Faster continuous (#847)

Re-organized fast body handling in solver, cutting the cost of
continuous in half.
Fixes for Linux samples.
b2World_GetAwakeBodyCount
Removed unused code from dynamic tree

benchmark/amd7950x_avx2/joint_grid.csv

@@ -1,9 +1,9 @@
 threads,fps
-1,200.145
-2,340.128
-3,496.707
-4,607.243
-5,698.173
-6,782.174
-7,856.148
-8,916.514
+1,197.415
+2,338.192
+3,485.363
+4,590.58
+5,678.894
+6,755.337
+7,830.03
+8,889.992

benchmark/amd7950x_avx2/large_pyramid.csv

@@ -1,9 +1,9 @@
 threads,fps
-1,325.321
-2,615.892
-3,878.262
-4,1111.8
-5,1322.39
-6,1511.06
-7,1719
-8,1738.61
+1,337.398
+2,599.766
+3,853.652
+4,1095.24
+5,1288.22
+6,1467.1
+7,1669.81
+8,1704.72

benchmark/amd7950x_avx2/many_pyramids.csv

@@ -1,9 +1,9 @@
 threads,fps
-1,83.1212
-2,162.42
-3,238.035
-4,307.18
-5,374.76
-6,432.28
-7,501.138
-8,544.638
+1,82.5412
+2,156.573
+3,233.64
+4,294.31
+5,357.121
+6,420.298
+7,487.163
+8,537.355

benchmark/amd7950x_avx2/rain.csv

@@ -1,9 +1,9 @@
 threads,fps
-1,154.443
-2,249.524
-3,323.318
-4,384.953
-5,434.035
-6,483.406
-7,527.558
-8,556.645
+1,158.379
+2,253.381
+3,337.197
+4,408.427
+5,475.14
+6,530.178
+7,583.798
+8,633.102

benchmark/amd7950x_avx2/smash.csv

@@ -1,9 +1,9 @@
 threads,fps
-1,197.528
-2,300.994
-3,385.163
-4,455.355
-5,508.735
-6,553.648
-7,589.206
-8,614.934
+1,194.531
+2,299.76
+3,389.501
+4,469.022
+5,532.045
+6,580.977
+7,620.559
+8,655.217

benchmark/amd7950x_avx2/spinner.csv

@@ -1,9 +1,9 @@
 threads,fps
-1,346.759
-2,538.038
-3,710.87
-4,871.698
-5,983.994
-6,1084.08
-7,1178.67
-8,1242.13
+1,342.697
+2,549.335
+3,711.477
+4,889.473
+5,1011.82
+6,1141.73
+7,1261.39
+8,1338.51

benchmark/amd7950x_avx2/tumbler.csv

@@ -1,9 +1,9 @@
 threads,fps
-1,464.164
-2,705.539
-3,932.064
-4,1133.55
-5,1297.75
-6,1444.39
-7,1563.34
-8,1633.51
+1,461.248
+2,726.229
+3,947.347
+4,1152.31
+5,1335.66
+6,1483.59
+7,1613.59
+8,1715.86

benchmark/amd7950x_avx2_nc/joint_grid.csv

@@ -1,9 +1,9 @@
 threads,fps
-1,196.348
-2,335.206
-3,493.878
-4,598.194
-5,691.697
-6,776.717
-7,848.272
-8,910.073
+1,191.305
+2,335.535
+3,485.589
+4,591.62
+5,682.079
+6,762.346
+7,834.071
+8,896.952

benchmark/amd7950x_avx2_nc/large_pyramid.csv

@@ -1,9 +1,9 @@
 threads,fps
-1,326.839
-2,605.397
-3,874.098
-4,1108.62
-5,1311.67
-6,1500.33
-7,1684.32
-8,1736.04
+1,337.272
+2,606.512
+3,867.572
+4,1096.58
+5,1297.5
+6,1495.5
+7,1668.46
+8,1700.88

benchmark/amd7950x_avx2_nc/many_pyramids.csv

@@ -1,9 +1,9 @@
 threads,fps
-1,84.1153
-2,163.509
-3,240.109
-4,302.481
-5,368.738
-6,428.312
-7,498.302
-8,545.101
+1,82.3111
+2,160.585
+3,236.095
+4,300.733
+5,362.227
+6,423.2
+7,496.346
+8,544.392

benchmark/amd7950x_avx2_nc/rain.csv

@@ -1,9 +1,9 @@
 threads,fps
-1,161.863
-2,276.688
-3,358.757
-4,434.589
-5,500.599
-6,563.324
-7,620.21
-8,672.701
+1,161.455
+2,275.894
+3,354.968
+4,429.525
+5,495.056
+6,556.892
+7,612.256
+8,664.778

benchmark/amd7950x_avx2_nc/smash.csv

@@ -1,9 +1,9 @@
 threads,fps
-1,190.875
-2,302.452
-3,391.787
-4,472.507
-5,531.519
-6,582.197
-7,623.562
-8,654.127
+1,191.697
+2,296.907
+3,387.265
+4,466.151
+5,524.415
+6,571.353
+7,614.959
+8,650.434

benchmark/amd7950x_avx2_nc/spinner.csv

@@ -1,9 +1,9 @@
 threads,fps
-1,344.368
-2,562.345
-3,738.154
-4,916.335
-5,1049.29
-6,1167.4
-7,1280.44
-8,1362.46
+1,339.941
+2,558.533
+3,730.695
+4,913.45
+5,1038.29
+6,1153.58
+7,1265.97
+8,1355.59

benchmark/amd7950x_avx2_nc/tumbler.csv

@@ -1,9 +1,9 @@
 threads,fps
-1,449.556
-2,717.776
-3,934.15
-4,1136.18
-5,1315.11
-6,1470.1
-7,1599.57
-8,1700.46
+1,440.223
+2,707.306
+3,929.712
+4,1127.08
+5,1313.66
+6,1461.68
+7,1588.84
+8,1690.27

benchmark/main.c

@@ -62,6 +62,8 @@ int GetNumberOfCores()
 	return (int)sysconf( _SC_NPROCESSORS_ONLN );
 #elif defined( __linux__ )
 	return (int)sysconf( _SC_NPROCESSORS_ONLN );
+#elif defined( __EMSCRIPTEN__ )
+	return (int)sysconf( _SC_NPROCESSORS_ONLN );
 #else
 	return 1;
 #endif

build.sh

@@ -4,5 +4,8 @@
 rm -rf build
 mkdir build
 cd build
-cmake -DBOX2D_BUILD_DOCS=OFF ..
+
+# I haven't been able to get Wayland working on WSL but X11 works.
+# https://www.glfw.org/docs/latest/compile.html
+cmake -DBOX2D_BUILD_DOCS=OFF -DGLFW_BUILD_WAYLAND=OFF ..
 cmake --build .

include/box2d/base.h

@@ -15,7 +15,7 @@
 	// using the Windows DLL
 	#define BOX2D_EXPORT __declspec( dllimport )
 #elif defined( box2d_EXPORTS )
-	// building or using the Box2D shared library
+	// building or using the shared library
 	#define BOX2D_EXPORT __attribute__( ( visibility( "default" ) ) )
 #else
 	// static library
@@ -66,6 +66,30 @@ B2_API int b2GetByteCount( void );
 /// @param assertFcn a non-null assert callback
 B2_API void b2SetAssertFcn( b2AssertFcn* assertFcn );
 
+// see https://github.com/scottt/debugbreak
+#if defined( _MSC_VER )
+#define B2_BREAKPOINT __debugbreak()
+#elif defined( __GNUC__ ) || defined( __clang__ )
+#define B2_BREAKPOINT __builtin_trap()
+#else
+// Unknown compiler
+#include <assert.h>
+#define B2_BREAKPOINT assert( 0 )
+#endif
+
+#if !defined( NDEBUG ) || defined( B2_ENABLE_ASSERT )
+B2_API int b2InternalAssertFcn( const char* condition, const char* fileName, int lineNumber );
+#define B2_ASSERT( condition )                                                                                                   \
+	do                                                                                                                           \
+	{                                                                                                                            \
+		if ( !( condition ) && b2InternalAssertFcn( #condition, __FILE__, (int)__LINE__ ) )                                          \
+			B2_BREAKPOINT;                                                                                                       \
+	}                                                                                                                            \
+	while ( 0 )
+#else
+#define B2_ASSERT( ... ) ( (void)0 )
+#endif
+
 /// Version numbering scheme.
 /// See https://semver.org/
 typedef struct b2Version

include/box2d/box2d.h

@@ -183,6 +183,9 @@ B2_API void b2World_EnableWarmStarting( b2WorldId worldId, bool flag );
 /// Is constraint warm starting enabled?
 B2_API bool b2World_IsWarmStartingEnabled( b2WorldId worldId );
 
+/// Get the number of awake bodies.
+B2_API int b2World_GetAwakeBodyCount( b2WorldId worldId );
+
 /// Get the current world performance profile
 B2_API b2Profile b2World_GetProfile( b2WorldId worldId );
 
@@ -444,7 +447,10 @@ B2_API int b2Body_GetJoints( b2BodyId bodyId, b2JointId* jointArray, int capacit
 /// Get the maximum capacity required for retrieving all the touching contacts on a body
 B2_API int b2Body_GetContactCapacity( b2BodyId bodyId );
 
-/// Get the touching contact data for a body
+/// Get the touching contact data for a body.
+/// @note Box2D uses speculative collision so some contact points may be separated.
+/// @returns the number of elements filled in the provided array
+/// @warning do not ignore the return value, it specifies the valid number of elements
 B2_API int b2Body_GetContactData( b2BodyId bodyId, b2ContactData* contactData, int capacity );
 
 /// Get the current world AABB that contains all the attached shapes. Note that this may not encompass the body origin.

include/box2d/collision.h

@@ -492,9 +492,11 @@ B2_API b2TOIOutput b2TimeOfImpact( const b2TOIInput* input );
  * @{
  */
 
-/// A manifold point is a contact point belonging to a contact
-/// manifold. It holds details related to the geometry and dynamics
-/// of the contact points.
+/// A manifold point is a contact point belonging to a contact manifold.
+/// It holds details related to the geometry and dynamics of the contact points.
+/// Box2D uses speculative collision so some contact points may be separated.
+/// You may use the maxNormalImpulse to determine if there was an interaction during
+/// the time step.
 typedef struct b2ManifoldPoint
 {
 	/// Location of the contact point in world space. Subject to precision loss at large coordinates.
@@ -518,8 +520,8 @@ typedef struct b2ManifoldPoint
 	/// The friction impulse
 	float tangentImpulse;
 
-	/// The maximum normal impulse applied during sub-stepping
-	/// This could be a bool to indicate the point is confirmed (may be a speculative point)
+	/// The maximum normal impulse applied during sub-stepping. This is important
+	/// to identify speculative contact points that had an interaction in the time step.
 	float maxNormalImpulse;
 
 	/// Relative normal velocity pre-solve. Used for hit events. If the normal impulse is
@@ -533,7 +535,8 @@ typedef struct b2ManifoldPoint
 	bool persisted;
 } b2ManifoldPoint;
 
-/// A contact manifold describes the contact points between colliding shapes
+/// A contact manifold describes the contact points between colliding shapes.
+/// @note Box2D uses speculative collision so some contact points may be separated.
 typedef struct b2Manifold
 {
 	/// The manifold points, up to two are possible in 2D
@@ -779,27 +782,15 @@ B2_API void b2DynamicTree_Validate( const b2DynamicTree* tree );
 /// called often.
 B2_API int b2DynamicTree_GetHeight( const b2DynamicTree* tree );
 
-/// Get the maximum balance of the tree. The balance is the difference in height of the two children of a node.
-B2_API int b2DynamicTree_GetMaxBalance( const b2DynamicTree* tree );
-
 /// Get the ratio of the sum of the node areas to the root area.
 B2_API float b2DynamicTree_GetAreaRatio( const b2DynamicTree* tree );
 
-/// Build an optimal tree. Very expensive. For testing.
-B2_API void b2DynamicTree_RebuildBottomUp( b2DynamicTree* tree );
-
 /// Get the number of proxies created
 B2_API int b2DynamicTree_GetProxyCount( const b2DynamicTree* tree );
 
 /// Rebuild the tree while retaining subtrees that haven't changed. Returns the number of boxes sorted.
 B2_API int b2DynamicTree_Rebuild( b2DynamicTree* tree, bool fullBuild );
 
-/// Shift the world origin. Useful for large worlds.
-/// The shift formula is: position -= newOrigin
-/// @param tree the tree to shift
-/// @param newOrigin the new origin with respect to the old origin
-B2_API void b2DynamicTree_ShiftOrigin( b2DynamicTree* tree, b2Vec2 newOrigin );
-
 /// Get the number of bytes used by this tree
 B2_API int b2DynamicTree_GetByteCount( const b2DynamicTree* tree );