content: perf tests for GpuMemoryBuffers mapping and data coherency

content/child/child_thread_impl_perftest.cc

Index: content/child/child_thread_impl_perftest.cc
diff --git a/content/child/child_thread_impl_perftest.cc b/content/child/child_thread_impl_perftest.cc
new file mode 100644
index 0000000000000000000000000000000000000000..0e565daa7247ba207014d779b7dffaefac87da2e
--- /dev/null
+++ b/content/child/child_thread_impl_perftest.cc
@@ -0,0 +1,185 @@
+// Copyright 2015 The Chromium Authors. All rights reserved.

[reveman, 2015/06/18 21:08:57]

hm, this is a browser perf test. is _perftest.cc the correct prefix for that?

[vignatti (out of this project), 2015/06/18 21:59:45]

I don't know where to add this test to be honest. I just thought that it should
end with _perftest.cc prefix given it is a perf test :) Any idea where is the
best place to fit this in?

[reveman, 2015/06/22 20:41:57]

Does this need to be an integration test (browsertest)? Can you implement these
perf tests as unit tests similar to gpu_memory_buffer_impl_unittests.cc instead?

+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "base/bind.h"
+#include "base/command_line.h"
+#include "base/memory/scoped_vector.h"
+#include "base/time/time.h"
+#include "content/child/child_gpu_memory_buffer_manager.h"
+#include "content/child/child_thread_impl.h"
+#include "content/common/gpu/client/gpu_memory_buffer_impl.h"
+#include "content/public/common/content_switches.h"
+#include "content/public/test/content_browser_test.h"
+#include "content/public/test/content_browser_test_utils.h"
+#include "content/shell/browser/shell.h"
+#include "testing/perf/perf_test.h"
+#include "url/gurl.h"
+
+namespace content {
+namespace {
+
+static const int kNumRuns = 30;
+
+enum NativeBufferFlag { kDisableNativeBuffers, kEnableNativeBuffers };
+
+enum MemoryOperation { kMemoryOperationWrite, kMemoryOperationNoop };

[reveman, 2015/06/18 21:08:57]

would this make more sense as separate tests? Map, Write, ReadWrite, etc.

[vignatti (out of this project), 2015/06/18 21:59:46]

Acknowledged.

+
+class ChildThreadImplBrowserTest : public ContentBrowserTest {
+ public:
+  ChildThreadImplBrowserTest()
+      : child_gpu_memory_buffer_manager_(nullptr) {}
+
+  // Overridden from BrowserTestBase:
+  void SetUpCommandLine(base::CommandLine* command_line) override {
+    command_line->AppendSwitch(switches::kSingleProcess);
+  }
+  void SetUpOnMainThread() override {
+    NavigateToURL(shell(), GURL(url::kAboutBlankURL));
+    PostTaskToInProcessRendererAndWait(
+        base::Bind(&ChildThreadImplBrowserTest::SetUpOnChildThread, this));
+  }
+
+  ChildGpuMemoryBufferManager* child_gpu_memory_buffer_manager() {
+    return child_gpu_memory_buffer_manager_;
+  }
+
+  NativeBufferFlag native_buffer_flag_;

[reveman, 2015/06/18 21:08:58]

do we need this as a member?

[vignatti (out of this project), 2015/06/18 21:59:46]

Acknowledged.

+
+ private:
+  void SetUpOnChildThread() {
+    child_gpu_memory_buffer_manager_ =
+        ChildThreadImpl::current()->gpu_memory_buffer_manager();
+  }
+
+  ChildGpuMemoryBufferManager* child_gpu_memory_buffer_manager_;
+};
+
+class ChildThreadImplGpuMemoryBufferPerfTest
+    : public ChildThreadImplBrowserTest,
+      public testing::WithParamInterface<
+          ::testing::tuple<NativeBufferFlag,
+                           gfx::GpuMemoryBuffer::Format,
+                           MemoryOperation>> {
+ public:
+  // Overridden from BrowserTestBase:
+  void SetUpCommandLine(base::CommandLine* command_line) override {
+    ChildThreadImplBrowserTest::SetUpCommandLine(command_line);
+    native_buffer_flag_ = ::testing::get<0>(GetParam());
+    switch (native_buffer_flag_) {
+      case kEnableNativeBuffers:
+        command_line->AppendSwitch(switches::kEnableNativeGpuMemoryBuffers);
+        break;
+      case kDisableNativeBuffers:
+        break;
+    }
+  }
+};
+
+std::string NativeBufferFlagName(NativeBufferFlag flag) {
+  switch (flag) {
+    case kDisableNativeBuffers:
+      return "fallback";

[reveman, 2015/06/18 21:08:57]

"fallback" can be confusing as kEnableNativeBuffers doesn't necessarily mean
that native buffers are used. We might fallback if not available. maybe just ""
and "native"

[vignatti (out of this project), 2015/06/18 21:59:46]

Acknowledged.

+    case kEnableNativeBuffers:
+      return "native";
+  }
+
+  NOTREACHED();
+  return "";
+}
+
+std::string MemoryOperationName(MemoryOperation operation) {
+  switch (operation) {
+    case kMemoryOperationWrite:
+      return "write";
+    case kMemoryOperationNoop:
+      return "noop";
+  }
+
+  NOTREACHED();
+  return "";
+}
+
+IN_PROC_BROWSER_TEST_P(ChildThreadImplGpuMemoryBufferPerfTest,
+                       Map) {
+  gfx::GpuMemoryBuffer::Format format = ::testing::get<1>(GetParam());
+  gfx::Size buffer_size(4, 4);
+  std::string flag_name = NativeBufferFlagName(native_buffer_flag_);
+  MemoryOperation operation = ::testing::get<2>(GetParam());
+  std::string operation_name = MemoryOperationName(operation);
+
+  scoped_ptr<gfx::GpuMemoryBuffer> buffer =
+      child_gpu_memory_buffer_manager()->AllocateGpuMemoryBuffer(
+          buffer_size, format, gfx::GpuMemoryBuffer::MAP);
+  ASSERT_TRUE(buffer);
+  EXPECT_EQ(format, buffer->GetFormat());
+
+  size_t num_planes =
+      GpuMemoryBufferImpl::NumberOfPlanesForGpuMemoryBufferFormat(format);
+
+  // Map buffer planes.
+  scoped_ptr<void* []> planes(new void* [num_planes]);
+
+  for (int i = 0; i < kNumRuns; ++i) {
+    base::TimeTicks start = base::TimeTicks::Now();
+    bool rv = buffer->Map(planes.get());
+    base::TimeTicks end = base::TimeTicks::Now();
+    ASSERT_TRUE(rv);
+    EXPECT_TRUE(buffer->IsMapped());
+
+    // TODO(vignatti): get the mean time and print to stdout only once. At the
+    // moment it's being useful to check individual runs though cause for
+    // example VGEM has way worse performance on its first runs (got check why).
+    perf_test::PrintResult(
+        "time_to_execute_map_",
+        flag_name,
+        operation_name,
+        static_cast<size_t>((end - start).InMicroseconds()),
+        "us", true);

[reveman, 2015/06/18 21:08:57]

can we just have one result per test? Map shows time to execute map, Write shows
time to write into buffer, etc.

[vignatti (out of this project), 2015/06/18 21:59:46]

as I said in the comment, I think right now I prefer spam everything on the
screen to understand why VGEM for example is recording quite different numbers
(https://gist.githubusercontent.com/tiagovignatti/02497f1f3d81e20649ab/raw/d41...).

Alternatively I think we could print out the mean and the standard deviation
capture this dispersion. wdyt?

+
+    // Get stride.
+    scoped_ptr<int[]> strides(new int[num_planes]);
+    buffer->GetStride(strides.get());
+
+    for (size_t plane = 0; plane < num_planes; ++plane) {
+      size_t row_size_in_bytes = 0;
+      EXPECT_TRUE(GpuMemoryBufferImpl::RowSizeInBytes(buffer_size.width(), format,
+            plane, &row_size_in_bytes));
+
+      scoped_ptr<char[]> data(new char[row_size_in_bytes]);
+      memset(data.get(), 0x2a + plane, row_size_in_bytes);
+
+      size_t height = buffer_size.height() /
+        GpuMemoryBufferImpl::SubsamplingFactor(format, plane);
+      for (size_t y = 0; y < height; ++y) {
+        switch (operation) {
+          case kMemoryOperationWrite:
+            // Copy |data| to row |y| of |plane| and verify result.
+            memcpy(static_cast<char*>(planes[plane]) + y * strides[plane], data.get(),
+                row_size_in_bytes);
+            EXPECT_EQ(memcmp(static_cast<char*>(planes[plane]) + y * strides[plane],

[reveman, 2015/06/18 21:08:58]

this does more than write..

[vignatti (out of this project), 2015/06/18 21:59:45]

Acknowledged.

+                  data.get(), row_size_in_bytes),
+                0);
+            break;
+          case kMemoryOperationNoop:
+            break;
+        }
+      }
+    }
+
+    buffer->Unmap();
+    EXPECT_FALSE(buffer->IsMapped());
+  }
+}
+
+INSTANTIATE_TEST_CASE_P(
+    ChildThreadImplGpuMemoryBufferPerfTests,
+    ChildThreadImplGpuMemoryBufferPerfTest,
+    ::testing::Combine(::testing::Values(kDisableNativeBuffers,
+                                         kEnableNativeBuffers),
+                       ::testing::Values(gfx::GpuMemoryBuffer::BGRA_8888),
+                       ::testing::Values(kMemoryOperationWrite,
+                                         kMemoryOperationNoop)));
+
+}  // namespace
+}  // namespace content