Fishbone diagrams to help debug CPU/GPU bound issues based on Bruce Waggoner saving voyager 1 presentation

CPU-GPU Bound Debug.txt

// Inspired By: https://www.youtube.com/watch?v=E6TS1c8KWFA
CPU‐Bound Performance Issues
 \
  \  **Draw Call Count**
   \    \
    \    \  Too many state changes per frame
     \    \  Small batches (lots of tiny draw calls)
      \
       \  **Synchronization**
        \    \
         \    \  CPU stalls waiting on GPU (vkQueueWaitIdle, fences)
          \    \  Overly broad pipeline barriers
           \
            \  **Resource Management Overhead**
             \    \
              \    \  Frequent vkUpdateDescriptorSets / rebinding
               \    \  Excessive vkMapMemory / vkUnmapMemory
                \    \  **Constant / Push‐Constant Writes**
                 \    \    \
                  \    \    \  vkCmdPushConstants calls in inner loop
                   \    \    \  Uniform‐buffer ring stalls (driver waits for in‐flight data)
                    \
                     \  **Wine / Translation Overhead**
                      \    \
                       \    \  Syscall marshalling cost
                        \    \  Translation of D3D calls to Vulkan
                         \
                          \  **Driver Validation / Tracking**
                           \    \
                            \    \  Command buffer recording validation
                             \    \  State‐tracking bookkeeping overhead
                              \
                               \  **Poor Command Buffer Recording**
                                    \
                                     \  Non‐batched barrier/transition calls

================================================================================================

GPU‐Bound Performance Issues
 \
  \  **Shader Complexity**
   \    \
    \    \  Poor LOD / ALU‐heavy ops
     \    \  Divergence / branching
      \
       \  **Memory Bandwidth**
        \    \
         \    \  Texture sampling stalls
          \    \  VRAM contention / tiling issues
           \
            \  **Overdraw & Blending**
             \    \
              \    \  Excessive transparency / blending
               \    \  Too much overdraw (fill‐rate bound)
                \
                 \  **Rasterization Overhead**
                  \    \
                   \    \  Complex multisample / anti‐aliasing
                    \    \  Heavy fragment shading per pixel
                     \
                      \  **Synchronization**
                       \    \
                        \    \  GPU waits on CPU (fences, semaphores)
                         \    \  Render‐pass / subpass barriers
                          \
                           \  **Driver Resource Mgmt**
                            \    \
                             \    \  Inefficient resource transitions
                              \    \  Poor pipeline‐state creation / hashing
                               \    \  Descriptor‐set update overhead
                                \
                                 \  **Constant / Push‐Constant Updates**
                                  \    \
                                   \    \  Push‐constant cache‐flush cost
                                    \    \  GPU stalls waiting on new uniform data
                                     \    \  Hidden pipeline stalls from small constant changes
                                      \
                                       \  **Inefficient Transitions**
                                            (
                                             e.g. too‐fine‐grained vkCmdPipelineBarrier
                                            )