Fix Tracy memory profiling overloads for aligned allocations

Fix disabling renderdoc support
Improve ll_aligned_alloc functions on darwin for 32 and 64byte aligned by utilizing posix_memalign

8 files changed, 58 insertions, 31 deletions

diff --git a/indra/newview/llappviewer.cpp b/indra/newview/llappviewer.cpp
index 9889765fff..84cce2348a 100644
--- a/indra/newview/llappviewer.cpp
+++ b/indra/newview/llappviewer.cpp
@@ -1324,6 +1324,7 @@ bool LLAppViewer::frame()
 bool LLAppViewer::doFrame()
 {
     LL_RECORD_BLOCK_TIME(FTM_FRAME);
+    LL_PROFILE_GPU_ZONE("Frame");
     {
     // and now adjust the visuals from previous frame.
     if(LLPerfStats::tunables.userAutoTuneEnabled && LLPerfStats::tunables.tuningFlag != LLPerfStats::Tunables::Nothing)
@@ -1413,24 +1414,26 @@ bool LLAppViewer::doFrame()
 
         if (!LLApp::isExiting())
         {
-            LL_PROFILE_ZONE_NAMED_CATEGORY_APP("df JoystickKeyboard");
-            pingMainloopTimeout("Main:JoystickKeyboard");
-
-            // Scan keyboard for movement keys.  Command keys and typing
-            // are handled by windows callbacks.  Don't do this until we're
-            // done initializing.  JC
-            if (gViewerWindow
-                && (gHeadlessClient || gViewerWindow->getWindow()->getVisible())
-                && gViewerWindow->getActive()
-                && !gViewerWindow->getWindow()->getMinimized()
-                && LLStartUp::getStartupState() == STATE_STARTED
-                && (gHeadlessClient || !gViewerWindow->getShowProgress())
-                && !gFocusMgr.focusLocked())
             {
-                LLPerfStats::RecordSceneTime T (LLPerfStats::StatType_t::RENDER_IDLE);
-                joystick->scanJoystick();
-                gKeyboard->scanKeyboard();
-                gViewerInput.scanMouse();
+                LL_PROFILE_ZONE_NAMED_CATEGORY_APP("df JoystickKeyboard");
+                pingMainloopTimeout("Main:JoystickKeyboard");
+
+                // Scan keyboard for movement keys.  Command keys and typing
+                // are handled by windows callbacks.  Don't do this until we're
+                // done initializing.  JC
+                if (gViewerWindow
+                    && (gHeadlessClient || gViewerWindow->getWindow()->getVisible())
+                    && gViewerWindow->getActive()
+                    && !gViewerWindow->getWindow()->getMinimized()
+                    && LLStartUp::getStartupState() == STATE_STARTED
+                    && (gHeadlessClient || !gViewerWindow->getShowProgress())
+                    && !gFocusMgr.focusLocked())
+                {
+                    LLPerfStats::RecordSceneTime T(LLPerfStats::StatType_t::RENDER_IDLE);
+                    joystick->scanJoystick();
+                    gKeyboard->scanKeyboard();
+                    gViewerInput.scanMouse();
+                }
             }
 
             // Update state based on messages, user input, object idle.
diff --git a/indra/newview/llappviewerlinux.cpp b/indra/newview/llappviewerlinux.cpp
index 1709970156..89d19d180b 100644
--- a/indra/newview/llappviewerlinux.cpp
+++ b/indra/newview/llappviewerlinux.cpp
@@ -73,6 +73,11 @@ static void exceptionTerminateHandler()
 
 int main( int argc, char **argv )
 {
+    // Call Tracy first thing to have it allocate memory
+    // https://github.com/wolfpld/tracy/issues/196
+    LL_PROFILER_FRAME_END;
+    LL_PROFILER_SET_THREAD_NAME("App");
+
     gArgC = argc;
     gArgV = argv;
 
diff --git a/indra/newview/llappviewermacosx.cpp b/indra/newview/llappviewermacosx.cpp
index 4162c0479a..f497a3cdf3 100644
--- a/indra/newview/llappviewermacosx.cpp
+++ b/indra/newview/llappviewermacosx.cpp
@@ -231,6 +231,11 @@ void infos(const std::string& message)
 
 int main( int argc, char **argv )
 {
+    // Call Tracy first thing to have it allocate memory
+    // https://github.com/wolfpld/tracy/issues/196
+    LL_PROFILER_FRAME_END;
+    LL_PROFILER_SET_THREAD_NAME("App");
+
     // Store off the command line args for use later.
     gArgC = argc;
     gArgV = argv;
diff --git a/indra/newview/lldrawpool.h b/indra/newview/lldrawpool.h
index 1c8864a9df..46696fc4a4 100644
--- a/indra/newview/lldrawpool.h
+++ b/indra/newview/lldrawpool.h
@@ -204,7 +204,7 @@ public:
         NUM_RENDER_TYPES,
     };
 
-    #ifdef LL_PROFILER_ENABLE_RENDER_DOC
+    #if LL_PROFILER_ENABLE_RENDER_DOC
     static inline const char* lookupPassName(U32 pass)
     {
         switch (pass)
@@ -340,7 +340,7 @@ public:
         }
     }
     #else
-    static inline const char* lookupPass(U32 pass) { return ""; }
+    static inline const char* lookupPassName(U32 pass) { return ""; }
     #endif
 
     LLRenderPass(const U32 type);
diff --git a/indra/newview/llheroprobemanager.cpp b/indra/newview/llheroprobemanager.cpp
index e754de2fd1..2ec161c093 100644
--- a/indra/newview/llheroprobemanager.cpp
+++ b/indra/newview/llheroprobemanager.cpp
@@ -81,6 +81,7 @@ void LLHeroProbeManager::update()
     }
 
     LL_PROFILE_ZONE_SCOPED_CATEGORY_DISPLAY;
+    LL_PROFILE_GPU_ZONE("hero manager update");
     llassert(!gCubeSnapshot); // assert a snapshot is not in progress
     if (LLAppViewer::instance()->logoutRequestSent())
     {
@@ -282,6 +283,9 @@ void LLHeroProbeManager::renderProbes()
 // In effect this simulates single-bounce lighting.
 void LLHeroProbeManager::updateProbeFace(LLReflectionMap* probe, U32 face, bool is_dynamic, F32 near_clip)
 {
+    LL_PROFILE_ZONE_SCOPED_CATEGORY_DISPLAY;
+    LL_PROFILE_GPU_ZONE("hero probe update");
+
     // hacky hot-swap of camera specific render targets
     gPipeline.mRT = &gPipeline.mHeroProbeRT;
 
@@ -352,7 +356,7 @@ void LLHeroProbeManager::updateProbeFace(LLReflectionMap* probe, U32 face, bool
 
         for (int i = 0; i < mMipChain.size(); ++i)
         {
-            LL_PROFILE_GPU_ZONE("probe mip");
+            LL_PROFILE_GPU_ZONE("hero probe mip");
             mMipChain[i].bindTarget();
             if (i == 0)
             {
@@ -379,7 +383,7 @@ void LLHeroProbeManager::updateProbeFace(LLReflectionMap* probe, U32 face, bool
 
             if (mip >= 0)
             {
-                LL_PROFILE_GPU_ZONE("probe mip copy");
+                LL_PROFILE_GPU_ZONE("hero probe mip copy");
                 mTexture->bind(0);
 
                 glCopyTexSubImage3D(GL_TEXTURE_CUBE_MAP_ARRAY, mip, 0, 0, sourceIdx * 6 + face, 0, 0, res, res);
@@ -427,7 +431,7 @@ void LLHeroProbeManager::generateRadiance(LLReflectionMap* probe)
 
             for (int i = 0; i < mMipChain.size() / 4; ++i)
             {
-                LL_PROFILE_GPU_ZONE("probe radiance gen");
+                LL_PROFILE_GPU_ZONE("hero probe radiance gen");
                 static LLStaticHashedString sMipLevel("mipLevel");
                 static LLStaticHashedString sRoughness("roughness");
                 static LLStaticHashedString sWidth("u_width");
@@ -474,6 +478,7 @@ void LLHeroProbeManager::updateUniforms()
     }
 
     LL_PROFILE_ZONE_SCOPED_CATEGORY_DISPLAY;
+    LL_PROFILE_GPU_ZONE("hpmu - uniforms")
 
     LLMatrix4a modelview;
     modelview.loadu(gGLModelView);
diff --git a/indra/newview/llreflectionmapmanager.cpp b/indra/newview/llreflectionmapmanager.cpp
index 4760ab376e..ae5ade5f24 100644
--- a/indra/newview/llreflectionmapmanager.cpp
+++ b/indra/newview/llreflectionmapmanager.cpp
@@ -210,6 +210,7 @@ void LLReflectionMapManager::update()
     }
 
     LL_PROFILE_ZONE_SCOPED_CATEGORY_DISPLAY;
+    LL_PROFILE_GPU_ZONE("reflection manager update");
     llassert(!gCubeSnapshot); // assert a snapshot is not in progress
     if (LLAppViewer::instance()->logoutRequestSent())
     {
@@ -696,6 +697,8 @@ void LLReflectionMapManager::doProbeUpdate()
 // In effect this simulates single-bounce lighting.
 void LLReflectionMapManager::updateProbeFace(LLReflectionMap* probe, U32 face)
 {
+    LL_PROFILE_ZONE_SCOPED_CATEGORY_DISPLAY;
+    LL_PROFILE_GPU_ZONE("probe update");
     // hacky hot-swap of camera specific render targets
     gPipeline.mRT = &gPipeline.mAuxillaryRT;
 
@@ -1011,6 +1014,7 @@ void LLReflectionMapManager::updateUniforms()
     }
 
     LL_PROFILE_ZONE_SCOPED_CATEGORY_DISPLAY;
+    LL_PROFILE_GPU_ZONE("rmmu - uniforms")
 
     // structure for packing uniform buffer object
     // see class3/deferred/reflectionProbeF.glsl
diff --git a/indra/newview/llviewerdisplay.cpp b/indra/newview/llviewerdisplay.cpp
index 32019f860d..b34c8600f7 100644
--- a/indra/newview/llviewerdisplay.cpp
+++ b/indra/newview/llviewerdisplay.cpp
@@ -407,6 +407,7 @@ static void update_tp_display(bool minimized)
 void display(bool rebuild, F32 zoom_factor, int subfield, bool for_snapshot)
 {
     LL_PROFILE_ZONE_NAMED_CATEGORY_DISPLAY("Render");
+    LL_PROFILE_GPU_ZONE("Render");
 
     LLPerfStats::RecordSceneTime T (LLPerfStats::StatType_t::RENDER_DISPLAY); // render time capture - This is the main stat for overall rendering.
 
@@ -709,6 +710,7 @@ void display(bool rebuild, F32 zoom_factor, int subfield, bool for_snapshot)
         if (gPipeline.RenderMirrors && !gSnapshot)
         {
             LL_PROFILE_ZONE_NAMED_CATEGORY_DISPLAY("Update hero probes");
+            LL_PROFILE_GPU_ZONE("hero manager")
             gPipeline.mHeroProbeManager.update();
             gPipeline.mHeroProbeManager.renderProbes();
         }
diff --git a/indra/newview/pipeline.cpp b/indra/newview/pipeline.cpp
index 18dd694246..ddfb5a5621 100644
--- a/indra/newview/pipeline.cpp
+++ b/indra/newview/pipeline.cpp
@@ -7182,11 +7182,11 @@ extern LLPointer<LLImageGL> gEXRImage;
 
 void LLPipeline::tonemap(LLRenderTarget* src, LLRenderTarget* dst)
 {
+    LL_PROFILE_GPU_ZONE("tonemap");
+
     dst->bindTarget();
     // gamma correct lighting
     {
-        LL_PROFILE_GPU_ZONE("tonemap");
-
         static LLCachedControl<bool> buildNoPost(gSavedSettings, "RenderDisablePostProcessing", false);
 
         LLGLDepthTest depth(GL_FALSE, GL_FALSE);
@@ -7235,11 +7235,11 @@ void LLPipeline::tonemap(LLRenderTarget* src, LLRenderTarget* dst)
 
 void LLPipeline::gammaCorrect(LLRenderTarget* src, LLRenderTarget* dst)
 {
+    LL_PROFILE_GPU_ZONE("gamma correct");
+
     dst->bindTarget();
     // gamma correct lighting
     {
-        LL_PROFILE_GPU_ZONE("gamma correct");
-
         LLGLDepthTest depth(GL_FALSE, GL_FALSE);
 
         static LLCachedControl<bool> buildNoPost(gSavedSettings, "RenderDisablePostProcessing", false);
@@ -7290,9 +7290,9 @@ void LLPipeline::copyScreenSpaceReflections(LLRenderTarget* src, LLRenderTarget*
 
 void LLPipeline::generateGlow(LLRenderTarget* src)
 {
+    LL_PROFILE_GPU_ZONE("glow generate");
     if (sRenderGlow)
     {
-        LL_PROFILE_GPU_ZONE("glow");
         mGlow[2].bindTarget();
         mGlow[2].clear();
 
@@ -7401,6 +7401,7 @@ void LLPipeline::generateGlow(LLRenderTarget* src)
 void LLPipeline::applyCAS(LLRenderTarget* src, LLRenderTarget* dst)
 {
     static LLCachedControl<F32> cas_sharpness(gSavedSettings, "RenderCASSharpness", 0.4f);
+	LL_PROFILE_GPU_ZONE("cas");
     if (cas_sharpness == 0.0f || !gCASProgram.isComplete())
     {
         gPipeline.copyRenderTarget(src, dst);
@@ -7445,6 +7446,7 @@ void LLPipeline::applyCAS(LLRenderTarget* src, LLRenderTarget* dst)
 
 void LLPipeline::applyFXAA(LLRenderTarget* src, LLRenderTarget* dst)
 {
+	LL_PROFILE_GPU_ZONE("FXAA");
     {
         llassert(!gCubeSnapshot);
         bool multisample = RenderFSAAType == 1 && gFXAAProgram[0].isComplete() && mFXAAMap.isComplete();
@@ -7536,7 +7538,7 @@ void LLPipeline::generateSMAABuffers(LLRenderTarget* src)
     // Present everything.
     if (multisample)
     {
-        LL_PROFILE_GPU_ZONE("aa");
+        LL_PROFILE_GPU_ZONE("SMAA Edge");
         static LLCachedControl<U32> aa_quality(gSavedSettings, "RenderFSAASamples", 0U);
         U32 fsaa_quality = std::clamp(aa_quality(), 0U, 3U);
 
@@ -7648,13 +7650,13 @@ void LLPipeline::generateSMAABuffers(LLRenderTarget* src)
 
 void LLPipeline::applySMAA(LLRenderTarget* src, LLRenderTarget* dst)
 {
+	LL_PROFILE_GPU_ZONE("SMAA");
     llassert(!gCubeSnapshot);
     bool multisample = RenderFSAAType == 2 && gSMAAEdgeDetectProgram[0].isComplete() && mFXAAMap.isComplete() && mSMAABlendBuffer.isComplete();
 
     // Present everything.
     if (multisample)
     {
-        LL_PROFILE_GPU_ZONE("aa");
         static LLCachedControl<U32> aa_quality(gSavedSettings, "RenderFSAASamples", 0U);
         U32 fsaa_quality = std::clamp(aa_quality(), 0U, 3U);
 
@@ -7732,8 +7734,9 @@ void LLPipeline::copyRenderTarget(LLRenderTarget* src, LLRenderTarget* dst)
 
 void LLPipeline::combineGlow(LLRenderTarget* src, LLRenderTarget* dst)
 {
-    // Go ahead and do our glow combine here in our destination.  We blit this later into the front buffer.
+    LL_PROFILE_GPU_ZONE("glow combine");
 
+    // Go ahead and do our glow combine here in our destination.  We blit this later into the front buffer.
     dst->bindTarget();
 
     {
@@ -7752,6 +7755,7 @@ void LLPipeline::combineGlow(LLRenderTarget* src, LLRenderTarget* dst)
 
 void LLPipeline::renderDoF(LLRenderTarget* src, LLRenderTarget* dst)
 {
+	LL_PROFILE_GPU_ZONE("dof");
     {
         bool dof_enabled =
             (RenderDepthOfFieldInEditMode || !LLToolMgr::getInstance()->inBuildMode()) &&
@@ -7762,7 +7766,6 @@ void LLPipeline::renderDoF(LLRenderTarget* src, LLRenderTarget* dst)
 
         if (dof_enabled)
         {
-            LL_PROFILE_GPU_ZONE("dof");
             LLGLDisable blend(GL_BLEND);
 
             // depth of field focal plane calculations