taking into account unbalance ratios (tested programmatically by |
| 147 | randomly taking N entries from 256, calculating the distribution for |
| 148 | each bit (number of nodes for which bit is one/zero), and calculating |
| 149 | the difference in number of nodes for each bit, choosing the minimum |
| 150 | difference -- for millions of runs). |
| 151 | |
| 152 | tot entries unbalance largest linear array (stat. approx.) |
| 153 | --------------------------------------------------------------------- |
| 154 | 48 entries: 2 (98%) 24+1=25 (target ~50/2=25) |
| 155 | 54 entries: 2 (97%) 27+1=28 (target ~56/2=28) |
| 156 | |
| 157 | Note: there exists rare worse cases where the unbalance is larger, but |
| 158 | it happens _very_ rarely. But need to provide a fallback if the subclass |
| 159 | does not fit, but it does not need to be efficient. |
| 160 | |
| 161 | |
| 162 | For pool of size 4, we need to approximate what is the maximum unbalance |
| 163 | we can get for choice of distributions grouped by pairs of bits. |
| 164 | |
| 165 | tot entries unbalance largest linear array (stat. approx.) |
| 166 | --------------------------------------------------------------------- |
| 167 | 92 entries: 8 (99%) 23+2=25 (target: ~100/4=25) |
| 168 | 104 entries: 8 (99%) 26+2=28 (target: ~112/4=28) |
| 169 | |
| 170 | |
| 171 | Note: there exists rare worse cases where the unbalance is larger, but |
| 172 | it happens _very_ rarely. But need to provide a fallback if the subclass |
| 173 | does not fit, but it does not need to be efficient. |
| 174 | |
| 175 | |
| 176 | * Population "does not fit" and distribution fallback |
| 177 | |
| 178 | When adding a child to a distribution node, if the child does not fit, |
| 179 | we recalculate the best distribution. If it does not fit in that |
| 180 | distribution neither, we need to expand the node type. |
| 181 | |
| 182 | When removing a child, if the node child count is brought to the number |
| 183 | of entries expected to statistically fit in the lower order node, we try |
| 184 | to shrink. However, if we notice that the distribution does not actually |
| 185 | fit in that shrinked node, we abort the shrink operation. If shrink |
| 186 | fails, we keep a counter of insertion/removal operations on the node |
| 187 | before we allow the shrink to be attempted again. |
| 188 | |
| 189 | |
| 190 | - Type C: pigeon-hole array |
| 191 | |
| 192 | Filled at 47.2%/48.8% or more (32-bit: 121 entries+, 64-bit: 125 entries+) |
| 193 | Array of children node pointers. Pointers NULL if no child at index. |
| 194 | 32-bit: 4*256 = 1024 bytes |
| 195 | 64-bit: 8*256 = 2048 bytes |
| 196 | |
| 197 | |
| 198 | * Analysis of the thresholds: |
| 199 | |
| 200 | Analysis of number of cache-lines touched for each node, per-node-type, |
| 201 | depending on the number of children per node, as we increment the number |
| 202 | of children from 0 to 256. Through this, we choose number of children |
| 203 | thresholds at which it is worthwhile to use a different node type. |
| 204 | |
| 205 | Per node: |
| 206 | |
| 207 | - ALWAYS 1 cache line hit for lookup failure (all cases) |
| 208 | |
| 209 | 32-bit |
| 210 | |
| 211 | - Unexisting |
| 212 | |
| 213 | 0 children |
| 214 | |
| 215 | - Type A: sequential search in value and pointer arrays |
| 216 | - 1 cache line hit for lookup success |
| 217 | - 32 bytes storage |
| 218 | |
| 219 | up to 6 children |
| 220 | |
| 221 | - 2 cache line hit for lookup success |
| 222 | - 64 bytes storage |
| 223 | |
| 224 | up to 12 children |
| 225 | |
| 226 | - 128 bytes storage |
| 227 | |
| 228 | up to 25 children |
| 229 | |
| 230 | - Type B: pool |
| 231 | |
| 232 | - 256 bytes storage |
| 233 | |
| 234 | up to 50 children |
| 235 | |
| 236 | - 512 bytes storage |
| 237 | up to 100 children |
| 238 | |
| 239 | - Type C: pigeon-hole array |
| 240 | - 1 cache line hit for lookup success |
| 241 | - 1024 bytes storage |
| 242 | |
| 243 | up to 256 children |
| 244 | |
| 245 | |
| 246 | 64-bit |
| 247 | |
| 248 | - Unexisting |
| 249 | |
| 250 | 0 children |
| 251 | |
| 252 | - Type A: sequential search in value and pointer arrays |
| 253 | - 1 cache line hit for lookup success |
| 254 | - 32 bytes storage |
| 255 | |
| 256 | up to 3 children |
| 257 | |
| 258 | - 2 cache line hit for lookup success |
| 259 | - 64 bytes storage |
| 260 | |
| 261 | up to 7 children |
| 262 | |
| 263 | - 128 bytes storage |
| 264 | |
| 265 | up to 14 children |
| 266 | |
| 267 | - 256 bytes storage |
| 268 | |
| 269 | up to 28 children |
| 270 | |
| 271 | - Type B: pool |
| 272 | |
| 273 | - 512 bytes storage |
| 274 | up to 56 children |
| 275 | |
| 276 | - 1024 bytes storage |
| 277 | up to 112 children |
| 278 | |
| 279 | - Type C: pigeon-hole array |
| 280 | - 1 cache line hit for lookup success |
| 281 | - 2048 bytes storage |
| 282 | |
| 283 | up to 256 children |
| 284 | |
| 285 | |
| 286 | * Analysis of node type encoding and node pointers: |
| 287 | |
| 288 | Lookups are _always_ from the top of the tree going down. This |
| 289 | facilitates RCU replacement as we only keep track of pointers going |
| 290 | downward. |
| 291 | |
| 292 | Type of node encoded in the parent's pointer. Need to reserve 2 |
| 293 | least-significant bits. |
| 294 | |
| 295 | Types of children: |
| 296 | |
| 297 | enum child_type { |
| 298 | RCU_JA_LINEAR = 0, /* Type A */ |
| 299 | /* 32-bit: 1 to 25 children, 8 to 128 bytes */ |
| 300 | /* 64-bit: 1 to 28 children, 16 to 256 bytes */ |
| 301 | RCU_JA_POOL = 1, /* Type B */ |
| 302 | /* 32-bit: 26 to 100 children, 256 to 512 bytes */ |
| 303 | /* 64-bit: 29 to 112 children, 512 to 1024 bytes */ |
| 304 | RCU_JA_PIGEON = 2, /* Type C */ |
| 305 | /* 32-bit: 101 to 256 children, 1024 bytes */ |
| 306 | /* 64-bit: 113 to 256 children, 2048 bytes */ |
| 307 | /* Leaf nodes are implicit from their height in the tree */ |
| 308 | }; |
| 309 | |
| 310 | If entire pointer is NULL, children is empty. |
| 311 | |
| 312 | |
| 313 | * Lookup and Update Algorithms |
| 314 | |
| 315 | Let's propose a quite simple scheme that uses a mutex on nodes to manage |
| 316 | update concurrency. It's certainly not optimal in terms of concurrency |
| 317 | management within a node, but it has the advantage of being simple to |
| 318 | implement and understand. |
| 319 | |
| 320 | We need to keep a count of the number of children nodes (for each node), |
| 321 | to keep track of when the node type thresholds are reached. It would be |
| 322 | important to put an hysteresis loop so we don't change between node |
| 323 | types too often for a loop on add/removal of the same node. |
| 324 | |
| 325 | We acquire locks from child to parent, nested. We take all locks |
| 326 | required to perform a given update in the tree (but no more) to keep it |
| 327 | consistent with respect to number of children per node. |
| 328 | |
| 329 | If check for node being gc'd (always under node lock) fails, we simply |
| 330 | need to release the lock and lookup the node again. |
| 331 | |
| 332 | |
| 333 | - Leaf lookup |
| 334 | |
| 335 | rcu_read_lock() |
| 336 | |
| 337 | RCU-lookup each level of the tree. If level is not populated, fail. |
| 338 | Until we reach the leaf node. |
| 339 | |
| 340 | rcu_read_unlock() |
| 341 | |
| 342 | |
| 343 | - Leaf insertion |
| 344 | |
| 345 | A) Lookup |
| 346 | |
| 347 | rcu_read_lock() |
| 348 | RCU-lookup insert position. Find location in tree where nodes are |
| 349 | missing for this insertion. If leaf is already present, insert fails, |
| 350 | releasing the rcu read lock. The insert location consists of a parent |
| 351 | node to which we want to attach a new node. |
| 352 | |
| 353 | B) Lock |
| 354 | |
| 355 | RCU-lookup parent node. Take the parent lock. If the parent needs to be |
| 356 | reallocated to make room for this insertion, RCU-lookup parent-parent |
| 357 | node and take the parent-parent lock. For each lock taken, check if |
| 358 | node is being gc'd. If gc'd, release lock, re-RCU-lookup this node, and |
| 359 | retry. |
| 360 | |
| 361 | C) Create |
| 362 | |
| 363 | Construct the whole branch from the new topmost intermediate node down |
| 364 | to the new leaf node we are inserting. |
| 365 | |
| 366 | D) Populate: |
| 367 | - If parent node reallocation is needed: |
| 368 | Reallocate the parent node, adding the new branch to it, and |
| 369 | increment its node count. |
| 370 | set gc flag in old nodes. |
| 371 | call_rcu free for all old nodes. |
| 372 | Populate new parent node with rcu_assign_pointer. |
| 373 | - Else: |
| 374 | Increment parent node count. |
| 375 | Use rcu_assign_pointer to populate this new branch into the parent |
| 376 | node. |
| 377 | |
| 378 | E) Locks |
| 379 | |
| 380 | Release parent and (if taken) parent-parent locks. |
| 381 | rcu_read_unlock() |
| 382 | |
| 383 | |
| 384 | - Leaf removal |
| 385 | |
| 386 | A) Lookup |
| 387 | |
| 388 | rcu_read_lock() |
| 389 | RCU-lookup leaf to remove. If leaf is missing, fail and release rcu |
| 390 | read lock. |
| 391 | |
| 392 | B) Lock |
| 393 | |
| 394 | RCU-lookup parent. Take the parent lock. If the parent needs to be |
| 395 | reallocated because it would be too large for the decremented number of |
| 396 | children, RCU-lookup parent-parent and take the parent-parent lock. Do |
| 397 | so recursively until no node reallocation is needed, or until root is |
| 398 | reached. |
| 399 | |
| 400 | For each lock taken, check if node is being gc'd. If gc'd, release lock, |
| 401 | re-RCU-lookup this node, and retry. |
| 402 | |
| 403 | C) Create |
| 404 | |
| 405 | The branch (or portion of branch) consisting of taken locks necessarily |
| 406 | has a simple node removal or update as operation to do on its top node. |
| 407 | |
| 408 | If the operation is a node removal, then, necessarily, the entire branch |
| 409 | under the node removal operation will simply disappear. No node |
| 410 | allocation is needed. |
| 411 | |
| 412 | Else, if the operation is a child node reallocation, the child node will |
| 413 | necessarily do a node removal. So _its_ entire child branch will |
| 414 | disappear. So reallocate this child node without the removed branch |
| 415 | (remember to decrement its nr children count). |
| 416 | |
| 417 | D) Populate |
| 418 | |
| 419 | No reallocation case: simply set the appropriate child pointer in the |
| 420 | topmost locked node to NULL. Decrement its nr children count. |
| 421 | |
| 422 | Reallocation case: set the child pointer in the topmost locked node to |
| 423 | the newly allocated node. |
| 424 | set old nodes gc flag. |
| 425 | call_rcu free for all old nodes. |
| 426 | |
| 427 | E) Locks |
| 428 | |
| 429 | Release all locks. |
| 430 | rcu_read_unlock() |
| 431 | |
| 432 | |
| 433 | For the various types of nodes: |
| 434 | |
| 435 | - sequential search (type A) |
| 436 | - RCU replacement: mutex |
| 437 | - Entry update: mutex |
| 438 | |
| 439 | - bitmap followed by pointer array (type B) |
| 440 | - RCU replacement: mutex |
| 441 | - Entry update: mutex |
| 442 | |
| 443 | - pigeon hole array (type C) |
| 444 | - RCU replacement: mutex |
| 445 | - Entry update: mutex |