zircon/third_party/ulib/scudo/local_cache.h

//===-- local_cache.h -------------------------------------------*- C++ -*-===//
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//

#ifndef SCUDO_LOCAL_CACHE_H_
#define SCUDO_LOCAL_CACHE_H_

#include "internal_defs.h"
#include "stats.h"

namespace scudo {

template <class SizeClassAllocator> struct SizeClassAllocatorLocalCache {
  typedef typename SizeClassAllocator::SizeClassMap SizeClassMap;

  struct TransferBatch {
    static const u32 MaxNumCached = SizeClassMap::MaxNumCachedHint;
    void setFromArray(void **Array, u32 N) {
      DCHECK_LE(N, MaxNumCached);
      memcpy(Batch, Array, N * sizeof(void *));
      Count = N;
    }
    void clear() { Count = 0; }
    void add(void *P) {
      Batch[Count++] = P;
      DCHECK_LE(Count, MaxNumCached);
    }
    void copyToArray(void **Array) const {
      memcpy(Array, Batch, Count * sizeof(void *));
    }
    u32 getCount() const { return Count; }
    void *get(u32 I) const {
      DCHECK_LE(I, Count);
      return Batch[I];
    }
    static u32 MaxCached(uptr Size) {
      return Min(MaxNumCached, SizeClassMap::getMaxCachedHint(Size));
    }
    TransferBatch *Next;

  private:
    u32 Count;
    void *Batch[MaxNumCached];
  };

  void initLinkerInitialized(GlobalStats *S) {
    Stats.initLinkerInitialized();
    if (S)
      S->link(&Stats);
  }

  void init(GlobalStats *S) {
    memset(this, 0, sizeof(*this));
    initLinkerInitialized(S);
  }

  void destroy(SizeClassAllocator *Allocator, GlobalStats *S) {
    drain(Allocator);
    if (S)
      S->unlink(&Stats);
  }

  void *allocate(SizeClassAllocator *Allocator, uptr ClassId) {
    CHECK_LT(ClassId, NumClasses);
    PerClass *C = &PerClassArray[ClassId];
    if (C->Count == 0) {
      if (UNLIKELY(!refill(C, Allocator, ClassId)))
        return nullptr;
      DCHECK_GT(C->Count, 0);
    }
    const uptr ClassSize = C->ClassSize;
    void *P = C->Chunks[--C->Count];
    // The jury is still out as to whether any kind of PREFETCH here increases
    // performance. It definitely decreases performance on Android though.
    // if (!SCUDO_ANDROID && C->Count) PREFETCH(C->Chunks[C->Count - 1]);
    Stats.add(StatAllocated, ClassSize);
    return P;
  }

  void deallocate(SizeClassAllocator *Allocator, uptr ClassId, void *P) {
    CHECK_LT(ClassId, NumClasses);
    PerClass *C = &PerClassArray[ClassId];
    // We still have to initialize the cache in the event that the first heap
    // operation in a thread is a deallocation.
    initCacheMaybe(C);
    if (C->Count == C->MaxCount)
      drain(C, Allocator, ClassId);
    const uptr ClassSize = C->ClassSize;
    C->Chunks[C->Count++] = P;
    Stats.sub(StatAllocated, ClassSize);
  }

  void drain(SizeClassAllocator *Allocator) {
    for (uptr I = 0; I < NumClasses; I++) {
      PerClass *C = &PerClassArray[I];
      while (C->Count > 0)
        drain(C, Allocator, I);
    }
  }

  TransferBatch *createBatch(uptr ClassId, SizeClassAllocator *Allocator,
                             void *B) {
    if (ClassId != SizeClassMap::BatchClassId)
      B = allocate(Allocator, SizeClassMap::BatchClassId);
    return reinterpret_cast<TransferBatch *>(B);
  }

  void destroyBatch(uptr ClassId, SizeClassAllocator *Allocator, void *B) {
    if (ClassId != SizeClassMap::BatchClassId)
      deallocate(Allocator, SizeClassMap::BatchClassId, B);
  }

private:
  static const uptr NumClasses = SizeClassMap::NumClasses;
  struct PerClass {
    u32 Count;
    u32 MaxCount;
    uptr ClassSize;
    void *Chunks[2 * TransferBatch::MaxNumCached];
  };
  PerClass PerClassArray[NumClasses];
  LocalStats Stats;

  void initCacheMaybe(PerClass *C) {
    if (LIKELY(C->MaxCount))
      return;
    initCache();
    DCHECK_NE(C->MaxCount, 0UL);
  }

  NOINLINE void initCache() {
    for (uptr I = 0; I < NumClasses; I++) {
      PerClass *P = &PerClassArray[I];
      const uptr Size = SizeClassAllocator::getSizeByClassId(I);
      P->MaxCount = 2 * TransferBatch::MaxCached(Size);
      P->ClassSize = Size;
    }
  }

  NOINLINE bool refill(PerClass *C, SizeClassAllocator *Allocator,
                       uptr ClassId) {
    initCacheMaybe(C);
    TransferBatch *B = Allocator->popBatch(&Stats, this, ClassId);
    DCHECK_GT(B->getCount(), 0);
    B->copyToArray(C->Chunks);
    C->Count = B->getCount();
    destroyBatch(ClassId, Allocator, B);
    return true;
  }

  NOINLINE void drain(PerClass *C, SizeClassAllocator *Allocator,
                      uptr ClassId) {
    const u32 Count = Min(C->MaxCount / 2, C->Count);
    const uptr FirstIndexToDrain = C->Count - Count;
    TransferBatch *B =
        createBatch(ClassId, Allocator, C->Chunks[FirstIndexToDrain]);
    B->setFromArray(&C->Chunks[FirstIndexToDrain], Count);
    C->Count -= Count;
    Allocator->pushBatch(ClassId, B);
  }
};

} // namespace scudo

#endif // SCUDO_LOCAL_CACHE_H_