| 1 | // Licensed to the .NET Foundation under one or more agreements. |
| 2 | // The .NET Foundation licenses this file to you under the MIT license. |
| 3 | // See the LICENSE file in the project root for more information. |
| 4 | // |
| 5 | |
| 6 | // |
| 7 | // NgenHash is an abstract base class (actually a templated base class) designed to factor out the |
| 8 | // functionality common to hashes persisted into ngen images. |
| 9 | // |
| 10 | // SEMANTICS |
| 11 | // |
| 12 | // * Arbitrary entry payload via payload. |
| 13 | // * 32-bit hash code for entries. |
| 14 | // * Separate entry allocation and insertion (allowing reliable insertion if required). |
| 15 | // * Enumerate all entries or entries matching a particular hash. |
| 16 | // * No entry deletion. |
| 17 | // * Base logic to efficiently serialize hash contents at ngen time. Hot/cold splitting of entries is |
| 18 | // supported (along with the ability to tweak the Save and Fixup stages of each entry if needed). |
| 19 | // * Base logic to support DAC memory enumeration of the hash (including per-entry tweaks as needed). |
| 20 | // * Lock free lookup (the caller must follow the protocol laid out below under USER REQUIREMENTS). |
| 21 | // * Automatic hash expansion (with dialable scale factor). |
| 22 | // * Hash insertion is supported at runtime even when an ngen image is loaded with previously serialized hash |
| 23 | // entries. |
| 24 | // * Base logic to support formatting hashes in the nidump tool (only need to supply code for the unique |
| 25 | // aspects of your hash). |
| 26 | // |
| 27 | // BENEFITS |
| 28 | // |
| 29 | // * Removes next pointer from all persisted hash entries: |
| 30 | // o Reduces data footprint of each entry. |
| 31 | // o Increases density of entries. |
| 32 | // o Removes a base relocation entry. |
| 33 | // o Removes a runtime write to each entry (from the relocation above). |
| 34 | // * Serializes all hot/cold hash entries contigiuously: |
| 35 | // o Helps keeps hash entries in the same bucket in the same cache line. |
| 36 | // * Compresses persisted bucket list and removes the use of pointers: |
| 37 | // o Reduces working set hit of reading hash table (especially on 64-bit systems). |
| 38 | // o Allows bucket list to be saved in read-only memory and thus use shared rather than private pages. |
| 39 | // * Factors out common code: |
| 40 | // o Less chance of bugs, one place to make fixes. |
| 41 | // o Less code overall. |
| 42 | // |
| 43 | // SUB-CLASSING REQUIREMENTS |
| 44 | // |
| 45 | // To author a new NgenHash-based hashtable, the following steps are required: |
| 46 | // 1) In most cases (where each hash entry will have multiple fields) a structure defining the hash entry |
| 47 | // should be declared (see EEClassHashEntry in ClassHash.h for an example). This structure need not |
| 48 | // include a field for the hash code or pointer to the next entry in the hash bucket; these are taken care |
| 49 | // of automatically by the base class. If the entry must reference another entry in the hash (this should |
| 50 | // be rare) the NgenHashEntryRef<> template class should be used to abstract the reference (this class |
| 51 | // hides some of the transformation work that must take place when entries are re-ordered during ngen |
| 52 | // serialization). |
| 53 | // 2) Declare your new hash class deriving from NgenHash and providing the following template parameters: |
| 54 | // FINAL_CLASS : The class you're declaring (this is used by the base class to locate certain helper |
| 55 | // methods in your class used to tweak hash behavior). |
| 56 | // VALUE : The type of your hash entries (the class defined in the previous step). |
| 57 | // SCALE_FACTOR : A multipler on bucket count every time the hash table is grown (currently once the |
| 58 | // number of hash entries exceeds twice the number of buckets). A value of 2 would double |
| 59 | // the number of buckets on each grow operation for example. |
| 60 | // 3) Define a constructor that invokes the base class constructor with various setup parameters (see |
| 61 | // NgenHash constructor in this header). If your hash table is created via a static method rather than |
| 62 | // direct construction (common) then call your constructor using an in-place new inside the static method |
| 63 | // (see EEClassHashTable::Create in ClassHash.cpp for an example). |
| 64 | // 4) Define your basic hash functionality (creation, insertion, lookup, enumeration, ngen Save/Fixup and DAC |
| 65 | // memory enumeration) using the Base* methods provided by NgenHash. |
| 66 | // 5) Tweak the operation of BaseSave, BaseFixup and BaseEnumMemoryRegions by providing definitions of the |
| 67 | // following methods (note that all methods must be defined though they may be no-ops): |
| 68 | // |
| 69 | // bool ShouldSave(DataImage *pImage, VALUE *pEntry); |
| 70 | // Return true if the given entry should be persisted into the ngen image (otherwise it won't be |
| 71 | // saved with the rest). |
| 72 | // |
| 73 | // bool IsHotEntry(VALUE *pEntry, CorProfileData *pProfileData); |
| 74 | // Return true is the entry is considered hot given the profiling data. |
| 75 | // |
| 76 | // bool SaveEntry(DataImage *pImage, CorProfileData *pProfileData, VALUE *pOldEntry, VALUE *pNewEntry, EntryMappingTable *pMap); |
| 77 | // Gives your hash class a chance to save any additional data needed into the ngen image during |
| 78 | // the Save phase or otherwise make entry updates prior to saving. The saving process creates a |
| 79 | // new copy of each hash entry and this method is passed pointers both to the original entry and |
| 80 | // the new version along with a mapping class that can translate any old entry address in the |
| 81 | // table into the corresponding new address. If you have inter-entry pointer fields this is your |
| 82 | // chance to fix up those fields with the new location of their target entries. |
| 83 | // |
| 84 | // void FixupEntry(DataImage *pImage, VALUE *pEntry, void *pFixupBase, DWORD cbFixupOffset); |
| 85 | // Similar to SaveEntry but called during BaseFixup. This is your chance to register fixups for |
| 86 | // any pointer type fields in your entry. Due to the way hash entries are packed during ngen |
| 87 | // serialization individual hash entries are not saved as separate ngen zap nodes. So this method |
| 88 | // is passed a pointer to the enclosing zapped data structure (pFixupBase) and the offset of the |
| 89 | // entry from this base (cbFixupOffset). When calling pImage->FixupPointerField(...) for |
| 90 | // instance, pass pFixupBase as the first parameter and cbFixupOffset + offsetof(YourEntryClass, |
| 91 | // yourField) as the second parameter. |
| 92 | // |
| 93 | // void EnumMemoryRegionsForEntry(EEClassHashEntry_t *pEntry, CLRDataEnumMemoryFlags flags); |
| 94 | // Called during BaseEnumMemoryRegions for each entry in the hash. Use to enumerate any memory |
| 95 | // referenced by the entry (but not the entry itself). |
| 96 | // |
| 97 | // USER REQUIREMENTS |
| 98 | // |
| 99 | // Synchronization: It is permissable to read data from the hash without taking a lock as long as: |
| 100 | // 1) Any hash modifications are performed under a lock or otherwise serialized. |
| 101 | // 2) Any miss on a lookup is handled by taking a lock are retry-ing the lookup. |
| 102 | // |
| 103 | // OVERALL DESIGN |
| 104 | // |
| 105 | // The hash contains up to three groups of hash entries. These consist of two groups of entries persisted to |
| 106 | // disk at ngen time (split into hot and cold based on profile data) and live entries added at runtime (or |
| 107 | // during the ngen process itself, prior to the save operation). |
| 108 | // |
| 109 | // The persisted entries are tightly packed together and can eliminate some pointers and other metadata since |
| 110 | // we statically know about every entry at the time we format the hash entries (the save phase of ngen |
| 111 | // generation). |
| 112 | // |
| 113 | // Each persisted entry is assigned to a bucket based on its hash code and all entries that collide on a given |
| 114 | // bucket are placed contiguously in memory. The bucket list itself therefore consists of an array or pairs, |
| 115 | // each pair containing the count of entries in the bucket and the location of the first entry in the chain. |
| 116 | // Since all entries are allocated contiguously entry location can be specified by an index into the array of |
| 117 | // entries. |
| 118 | // |
| 119 | // Separate bucket lists and entry arrays are stored for hot and cold entries. |
| 120 | // |
| 121 | // The live entries (referred to here as volatile or warm entries) follow a more traditional hash |
| 122 | // implementation where entries are allocated individually from a loader heap and are chained together with a |
| 123 | // singly linked list if they collide. Here the bucket list is a simple array of pointers to the first entry |
| 124 | // in each chain (if any). |
| 125 | // |
| 126 | // Unlike the persisted entris the warm section of the table must cope with entry insertions and growing the |
| 127 | // bucket list when the table becomes too loaded (too many entries causing excessive bucket collisions). This |
| 128 | // happens when an entry insertion notes that there are twice as many entries as buckets. The bucket list is |
| 129 | // then reallocated (from a loader heap, consequently the old one is leaked) and resized based on a scale |
| 130 | // factor supplied by the hash sub-class. |
| 131 | // |
| 132 | // At runtime we lookup or enumerate entries by visiting all three sets of entries in the order Hot, Warm and |
| 133 | // Cold. This imposes a slight but constant time overhead. |
| 134 | // |
| 135 | |
| 136 | #ifndef __NGEN_HASH_INCLUDED |
| 137 | #define __NGEN_HASH_INCLUDED |
| 138 | |
| 139 | #ifdef FEATURE_PREJIT |
| 140 | #include "corcompile.h" |
| 141 | #endif |
| 142 | |
| 143 | // The type used to contain an entry hash value. This is not customizable on a per-hash class basis: all |
| 144 | // NgenHash derived hashes will share the same definition. Note that we only care about the data size, and the |
| 145 | // fact that it is an unsigned integer value (so we can take a modulus for bucket computation and use bitwise |
| 146 | // equality checks). The base class does not care about or participate in how these hash values are calculated. |
| 147 | typedef DWORD NgenHashValue; |
| 148 | |
| 149 | // The following code (and code in NgenHash.inl) has to replicate the base class template parameters (and in |
| 150 | // some cases the arguments) many many times. In the interests of brevity (and to make it a whole lot easier |
| 151 | // to modify these parameters in the future) we define macro shorthands for them here. Scan through the code |
| 152 | // to see how these are used. |
| 153 | #define NGEN_HASH_PARAMS typename FINAL_CLASS, typename VALUE, int SCALE_FACTOR |
| 154 | #define NGEN_HASH_ARGS FINAL_CLASS, VALUE, SCALE_FACTOR |
| 155 | |
| 156 | // Forward definition of NgenHashEntryRef (it takes the same template parameters as NgenHash and simplifies |
| 157 | // hash entries that need to refer to other hash entries). |
| 158 | template <NGEN_HASH_PARAMS> |
| 159 | class NgenHashEntryRef; |
| 160 | |
| 161 | // The base hash class itself. It's abstract and exposes its functionality via protected members (nothing is |
| 162 | // public). |
| 163 | template <NGEN_HASH_PARAMS> |
| 164 | class NgenHashTable |
| 165 | { |
| 166 | // NgenHashEntryRef needs access to the base table internal during Fixup in order to compute zap node |
| 167 | // bases. |
| 168 | friend class NgenHashEntryRef<NGEN_HASH_ARGS>; |
| 169 | |
| 170 | #ifdef DACCESS_COMPILE |
| 171 | // Nidump knows how to walk this data structure. |
| 172 | friend class NativeImageDumper; |
| 173 | #endif |
| 174 | |
| 175 | protected: |
| 176 | // This opaque structure provides enumeration context when walking the set of entries which share a common |
| 177 | // hash code. Initialized by BaseFindFirstEntryByHash and read/updated by BaseFindNextEntryByHash. |
| 178 | class LookupContext |
| 179 | { |
| 180 | friend class NgenHashTable<NGEN_HASH_ARGS>; |
| 181 | |
| 182 | TADDR m_pEntry; // The entry the caller is currently looking at (or NULL to begin |
| 183 | // with). This is a VolatileEntry* or PersistedEntry* (depending on |
| 184 | // m_eType below) and should always be a target address not a DAC |
| 185 | // PTR_. |
| 186 | DWORD m_eType; // The entry types we're currently walking (Hot, Warm, Cold in that order) |
| 187 | DWORD m_cRemainingEntries; // The remaining entries in the bucket chain (Hot or Cold entries only) |
| 188 | }; |
| 189 | |
| 190 | // This opaque structure provides enumeration context when walking all entries in the table. Initialized |
| 191 | // by BaseInitIterator and updated via the BaseIterator::Next. Note that this structure is somewhat |
| 192 | // similar to LookupContext above (though it requires a bit more state). It's possible we could factor |
| 193 | // these two iterators into some common base code but the actual implementations have enough differing |
| 194 | // requirements that the resultant code could be less readable (and slightly less performant). |
| 195 | class BaseIterator |
| 196 | { |
| 197 | public: |
| 198 | // Returns a pointer to the next entry in the hash table or NULL once all entries have been |
| 199 | // enumerated. Once NULL has been return the only legal operation is to re-initialize the iterator |
| 200 | // with BaseInitIterator. |
| 201 | DPTR(VALUE) Next(); |
| 202 | |
| 203 | private: |
| 204 | friend class NgenHashTable<NGEN_HASH_ARGS>; |
| 205 | |
| 206 | DPTR(NgenHashTable<NGEN_HASH_ARGS>) m_pTable; // Pointer back to the table being enumerated. |
| 207 | TADDR m_pEntry; // The entry the caller is currently looking at (or |
| 208 | // NULL to begin with). This is a VolatileEntry* or |
| 209 | // PersistedEntry* (depending on m_eType below) and |
| 210 | // should always be a target address not a DAC PTR_. |
| 211 | DWORD m_eType; // The entry types we're currently walking (Hot, Warm, |
| 212 | // Cold in that order). |
| 213 | union |
| 214 | { |
| 215 | DWORD m_dwBucket; // Index of bucket we're currently walking (Warm). |
| 216 | DWORD m_cRemainingEntries; // Number of entries remaining in hot/cold section |
| 217 | // (Hot, Cold). |
| 218 | }; |
| 219 | }; |
| 220 | |
| 221 | #ifndef DACCESS_COMPILE |
| 222 | // Base constructor. Call this from your derived constructor to provide the owning module, loader heap and |
| 223 | // initial number of buckets (which must be non-zero). Module must be provided if this hash is to be |
| 224 | // serialized into an ngen image. It is exposed to the derived hash class (many need it) but otherwise is |
| 225 | // only used to locate a loader heap for allocating bucket lists and entries unless an alternative heap is |
| 226 | // provided. Note that the heap provided is not serialized (so you'll allocate from that heap at |
| 227 | // ngen-time, but revert to allocating from the module's heap at runtime). If no Module pointer is |
| 228 | // supplied (non-ngen'd hash table) you must provide a direct heap pointer. |
| 229 | NgenHashTable(Module *pModule, LoaderHeap *pHeap, DWORD cInitialBuckets); |
| 230 | |
| 231 | // Allocate an uninitialized entry for the hash table (it's not inserted). The AllocMemTracker is optional |
| 232 | // and may be specified as NULL for untracked allocations. This is split from the hash insertion logic so |
| 233 | // that callers can pre-allocate entries and then perform insertions which cannot fault. |
| 234 | VALUE *BaseAllocateEntry(AllocMemTracker *pamTracker); |
| 235 | |
| 236 | // Insert an entry previously allocated via BaseAllocateEntry (you cannot allocated entries in any other |
| 237 | // manner) and associated with the given hash value. The entry should have been initialized prior to |
| 238 | // insertion. |
| 239 | void BaseInsertEntry(NgenHashValue iHash, VALUE *pEntry); |
| 240 | #endif // !DACCESS_COMPILE |
| 241 | |
| 242 | // Return the number of entries held in the table (does not include entries allocated but not inserted |
| 243 | // yet). |
| 244 | DWORD BaseGetElementCount(); |
| 245 | |
| 246 | // Initializes the iterator context passed by the caller to make it ready to walk every entry in the table |
| 247 | // in an arbitrary order. Call pIterator->Next() to retrieve the first entry. |
| 248 | void BaseInitIterator(BaseIterator *pIterator); |
| 249 | |
| 250 | // Find first entry matching a given hash value (returns NULL on no match). Call BaseFindNextEntryByHash |
| 251 | // to iterate the remaining matches (until it returns NULL). The LookupContext supplied by the caller is |
| 252 | // initialized by BaseFindFirstEntryByHash and read/updated by BaseFindNextEntryByHash to keep track of |
| 253 | // where we are. |
| 254 | DPTR(VALUE) BaseFindFirstEntryByHash(NgenHashValue iHash, LookupContext *pContext); |
| 255 | DPTR(VALUE) BaseFindNextEntryByHash(LookupContext *pContext); |
| 256 | |
| 257 | #if defined(FEATURE_PREJIT) && !defined(DACCESS_COMPILE) |
| 258 | // Call during ngen to save hash table data structures into the ngen image. Calls derived-class |
| 259 | // implementations of ShouldSave to determine which entries should be serialized, IsHotEntry to hot/cold |
| 260 | // split the entries and SaveEntry to allow per-entry extension of the saving process. |
| 261 | void BaseSave(DataImage *pImage, CorProfileData *pProfileData); |
| 262 | |
| 263 | // Call during ngen to register fixups for hash table data structure fields. Calls derived-class |
| 264 | // implementation of FixupEntry to allow per-entry extension of the fixup process. |
| 265 | void BaseFixup(DataImage *pImage); |
| 266 | |
| 267 | // Opaque structure used to store state will BaseSave is re-arranging hash entries and also passed to |
| 268 | // sub-classes' SaveEntry method to facilitate mapping old entry addresses to their new locations. |
| 269 | class EntryMappingTable |
| 270 | { |
| 271 | public: |
| 272 | ~EntryMappingTable(); |
| 273 | |
| 274 | // Given an old entry address (pre-BaseSave) return the address of the entry relocated ready for |
| 275 | // saving to disk. Note that this address is the (ngen) runtime address, not the disk image address |
| 276 | // you can further obtain by calling DataImage::GetImagePointer(). |
| 277 | VALUE *GetNewEntryAddress(VALUE *pOldEntry); |
| 278 | |
| 279 | private: |
| 280 | friend class NgenHashTable<NGEN_HASH_ARGS>; |
| 281 | |
| 282 | // Each Entry holds the mapping from one old-to-new hash entry. |
| 283 | struct Entry |
| 284 | { |
| 285 | VALUE *m_pOldEntry; // Pointer to the user part of the old entry |
| 286 | VALUE *m_pNewEntry; // Pointer to the user part of the new version |
| 287 | NgenHashValue m_iHashValue; // The hash code of the entry |
| 288 | DWORD m_dwNewBucket; // The new bucket index of the entry |
| 289 | DWORD m_dwChainOrdinal; // The 0-based position within the chain of that bucket |
| 290 | bool m_fHot; // If true this entry was identified as hot by the sub-class |
| 291 | }; |
| 292 | |
| 293 | Entry *m_pEntries; // Pointer to array of Entries |
| 294 | DWORD m_cEntries; // Count of valid entries in the above (may be smaller than |
| 295 | // allocated size) |
| 296 | }; |
| 297 | #endif // FEATURE_PREJIT && !DACCESS_COMPILE |
| 298 | |
| 299 | #ifdef DACCESS_COMPILE |
| 300 | // Call during DAC enumeration of memory regions to save in mini-dump to enumerate all hash table data |
| 301 | // structures. Calls derived-class implementation of EnumMemoryRegionsForEntry to allow additional |
| 302 | // per-entry memory to be reported. |
| 303 | void BaseEnumMemoryRegions(CLRDataEnumMemoryFlags flags); |
| 304 | #endif // DACCESS_COMPILE |
| 305 | |
| 306 | PTR_Module GetModule() |
| 307 | { |
| 308 | return ReadPointerMaybeNull(this, &NgenHashTable<NGEN_HASH_ARGS>::m_pModule); |
| 309 | } |
| 310 | |
| 311 | // Owning module set at hash creation time (possibly NULL if this hash instance is not to be ngen'd). |
| 312 | RelativePointer<PTR_Module> m_pModule; |
| 313 | |
| 314 | private: |
| 315 | // Internal implementation details. Nothing of interest to sub-classers for here on. |
| 316 | |
| 317 | #ifdef FEATURE_PREJIT |
| 318 | // This is the format of each hash entry that is persisted to disk (Hot or Cold entries). |
| 319 | struct PersistedEntry |
| 320 | { |
| 321 | VALUE m_sValue; // The sub-class supplied entry layout |
| 322 | NgenHashValue m_iHashValue; // The hash code associated with the entry (comes after m_sValue to |
| 323 | // minimize chance of pad bytes on a 64-bit system). |
| 324 | }; |
| 325 | typedef DPTR(PersistedEntry) PTR_PersistedEntry; |
| 326 | typedef ArrayDPTR(PersistedEntry) APTR_PersistedEntry; |
| 327 | |
| 328 | // This class encapsulates a bucket list identifying chains of related persisted entries. It's compressed |
| 329 | // rather than being a simple array, hence the encapsulation. |
| 330 | // A bucket list represents a non-zero sequence of buckets, each bucket identified by a zero-based index. |
| 331 | // Each bucket holds the index of the entry at the start of the bucket chain and a count of entries in |
| 332 | // that chain. (In persisted form hash entries are collated into hot and cold entries which are then |
| 333 | // allocated in contiguous blocks: this allows entries to be identified by an index into the entry block). |
| 334 | // Buckets with zero entries have an undefined start index (and a zero count obviously). |
| 335 | class PersistedBucketList |
| 336 | { |
| 337 | #ifdef DACCESS_COMPILE |
| 338 | friend class NativeImageDumper; |
| 339 | #endif |
| 340 | |
| 341 | public: |
| 342 | // Allocate and initialize a new list with the given count of buckets and configured to hold no more |
| 343 | // than the given number of entries or have a bucket chain longer than the specified maximum. These |
| 344 | // two maximums allow the implementation to choose an optimal data format for the bucket list at |
| 345 | // runtime and are enforced by asserts in the debug build. |
| 346 | static PersistedBucketList *CreateList(DWORD cBuckets, DWORD cEntries, DWORD cMaxEntriesInBucket); |
| 347 | |
| 348 | // For the given bucket set the index of the initial entry and the count of entries in the chain. If |
| 349 | // the count is zero the initial entry index is meaningless and ignored. |
| 350 | void SetBucket(DWORD dwIndex, DWORD dwFirstEntry, DWORD cEntries); |
| 351 | |
| 352 | // Get the size in bytes of this entire bucket list (need to pass in the bucket count since we save |
| 353 | // space by not storing it here, but we do validate this in debug mode). |
| 354 | size_t GetSize(DWORD cBuckets); |
| 355 | |
| 356 | // Get the initial entry index and entry count for the given bucket. Initial entry index value is |
| 357 | // undefined when count comes back as zero. |
| 358 | void GetBucket(DWORD dwIndex, DWORD *pdwFirstEntry, DWORD *pdwCount); |
| 359 | |
| 360 | // Simplified initial entry index when you don't need the count (don't call this for buckets with zero |
| 361 | // entries). |
| 362 | DWORD GetInitialEntry(DWORD dwIndex); |
| 363 | |
| 364 | private: |
| 365 | // Return the number of bits required to express a unique ID for the number of entities given. |
| 366 | static DWORD BitsRequired(DWORD cEntities); |
| 367 | |
| 368 | // Return the minimum size (in bytes) of each bucket list entry that can express all buckets given the |
| 369 | // max count of entries and entries in a single bucket chain. |
| 370 | static DWORD GetBucketSize(DWORD cEntries, DWORD cMaxEntriesInBucket); |
| 371 | |
| 372 | // Each bucket is represented by a variable sized bitfield (16, 32 or 64 bits) whose low-order bits |
| 373 | // contain the index of the first entry in the chain and higher-order (just above the initial entry |
| 374 | // bits) contain the count of entries in the chain. |
| 375 | |
| 376 | DWORD m_cbBucket; // The size in bytes of each bucket descriptor (2, 4 or 8) |
| 377 | DWORD m_dwInitialEntryMask; // The bitmask used to extract the initial entry index from a bucket |
| 378 | DWORD m_dwEntryCountShift; // The bit shift used to extract the entry count from a bucket |
| 379 | |
| 380 | #ifdef _DEBUG |
| 381 | // In debug mode we remember more initial state to catch common errors. |
| 382 | DWORD m_cBuckets; // Number of buckets in the list |
| 383 | DWORD m_cEntries; // Total number of entries mapped by the list |
| 384 | DWORD m_cMaxEntriesInBucket; // Largest bucket chain in the list |
| 385 | #endif |
| 386 | }; |
| 387 | typedef DPTR(PersistedBucketList) PTR_PersistedBucketList; |
| 388 | |
| 389 | // Pointers and counters for entries and buckets persisted to disk during ngen. Collected into a structure |
| 390 | // because this logic is replicated for Hot and Cold entries so we can factor some common code. |
| 391 | struct PersistedEntries |
| 392 | { |
| 393 | RelativePointer<APTR_PersistedEntry> m_pEntries; // Pointer to a contiguous block of PersistedEntry structures |
| 394 | // (NULL if zero entries) |
| 395 | RelativePointer<PTR_PersistedBucketList> m_pBuckets; // Pointer to abstracted bucket list mapping above entries |
| 396 | // into a hash (NULL if zero buckets, which is iff zero |
| 397 | // entries) |
| 398 | DWORD m_cEntries; // Count of entries in the above block |
| 399 | DWORD m_cBuckets; // Count of buckets in the above bucket list |
| 400 | }; |
| 401 | #endif // FEATURE_PREJIT |
| 402 | |
| 403 | // This is the format of a Warm entry, defined for our purposes to be a non-persisted entry (i.e. those |
| 404 | // created at runtime or during the creation of the ngen image itself). |
| 405 | struct VolatileEntry; |
| 406 | typedef DPTR(struct VolatileEntry) PTR_VolatileEntry; |
| 407 | struct VolatileEntry |
| 408 | { |
| 409 | VALUE m_sValue; // The derived-class format of an entry |
| 410 | PTR_VolatileEntry m_pNextEntry; // Pointer to the next entry in the bucket chain (or NULL) |
| 411 | NgenHashValue m_iHashValue; // The hash value associated with the entry |
| 412 | }; |
| 413 | |
| 414 | // Types of hash entry. |
| 415 | enum EntryType |
| 416 | { |
| 417 | Cold, // Persisted, profiling suggests this data is not read typically |
| 418 | Warm, // Volatile (in-memory) |
| 419 | Hot // Persisted, profiling suggests this data is probably read (or no profiling data was available) |
| 420 | }; |
| 421 | |
| 422 | #ifdef FEATURE_PREJIT |
| 423 | // Find the first persisted entry (hot or cold based on pEntries) that matches the given hash. Looks only |
| 424 | // in the persisted block given (i.e. searches only hot *or* cold entries). Returns NULL on failure. |
| 425 | // Otherwise returns pointer to the derived class portion of the entry and initializes the provided |
| 426 | // LookupContext to allow enumeration of any further matches. |
| 427 | DPTR(VALUE) FindPersistedEntryByHash(PersistedEntries *pEntries, NgenHashValue iHash, LookupContext *pContext); |
| 428 | #endif // FEATURE_PREJIT |
| 429 | |
| 430 | // Find the first volatile (warm) entry that matches the given hash. Looks only at warm entries. Returns |
| 431 | // NULL on failure. Otherwise returns pointer to the derived class portion of the entry and initializes |
| 432 | // the provided LookupContext to allow enumeration of any further matches. |
| 433 | DPTR(VALUE) FindVolatileEntryByHash(NgenHashValue iHash, LookupContext *pContext); |
| 434 | |
| 435 | #ifndef DACCESS_COMPILE |
| 436 | // Determine loader heap to be used for allocation of entries and bucket lists. |
| 437 | LoaderHeap *GetHeap(); |
| 438 | |
| 439 | // Increase the size of the bucket list in order to reduce the size of bucket chains. Does nothing on |
| 440 | // failure to allocate (since this impacts perf, not correctness). |
| 441 | void GrowTable(); |
| 442 | |
| 443 | // Returns the next prime larger (or equal to) than the number given. |
| 444 | DWORD NextLargestPrime(DWORD dwNumber); |
| 445 | #endif // !DACCESS_COMPILE |
| 446 | |
| 447 | DPTR(PTR_VolatileEntry) GetWarmBuckets() |
| 448 | { |
| 449 | SUPPORTS_DAC; |
| 450 | |
| 451 | return ReadPointer(this, &NgenHashTable<NGEN_HASH_ARGS>::m_pWarmBuckets); |
| 452 | } |
| 453 | |
| 454 | #ifdef FEATURE_PREJIT |
| 455 | APTR_PersistedEntry GetPersistedHotEntries() |
| 456 | { |
| 457 | SUPPORTS_DAC; |
| 458 | |
| 459 | return ReadPointerMaybeNull(this, |
| 460 | &NgenHashTable<NGEN_HASH_ARGS>::m_sHotEntries, |
| 461 | &decltype(NgenHashTable<NGEN_HASH_ARGS>::m_sHotEntries)::m_pEntries); |
| 462 | } |
| 463 | |
| 464 | PTR_PersistedBucketList GetPersistedHotBuckets() |
| 465 | { |
| 466 | SUPPORTS_DAC; |
| 467 | |
| 468 | return ReadPointerMaybeNull(this, |
| 469 | &NgenHashTable<NGEN_HASH_ARGS>::m_sHotEntries, |
| 470 | &decltype(NgenHashTable<NGEN_HASH_ARGS>::m_sHotEntries)::m_pBuckets); |
| 471 | } |
| 472 | |
| 473 | APTR_PersistedEntry GetPersistedColdEntries() |
| 474 | { |
| 475 | SUPPORTS_DAC; |
| 476 | |
| 477 | return ReadPointerMaybeNull(this, |
| 478 | &NgenHashTable<NGEN_HASH_ARGS>::m_sColdEntries, |
| 479 | &decltype(NgenHashTable<NGEN_HASH_ARGS>::m_sColdEntries)::m_pEntries); |
| 480 | } |
| 481 | |
| 482 | PTR_PersistedBucketList GetPersistedColdBuckets() |
| 483 | { |
| 484 | SUPPORTS_DAC; |
| 485 | |
| 486 | return ReadPointerMaybeNull(this, |
| 487 | &NgenHashTable<NGEN_HASH_ARGS>::m_sColdEntries, |
| 488 | &decltype(NgenHashTable<NGEN_HASH_ARGS>::m_sColdEntries)::m_pBuckets); |
| 489 | } |
| 490 | |
| 491 | #ifdef DACCESS_COMPILE |
| 492 | APTR_PersistedEntry GetPersistedEntries(DPTR(PersistedEntries) pEntries) |
| 493 | { |
| 494 | SUPPORTS_DAC; |
| 495 | |
| 496 | TADDR hotEntriesAddr = dac_cast<TADDR>(this) + offsetof(NgenHashTable<NGEN_HASH_ARGS>, m_sHotEntries); |
| 497 | TADDR coldEntriesAddr = dac_cast<TADDR>(this) + offsetof(NgenHashTable<NGEN_HASH_ARGS>, m_sColdEntries); |
| 498 | |
| 499 | if (hotEntriesAddr == dac_cast<TADDR>(pEntries)) |
| 500 | { |
| 501 | return GetPersistedHotEntries(); |
| 502 | } |
| 503 | else |
| 504 | { |
| 505 | _ASSERTE(hotEntriesAddr == dac_cast<TADDR>(pEntries)); |
| 506 | |
| 507 | return GetPersistedColdEntries(); |
| 508 | } |
| 509 | } |
| 510 | |
| 511 | PTR_PersistedBucketList GetPersistedBuckets(DPTR(PersistedEntries) pEntries) |
| 512 | { |
| 513 | SUPPORTS_DAC; |
| 514 | |
| 515 | TADDR hotEntriesAddr = dac_cast<TADDR>(this) + offsetof(NgenHashTable<NGEN_HASH_ARGS>, m_sHotEntries); |
| 516 | TADDR coldEntriesAddr = dac_cast<TADDR>(this) + offsetof(NgenHashTable<NGEN_HASH_ARGS>, m_sColdEntries); |
| 517 | |
| 518 | if (hotEntriesAddr == dac_cast<TADDR>(pEntries)) |
| 519 | { |
| 520 | return GetPersistedHotBuckets(); |
| 521 | } |
| 522 | else |
| 523 | { |
| 524 | _ASSERTE(hotEntriesAddr == dac_cast<TADDR>(pEntries)); |
| 525 | |
| 526 | return GetPersistedColdBuckets(); |
| 527 | } |
| 528 | } |
| 529 | #endif // DACCESS_COMPILE |
| 530 | #endif // FEATURE_PREJIT |
| 531 | |
| 532 | // Loader heap provided at construction time. May be NULL (in which case m_pModule must *not* be NULL). |
| 533 | LoaderHeap *m_pHeap; |
| 534 | |
| 535 | // Fields related to the runtime (volatile or warm) part of the hash. |
| 536 | RelativePointer<DPTR(PTR_VolatileEntry)> m_pWarmBuckets; // Pointer to a simple bucket list (array of VolatileEntry pointers) |
| 537 | DWORD m_cWarmBuckets; // Count of buckets in the above array (always non-zero) |
| 538 | DWORD m_cWarmEntries; // Count of elements in the warm section of the hash |
| 539 | |
| 540 | #ifdef FEATURE_PREJIT |
| 541 | PersistedEntries m_sHotEntries; // Hot persisted hash entries (if any) |
| 542 | PersistedEntries m_sColdEntries; // Cold persisted hash entries (if any) |
| 543 | |
| 544 | DWORD m_cInitialBuckets; // Initial number of warm buckets we started with. Only used |
| 545 | // to reset warm bucket count in ngen-persisted table. |
| 546 | #endif // FEATURE_PREJIT |
| 547 | }; |
| 548 | |
| 549 | // Abstraction around cross-hash entry references (e.g. EEClassHashTable, where entries for nested types point |
| 550 | // to entries for their enclosing types). Under the covers we use a relative pointer which avoids the need to |
| 551 | // allocate a base relocation fixup and the resulting write into the entry at load time. The abstraction hides |
| 552 | // some of the complexity needed to achieve this. |
| 553 | template <NGEN_HASH_PARAMS> |
| 554 | class NgenHashEntryRef |
| 555 | { |
| 556 | public: |
| 557 | // Get a pointer to the referenced entry. |
| 558 | DPTR(VALUE) Get(); |
| 559 | |
| 560 | #ifndef DACCESS_COMPILE |
| 561 | // Set the reference to point to the given entry. |
| 562 | void Set(VALUE *pEntry); |
| 563 | |
| 564 | #ifdef FEATURE_PREJIT |
| 565 | // Call this during the ngen Fixup phase to adjust the relative pointer to account for ngen image layout. |
| 566 | void Fixup(DataImage *pImage, NgenHashTable<NGEN_HASH_ARGS> *pTable); |
| 567 | #endif // FEATURE_PREJIT |
| 568 | |
| 569 | NgenHashEntryRef<NGEN_HASH_ARGS>& operator = (const NgenHashEntryRef<NGEN_HASH_ARGS> &src) |
| 570 | { |
| 571 | src.m_rpEntryRef.BitwiseCopyTo(m_rpEntryRef); |
| 572 | |
| 573 | return *this; |
| 574 | } |
| 575 | #endif // !DACCESS_COMPILE |
| 576 | |
| 577 | private: |
| 578 | RelativePointer<DPTR(VALUE)> m_rpEntryRef; // Entry ref encoded as a delta from this field's location. |
| 579 | }; |
| 580 | |
| 581 | #endif // __NGEN_HASH_INCLUDED |
| 582 | |