Fix Bug 753: Stuttering caused by lack of GL resource synchronization between Java GL- and CALayer thread ; Simplify / Fix waitUntilRenderSignal().

Stuttering caused by lack of GL resource synchronization between Java GL- and CALayer thread
+ // Required(?) to finish previous GL rendering to give CALayer proper result,
+ // i.e. synchronize both threads each w/ their GLContext sharing same resources.
+ //
+ // FIXME: IMHO this synchronization should be implicitly performed via 'CGL.flushBuffer(contextHandle)' above,
+ // in case this will be determined a driver bug - use a QUIRK entry in GLRendererQuirks!
+ gl.glFinish();

Simplify / Fix waitUntilRenderSignal()
- remove loop and 'ready' condition -> nonsense
- if too later, i.e. lastWaitTime+TO < now, use now+TO as max. vsync waiting time

Bug735 Tests:
  - Make vsync, wait and ECT (exclusive context thread) configurable via main args.
  - Inv2*, Inv3* and Inv4*: Fluent Animation
  - Works w/ ECT

2 files changed, 82 insertions, 33 deletions

diff --git a/src/jogl/classes/jogamp/opengl/macosx/cgl/MacOSXCGLContext.java b/src/jogl/classes/jogamp/opengl/macosx/cgl/MacOSXCGLContext.java
index 976056891..6b086ce44 100644
--- a/src/jogl/classes/jogamp/opengl/macosx/cgl/MacOSXCGLContext.java
+++ b/src/jogl/classes/jogamp/opengl/macosx/cgl/MacOSXCGLContext.java
@@ -81,7 +81,7 @@ import com.jogamp.opengl.util.glsl.ShaderCode;
 import com.jogamp.opengl.util.glsl.ShaderProgram;
 
 public class MacOSXCGLContext extends GLContextImpl
-{
+{  
   // Abstract interface for implementation of this context (either
   // NSOpenGL-based or CGL-based)
   protected interface GLBackendImpl {
@@ -1000,6 +1000,8 @@ public class MacOSXCGLContext extends GLContextImpl
               CGL.setNSOpenGLLayerSwapInterval(l, interval);
               if( 0 < interval ) {
                   vsyncTimeout = interval * screenVSyncTimeout + 1000; // +1ms
+              } else {
+                  vsyncTimeout = 1 * screenVSyncTimeout + 1000; // +1ms
               }
               if(DEBUG) { System.err.println("NS setSwapInterval: "+interval+" -> "+vsyncTimeout+" micros"); }
           }
@@ -1008,6 +1010,14 @@ public class MacOSXCGLContext extends GLContextImpl
       }
       
       private int skipSync=0;
+      /** TODO: Remove after discussion
+      private boolean perfIterReset = false;
+      private int perfIter = 0;
+      private long waitGLS = 0;
+      private long finishGLS = 0;
+      private long frameXS = 0;
+      private long lastFrameStart = 0;
+      */
       
       @Override
       public boolean swapBuffers() {
@@ -1036,6 +1046,45 @@ public class MacOSXCGLContext extends GLContextImpl
                           res = CGL.flushBuffer(contextHandle);
                           if(res) {
                               if(0 == skipSync) {
+                                  /** TODO: Remove after discussion
+                                  perfIter++;
+                                  if( !perfIterReset && 100 == perfIter ) {
+                                      perfIterReset = true;
+                                      perfIter = 1;
+                                      waitGLS = 0;
+                                      finishGLS = 0;
+                                      frameXS = 0;
+                                  }
+                                  final long lastFramePeriod0 = TimeUnit.NANOSECONDS.toMicros(System.nanoTime()) - lastFrameStart;
+                                  gl.glFinish(); // Require to finish previous GL rendering to give CALayer proper result
+                                  final long lastFramePeriod1 = TimeUnit.NANOSECONDS.toMicros(System.nanoTime()) - lastFrameStart;
+                                  
+                                  // If v-sync is disabled, frames will be drawn as quickly as possible w/o delay, 
+                                  // while still synchronizing w/ CALayer.
+                                  // If v-sync is enabled wait until next swap interval (v-sync).
+                                  CGL.waitUntilNSOpenGLLayerIsReady(cmd.nsOpenGLLayer, vsyncTimeout);
+                                  final long lastFramePeriodX = TimeUnit.NANOSECONDS.toMicros(System.nanoTime()) - lastFrameStart;
+                                  
+                                  final long finishGL = lastFramePeriod1 - lastFramePeriod0;
+                                  final long waitGL = lastFramePeriodX - lastFramePeriod1;
+                                  finishGLS += finishGL;
+                                  waitGLS += waitGL;
+                                  frameXS += lastFramePeriodX;
+                                  
+                                  System.err.println("XXX["+perfIter+"] TO "+vsyncTimeout/1000+" ms, "+
+                                                     "lFrame0 "+lastFramePeriod0/1000+" ms, "+
+                                                     "lFrameX "+lastFramePeriodX/1000+" / "+frameXS/1000+" ~"+(frameXS/perfIter)/1000.0+" ms, "+
+                                                     "finishGL "+finishGL/1000+" / "+finishGLS/1000+" ~"+(finishGLS/perfIter)/1000.0+" ms, "+
+                                                     "waitGL "+waitGL/1000+" / "+waitGLS/1000+" ~"+(waitGLS/perfIter)/1000.0+" ms");
+                                  */
+                                  //
+                                  // Required(?) to finish previous GL rendering to give CALayer proper result,
+                                  // i.e. synchronize both threads each w/ their GLContext sharing same resources.
+                                  //
+                                  // FIXME: IMHO this synchronization should be implicitly performed via 'CGL.flushBuffer(contextHandle)' above,
+                                  // in case this will be determined a driver bug - use a QUIRK entry in GLRendererQuirks!
+                                  gl.glFinish();
+                                  
                                   // If v-sync is disabled, frames will be drawn as quickly as possible w/o delay, 
                                   // while still synchronizing w/ CALayer.
                                   // If v-sync is enabled wait until next swap interval (v-sync).
@@ -1050,6 +1099,7 @@ public class MacOSXCGLContext extends GLContextImpl
                                   // trigger CALayer to update incl. possible surface change (new pbuffer handle)
                                   CGL.setNSOpenGLLayerNeedsDisplayPBuffer(cmd.nsOpenGLLayer, drawable.getHandle());                          
                               }
+                              // lastFrameStart = TimeUnit.NANOSECONDS.toMicros(System.nanoTime());
                           }
                       } else {
                           res = true;
diff --git a/src/jogl/native/macosx/MacOSXWindowSystemInterface-calayer.m b/src/jogl/native/macosx/MacOSXWindowSystemInterface-calayer.m
index 6738364a7..046171efc 100644
--- a/src/jogl/native/macosx/MacOSXWindowSystemInterface-calayer.m
+++ b/src/jogl/native/macosx/MacOSXWindowSystemInterface-calayer.m
@@ -818,40 +818,39 @@ static const GLfloat gl_verts[] = {
 
 - (void)waitUntilRenderSignal: (long) to_micros
 {
-    BOOL ready = NO;
-    int wr = 0;
+    struct timespec t0, to_until;
+    BOOL tooLate;
+    int wr;
+    if( 0 >= to_micros ) {
+        to_micros = 16666 + 1000; // defaults to 1/60s + 1ms
+        NSLog(@"MyNSOpenGLContext::waitUntilRenderSignal: to_micros was zero, using defaults");
+    }
     pthread_mutex_lock(&renderLock);
-    SYNC_PRINT("{W %ld us", to_micros);
-    do {
-        if(0 >= swapInterval) {
-            ready = YES;
-        }
-        if(NO == ready) {
-            #ifdef DBG_SYNC
-                struct timespec t0, t1, td, td2;
-                timespec_now(&t0);
-            #endif
-            if( 0 >= to_micros ) {
-                to_micros = 16666 + 1000; // defaults to 1/60s + 1ms
-                NSLog(@"MyNSOpenGLContext::waitUntilRenderSignal: to_micros was zero, using defaults");
-            }
-            struct timespec to_abs = lastWaitTime;
-            timespec_addmicros(&to_abs, to_micros);
-            #ifdef DBG_SYNC
-                timespec_subtract(&td, &to_abs, &t0);
-                fprintf(stderr, ", (%ld) / ", timespec_milliseconds(&td));
-            #endif
-            wr = pthread_cond_timedwait(&renderSignal, &renderLock, &to_abs);
-            #ifdef DBG_SYNC
-                timespec_now(&t1);
-                timespec_subtract(&td, &t1, &t0);
-                timespec_subtract(&td2, &t1, &lastWaitTime);
-                fprintf(stderr, "(%ld) / (%ld) ms", timespec_milliseconds(&td), timespec_milliseconds(&td2));
-            #endif
-            ready = YES;
+    timespec_now(&t0);
+    to_until = lastWaitTime;
+    timespec_addmicros(&to_until, to_micros);
+    tooLate = timespec_compare(&to_until, &t0) < 0;
+    #ifdef DBG_SYNC
+        struct timespec td_until;
+        timespec_subtract(&td_until, &to_until, &t0);
+        SYNC_PRINT("{W %ld ms, to %ld ms, late %d", to_micros/1000, timespec_milliseconds(&td_until), tooLate);
+    #endif
+    if( 0 < swapInterval ) {
+        if( tooLate ) {
+            // adjust!
+            to_until = t0;
+            timespec_addmicros(&to_until, to_micros);
         }
-    } while (NO == ready && 0 == wr) ;
-    SYNC_PRINT("-%d-%d-%d}", shallDraw, wr, ready);
+        wr = pthread_cond_timedwait(&renderSignal, &renderLock, &to_until);
+        #ifdef DBG_SYNC
+            struct timespec t1, td, td2;
+            timespec_now(&t1);
+            timespec_subtract(&td, &t1, &t0);
+            timespec_subtract(&td2, &t1, &lastWaitTime);
+            fprintf(stderr, "(%ld) / (%ld) ms", timespec_milliseconds(&td), timespec_milliseconds(&td2));
+        #endif
+    }
+    SYNC_PRINT("-%d-%d}\n", shallDraw, wr);
     timespec_now(&lastWaitTime);
     pthread_mutex_unlock(&renderLock);
 }