commit 8c5cbcfedc455259e9f199cfaa2679424c5f8362
author Jon Ashburn <jon@lunarg.com> Thu May 07 10:27:37 2015 -0600
committer Courtney Goeltzenleuchter <courtney@LunarG.com> Wed Jun 17 19:47:01 2015 -0600
tree 27cf36a79552a5c503cc11e2a1df4b764f1ca7c1
parent 857f2441ddf5cb6c9d1ed1c87d8102082357603a

layers: Add initialization of instance dispatch table

Make the layer init  separate from either device or instance dispatch
table init, since these are done at different times.

layers/mem_tracker.cpp

diff --git a/layers/mem_tracker.cpp b/layers/mem_tracker.cpp
index 2e9307d..20a69c3 100644
--- a/layers/mem_tracker.cpp
+++ b/layers/mem_tracker.cpp
@@ -43,8 +43,12 @@
 #include "layers_msg.h"
 
 static VkLayerDispatchTable nextTable;
+static VkLayerInstanceDispatchTable nextInstanceTable;
 static VkBaseLayerObject *pCurObj;
 static LOADER_PLATFORM_THREAD_ONCE_DECLARATION(g_initOnce);
+static LOADER_PLATFORM_THREAD_ONCE_DECLARATION(g_tabDeviceOnce);
+static LOADER_PLATFORM_THREAD_ONCE_DECLARATION(g_tabInstanceOnce);
+
 // TODO : This can be much smarter, using separate locks for separate global data
 static int globalLockInitialized = 0;
 static loader_platform_thread_mutex globalLock;
@@ -771,6 +775,23 @@
     }
 }
 
+// TODO handle multiple GPUs/instances for both instance and device dispatch tables
+static void initDeviceTable(void)
+{
+    PFN_vkGetProcAddr fpNextGPA;
+    fpNextGPA = (PFN_vkGetProcAddr) pCurObj->pGPA;
+    assert(fpNextGPA);
+    layer_initialize_dispatch_table(&nextTable, fpNextGPA, (VkPhysicalDevice) pCurObj->nextObject);
+}
+
+static void initInstanceTable(void)
+{
+    PFN_vkGetInstanceProcAddr fpNextGPA;
+    fpNextGPA = (PFN_vkGetInstanceProcAddr) pCurObj->pGPA;
+    assert(fpNextGPA);
+    layer_init_instance_dispatch_table(&nextInstanceTable, fpNextGPA, (VkInstance) pCurObj->nextObject);
+}
+
 static void initMemTracker(
     void)
 {
@@ -790,14 +811,6 @@
         }
     }
 
-    // initialize Layer dispatch table
-    // TODO handle multiple GPUs
-    PFN_vkGetProcAddr fpNextGPA;
-    fpNextGPA = (PFN_vkGetProcAddr) pCurObj->pGPA;
-    assert(fpNextGPA);
-
-    layer_initialize_dispatch_table(&nextTable, fpNextGPA, (VkPhysicalDevice) pCurObj->nextObject);
-
     if (!globalLockInitialized)
     {
         // TODO/TBD: Need to delete this mutex sometime.  How???  One
@@ -815,8 +828,6 @@
     const VkDeviceCreateInfo *pCreateInfo,
     VkDevice                 *pDevice)
 {
-    pCurObj = (VkBaseLayerObject *) gpu;
-    loader_platform_thread_once(&g_initOnce, initMemTracker);
     VkResult result = nextTable.CreateDevice(gpu, pCreateInfo, pDevice);
     // Save off device in case we need it to create Fences
     globalDevice = *pDevice;
@@ -925,6 +936,7 @@
     {
         pCurObj = (VkBaseLayerObject *)  gpu;
         loader_platform_thread_once(&g_initOnce, initMemTracker);
+        loader_platform_thread_once(&g_tabDeviceOnce, initDeviceTable);
         VkResult result = nextTable.EnumerateLayers(gpu,
             maxStringSize, pLayerCount, pOutLayers, pReserved);
         return result;
@@ -2149,6 +2161,7 @@
     }
     pCurObj = gpuw;
     loader_platform_thread_once(&g_initOnce, initMemTracker);
+    loader_platform_thread_once(&g_tabDeviceOnce, initDeviceTable);
 
     if (!strcmp(funcName, "vkGetProcAddr"))
         return (void *) vkGetProcAddr;
@@ -2311,9 +2324,10 @@
     if (instance == NULL) {
         return NULL;
     }
-    //TODO
-    //pCurObj = instw;
-    //loader_platform_thread_once(&g_initInstanceOnce, initInstanceMemTracker);
+
+    pCurObj = instw;
+    loader_platform_thread_once(&g_initOnce, initMemTracker);
+    loader_platform_thread_once(&g_tabInstanceOnce, initInstanceTable);
 
     if (!strcmp(funcName, "vkGetProcAddr"))
         return (void *) vkGetProcAddr;