Add support for caching small reads.

Add benchmarking to verify this is faster.

Test: Ran unit tests.
Change-Id: I1487114331f4581ec2368e56c4f18c6e3e6bcc7d

libunwindstack/Android.bp

@@ -184,6 +184,7 @@ cc_test {
         "tests/MapInfoGetLoadBiasTest.cpp",
         "tests/MapsTest.cpp",
         "tests/MemoryBufferTest.cpp",
+        "tests/MemoryCacheTest.cpp",
         "tests/MemoryFake.cpp",
         "tests/MemoryFileTest.cpp",
         "tests/MemoryLocalTest.cpp",
@@ -310,6 +311,28 @@ cc_binary {
     ],
 }
 
+//-------------------------------------------------------------------------
+// Benchmarks
+//-------------------------------------------------------------------------
+cc_benchmark {
+    name: "unwind_benchmarks",
+    host_supported: true,
+    defaults: ["libunwindstack_flags"],
+
+    // Disable optimizations so that all of the calls are not optimized away.
+    cflags: [
+        "-O0",
+    ],
+
+    srcs: [
+        "benchmarks/unwind_benchmarks.cpp",
+    ],
+
+    shared_libs: [
+        "libunwindstack",
+    ],
+}
+
 // Generates the elf data for use in the tests for .gnu_debugdata frames.
 // Once these files are generated, use the xz command to compress the data.
 cc_binary_host {

libunwindstack/Memory.cpp

@@ -174,6 +174,13 @@ std::shared_ptr<Memory> Memory::CreateProcessMemory(pid_t pid) {
   return std::shared_ptr<Memory>(new MemoryRemote(pid));
 }
 
+std::shared_ptr<Memory> Memory::CreateProcessMemoryCached(pid_t pid) {
+  if (pid == getpid()) {
+    return std::shared_ptr<Memory>(new MemoryCache(new MemoryLocal()));
+  }
+  return std::shared_ptr<Memory>(new MemoryCache(new MemoryRemote(pid)));
+}
+
 size_t MemoryBuffer::Read(uint64_t addr, void* dst, size_t size) {
   if (addr >= raw_.size()) {
     return 0;
@@ -398,4 +405,50 @@ size_t MemoryOfflineParts::Read(uint64_t addr, void* dst, size_t size) {
   return 0;
 }
 
+size_t MemoryCache::Read(uint64_t addr, void* dst, size_t size) {
+  // Only bother caching and looking at the cache if this is a small read for now.
+  if (size > 64) {
+    return impl_->Read(addr, dst, size);
+  }
+
+  uint64_t addr_page = addr >> kCacheBits;
+  auto entry = cache_.find(addr_page);
+  uint8_t* cache_dst;
+  if (entry != cache_.end()) {
+    cache_dst = entry->second;
+  } else {
+    cache_dst = cache_[addr_page];
+    if (!impl_->ReadFully(addr_page << kCacheBits, cache_dst, kCacheSize)) {
+      // Erase the entry.
+      cache_.erase(addr_page);
+      return impl_->Read(addr, dst, size);
+    }
+  }
+  size_t max_read = ((addr_page + 1) << kCacheBits) - addr;
+  if (size <= max_read) {
+    memcpy(dst, &cache_dst[addr & kCacheMask], size);
+    return size;
+  }
+
+  // The read crossed into another cached entry, since a read can only cross
+  // into one extra cached page, duplicate the code rather than looping.
+  memcpy(dst, &cache_dst[addr & kCacheMask], max_read);
+  dst = &reinterpret_cast<uint8_t*>(dst)[max_read];
+  addr_page++;
+
+  entry = cache_.find(addr_page);
+  if (entry != cache_.end()) {
+    cache_dst = entry->second;
+  } else {
+    cache_dst = cache_[addr_page];
+    if (!impl_->ReadFully(addr_page << kCacheBits, cache_dst, kCacheSize)) {
+      // Erase the entry.
+      cache_.erase(addr_page);
+      return impl_->Read(addr_page << kCacheBits, dst, size - max_read) + max_read;
+    }
+  }
+  memcpy(dst, cache_dst, size - max_read);
+  return size;
+}
+
 }  // namespace unwindstack

libunwindstack/benchmarks/unwind_benchmarks.cpp

@@ -0,0 +1,83 @@
+/*
+ * Copyright (C) 2018 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <stdint.h>
+
+#include <memory>
+
+#include <benchmark/benchmark.h>
+
+#include <unwindstack/Maps.h>
+#include <unwindstack/Memory.h>
+#include <unwindstack/Regs.h>
+#include <unwindstack/RegsGetLocal.h>
+#include <unwindstack/Unwinder.h>
+
+size_t Call6(std::shared_ptr<unwindstack::Memory>& process_memory, unwindstack::Maps* maps) {
+  std::unique_ptr<unwindstack::Regs> regs(unwindstack::Regs::CreateFromLocal());
+  unwindstack::RegsGetLocal(regs.get());
+  unwindstack::Unwinder unwinder(32, maps, regs.get(), process_memory);
+  unwinder.Unwind();
+  return unwinder.NumFrames();
+}
+
+size_t Call5(std::shared_ptr<unwindstack::Memory>& process_memory, unwindstack::Maps* maps) {
+  return Call6(process_memory, maps);
+}
+
+size_t Call4(std::shared_ptr<unwindstack::Memory>& process_memory, unwindstack::Maps* maps) {
+  return Call5(process_memory, maps);
+}
+
+size_t Call3(std::shared_ptr<unwindstack::Memory>& process_memory, unwindstack::Maps* maps) {
+  return Call4(process_memory, maps);
+}
+
+size_t Call2(std::shared_ptr<unwindstack::Memory>& process_memory, unwindstack::Maps* maps) {
+  return Call3(process_memory, maps);
+}
+
+size_t Call1(std::shared_ptr<unwindstack::Memory>& process_memory, unwindstack::Maps* maps) {
+  return Call2(process_memory, maps);
+}
+
+static void BM_uncached_unwind(benchmark::State& state) {
+  auto process_memory = unwindstack::Memory::CreateProcessMemory(getpid());
+  unwindstack::LocalMaps maps;
+  if (!maps.Parse()) {
+    state.SkipWithError("Failed to parse local maps.");
+  }
+
+  for (auto _ : state) {
+    benchmark::DoNotOptimize(Call1(process_memory, &maps));
+  }
+}
+BENCHMARK(BM_uncached_unwind);
+
+static void BM_cached_unwind(benchmark::State& state) {
+  auto process_memory = unwindstack::Memory::CreateProcessMemoryCached(getpid());
+  unwindstack::LocalMaps maps;
+  if (!maps.Parse()) {
+    state.SkipWithError("Failed to parse local maps.");
+  }
+
+  for (auto _ : state) {
+    benchmark::DoNotOptimize(Call1(process_memory, &maps));
+  }
+}
+BENCHMARK(BM_cached_unwind);
+
+BENCHMARK_MAIN();

libunwindstack/include/unwindstack/Memory.h

@@ -25,6 +25,7 @@
 #include <map>
 #include <memory>
 #include <string>
+#include <unordered_map>
 #include <vector>
 
 namespace unwindstack {
@@ -35,9 +36,12 @@ class Memory {
   virtual ~Memory() = default;
 
   static std::shared_ptr<Memory> CreateProcessMemory(pid_t pid);
+  static std::shared_ptr<Memory> CreateProcessMemoryCached(pid_t pid);
 
   virtual bool ReadString(uint64_t addr, std::string* string, uint64_t max_read = UINT64_MAX);
 
+  virtual void Clear() {}
+
   virtual size_t Read(uint64_t addr, void* dst, size_t size) = 0;
 
   bool ReadFully(uint64_t addr, void* dst, size_t size);
@@ -51,6 +55,24 @@ class Memory {
   }
 };
 
+class MemoryCache : public Memory {
+ public:
+  MemoryCache(Memory* memory) : impl_(memory) {}
+  virtual ~MemoryCache() = default;
+
+  size_t Read(uint64_t addr, void* dst, size_t size) override;
+
+  void Clear() override { cache_.clear(); }
+
+ private:
+  constexpr static size_t kCacheBits = 12;
+  constexpr static size_t kCacheMask = (1 << kCacheBits) - 1;
+  constexpr static size_t kCacheSize = 1 << kCacheBits;
+  std::unordered_map<uint64_t, uint8_t[kCacheSize]> cache_;
+
+  std::unique_ptr<Memory> impl_;
+};
+
 class MemoryBuffer : public Memory {
  public:
   MemoryBuffer() = default;

libunwindstack/tests/MemoryCacheTest.cpp

@@ -0,0 +1,143 @@
+/*
+ * Copyright (C) 2018 The Android Open Source Project
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *      http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <stdint.h>
+
+#include <vector>
+
+#include <gtest/gtest.h>
+
+#include <unwindstack/Memory.h>
+
+#include "MemoryFake.h"
+
+namespace unwindstack {
+
+class MemoryCacheTest : public ::testing::Test {
+ protected:
+  void SetUp() override {
+    memory_ = new MemoryFake;
+    memory_cache_.reset(new MemoryCache(memory_));
+
+    memory_->SetMemoryBlock(0x8000, 4096, 0xab);
+    memory_->SetMemoryBlock(0x9000, 4096, 0xde);
+    memory_->SetMemoryBlock(0xa000, 3000, 0x50);
+  }
+
+  MemoryFake* memory_;
+  std::unique_ptr<MemoryCache> memory_cache_;
+
+  constexpr static size_t kMaxCachedSize = 64;
+};
+
+TEST_F(MemoryCacheTest, cached_read) {
+  for (size_t i = 1; i <= kMaxCachedSize; i++) {
+    std::vector<uint8_t> buffer(i);
+    ASSERT_TRUE(memory_cache_->ReadFully(0x8000 + i, buffer.data(), i))
+        << "Read failed at size " << i;
+    ASSERT_EQ(std::vector<uint8_t>(i, 0xab), buffer) << "Failed at size " << i;
+  }
+
+  // Verify the cached data is used.
+  memory_->SetMemoryBlock(0x8000, 4096, 0xff);
+  for (size_t i = 1; i <= kMaxCachedSize; i++) {
+    std::vector<uint8_t> buffer(i);
+    ASSERT_TRUE(memory_cache_->ReadFully(0x8000 + i, buffer.data(), i))
+        << "Read failed at size " << i;
+    ASSERT_EQ(std::vector<uint8_t>(i, 0xab), buffer) << "Failed at size " << i;
+  }
+}
+
+TEST_F(MemoryCacheTest, no_cached_read_after_clear) {
+  for (size_t i = 1; i <= kMaxCachedSize; i++) {
+    std::vector<uint8_t> buffer(i);
+    ASSERT_TRUE(memory_cache_->ReadFully(0x8000 + i, buffer.data(), i))
+        << "Read failed at size " << i;
+    ASSERT_EQ(std::vector<uint8_t>(i, 0xab), buffer) << "Failed at size " << i;
+  }
+
+  // Verify the cached data is not used after a reset.
+  memory_cache_->Clear();
+  memory_->SetMemoryBlock(0x8000, 4096, 0xff);
+  for (size_t i = 1; i <= kMaxCachedSize; i++) {
+    std::vector<uint8_t> buffer(i);
+    ASSERT_TRUE(memory_cache_->ReadFully(0x8000 + i, buffer.data(), i))
+        << "Read failed at size " << i;
+    ASSERT_EQ(std::vector<uint8_t>(i, 0xff), buffer) << "Failed at size " << i;
+  }
+}
+
+TEST_F(MemoryCacheTest, cached_read_across_caches) {
+  std::vector<uint8_t> expect(16, 0xab);
+  expect.resize(32, 0xde);
+
+  std::vector<uint8_t> buffer(32);
+  ASSERT_TRUE(memory_cache_->ReadFully(0x8ff0, buffer.data(), 32));
+  ASSERT_EQ(expect, buffer);
+
+  // Verify the cached data is used.
+  memory_->SetMemoryBlock(0x8000, 4096, 0xff);
+  memory_->SetMemoryBlock(0x9000, 4096, 0xff);
+  ASSERT_TRUE(memory_cache_->ReadFully(0x8ff0, buffer.data(), 32));
+  ASSERT_EQ(expect, buffer);
+}
+
+TEST_F(MemoryCacheTest, no_cache_read) {
+  for (size_t i = kMaxCachedSize + 1; i < 2 * kMaxCachedSize; i++) {
+    std::vector<uint8_t> buffer(i);
+    ASSERT_TRUE(memory_cache_->ReadFully(0x8000 + i, buffer.data(), i))
+        << "Read failed at size " << i;
+    ASSERT_EQ(std::vector<uint8_t>(i, 0xab), buffer) << "Failed at size " << i;
+  }
+
+  // Verify the cached data is not used.
+  memory_->SetMemoryBlock(0x8000, 4096, 0xff);
+  for (size_t i = kMaxCachedSize + 1; i < 2 * kMaxCachedSize; i++) {
+    std::vector<uint8_t> buffer(i);
+    ASSERT_TRUE(memory_cache_->ReadFully(0x8000 + i, buffer.data(), i))
+        << "Read failed at size " << i;
+    ASSERT_EQ(std::vector<uint8_t>(i, 0xff), buffer) << "Failed at size " << i;
+  }
+}
+
+TEST_F(MemoryCacheTest, read_for_cache_fail) {
+  std::vector<uint8_t> buffer(kMaxCachedSize);
+  ASSERT_TRUE(memory_cache_->ReadFully(0xa010, buffer.data(), kMaxCachedSize));
+  ASSERT_EQ(std::vector<uint8_t>(kMaxCachedSize, 0x50), buffer);
+
+  // Verify the cached data is not used.
+  memory_->SetMemoryBlock(0xa000, 3000, 0xff);
+  ASSERT_TRUE(memory_cache_->ReadFully(0xa010, buffer.data(), kMaxCachedSize));
+  ASSERT_EQ(std::vector<uint8_t>(kMaxCachedSize, 0xff), buffer);
+}
+
+TEST_F(MemoryCacheTest, read_for_cache_fail_cross) {
+  std::vector<uint8_t> expect(16, 0xde);
+  expect.resize(32, 0x50);
+
+  std::vector<uint8_t> buffer(32);
+  ASSERT_TRUE(memory_cache_->ReadFully(0x9ff0, buffer.data(), 32));
+  ASSERT_EQ(expect, buffer);
+
+  // Verify the cached data is not used for the second half but for the first.
+  memory_->SetMemoryBlock(0xa000, 3000, 0xff);
+  ASSERT_TRUE(memory_cache_->ReadFully(0x9ff0, buffer.data(), 32));
+  expect.resize(16);
+  expect.resize(32, 0xff);
+  ASSERT_EQ(expect, buffer);
+}
+
+}  // namespace unwindstack