Defer normal map generation to reduce GPU model upload stalls by ~50%

Each loadTexture call was generating a normal/height map inline (3 full-image
passes: luminance + blur + Sobel). For models with 15-20 textures this added
30-40ms to the 70ms model upload. Now deferred to a per-frame budget (2/frame
in-game, 10/frame during load screen). Models render without POM until their
normal maps are ready.

include/rendering/character_renderer.hpp

@@ -12,6 +12,7 @@
 #include <string>
 #include <utility>
 #include <future>
+#include <deque>
 
 namespace wowee {
 namespace pipeline { class AssetManager; }
@@ -278,6 +279,7 @@ private:
         uint64_t lastUse = 0;
         bool hasAlpha = false;
         bool colorKeyBlack = false;
+        bool normalMapPending = false;  // deferred normal map generation
     };
     std::unordered_map<std::string, TextureCacheEntry> textureCache;
     std::unordered_map<VkTexture*, bool> textureHasAlphaByPtr_;
@@ -302,6 +304,17 @@ private:
     std::unique_ptr<VkTexture> generateNormalHeightMap(
         const uint8_t* pixels, uint32_t width, uint32_t height, float& outVariance);
 
+    // Deferred normal map generation — avoids stalling loadModel
+    struct PendingNormalMap {
+        std::string cacheKey;
+        std::vector<uint8_t> pixels;  // RGBA pixel data
+        uint32_t width, height;
+    };
+    std::deque<PendingNormalMap> pendingNormalMaps_;
+public:
+    void processPendingNormalMaps(int budget = 2);
+private:
+
     // Normal mapping / POM settings
     bool normalMappingEnabled_ = true;
     float normalMapStrength_ = 0.8f;

src/core/application.cpp

@@ -922,14 +922,20 @@ void Application::update(float deltaTime) {
                 auto t3 = std::chrono::steady_clock::now();
                 processDeferredEquipmentQueue();
                 auto t4 = std::chrono::steady_clock::now();
+                // Process deferred normal maps (2 per frame to spread CPU cost)
+                if (auto* cr = renderer ? renderer->getCharacterRenderer() : nullptr) {
+                    cr->processPendingNormalMaps(2);
+                }
+                auto t5 = std::chrono::steady_clock::now();
                 float pMs = std::chrono::duration<float, std::milli>(t1 - t0).count();
                 float cMs = std::chrono::duration<float, std::milli>(t2 - t1).count();
                 float nMs = std::chrono::duration<float, std::milli>(t3 - t2).count();
                 float eMs = std::chrono::duration<float, std::milli>(t4 - t3).count();
-                float total = pMs + cMs + nMs + eMs;
+                float nmMs = std::chrono::duration<float, std::milli>(t5 - t4).count();
+                float total = pMs + cMs + nMs + eMs + nmMs;
                 if (total > 4.0f) {
                     LOG_WARNING("spawn/equip breakdown: player=", pMs, "ms creature=", cMs,
-                                "ms npcComposite=", nMs, "ms equip=", eMs, "ms");
+                                "ms npcComposite=", nMs, "ms equip=", eMs, "ms normalMaps=", nmMs, "ms");
                 }
             });
             // Self-heal missing creature visuals: if a nearby UNIT exists in
@@ -4250,6 +4256,9 @@ void Application::loadOnlineWorldTerrain(uint32_t mapId, float x, float y, float
             processCreatureSpawnQueue();
             processAsyncNpcCompositeResults();
             processDeferredEquipmentQueue();
+            if (auto* cr = renderer ? renderer->getCharacterRenderer() : nullptr) {
+                cr->processPendingNormalMaps(10);  // higher budget during load screen
+            }
 
             // Process ALL pending game object spawns (no 1-per-frame cap during load screen).
             while (!pendingGameObjectSpawns_.empty()) {

src/rendering/character_renderer.cpp

@@ -687,13 +687,16 @@ VkTexture* CharacterRenderer::loadTexture(const std::string& path) {
     e.hasAlpha = hasAlpha;
     e.colorKeyBlack = colorKeyBlackHint;
 
-    // Generate normal/height map from diffuse texture
-    float nhVariance = 0.0f;
-    auto nhMap = generateNormalHeightMap(blpImage.data.data(), blpImage.width, blpImage.height, nhVariance);
-    if (nhMap) {
-        e.heightMapVariance = nhVariance;
-        e.approxBytes += approxTextureBytesWithMips(blpImage.width, blpImage.height);
-        e.normalHeightMap = std::move(nhMap);
+    // Defer normal/height map generation to avoid stalling loadModel.
+    // Normal maps are generated in processPendingNormalMaps() at a per-frame budget.
+    if (blpImage.width >= 32 && blpImage.height >= 32) {
+        PendingNormalMap pending;
+        pending.cacheKey = key;
+        pending.pixels.assign(blpImage.data.begin(), blpImage.data.end());
+        pending.width = blpImage.width;
+        pending.height = blpImage.height;
+        pendingNormalMaps_.push_back(std::move(pending));
+        e.normalMapPending = true;
     }
 
     textureCacheBytes_ += e.approxBytes;
@@ -705,6 +708,34 @@ VkTexture* CharacterRenderer::loadTexture(const std::string& path) {
     return texPtr;
 }
 
+void CharacterRenderer::processPendingNormalMaps(int budget) {
+    if (pendingNormalMaps_.empty() || !vkCtx_) return;
+
+    int processed = 0;
+    while (!pendingNormalMaps_.empty() && processed < budget) {
+        auto pending = std::move(pendingNormalMaps_.front());
+        pendingNormalMaps_.pop_front();
+
+        auto it = textureCache.find(pending.cacheKey);
+        if (it == textureCache.end()) continue;  // texture was evicted
+
+        float nhVariance = 0.0f;
+        vkCtx_->beginUploadBatch();
+        auto nhMap = generateNormalHeightMap(pending.pixels.data(),
+            pending.width, pending.height, nhVariance);
+        vkCtx_->endUploadBatch();
+
+        if (nhMap) {
+            it->second.heightMapVariance = nhVariance;
+            it->second.approxBytes += approxTextureBytesWithMips(pending.width, pending.height);
+            textureCacheBytes_ += approxTextureBytesWithMips(pending.width, pending.height);
+            it->second.normalHeightMap = std::move(nhMap);
+        }
+        it->second.normalMapPending = false;
+        processed++;
+    }
+}
+
 // Alpha-blend overlay onto composite at (dstX, dstY)
 static void blitOverlay(std::vector<uint8_t>& composite, int compW, int compH,
                          const pipeline::BLPImage& overlay, int dstX, int dstY) {