Common: Persistent-mapped stream buffer implementation

src/common/gl/stream_buffer.cpp

@@ -1,9 +1,12 @@
 #include "stream_buffer.h"
+#include "YBaseLib/Assert.h"
+#include <array>
+#include <cstdio>
 
 namespace GL {
 
 StreamBuffer::StreamBuffer(GLenum target, GLuint buffer_id, u32 size)
-  : m_target(target), m_buffer_id(buffer_id), m_size(size), m_cpu_buffer(size)
+  : m_target(target), m_buffer_id(buffer_id), m_size(size)
 {
 }
 
@@ -22,22 +25,30 @@ void StreamBuffer::Unbind()
   glBindBuffer(m_target, 0);
 }
 
-StreamBuffer::MappingResult StreamBuffer::Map(u32 alignment, u32 min_size)
+namespace detail {
-{
-  return MappingResult{static_cast<void*>(m_cpu_buffer.data()), 0, 0, m_size / alignment};
-}
 
-void StreamBuffer::Unmap(u32 used_size)
+// Uses glBufferSubData() to update. Preferred for drivers which don't support {ARB,EXT}_buffer_storage.
+class BufferSubDataStreamBuffer final : public StreamBuffer
 {
+public:
+  ~BufferSubDataStreamBuffer() override = default;
+
+  MappingResult Map(u32 alignment, u32 min_size) override
+  {
+    return MappingResult{static_cast<void*>(m_cpu_buffer.data()), 0, 0, m_size / alignment};
+  }
+
+  void Unmap(u32 used_size) override
+  {
     if (used_size == 0)
       return;
 
     glBindBuffer(m_target, m_buffer_id);
     glBufferSubData(m_target, 0, used_size, m_cpu_buffer.data());
-}
+  }
 
-std::unique_ptr<StreamBuffer> StreamBuffer::Create(GLenum target, u32 size)
+  static std::unique_ptr<StreamBuffer> Create(GLenum target, u32 size)
-{
+  {
     glGetError();
 
     GLuint buffer_id;
@@ -52,7 +63,236 @@ std::unique_ptr<StreamBuffer> StreamBuffer::Create(GLenum target, u32 size)
       return {};
     }
 
-  return std::unique_ptr<StreamBuffer>(new StreamBuffer(target, buffer_id, size));
+    return std::unique_ptr<StreamBuffer>(new BufferSubDataStreamBuffer(target, buffer_id, size));
+  }
+
+private:
+  BufferSubDataStreamBuffer(GLenum target, GLuint buffer_id, u32 size)
+    : StreamBuffer(target, buffer_id, size), m_cpu_buffer(size)
+  {
+  }
+
+  std::vector<u8> m_cpu_buffer;
+};
+
+// Uses BufferData() to orphan the buffer after every update. Used on Mali where BufferSubData forces a sync.
+class BufferDataStreamBuffer final : public StreamBuffer
+{
+public:
+  ~BufferDataStreamBuffer() override = default;
+
+  MappingResult Map(u32 alignment, u32 min_size) override
+  {
+    return MappingResult{static_cast<void*>(m_cpu_buffer.data()), 0, 0, m_size / alignment};
+  }
+
+  void Unmap(u32 used_size) override
+  {
+    if (used_size == 0)
+      return;
+
+    glBindBuffer(m_target, m_buffer_id);
+    glBufferData(m_target, used_size, m_cpu_buffer.data(), GL_STREAM_DRAW);
+  }
+
+  static std::unique_ptr<StreamBuffer> Create(GLenum target, u32 size)
+  {
+    glGetError();
+
+    GLuint buffer_id;
+    glGenBuffers(1, &buffer_id);
+    glBindBuffer(target, buffer_id);
+    glBufferData(target, size, nullptr, GL_STREAM_DRAW);
+
+    GLenum err = glGetError();
+    if (err != GL_NO_ERROR)
+    {
+      glDeleteBuffers(1, &buffer_id);
+      return {};
+    }
+
+    return std::unique_ptr<StreamBuffer>(new BufferDataStreamBuffer(target, buffer_id, size));
+  }
+
+private:
+  BufferDataStreamBuffer(GLenum target, GLuint buffer_id, u32 size)
+    : StreamBuffer(target, buffer_id, size), m_cpu_buffer(size)
+  {
+  }
+
+  std::vector<u8> m_cpu_buffer;
+};
+
+// Base class for implementations which require syncing.
+class SyncingStreamBuffer : public StreamBuffer
+{
+public:
+  enum : u32
+  {
+    NUM_SYNC_POINTS = 16
+  };
+
+  virtual ~SyncingStreamBuffer() override
+  {
+    for (u32 i = m_available_block_index; i <= m_used_block_index; i++)
+    {
+      DebugAssert(m_sync_objects[i]);
+      glDeleteSync(m_sync_objects[i]);
+    }
+  }
+
+protected:
+  SyncingStreamBuffer(GLenum target, GLuint buffer_id, u32 size)
+    : StreamBuffer(target, buffer_id, size), m_bytes_per_block((size + (NUM_SYNC_POINTS)-1) / NUM_SYNC_POINTS)
+  {
+  }
+
+  u32 GetSyncIndexForOffset(u32 offset) { return offset / m_bytes_per_block; }
+
+  void AddSyncsForOffset(u32 offset)
+  {
+    const u32 end = GetSyncIndexForOffset(offset);
+    for (; m_used_block_index < end; m_used_block_index++)
+    {
+      DebugAssert(!m_sync_objects[m_used_block_index]);
+      m_sync_objects[m_used_block_index] = glFenceSync(GL_SYNC_GPU_COMMANDS_COMPLETE, 0);
+    }
+  }
+
+  void WaitForSync(GLsync& sync)
+  {
+    glClientWaitSync(sync, GL_SYNC_FLUSH_COMMANDS_BIT, GL_TIMEOUT_IGNORED);
+    glDeleteSync(sync);
+    sync = nullptr;
+  }
+
+  void EnsureSyncsWaitedForOffset(u32 offset)
+  {
+    const u32 end = std::min<u32>(GetSyncIndexForOffset(offset) + 1, NUM_SYNC_POINTS);
+    for (; m_available_block_index < end; m_available_block_index++)
+    {
+      DebugAssert(m_sync_objects[m_available_block_index]);
+      WaitForSync(m_sync_objects[m_available_block_index]);
+    }
+  }
+
+  void AllocateSpace(u32 size)
+  {
+    // add sync objects for writes since the last allocation
+    AddSyncsForOffset(m_position);
+
+    // wait for sync objects for the space we want to use
+    EnsureSyncsWaitedForOffset(m_position + size);
+
+    // wrap-around?
+    if ((m_position + size) > m_size)
+    {
+      // current position ... buffer end
+      AddSyncsForOffset(m_size);
+
+      // rewind, and try again
+      m_position = 0;
+
+      // wait for the sync at the start of the buffer
+      WaitForSync(m_sync_objects[0]);
+      m_available_block_index = 1;
+
+      // and however much more we need to satisfy the allocation
+      EnsureSyncsWaitedForOffset(size);
+      m_used_block_index = 0;
+    }
+  }
+
+  u32 m_position = 0;
+  u32 m_used_block_index = 0;
+  u32 m_available_block_index = NUM_SYNC_POINTS;
+  u32 m_bytes_per_block;
+  std::array<GLsync, NUM_SYNC_POINTS> m_sync_objects{};
+};
+
+class BufferStorageStreamBuffer : public SyncingStreamBuffer
+{
+public:
+  ~BufferStorageStreamBuffer() override
+  {
+    glBindBuffer(m_target, m_buffer_id);
+    glUnmapBuffer(m_target);
+  }
+
+  MappingResult Map(u32 alignment, u32 min_size) override
+  {
+    if (m_position > 0)
+      m_position = Common::AlignUp(m_position, alignment);
+
+    AllocateSpace(min_size);
+    DebugAssert((m_position + min_size) <= (m_available_block_index * m_bytes_per_block));
+
+    const u32 free_space_in_block = ((m_available_block_index * m_bytes_per_block) - m_position);
+    return MappingResult{static_cast<void*>(m_mapped_ptr + m_position), m_position, m_position / alignment,
+                         free_space_in_block / alignment};
+  }
+
+  void Unmap(u32 used_size) override
+  {
+    DebugAssert((m_position + used_size) <= m_size);
+    m_position += used_size;
+  }
+
+  static std::unique_ptr<StreamBuffer> Create(GLenum target, u32 size)
+  {
+    glGetError();
+
+    GLuint buffer_id;
+    glGenBuffers(1, &buffer_id);
+    glBindBuffer(target, buffer_id);
+    if (GLAD_GL_VERSION_4_4 || GLAD_GL_ARB_buffer_storage)
+      glBufferStorage(target, size, nullptr, GL_MAP_WRITE_BIT | GL_MAP_PERSISTENT_BIT | GL_MAP_COHERENT_BIT);
+    else if (GLAD_GL_EXT_buffer_storage)
+      glBufferStorageEXT(target, size, nullptr, GL_MAP_WRITE_BIT | GL_MAP_PERSISTENT_BIT | GL_MAP_COHERENT_BIT);
+
+    GLenum err = glGetError();
+    if (err != GL_NO_ERROR)
+    {
+      glDeleteBuffers(1, &buffer_id);
+      return {};
+    }
+
+    u8* mapped_ptr = static_cast<u8*>(
+      glMapBufferRange(target, 0, size, GL_MAP_WRITE_BIT | GL_MAP_PERSISTENT_BIT | GL_MAP_COHERENT_BIT));
+    Assert(mapped_ptr);
+
+    return std::unique_ptr<StreamBuffer>(new BufferStorageStreamBuffer(target, buffer_id, size, mapped_ptr));
+  }
+
+private:
+  BufferStorageStreamBuffer(GLenum target, GLuint buffer_id, u32 size, u8* mapped_ptr)
+    : SyncingStreamBuffer(target, buffer_id, size), m_mapped_ptr(mapped_ptr)
+  {
+  }
+
+  u8* m_mapped_ptr;
+};
+
+} // namespace detail
+
+std::unique_ptr<StreamBuffer> StreamBuffer::Create(GLenum target, u32 size)
+{
+  std::unique_ptr<StreamBuffer> buf;
+  if (GLAD_GL_VERSION_4_4 || GLAD_GL_ARB_buffer_storage || GLAD_GL_EXT_buffer_storage)
+  {
+    buf = detail::BufferStorageStreamBuffer::Create(target, size);
+    if (buf)
+      return buf;
+  }
+
+  const char* vendor = reinterpret_cast<const char*>(glGetString(GL_VENDOR));
+  if (std::strcmp(vendor, "ARM") == 0 || std::strcmp(vendor, "Qualcomm") == 0)
+  {
+    // Mali and Adreno drivers can't do sub-buffer tracking...
+    return detail::BufferDataStreamBuffer::Create(target, size);
+  }
+
+  return detail::BufferSubDataStreamBuffer::Create(target, size);
 }
 
 } // namespace GL

src/common/gl/stream_buffer.h

@@ -6,11 +6,10 @@
 #include <vector>
 
 namespace GL {
-// TODO: Persistent mapping-based implementation
 class StreamBuffer
 {
 public:
-  ~StreamBuffer();
+  virtual ~StreamBuffer();
 
   ALWAYS_INLINE GLuint GetGLBufferId() const { return m_buffer_id; }
   ALWAYS_INLINE GLenum GetGLTarget() const { return m_target; }
@@ -27,18 +26,16 @@ public:
     u32 space_aligned; // remaining space / alignment
   };
 
-  MappingResult Map(u32 alignment, u32 min_size);
+  virtual MappingResult Map(u32 alignment, u32 min_size) = 0;
-  void Unmap(u32 used_size);
+  virtual void Unmap(u32 used_size) = 0;
 
   static std::unique_ptr<StreamBuffer> Create(GLenum target, u32 size);
 
-private:
+protected:
   StreamBuffer(GLenum target, GLuint buffer_id, u32 size);
 
   GLenum m_target;
   GLuint m_buffer_id;
   u32 m_size;
-
-  std::vector<u8> m_cpu_buffer;
 };
 } // namespace GL

src/core/gpu_hw_opengl.cpp

@@ -756,14 +756,14 @@ void GPU_HW_OpenGL::FlushRender()
   if (m_batch.NeedsTwoPassRendering())
   {
     SetDrawState(BatchRenderMode::OnlyTransparent);
-    glDrawArrays(gl_primitives[static_cast<u8>(m_batch.primitive)], 0, vertex_count);
+    glDrawArrays(gl_primitives[static_cast<u8>(m_batch.primitive)], m_batch_base_vertex, vertex_count);
     SetDrawState(BatchRenderMode::OnlyOpaque);
-    glDrawArrays(gl_primitives[static_cast<u8>(m_batch.primitive)], 0, vertex_count);
+    glDrawArrays(gl_primitives[static_cast<u8>(m_batch.primitive)], m_batch_base_vertex, vertex_count);
   }
   else
   {
     SetDrawState(m_batch.GetRenderMode());
-    glDrawArrays(gl_primitives[static_cast<u8>(m_batch.primitive)], 0, vertex_count);
+    glDrawArrays(gl_primitives[static_cast<u8>(m_batch.primitive)], m_batch_base_vertex, vertex_count);
   }
 }