chore: correct diagram

Author: mrshaw01

ggml/7_VulkandBackend.md

@@ -201,57 +201,57 @@ So even though Vulkan is the transport mechanism, the real architectural center
 
 ```mermaid
 flowchart TD
-    A[Application / GGML frontend] --> B[Vulkan backend API]
-    B --> B1[Init backend by device index]
-    B --> B2[Query device count / description / memory]
-    B --> B3[Get Vulkan buffer type]
-    B --> B4[Get pinned host buffer type]
-
-    B --> C[GGML backend registry]
-    C --> D[Vulkan device runtime]
-
-    D --> D1[Physical device + logical device]
-    D --> D2[Vendor / driver / architecture classification]
-    D --> D3[Capability flags]
-    D --> D4[Queue topology]
-    D --> D5[Descriptor set layout]
-    D --> D6[Pipeline collections]
-    D --> D7[Fence + staging resources]
-    D --> D8[Pinned host memory tracking]
-
-    D3 --> E1[FP16 / BF16]
-    D3 --> E2[Integer dot product]
-    D3 --> E3[Subgroup operations]
-    D3 --> E4[Cooperative matrix support]
-    D3 --> E5[UMA / host-memory preference]
-
-    F[Shader sources (*.comp)] --> G[glslc + vulkan-shaders-gen]
-    G --> H[Generated shader header]
-    G --> I[Generated per-shader C++ sources]
-
-    H --> J[Shader modules]
+    A["Application / GGML frontend"] --> B["Vulkan backend API"]
+    B --> B1["Init backend by device index"]
+    B --> B2["Query device count / description / memory"]
+    B --> B3["Get Vulkan buffer type"]
+    B --> B4["Get pinned host buffer type"]
+
+    B --> C["GGML backend registry"]
+    C --> D["Vulkan device runtime"]
+
+    D --> D1["Physical device + logical device"]
+    D --> D2["Vendor / driver / architecture classification"]
+    D --> D3["Capability flags"]
+    D --> D4["Queue topology"]
+    D --> D5["Descriptor set layout"]
+    D --> D6["Pipeline collections"]
+    D --> D7["Fence + staging resources"]
+    D --> D8["Pinned host memory tracking"]
+
+    D3 --> E1["FP16 / BF16"]
+    D3 --> E2["Integer dot product"]
+    D3 --> E3["Subgroup operations"]
+    D3 --> E4["Cooperative matrix support"]
+    D3 --> E5["UMA / host-memory preference"]
+
+    F["Shader sources (*.comp)"] --> G["glslc + vulkan-shaders-gen"]
+    G --> H["Generated shader header"]
+    G --> I["Generated per-shader C++ sources"]
+
+    H --> J["Shader modules"]
     I --> J
-    J --> K[Compute pipelines]
-
-    K --> K1[Generic operator pipelines]
-    K --> K2[Matmul pipelines]
-    K --> K3[Dequant + matmul pipelines]
-    K --> K4[Vendor / feature-specific variants]
-
-    D --> L[Buffer system]
-    L --> L1[Device-local Vulkan buffers]
-    L --> L2[Pinned host buffers]
-    L --> L3[Subbuffers / views]
-
-    D --> M[Queue + command subsystem]
-    M --> M1[Compute queue]
-    M --> M2[Transfer queue]
-    M --> M3[Command pools]
-    M --> M4[Command buffers]
-    M --> M5[Submission + synchronization]
-
-    A --> N[GGML computation graph]
-    N --> O[Vulkan graph execution]
+    J --> K["Compute pipelines"]
+
+    K --> K1["Generic operator pipelines"]
+    K --> K2["Matmul pipelines"]
+    K --> K3["Dequant + matmul pipelines"]
+    K --> K4["Vendor / feature-specific variants"]
+
+    D --> L["Buffer system"]
+    L --> L1["Device-local Vulkan buffers"]
+    L --> L2["Pinned host buffers"]
+    L --> L3["Subbuffers / views"]
+
+    D --> M["Queue + command subsystem"]
+    M --> M1["Compute queue"]
+    M --> M2["Transfer queue"]
+    M --> M3["Command pools"]
+    M --> M4["Command buffers"]
+    M --> M5["Submission + synchronization"]
+
+    A --> N["GGML computation graph"]
+    N --> O["Vulkan graph execution"]
     O --> K
     O --> L
     O --> M