v8/test/unittests/heap/spaces-unittest.cc
Leszek Swirski e659917aa3 [offthread] Add OffThreadFactory
Introduce OffThreadFactory with initial string construction support.

The OffThreadFactory shares with Factory a new CRTP base class, called
FactoryBase. Methods in FactoryBase return a FactoryHandle<Factory, T>
alias, which is Handle<T> for normal Factory and a new OffThreadHandle<T>
for OffThreadFactory. OffThreadHandle<T> behaves like Handle<T>, except
it stores the object in-line rather than needing external storage.

Any shared factory methods are moved into FactoryBase, which uses CRTP
to call the sub-class's AllocateRaw method (plus a few more customization
points which need Isolate access on the main thread).

Methods that used to take an Isolate or Factory, and are needed off the
main thread, are now expected to be templated on the factory type and
to use the appropriate handle.

Once an OffThreadFactory has finished being used (e.g. off-thread
compilation completed) its pages are "Published" into the main-thread
Heap. To deal with string internalization without creating a bunch of
ThinStrings, this is done in two stages:

  1. 'FinishOffThread': The off-thread pages are walked to
     collect all slots pointing to "internalized" strings. After this is
     called it is invalid to allocate any more objects with the factory.
  2. 'Publish': On the main thread, we transform these slots into
     <Handle to holder, offset> pairs, then for each saved slot
     re-internalize its string and update the slot to point to the
     internalized string.

Bug: chromium:1011762
Change-Id: I008a694da3c357de34362bd86fe7e1f46b535d5e
Reviewed-on: https://chromium-review.googlesource.com/c/v8/v8/+/1992434
Commit-Queue: Leszek Swirski <leszeks@chromium.org>
Reviewed-by: Ulan Degenbaev <ulan@chromium.org>
Reviewed-by: Toon Verwaest <verwaest@chromium.org>
Cr-Commit-Position: refs/heads/master@{#65787}
2020-01-15 12:38:29 +00:00

516 lines
19 KiB
C++

// Copyright 2017 the V8 project authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file.
#include "src/heap/spaces.h"
#include <memory>
#include "src/common/globals.h"
#include "src/execution/isolate.h"
#include "src/heap/heap-inl.h"
#include "src/heap/heap-write-barrier-inl.h"
#include "src/heap/heap.h"
#include "src/heap/spaces-inl.h"
#include "test/unittests/test-utils.h"
namespace v8 {
namespace internal {
using SpacesTest = TestWithIsolate;
TEST_F(SpacesTest, CompactionSpaceMerge) {
Heap* heap = i_isolate()->heap();
OldSpace* old_space = heap->old_space();
EXPECT_TRUE(old_space != nullptr);
CompactionSpace* compaction_space =
new CompactionSpace(heap, OLD_SPACE, NOT_EXECUTABLE,
LocalSpaceKind::kCompactionSpaceForMarkCompact);
EXPECT_TRUE(compaction_space != nullptr);
for (Page* p : *old_space) {
// Unlink free lists from the main space to avoid reusing the memory for
// compaction spaces.
old_space->UnlinkFreeListCategories(p);
}
// Cannot loop until "Available()" since we initially have 0 bytes available
// and would thus neither grow, nor be able to allocate an object.
const int kNumObjects = 10;
const int kNumObjectsPerPage =
compaction_space->AreaSize() / kMaxRegularHeapObjectSize;
const int kExpectedPages =
(kNumObjects + kNumObjectsPerPage - 1) / kNumObjectsPerPage;
for (int i = 0; i < kNumObjects; i++) {
HeapObject object =
compaction_space->AllocateRawUnaligned(kMaxRegularHeapObjectSize)
.ToObjectChecked();
heap->CreateFillerObjectAt(object.address(), kMaxRegularHeapObjectSize,
ClearRecordedSlots::kNo);
}
int pages_in_old_space = old_space->CountTotalPages();
int pages_in_compaction_space = compaction_space->CountTotalPages();
EXPECT_EQ(kExpectedPages, pages_in_compaction_space);
old_space->MergeLocalSpace(compaction_space);
EXPECT_EQ(pages_in_old_space + pages_in_compaction_space,
old_space->CountTotalPages());
delete compaction_space;
}
class OffThreadAllocationThread final : public base::Thread {
public:
explicit OffThreadAllocationThread(Heap* heap)
: Thread(Options("OffThreadAllocationThread")), heap_(heap) {}
void Run() override {
off_thread_space_ = std::make_unique<OffThreadSpace>(heap_);
EXPECT_TRUE(off_thread_space_ != nullptr);
// Cannot loop until "Available()" since we initially have 0 bytes available
// and would thus neither grow, nor be able to allocate an object.
const int kNumObjects = 10;
const int kNumObjectsPerPage =
off_thread_space_->AreaSize() / kMaxRegularHeapObjectSize;
const int kExpectedPages =
(kNumObjects + kNumObjectsPerPage - 1) / kNumObjectsPerPage;
for (int i = 0; i < kNumObjects; i++) {
HeapObject object =
off_thread_space_->AllocateRawUnaligned(kMaxRegularHeapObjectSize)
.ToObjectChecked();
heap_->CreateFillerObjectAt(object.address(), kMaxRegularHeapObjectSize,
ClearRecordedSlots::kNo);
}
int pages_in_off_thread_space = off_thread_space_->CountTotalPages();
EXPECT_EQ(kExpectedPages, pages_in_off_thread_space);
}
OffThreadSpace* space() { return off_thread_space_.get(); }
private:
Heap* heap_;
std::unique_ptr<OffThreadSpace> off_thread_space_;
};
TEST_F(SpacesTest, OffThreadSpaceAllocate) {
Heap* heap = i_isolate()->heap();
static const int kNumThreads = 10;
std::unique_ptr<OffThreadAllocationThread> threads[10];
for (int i = 0; i < kNumThreads; ++i) {
threads[i] = std::make_unique<OffThreadAllocationThread>(heap);
}
for (int i = 0; i < kNumThreads; ++i) {
CHECK(threads[i]->Start());
}
for (int i = 0; i < kNumThreads; ++i) {
threads[i]->Join();
}
}
TEST_F(SpacesTest, OffThreadSpaceMerge) {
Heap* heap = i_isolate()->heap();
OldSpace* old_space = heap->old_space();
EXPECT_TRUE(old_space != nullptr);
static const int kNumThreads = 10;
std::unique_ptr<OffThreadAllocationThread> threads[10];
for (int i = 0; i < kNumThreads; ++i) {
threads[i] = std::make_unique<OffThreadAllocationThread>(heap);
}
for (int i = 0; i < kNumThreads; ++i) {
CHECK(threads[i]->Start());
}
for (int i = 0; i < kNumThreads; ++i) {
threads[i]->Join();
}
int pages_in_old_space = old_space->CountTotalPages();
int expected_merged_pages = 0;
for (int i = 0; i < kNumThreads; ++i) {
int pages_in_off_thread_space = threads[i]->space()->CountTotalPages();
old_space->MergeLocalSpace(threads[i]->space());
expected_merged_pages += pages_in_off_thread_space;
}
EXPECT_EQ(pages_in_old_space + expected_merged_pages,
old_space->CountTotalPages());
}
TEST_F(SpacesTest, OffThreadSpaceMergeDuringIncrementalMarking) {
Heap* heap = i_isolate()->heap();
OldSpace* old_space = heap->old_space();
EXPECT_TRUE(old_space != nullptr);
static const int kNumThreads = 10;
std::unique_ptr<OffThreadAllocationThread> threads[10];
for (int i = 0; i < kNumThreads; ++i) {
threads[i] = std::make_unique<OffThreadAllocationThread>(heap);
}
for (int i = 0; i < kNumThreads; ++i) {
CHECK(threads[i]->Start());
}
for (int i = 0; i < kNumThreads; ++i) {
threads[i]->Join();
}
heap->StartIncrementalMarking(Heap::kNoGCFlags,
GarbageCollectionReason::kTesting);
int pages_in_old_space = old_space->CountTotalPages();
int expected_merged_pages = 0;
for (int i = 0; i < kNumThreads; ++i) {
int pages_in_off_thread_space = threads[i]->space()->CountTotalPages();
old_space->MergeLocalSpace(threads[i]->space());
expected_merged_pages += pages_in_off_thread_space;
}
heap->FinalizeIncrementalMarkingAtomically(GarbageCollectionReason::kTesting);
EXPECT_EQ(pages_in_old_space + expected_merged_pages,
old_space->CountTotalPages());
}
class LargeOffThreadAllocationThread final : public base::Thread {
public:
explicit LargeOffThreadAllocationThread(Heap* heap)
: Thread(Options("LargeOffThreadAllocationThread")), heap_(heap) {}
void Run() override {
off_thread_lo_space_ = std::make_unique<OffThreadLargeObjectSpace>(heap_);
EXPECT_TRUE(off_thread_lo_space_ != nullptr);
const int kNumObjects = 10;
const int kExpectedPages = kNumObjects;
for (int i = 0; i < kNumObjects; i++) {
HeapObject object =
off_thread_lo_space_
->AllocateRaw(kMaxRegularHeapObjectSize + kTaggedSize)
.ToObjectChecked();
heap_->CreateFillerObjectAt(object.address(),
kMaxRegularHeapObjectSize + kTaggedSize,
ClearRecordedSlots::kNo);
}
int pages_in_off_thread_space = off_thread_lo_space_->PageCount();
EXPECT_EQ(kExpectedPages, pages_in_off_thread_space);
}
OffThreadLargeObjectSpace* space() { return off_thread_lo_space_.get(); }
private:
Heap* heap_;
std::unique_ptr<OffThreadLargeObjectSpace> off_thread_lo_space_;
};
TEST_F(SpacesTest, OffThreadLargeObjectSpaceAllocate) {
Heap* heap = i_isolate()->heap();
static const int kNumThreads = 10;
std::unique_ptr<LargeOffThreadAllocationThread> threads[10];
for (int i = 0; i < kNumThreads; ++i) {
threads[i] = std::make_unique<LargeOffThreadAllocationThread>(heap);
}
for (int i = 0; i < kNumThreads; ++i) {
CHECK(threads[i]->Start());
}
for (int i = 0; i < kNumThreads; ++i) {
threads[i]->Join();
}
}
TEST_F(SpacesTest, OffThreadLargeObjectSpaceMerge) {
Heap* heap = i_isolate()->heap();
OldLargeObjectSpace* lo_space = heap->lo_space();
EXPECT_TRUE(lo_space != nullptr);
static const int kNumThreads = 10;
std::unique_ptr<LargeOffThreadAllocationThread> threads[10];
for (int i = 0; i < kNumThreads; ++i) {
threads[i] = std::make_unique<LargeOffThreadAllocationThread>(heap);
}
for (int i = 0; i < kNumThreads; ++i) {
CHECK(threads[i]->Start());
}
for (int i = 0; i < kNumThreads; ++i) {
threads[i]->Join();
}
int pages_in_old_space = lo_space->PageCount();
int expected_merged_pages = 0;
for (int i = 0; i < kNumThreads; ++i) {
int pages_in_off_thread_space = threads[i]->space()->PageCount();
lo_space->MergeOffThreadSpace(threads[i]->space());
expected_merged_pages += pages_in_off_thread_space;
}
EXPECT_EQ(pages_in_old_space + expected_merged_pages, lo_space->PageCount());
}
TEST_F(SpacesTest, OffThreadLargeObjectSpaceMergeDuringIncrementalMarking) {
Heap* heap = i_isolate()->heap();
OldLargeObjectSpace* lo_space = heap->lo_space();
EXPECT_TRUE(lo_space != nullptr);
static const int kNumThreads = 10;
std::unique_ptr<LargeOffThreadAllocationThread> threads[10];
for (int i = 0; i < kNumThreads; ++i) {
threads[i] = std::make_unique<LargeOffThreadAllocationThread>(heap);
}
for (int i = 0; i < kNumThreads; ++i) {
CHECK(threads[i]->Start());
}
for (int i = 0; i < kNumThreads; ++i) {
threads[i]->Join();
}
int pages_in_old_space = lo_space->PageCount();
int expected_merged_pages = 0;
for (int i = 0; i < kNumThreads; ++i) {
int pages_in_off_thread_space = threads[i]->space()->PageCount();
lo_space->MergeOffThreadSpace(threads[i]->space());
expected_merged_pages += pages_in_off_thread_space;
}
EXPECT_EQ(pages_in_old_space + expected_merged_pages, lo_space->PageCount());
}
TEST_F(SpacesTest, WriteBarrierFromHeapObject) {
constexpr Address address1 = Page::kPageSize;
HeapObject object1 = HeapObject::unchecked_cast(Object(address1));
MemoryChunk* chunk1 = MemoryChunk::FromHeapObject(object1);
heap_internals::MemoryChunk* slim_chunk1 =
heap_internals::MemoryChunk::FromHeapObject(object1);
EXPECT_EQ(static_cast<void*>(chunk1), static_cast<void*>(slim_chunk1));
constexpr Address address2 = 2 * Page::kPageSize - 1;
HeapObject object2 = HeapObject::unchecked_cast(Object(address2));
MemoryChunk* chunk2 = MemoryChunk::FromHeapObject(object2);
heap_internals::MemoryChunk* slim_chunk2 =
heap_internals::MemoryChunk::FromHeapObject(object2);
EXPECT_EQ(static_cast<void*>(chunk2), static_cast<void*>(slim_chunk2));
}
TEST_F(SpacesTest, WriteBarrierIsMarking) {
const size_t kSizeOfMemoryChunk = sizeof(MemoryChunk);
char memory[kSizeOfMemoryChunk];
memset(&memory, 0, kSizeOfMemoryChunk);
MemoryChunk* chunk = reinterpret_cast<MemoryChunk*>(&memory);
heap_internals::MemoryChunk* slim_chunk =
reinterpret_cast<heap_internals::MemoryChunk*>(&memory);
EXPECT_FALSE(chunk->IsFlagSet(MemoryChunk::INCREMENTAL_MARKING));
EXPECT_FALSE(slim_chunk->IsMarking());
chunk->SetFlag(MemoryChunk::INCREMENTAL_MARKING);
EXPECT_TRUE(chunk->IsFlagSet(MemoryChunk::INCREMENTAL_MARKING));
EXPECT_TRUE(slim_chunk->IsMarking());
chunk->ClearFlag(MemoryChunk::INCREMENTAL_MARKING);
EXPECT_FALSE(chunk->IsFlagSet(MemoryChunk::INCREMENTAL_MARKING));
EXPECT_FALSE(slim_chunk->IsMarking());
}
TEST_F(SpacesTest, WriteBarrierInYoungGenerationToSpace) {
const size_t kSizeOfMemoryChunk = sizeof(MemoryChunk);
char memory[kSizeOfMemoryChunk];
memset(&memory, 0, kSizeOfMemoryChunk);
MemoryChunk* chunk = reinterpret_cast<MemoryChunk*>(&memory);
heap_internals::MemoryChunk* slim_chunk =
reinterpret_cast<heap_internals::MemoryChunk*>(&memory);
EXPECT_FALSE(chunk->InYoungGeneration());
EXPECT_FALSE(slim_chunk->InYoungGeneration());
chunk->SetFlag(MemoryChunk::TO_PAGE);
EXPECT_TRUE(chunk->InYoungGeneration());
EXPECT_TRUE(slim_chunk->InYoungGeneration());
chunk->ClearFlag(MemoryChunk::TO_PAGE);
EXPECT_FALSE(chunk->InYoungGeneration());
EXPECT_FALSE(slim_chunk->InYoungGeneration());
}
TEST_F(SpacesTest, WriteBarrierInYoungGenerationFromSpace) {
const size_t kSizeOfMemoryChunk = sizeof(MemoryChunk);
char memory[kSizeOfMemoryChunk];
memset(&memory, 0, kSizeOfMemoryChunk);
MemoryChunk* chunk = reinterpret_cast<MemoryChunk*>(&memory);
heap_internals::MemoryChunk* slim_chunk =
reinterpret_cast<heap_internals::MemoryChunk*>(&memory);
EXPECT_FALSE(chunk->InYoungGeneration());
EXPECT_FALSE(slim_chunk->InYoungGeneration());
chunk->SetFlag(MemoryChunk::FROM_PAGE);
EXPECT_TRUE(chunk->InYoungGeneration());
EXPECT_TRUE(slim_chunk->InYoungGeneration());
chunk->ClearFlag(MemoryChunk::FROM_PAGE);
EXPECT_FALSE(chunk->InYoungGeneration());
EXPECT_FALSE(slim_chunk->InYoungGeneration());
}
TEST_F(SpacesTest, CodeRangeAddressReuse) {
CodeRangeAddressHint hint;
// Create code ranges.
Address code_range1 = hint.GetAddressHint(100);
Address code_range2 = hint.GetAddressHint(200);
Address code_range3 = hint.GetAddressHint(100);
// Since the addresses are random, we cannot check that they are different.
// Free two code ranges.
hint.NotifyFreedCodeRange(code_range1, 100);
hint.NotifyFreedCodeRange(code_range2, 200);
// The next two code ranges should reuse the freed addresses.
Address code_range4 = hint.GetAddressHint(100);
EXPECT_EQ(code_range4, code_range1);
Address code_range5 = hint.GetAddressHint(200);
EXPECT_EQ(code_range5, code_range2);
// Free the third code range and check address reuse.
hint.NotifyFreedCodeRange(code_range3, 100);
Address code_range6 = hint.GetAddressHint(100);
EXPECT_EQ(code_range6, code_range3);
}
// Tests that FreeListMany::SelectFreeListCategoryType returns what it should.
TEST_F(SpacesTest, FreeListManySelectFreeListCategoryType) {
FreeListMany free_list;
// Testing that all sizes below 256 bytes get assigned the correct category
for (size_t size = 0; size <= FreeListMany::kPreciseCategoryMaxSize; size++) {
FreeListCategoryType cat = free_list.SelectFreeListCategoryType(size);
if (cat == 0) {
// If cat == 0, then we make sure that |size| doesn't fit in the 2nd
// category.
EXPECT_LT(size, free_list.categories_min[1]);
} else {
// Otherwise, size should fit in |cat|, but not in |cat+1|.
EXPECT_LE(free_list.categories_min[cat], size);
EXPECT_LT(size, free_list.categories_min[cat + 1]);
}
}
// Testing every size above 256 would take long time, so test only some
// "interesting cases": picking some number in the middle of the categories,
// as well as at the categories' bounds.
for (int cat = kFirstCategory + 1; cat <= free_list.last_category_; cat++) {
std::vector<size_t> sizes;
// Adding size less than this category's minimum
sizes.push_back(free_list.categories_min[cat] - 8);
// Adding size equal to this category's minimum
sizes.push_back(free_list.categories_min[cat]);
// Adding size greater than this category's minimum
sizes.push_back(free_list.categories_min[cat] + 8);
// Adding size between this category's minimum and the next category
if (cat != free_list.last_category_) {
sizes.push_back(
(free_list.categories_min[cat] + free_list.categories_min[cat + 1]) /
2);
}
for (size_t size : sizes) {
FreeListCategoryType cat = free_list.SelectFreeListCategoryType(size);
if (cat == free_list.last_category_) {
// If cat == last_category, then we make sure that |size| indeeds fits
// in the last category.
EXPECT_LE(free_list.categories_min[cat], size);
} else {
// Otherwise, size should fit in |cat|, but not in |cat+1|.
EXPECT_LE(free_list.categories_min[cat], size);
EXPECT_LT(size, free_list.categories_min[cat + 1]);
}
}
}
}
// Tests that FreeListMany::GuaranteedAllocatable returns what it should.
TEST_F(SpacesTest, FreeListManyGuaranteedAllocatable) {
FreeListMany free_list;
for (int cat = kFirstCategory; cat < free_list.last_category_; cat++) {
std::vector<size_t> sizes;
// Adding size less than this category's minimum
sizes.push_back(free_list.categories_min[cat] - 8);
// Adding size equal to this category's minimum
sizes.push_back(free_list.categories_min[cat]);
// Adding size greater than this category's minimum
sizes.push_back(free_list.categories_min[cat] + 8);
if (cat != free_list.last_category_) {
// Adding size between this category's minimum and the next category
sizes.push_back(
(free_list.categories_min[cat] + free_list.categories_min[cat + 1]) /
2);
}
for (size_t size : sizes) {
FreeListCategoryType cat_free =
free_list.SelectFreeListCategoryType(size);
size_t guaranteed_allocatable = free_list.GuaranteedAllocatable(size);
if (cat_free == free_list.last_category_) {
// If |cat_free| == last_category, then guaranteed_allocatable must
// return the last category, because when allocating, the last category
// is searched entirely.
EXPECT_EQ(free_list.SelectFreeListCategoryType(guaranteed_allocatable),
free_list.last_category_);
} else if (size < free_list.categories_min[0]) {
// If size < free_list.categories_min[0], then the bytes are wasted, and
// guaranteed_allocatable should return 0.
EXPECT_EQ(guaranteed_allocatable, 0ul);
} else {
// Otherwise, |guaranteed_allocatable| is equal to the minimum of
// |size|'s category (|cat_free|);
EXPECT_EQ(free_list.categories_min[cat_free], guaranteed_allocatable);
}
}
}
}
// Tests that
// FreeListManyCachedFastPath::SelectFastAllocationFreeListCategoryType returns
// what it should.
TEST_F(SpacesTest,
FreeListManyCachedFastPathSelectFastAllocationFreeListCategoryType) {
FreeListManyCachedFastPath free_list;
for (int cat = kFirstCategory; cat <= free_list.last_category_; cat++) {
std::vector<size_t> sizes;
// Adding size less than this category's minimum
sizes.push_back(free_list.categories_min[cat] - 8);
// Adding size equal to this category's minimum
sizes.push_back(free_list.categories_min[cat]);
// Adding size greater than this category's minimum
sizes.push_back(free_list.categories_min[cat] + 8);
// Adding size between this category's minimum and the next category
if (cat != free_list.last_category_) {
sizes.push_back(
(free_list.categories_min[cat] + free_list.categories_min[cat + 1]) /
2);
}
for (size_t size : sizes) {
FreeListCategoryType cat =
free_list.SelectFastAllocationFreeListCategoryType(size);
if (size <= FreeListManyCachedFastPath::kTinyObjectMaxSize) {
// For tiny objects, the first category of the fast path should be
// chosen.
EXPECT_TRUE(cat == FreeListManyCachedFastPath::kFastPathFirstCategory);
} else if (size >= free_list.categories_min[free_list.last_category_] -
FreeListManyCachedFastPath::kFastPathOffset) {
// For objects close to the minimum of the last category, the last
// category is chosen.
EXPECT_EQ(cat, free_list.last_category_);
} else {
// For other objects, the chosen category must satisfy that its minimum
// is at least |size|+1.85k.
EXPECT_GE(free_list.categories_min[cat],
size + FreeListManyCachedFastPath::kFastPathOffset);
// And the smaller categoriy's minimum is less than |size|+1.85k
// (otherwise it would have been chosen instead).
EXPECT_LT(free_list.categories_min[cat - 1],
size + FreeListManyCachedFastPath::kFastPathOffset);
}
}
}
}
} // namespace internal
} // namespace v8