{maxwell_3d,buffer_cache}: Implement memory barriers using 3D registers

Drop MemoryBarrier from the buffer cache and use Maxwell3D's register
WaitForIdle.

To implement this on OpenGL we just call glMemoryBarrier with the
necessary bits.

Vulkan lacks this synchronization primitive, so we set an event and
immediately wait for it. This is not a pretty solution, but it's what
Vulkan can do without submitting the current command buffer to the queue
(which ends up being more expensive on the CPU).

src/video_core/renderer_opengl/gl_buffer_cache.cpp

@@ -51,10 +51,6 @@ Buffer OGLBufferCache::CreateBlock(VAddr cpu_addr, std::size_t size) {
     return std::make_shared<CachedBufferBlock>(cpu_addr, size);
 }
 
-void OGLBufferCache::WriteBarrier() {
-    glMemoryBarrier(GL_SHADER_STORAGE_BARRIER_BIT);
-}
-
 GLuint OGLBufferCache::ToHandle(const Buffer& buffer) {
     return buffer->GetHandle();
 }

src/video_core/renderer_opengl/gl_buffer_cache.h

@@ -59,8 +59,6 @@ protected:
 
     GLuint ToHandle(const Buffer& buffer) override;
 
-    void WriteBarrier() override;
-
     void UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
                          const u8* data) override;
 

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -725,6 +725,17 @@ void RasterizerOpenGL::FlushAndInvalidateRegion(VAddr addr, u64 size) {
     InvalidateRegion(addr, size);
 }
 
+void RasterizerOpenGL::WaitForIdle() {
+    // Place a barrier on everything that is not framebuffer related.
+    // This is related to another flag that is not currently implemented.
+    glMemoryBarrier(GL_VERTEX_ATTRIB_ARRAY_BARRIER_BIT | GL_ELEMENT_ARRAY_BARRIER_BIT |
+                    GL_UNIFORM_BARRIER_BIT | GL_TEXTURE_FETCH_BARRIER_BIT |
+                    GL_SHADER_IMAGE_ACCESS_BARRIER_BIT | GL_COMMAND_BARRIER_BIT |
+                    GL_PIXEL_BUFFER_BARRIER_BIT | GL_TEXTURE_UPDATE_BARRIER_BIT |
+                    GL_BUFFER_UPDATE_BARRIER_BIT | GL_TRANSFORM_FEEDBACK_BARRIER_BIT |
+                    GL_SHADER_STORAGE_BARRIER_BIT | GL_QUERY_BUFFER_BARRIER_BIT);
+}
+
 void RasterizerOpenGL::FlushCommands() {
     // Only flush when we have commands queued to OpenGL.
     if (num_queued_commands == 0) {

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -75,6 +75,7 @@ public:
     void SignalSyncPoint(u32 value) override;
     void ReleaseFences() override;
     void FlushAndInvalidateRegion(VAddr addr, u64 size) override;
+    void WaitForIdle() override;
     void FlushCommands() override;
     void TickFrame() override;
     bool AccelerateSurfaceCopy(const Tegra::Engines::Fermi2D::Regs::Surface& src,