GPUBatch: Implement multidrawindirect and indirect offset

7 files changed, 61 insertions, 13 deletions

diff --git a/source/blender/draw/intern/draw_debug.cc b/source/blender/draw/intern/draw_debug.cc
index b9d10302c1e..ab78db5d913 100644
--- a/source/blender/draw/intern/draw_debug.cc
+++ b/source/blender/draw/intern/draw_debug.cc
@@ -525,14 +525,14 @@ void DebugDraw::display_lines()
   if (gpu_draw_buf_used) {
     GPU_debug_group_begin("GPU");
     GPU_storagebuf_bind(gpu_draw_buf_, slot);
-    GPU_batch_draw_indirect(batch, gpu_draw_buf_);
+    GPU_batch_draw_indirect(batch, gpu_draw_buf_, 0);
     GPU_storagebuf_unbind(gpu_draw_buf_);
     GPU_debug_group_end();
   }
 
   GPU_debug_group_begin("CPU");
   GPU_storagebuf_bind(cpu_draw_buf_, slot);
-  GPU_batch_draw_indirect(batch, cpu_draw_buf_);
+  GPU_batch_draw_indirect(batch, cpu_draw_buf_, 0);
   GPU_storagebuf_unbind(cpu_draw_buf_);
   GPU_debug_group_end();
 
@@ -557,14 +557,14 @@ void DebugDraw::display_prints()
   if (gpu_print_buf_used) {
     GPU_debug_group_begin("GPU");
     GPU_storagebuf_bind(gpu_print_buf_, slot);
-    GPU_batch_draw_indirect(batch, gpu_print_buf_);
+    GPU_batch_draw_indirect(batch, gpu_print_buf_, 0);
     GPU_storagebuf_unbind(gpu_print_buf_);
     GPU_debug_group_end();
   }
 
   GPU_debug_group_begin("CPU");
   GPU_storagebuf_bind(cpu_print_buf_, slot);
-  GPU_batch_draw_indirect(batch, cpu_print_buf_);
+  GPU_batch_draw_indirect(batch, cpu_print_buf_, 0);
   GPU_storagebuf_unbind(cpu_print_buf_);
   GPU_debug_group_end();
 
diff --git a/source/blender/draw/intern/draw_manager_exec.c b/source/blender/draw/intern/draw_manager_exec.c
index 4dda0ceb2ef..0e39cc1d3b9 100644
--- a/source/blender/draw/intern/draw_manager_exec.c
+++ b/source/blender/draw/intern/draw_manager_exec.c
@@ -891,7 +891,7 @@ static void draw_call_indirect(DRWShadingGroup *shgroup,
   }
 
   GPU_batch_set_shader(batch, shgroup->shader);
-  GPU_batch_draw_indirect(batch, indirect_buf);
+  GPU_batch_draw_indirect(batch, indirect_buf, 0);
 }
 
 static void draw_call_batching_start(DRWCommandsState *state)
diff --git a/source/blender/gpu/GPU_batch.h b/source/blender/gpu/GPU_batch.h
index c085b592a77..867d2b5e1bd 100644
--- a/source/blender/gpu/GPU_batch.h
+++ b/source/blender/gpu/GPU_batch.h
@@ -180,7 +180,9 @@ void GPU_batch_draw_advanced(GPUBatch *batch, int v_first, int v_count, int i_fi
  * Issue a draw call using GPU computed arguments. The argument are expected to be valid for the
  * type of geometry drawn (index or non-indexed).
  */
-void GPU_batch_draw_indirect(GPUBatch *batch, GPUStorageBuf *indirect_buf);
+void GPU_batch_draw_indirect(GPUBatch *batch, GPUStorageBuf *indirect_buf, intptr_t offset);
+void GPU_batch_multi_draw_indirect(
+    GPUBatch *batch, GPUStorageBuf *indirect_buf, int count, intptr_t offset, intptr_t stride);
 
 #if 0 /* future plans */
 
diff --git a/source/blender/gpu/intern/gpu_batch.cc b/source/blender/gpu/intern/gpu_batch.cc
index 0b47a7b2952..897f21e8b54 100644
--- a/source/blender/gpu/intern/gpu_batch.cc
+++ b/source/blender/gpu/intern/gpu_batch.cc
@@ -270,13 +270,23 @@ void GPU_batch_draw_advanced(
   batch->draw(v_first, v_count, i_first, i_count);
 }
 
-void GPU_batch_draw_indirect(GPUBatch *gpu_batch, GPUStorageBuf *indirect_buf)
+void GPU_batch_draw_indirect(GPUBatch *gpu_batch, GPUStorageBuf *indirect_buf, intptr_t offset)
 {
   BLI_assert(Context::get()->shader != nullptr);
   BLI_assert(indirect_buf != nullptr);
   Batch *batch = static_cast<Batch *>(gpu_batch);
 
-  batch->draw_indirect(indirect_buf);
+  batch->draw_indirect(indirect_buf, offset);
+}
+
+void GPU_batch_multi_draw_indirect(
+    GPUBatch *gpu_batch, GPUStorageBuf *indirect_buf, int count, intptr_t offset, intptr_t stride)
+{
+  BLI_assert(Context::get()->shader != nullptr);
+  BLI_assert(indirect_buf != nullptr);
+  Batch *batch = static_cast<Batch *>(gpu_batch);
+
+  batch->multi_draw_indirect(indirect_buf, count, offset, stride);
 }
 
 /** \} */
diff --git a/source/blender/gpu/intern/gpu_batch_private.hh b/source/blender/gpu/intern/gpu_batch_private.hh
index 8ca19884fd7..59646925d68 100644
--- a/source/blender/gpu/intern/gpu_batch_private.hh
+++ b/source/blender/gpu/intern/gpu_batch_private.hh
@@ -29,7 +29,11 @@ class Batch : public GPUBatch {
   virtual ~Batch() = default;
 
   virtual void draw(int v_first, int v_count, int i_first, int i_count) = 0;
-  virtual void draw_indirect(GPUStorageBuf *indirect_buf) = 0;
+  virtual void draw_indirect(GPUStorageBuf *indirect_buf, intptr_t offset) = 0;
+  virtual void multi_draw_indirect(GPUStorageBuf *indirect_buf,
+                                   int count,
+                                   intptr_t offset,
+                                   intptr_t stride) = 0;
 
   /* Convenience casts. */
   IndexBuf *elem_() const
diff --git a/source/blender/gpu/opengl/gl_batch.cc b/source/blender/gpu/opengl/gl_batch.cc
index 4ec86b98cbe..ff8867fe3e6 100644
--- a/source/blender/gpu/opengl/gl_batch.cc
+++ b/source/blender/gpu/opengl/gl_batch.cc
@@ -327,12 +327,13 @@ void GLBatch::draw(int v_first, int v_count, int i_first, int i_count)
   }
 }
 
-void GLBatch::draw_indirect(GPUStorageBuf *indirect_buf)
+void GLBatch::draw_indirect(GPUStorageBuf *indirect_buf, intptr_t offset)
 {
   GL_CHECK_RESOURCES("Batch");
 
   this->bind(0);
 
+  /* TODO(fclem): Make the barrier and binding optional if consecutive draws are issued. */
   dynamic_cast<GLStorageBuf *>(unwrap(indirect_buf))->bind_as(GL_DRAW_INDIRECT_BUFFER);
   /* This barrier needs to be here as it only work on the currently bound indirect buffer. */
   glMemoryBarrier(GL_COMMAND_BARRIER_BIT);
@@ -341,10 +342,37 @@ void GLBatch::draw_indirect(GPUStorageBuf *indirect_buf)
   if (elem) {
     const GLIndexBuf *el = this->elem_();
     GLenum index_type = to_gl(el->index_type_);
-    glDrawElementsIndirect(gl_type, index_type, (GLvoid *)nullptr);
+    glDrawElementsIndirect(gl_type, index_type, (GLvoid *)offset);
   }
   else {
-    glDrawArraysIndirect(gl_type, (GLvoid *)nullptr);
+    glDrawArraysIndirect(gl_type, (GLvoid *)offset);
+  }
+  /* Unbind. */
+  glBindBuffer(GL_DRAW_INDIRECT_BUFFER, 0);
+}
+
+void GLBatch::multi_draw_indirect(GPUStorageBuf *indirect_buf,
+                                  int count,
+                                  intptr_t offset,
+                                  intptr_t stride)
+{
+  GL_CHECK_RESOURCES("Batch");
+
+  this->bind(0);
+
+  /* TODO(fclem): Make the barrier and binding optional if consecutive draws are issued. */
+  dynamic_cast<GLStorageBuf *>(unwrap(indirect_buf))->bind_as(GL_DRAW_INDIRECT_BUFFER);
+  /* This barrier needs to be here as it only work on the currently bound indirect buffer. */
+  glMemoryBarrier(GL_COMMAND_BARRIER_BIT);
+
+  GLenum gl_type = to_gl(prim_type);
+  if (elem) {
+    const GLIndexBuf *el = this->elem_();
+    GLenum index_type = to_gl(el->index_type_);
+    glMultiDrawElementsIndirect(gl_type, index_type, (GLvoid *)offset, count, stride);
+  }
+  else {
+    glMultiDrawArraysIndirect(gl_type, (GLvoid *)offset, count, stride);
   }
   /* Unbind. */
   glBindBuffer(GL_DRAW_INDIRECT_BUFFER, 0);
diff --git a/source/blender/gpu/opengl/gl_batch.hh b/source/blender/gpu/opengl/gl_batch.hh
index 0d7ea7c4a9e..714aa1220be 100644
--- a/source/blender/gpu/opengl/gl_batch.hh
+++ b/source/blender/gpu/opengl/gl_batch.hh
@@ -91,7 +91,11 @@ class GLBatch : public Batch {
 
  public:
   void draw(int v_first, int v_count, int i_first, int i_count) override;
-  void draw_indirect(GPUStorageBuf *indirect_buf) override;
+  void draw_indirect(GPUStorageBuf *indirect_buf, intptr_t offset) override;
+  void multi_draw_indirect(GPUStorageBuf *indirect_buf,
+                           int count,
+                           intptr_t offset,
+                           intptr_t stride) override;
   void bind(int i_first);
 
   /* Convenience getters. */