Bart/bullet3
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

Adds multithreading support for batch ray casts.

Browse Source 
To enable the feature, enable the BULLET2_MULTITHREADING option.

Increases the number of rays that can go in a batch request by storing
them in the shared memory stream instead of the shared memory command.
Adds the API b3RaycastBatchSetNumThreads to specify the number of
threads to use for the raycast batch, also adds the argument numThreads
to the pybullet function rayTestBatch.
Rays are distributed among the threads in a greedy fashion there's a shared
queue of work, once a thread finishes its task, it picks the next
available ray from the task. This works better than pre-distributing the
rays among threads, since there's a large variance in computation time per ray.

Some controversial changes:
- Added a pointer to PhysicsClient to the SharedMemoryCommand struct, this
was necessary to keep the C-API the same for b3RaycastBatchAddRay, while
adding the ray to the shared memory stream instead of the command
struct. I think this may be useful to simplify other APIs as well, that
take both a client handle and a command handle.
- Moved #define SHARED_MEMORY_MAX_STREAM_CHUNK_SIZE to
SharedMemoryPublic. This was necessary for the definition of
MAX_RAY_INTERSECTION_BATCH_SIZE.
This commit is contained in:
Tigran Gasparian
2018-06-15 16:47:04 +02:00
parent 997211650e
commit 08409cae9f
13 changed files with 213 additions and 75 deletions
Download Patch File Download Diff File Expand all files Collapse all files

173
examples/SharedMemory/PhysicsServerCommandProcessor.cpp
View File

@@ -1565,7 +1565,7 @@ struct PhysicsServerCommandProcessorInternalData
b3ResizablePool< InternalBodyHandle > m_bodyHandles;
b3ResizablePool<InternalCollisionShapeHandle> m_userCollisionShapeHandles;
b3ResizablePool<InternalVisualShapeHandle> m_userVisualShapeHandles;
b3PluginManager m_pluginManager;
@@ -1780,6 +1780,13 @@ PhysicsServerCommandProcessor::PhysicsServerCommandProcessor()
createEmptyDynamicsWorld();
if (btGetTaskScheduler() == 0) {
btITaskScheduler *scheduler = btCreateDefaultTaskScheduler();
if (scheduler == 0) {
scheduler = btGetSequentialTaskScheduler();
}
btSetTaskScheduler(scheduler);
}
}
PhysicsServerCommandProcessor::~PhysicsServerCommandProcessor()
@@ -4681,31 +4688,87 @@ bool PhysicsServerCommandProcessor::processRequestKeyboardEventsCommand(const st
return hasStatus;
}
bool PhysicsServerCommandProcessor::processRequestRaycastIntersectionsCommand(const struct SharedMemoryCommand& clientCmd, struct SharedMemoryStatus& serverStatusOut, char* bufferServerToClient, int bufferSizeInBytes)
{
bool hasStatus = true;
BT_PROFILE("CMD_REQUEST_RAY_CAST_INTERSECTIONS");
serverStatusOut.m_raycastHits.m_numRaycastHits = 0;
for (int ray=0;ray<clientCmd.m_requestRaycastIntersections.m_numRays;ray++)
#if __cplusplus >= 201103L
#include <atomic>
struct CastSyncInfo {
std::atomic<int> m_nextTaskNumber;
CastSyncInfo() : m_nextTaskNumber(0) {}
inline int getNextTask() {
return m_nextTaskNumber++;
}
};
#else // __cplusplus >= 201103L
struct CastSyncInfo {
volatile int m_nextTaskNumber;
btSpinMutex m_taskLock;
CastSyncInfo() : m_nextTaskNumber(0) {}
inline int getNextTask() {
m_taskLock.lock();
const int taskNr = m_nextTaskNumber++;
m_taskLock.unlock();
return taskNr;
}
};
#endif // __cplusplus >= 201103L
struct BatchRayCaster : public btIParallelForBody
{
CastSyncInfo *m_syncInfo;
const btCollisionWorld *m_world;
const b3RayData *m_rayInputBuffer;
b3RayHitInfo *m_hitInfoOutputBuffer;
int m_numRays;
BatchRayCaster(const btCollisionWorld* world, const b3RayData *rayInputBuffer, b3RayHitInfo *hitInfoOutputBuffer, int numRays)
: m_world(world), m_rayInputBuffer(rayInputBuffer), m_hitInfoOutputBuffer(hitInfoOutputBuffer), m_numRays(numRays) {
m_syncInfo = new CastSyncInfo;
}
~BatchRayCaster() {
delete m_syncInfo;
}
void castRays(int numWorkers) {
#if BT_THREADSAFE
btParallelFor(0, numWorkers, 1, *this);
#else // BT_THREADSAFE
for (int i = 0; i < m_numRays; i++) {
processRay(i);
}
#endif // BT_THREADSAFE
}
void forLoop( int iBegin, int iEnd ) const
{
btVector3 rayFromWorld(clientCmd.m_requestRaycastIntersections.m_rayFromPositions[ray][0],
clientCmd.m_requestRaycastIntersections.m_rayFromPositions[ray][1],
clientCmd.m_requestRaycastIntersections.m_rayFromPositions[ray][2]);
btVector3 rayToWorld(clientCmd.m_requestRaycastIntersections.m_rayToPositions[ray][0],
clientCmd.m_requestRaycastIntersections.m_rayToPositions[ray][1],
clientCmd.m_requestRaycastIntersections.m_rayToPositions[ray][2]);
while(true) {
const int taskNr = m_syncInfo->getNextTask();
if (taskNr >= m_numRays)
return;
processRay(taskNr);
}
}
void processRay(int ray) const {
const double *from = m_rayInputBuffer[ray].m_rayFromPosition;
const double *to = m_rayInputBuffer[ray].m_rayToPosition;
btVector3 rayFromWorld(from[0], from[1], from[2]);
btVector3 rayToWorld(to[0], to[1], to[2]);
btCollisionWorld::ClosestRayResultCallback rayResultCallback(rayFromWorld,rayToWorld);
rayResultCallback.m_flags |= btTriangleRaycastCallback::kF_UseGjkConvexCastRaytest;
m_data->m_dynamicsWorld->rayTest(rayFromWorld,rayToWorld,rayResultCallback);
int rayHits = serverStatusOut.m_raycastHits.m_numRaycastHits;
m_world->rayTest(rayFromWorld,rayToWorld,rayResultCallback);
b3RayHitInfo& hit = m_hitInfoOutputBuffer[ray];
if (rayResultCallback.hasHit())
{
serverStatusOut.m_raycastHits.m_rayHits[rayHits].m_hitFraction
= rayResultCallback.m_closestHitFraction;
hit.m_hitFraction = rayResultCallback.m_closestHitFraction;
int objectUniqueId = -1;
int linkIndex = -1;
@@ -4724,39 +4787,61 @@ bool PhysicsServerCommandProcessor::processRequestRaycastIntersectionsCommand(co
}
}
serverStatusOut.m_raycastHits.m_rayHits[rayHits].m_hitObjectUniqueId
= objectUniqueId;
serverStatusOut.m_raycastHits.m_rayHits[rayHits].m_hitObjectLinkIndex
= linkIndex;
hit.m_hitObjectUniqueId = objectUniqueId;
hit.m_hitObjectLinkIndex = linkIndex;
serverStatusOut.m_raycastHits.m_rayHits[rayHits].m_hitPositionWorld[0]
= rayResultCallback.m_hitPointWorld[0];
serverStatusOut.m_raycastHits.m_rayHits[rayHits].m_hitPositionWorld[1]
= rayResultCallback.m_hitPointWorld[1];
serverStatusOut.m_raycastHits.m_rayHits[rayHits].m_hitPositionWorld[2]
= rayResultCallback.m_hitPointWorld[2];
serverStatusOut.m_raycastHits.m_rayHits[rayHits].m_hitNormalWorld[0]
= rayResultCallback.m_hitNormalWorld[0];
serverStatusOut.m_raycastHits.m_rayHits[rayHits].m_hitNormalWorld[1]
= rayResultCallback.m_hitNormalWorld[1];
serverStatusOut.m_raycastHits.m_rayHits[rayHits].m_hitNormalWorld[2]
= rayResultCallback.m_hitNormalWorld[2];
hit.m_hitPositionWorld[0] = rayResultCallback.m_hitPointWorld[0];
hit.m_hitPositionWorld[1] = rayResultCallback.m_hitPointWorld[1];
hit.m_hitPositionWorld[2] = rayResultCallback.m_hitPointWorld[2];
hit.m_hitNormalWorld[0] = rayResultCallback.m_hitNormalWorld[0];
hit.m_hitNormalWorld[1] = rayResultCallback.m_hitNormalWorld[1];
hit.m_hitNormalWorld[2] = rayResultCallback.m_hitNormalWorld[2];
} else
{
serverStatusOut.m_raycastHits.m_rayHits[rayHits].m_hitFraction = 1;
serverStatusOut.m_raycastHits.m_rayHits[serverStatusOut.m_raycastHits.m_numRaycastHits].m_hitObjectUniqueId = -1;
serverStatusOut.m_raycastHits.m_rayHits[serverStatusOut.m_raycastHits.m_numRaycastHits].m_hitObjectLinkIndex = -1;
serverStatusOut.m_raycastHits.m_rayHits[rayHits].m_hitPositionWorld[0] = 0;
serverStatusOut.m_raycastHits.m_rayHits[rayHits].m_hitPositionWorld[1] = 0;
serverStatusOut.m_raycastHits.m_rayHits[rayHits].m_hitPositionWorld[2] = 0;
serverStatusOut.m_raycastHits.m_rayHits[rayHits].m_hitNormalWorld[0] = 0;
serverStatusOut.m_raycastHits.m_rayHits[rayHits].m_hitNormalWorld[1] = 0;
serverStatusOut.m_raycastHits.m_rayHits[rayHits].m_hitNormalWorld[2] = 0;
hit.m_hitFraction = 1;
hit.m_hitObjectUniqueId = -1;
hit.m_hitObjectLinkIndex = -1;
hit.m_hitPositionWorld[0] = 0;
hit.m_hitPositionWorld[1] = 0;
hit.m_hitPositionWorld[2] = 0;
hit.m_hitNormalWorld[0] = 0;
hit.m_hitNormalWorld[1] = 0;
hit.m_hitNormalWorld[2] = 0;
}
serverStatusOut.m_raycastHits.m_numRaycastHits++;
}
};
bool PhysicsServerCommandProcessor::processRequestRaycastIntersectionsCommand(const struct SharedMemoryCommand& clientCmd, struct SharedMemoryStatus& serverStatusOut, char* bufferServerToClient, int bufferSizeInBytes)
{
bool hasStatus = true;
BT_PROFILE("CMD_REQUEST_RAY_CAST_INTERSECTIONS");
serverStatusOut.m_raycastHits.m_numRaycastHits = 0;
const int numRays = clientCmd.m_requestRaycastIntersections.m_numRays;
const int numThreads = clientCmd.m_requestRaycastIntersections.m_numThreads;
b3RayData *rayInputBuffer = (b3RayData *)malloc(sizeof(b3RayData) * numRays);
memcpy(rayInputBuffer, bufferServerToClient, sizeof(b3RayData) * numRays);
BatchRayCaster batchRayCaster(m_data->m_dynamicsWorld, rayInputBuffer, (b3RayHitInfo *)bufferServerToClient, numRays);
if (numThreads == 0) {
// When 0 is specified, Bullet can decide how many threads to use.
// About 16 rays per thread seems to work reasonably well.
batchRayCaster.castRays(numRays / 32);
} else if (numThreads == 1) {
// Sequentially trace all rays:
for (int i = 0; i < numRays; i++) {
batchRayCaster.processRay(i);
}
} else {
// Otherwise, just use the user-specified number of threads. This is
// still limited by the number of virtual cores on the machine.
batchRayCaster.castRays(numThreads);
}
free(rayInputBuffer);
serverStatusOut.m_raycastHits.m_numRaycastHits = numRays;
serverStatusOut.m_type = CMD_REQUEST_RAY_CAST_INTERSECTIONS_COMPLETED;
return hasStatus;
}