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

remove dependency on Adl math stuff

Browse Source
This commit is contained in:
erwin coumans
2013-03-17 01:37:25 -07:00
parent d3c80fe160
commit fc5e2ad5ba
5 changed files with 81 additions and 129 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
demo/gpudemo/GpuDemo.h
View File

@@ -34,9 +34,9 @@ public:
:useOpenCL(true), :useOpenCL(true),
preferredOpenCLPlatformIndex(-1), preferredOpenCLPlatformIndex(-1),
preferredOpenCLDeviceIndex(-1), preferredOpenCLDeviceIndex(-1),
arraySizeX(33), arraySizeX(23),
arraySizeY(30 ), arraySizeY(23 ),
arraySizeZ(33), arraySizeZ(23),
m_useConcaveMesh(false), m_useConcaveMesh(false),
gapX(4.3), gapX(4.3),
gapY(2.0), gapY(2.0),

142
opencl/gpu_rigidbody/host/Solver.cpp
View File

@@ -15,7 +15,7 @@ subject to the following restrictions:
#include "Solver.h" #include "Solver.h"
#include "../Stubs/AdlMatrix3x3.h"
#define SOLVER_SETUP_KERNEL_PATH "opencl/gpu_rigidbody/kernels/solverSetup.cl" #define SOLVER_SETUP_KERNEL_PATH "opencl/gpu_rigidbody/kernels/solverSetup.cl"
#define SOLVER_SETUP2_KERNEL_PATH "opencl/gpu_rigidbody/kernels/solverSetup2.cl" #define SOLVER_SETUP2_KERNEL_PATH "opencl/gpu_rigidbody/kernels/solverSetup2.cl"
@@ -35,7 +35,7 @@ subject to the following restrictions:
#include "../kernels/batchingKernels.h" #include "../kernels/batchingKernels.h"
#include "BulletCommon/btQuickprof.h" #include "BulletCommon/btQuickprof.h"
#include "../../parallel_primitives/host/btLauncherCL.h" #include "../../parallel_primitives/host/btLauncherCL.h"
#include "BulletCommon/btVector3.h"
struct SolverDebugInfo struct SolverDebugInfo
{ {
@@ -91,7 +91,7 @@ Solver::Solver(cl_context ctx, cl_device_id device, cl_command_queue queue, int
m_scan = new btPrefixScanCL(ctx,device,queue,N_SPLIT*N_SPLIT); m_scan = new btPrefixScanCL(ctx,device,queue,N_SPLIT*N_SPLIT);
m_search = new btBoundSearchCL(ctx,device,queue,N_SPLIT*N_SPLIT); m_search = new btBoundSearchCL(ctx,device,queue,N_SPLIT*N_SPLIT);
const int sortSize = NEXTMULTIPLEOF( pairCapacity, 512 ); const int sortSize = BTNEXTMULTIPLEOF( pairCapacity, 512 );
m_sortDataBuffer = new btOpenCLArray<btSortData>(ctx,queue,sortSize); m_sortDataBuffer = new btOpenCLArray<btSortData>(ctx,queue,sortSize);
m_contactBuffer = new btOpenCLArray<btContact4>(ctx,queue); m_contactBuffer = new btOpenCLArray<btContact4>(ctx,queue);
@@ -266,21 +266,21 @@ Solver::~Solver()
static static
inline inline
float calcRelVel(const float4& l0, const float4& l1, const float4& a0, const float4& a1, float calcRelVel(const btVector3& l0, const btVector3& l1, const btVector3& a0, const btVector3& a1,
const float4& linVel0, const float4& angVel0, const float4& linVel1, const float4& angVel1) const btVector3& linVel0, const btVector3& angVel0, const btVector3& linVel1, const btVector3& angVel1)
{ {
return dot3F4(l0, linVel0) + dot3F4(a0, angVel0) + dot3F4(l1, linVel1) + dot3F4(a1, angVel1); return btDot(l0, linVel0) + btDot(a0, angVel0) + btDot(l1, linVel1) + btDot(a1, angVel1);
} }
static static
inline inline
void setLinearAndAngular(const float4& n, const float4& r0, const float4& r1, void setLinearAndAngular(const btVector3& n, const btVector3& r0, const btVector3& r1,
float4& linear, float4& angular0, float4& angular1) btVector3& linear, btVector3& angular0, btVector3& angular1)
{ {
linear = -n; linear = -n;
angular0 = -cross3(r0, n); angular0 = -btCross(r0, n);
angular1 = cross3(r1, n); angular1 = btCross(r1, n);
} }
@@ -288,14 +288,15 @@ template<bool JACOBI>
static static
__inline __inline
void solveContact(btGpuConstraint4& cs, void solveContact(btGpuConstraint4& cs,
const float4& posA, float4& linVelA, float4& angVelA, float invMassA, const Matrix3x3& invInertiaA, const btVector3& posA, btVector3& linVelA, btVector3& angVelA, float invMassA, const btMatrix3x3& invInertiaA,
const float4& posB, float4& linVelB, float4& angVelB, float invMassB, const Matrix3x3& invInertiaB, const btVector3& posB, btVector3& linVelB, btVector3& angVelB, float invMassB, const btMatrix3x3& invInertiaB,
float maxRambdaDt[4], float minRambdaDt[4]) float maxRambdaDt[4], float minRambdaDt[4])
{ {
float4 dLinVelA = make_float4(0.f);
float4 dAngVelA = make_float4(0.f); btVector3 dLinVelA; dLinVelA.setZero();
float4 dLinVelB = make_float4(0.f); btVector3 dAngVelA; dAngVelA.setZero();
float4 dAngVelB = make_float4(0.f); btVector3 dLinVelB; dLinVelB.setZero();
btVector3 dAngVelB; dAngVelB.setZero();
for(int ic=0; ic<4; ic++) for(int ic=0; ic<4; ic++)
{ {
@@ -303,12 +304,12 @@ void solveContact(btGpuConstraint4& cs,
if( cs.m_jacCoeffInv[ic] == 0.f ) continue; if( cs.m_jacCoeffInv[ic] == 0.f ) continue;
{ {
float4 angular0, angular1, linear; btVector3 angular0, angular1, linear;
btVector3 r0 = cs.m_worldPos[ic] - (btVector3&)posA; btVector3 r0 = cs.m_worldPos[ic] - (btVector3&)posA;
btVector3 r1 = cs.m_worldPos[ic] - (btVector3&)posB; btVector3 r1 = cs.m_worldPos[ic] - (btVector3&)posB;
setLinearAndAngular( (const float4 &)-cs.m_linear, (const float4 &)r0, (const float4 &)r1, linear, angular0, angular1 ); setLinearAndAngular( (const btVector3 &)-cs.m_linear, (const btVector3 &)r0, (const btVector3 &)r1, linear, angular0, angular1 );
float rambdaDt = calcRelVel((const float4 &)cs.m_linear,(const float4 &) -cs.m_linear, angular0, angular1, float rambdaDt = calcRelVel((const btVector3 &)cs.m_linear,(const btVector3 &) -cs.m_linear, angular0, angular1,
linVelA, angVelA, linVelB, angVelB ) + cs.m_b[ic]; linVelA, angVelA, linVelB, angVelB ) + cs.m_b[ic];
rambdaDt *= cs.m_jacCoeffInv[ic]; rambdaDt *= cs.m_jacCoeffInv[ic];
@@ -316,19 +317,19 @@ void solveContact(btGpuConstraint4& cs,
float prevSum = cs.m_appliedRambdaDt[ic]; float prevSum = cs.m_appliedRambdaDt[ic];
float updated = prevSum; float updated = prevSum;
updated += rambdaDt; updated += rambdaDt;
updated = max2( updated, minRambdaDt[ic] ); updated = btMax( updated, minRambdaDt[ic] );
updated = min2( updated, maxRambdaDt[ic] ); updated = btMin( updated, maxRambdaDt[ic] );
rambdaDt = updated - prevSum; rambdaDt = updated - prevSum;
cs.m_appliedRambdaDt[ic] = updated; cs.m_appliedRambdaDt[ic] = updated;
} }
float4 linImp0 = invMassA*linear*rambdaDt; btVector3 linImp0 = invMassA*linear*rambdaDt;
float4 linImp1 = invMassB*(-linear)*rambdaDt; btVector3 linImp1 = invMassB*(-linear)*rambdaDt;
float4 angImp0 = mtMul1(invInertiaA, angular0)*rambdaDt; btVector3 angImp0 = (invInertiaA* angular0)*rambdaDt;
float4 angImp1 = mtMul1(invInertiaB, angular1)*rambdaDt; btVector3 angImp1 = (invInertiaB* angular1)*rambdaDt;
#ifdef _WIN32 #ifdef _WIN32
btAssert(_finite(linImp0.x)); btAssert(_finite(linImp0.x()));
btAssert(_finite(linImp1.x)); btAssert(_finite(linImp1.x()));
#endif #endif
if( JACOBI ) if( JACOBI )
{ {
@@ -354,64 +355,40 @@ void solveContact(btGpuConstraint4& cs,
linVelB += dLinVelB; linVelB += dLinVelB;
angVelB += dAngVelB; angVelB += dAngVelB;
} }
} }
void btPlaneSpace1 (const float4* n, float4* p, float4* q)
{
if (btFabs(n->z) > SIMDSQRT12) {
// choose p in y-z plane
btScalar a = n->y*n->y + n->z*n->z;
btScalar k = btRecipSqrt (a);
p->x = 0;
p->y = -n->z*k;
p->z = n->y*k;
// set q = n x p
q->x = a*k;
q->y = -n->x*p->z;
q->z = n->x*p->y;
}
else {
// choose p in x-y plane
btScalar a = n->x*n->x + n->y*n->y;
btScalar k = btRecipSqrt (a);
p->x = -n->y*k;
p->y = n->x*k;
p->z = 0;
// set q = n x p
q->x = -n->z*p->y;
q->y = n->z*p->x;
q->z = a*k;
}
}
static static
__inline __inline
void solveFriction(btGpuConstraint4& cs, void solveFriction(btGpuConstraint4& cs,
const float4& posA, float4& linVelA, float4& angVelA, float invMassA, const Matrix3x3& invInertiaA, const btVector3& posA, btVector3& linVelA, btVector3& angVelA, float invMassA, const btMatrix3x3& invInertiaA,
const float4& posB, float4& linVelB, float4& angVelB, float invMassB, const Matrix3x3& invInertiaB, const btVector3& posB, btVector3& linVelB, btVector3& angVelB, float invMassB, const btMatrix3x3& invInertiaB,
float maxRambdaDt[4], float minRambdaDt[4]) float maxRambdaDt[4], float minRambdaDt[4])
{ {
if( cs.m_fJacCoeffInv[0] == 0 && cs.m_fJacCoeffInv[0] == 0 ) return; if( cs.m_fJacCoeffInv[0] == 0 && cs.m_fJacCoeffInv[0] == 0 ) return;
const float4& center = (const float4&)cs.m_center; const btVector3& center = (const btVector3&)cs.m_center;
float4 n = -(const float4&)cs.m_linear; btVector3 n = -(const btVector3&)cs.m_linear;
float4 tangent[2]; btVector3 tangent[2];
#if 1 #if 1
btPlaneSpace1 (&n, &tangent[0],&tangent[1]); btPlaneSpace1 (n, tangent[0],tangent[1]);
#else #else
float4 r = cs.m_worldPos[0]-center; btVector3 r = cs.m_worldPos[0]-center;
tangent[0] = cross3( n, r ); tangent[0] = cross3( n, r );
tangent[1] = cross3( tangent[0], n ); tangent[1] = cross3( tangent[0], n );
tangent[0] = normalize3( tangent[0] ); tangent[0] = normalize3( tangent[0] );
tangent[1] = normalize3( tangent[1] ); tangent[1] = normalize3( tangent[1] );
#endif #endif
float4 angular0, angular1, linear; btVector3 angular0, angular1, linear;
float4 r0 = center - posA; btVector3 r0 = center - posA;
float4 r1 = center - posB; btVector3 r1 = center - posB;
for(int i=0; i<2; i++) for(int i=0; i<2; i++)
{ {
setLinearAndAngular( tangent[i], r0, r1, linear, angular0, angular1 ); setLinearAndAngular( tangent[i], r0, r1, linear, angular0, angular1 );
@@ -423,19 +400,19 @@ void solveContact(btGpuConstraint4& cs,
float prevSum = cs.m_fAppliedRambdaDt[i]; float prevSum = cs.m_fAppliedRambdaDt[i];
float updated = prevSum; float updated = prevSum;
updated += rambdaDt; updated += rambdaDt;
updated = max2( updated, minRambdaDt[i] ); updated = btMax( updated, minRambdaDt[i] );
updated = min2( updated, maxRambdaDt[i] ); updated = btMin( updated, maxRambdaDt[i] );
rambdaDt = updated - prevSum; rambdaDt = updated - prevSum;
cs.m_fAppliedRambdaDt[i] = updated; cs.m_fAppliedRambdaDt[i] = updated;
} }
float4 linImp0 = invMassA*linear*rambdaDt; btVector3 linImp0 = invMassA*linear*rambdaDt;
float4 linImp1 = invMassB*(-linear)*rambdaDt; btVector3 linImp1 = invMassB*(-linear)*rambdaDt;
float4 angImp0 = mtMul1(invInertiaA, angular0)*rambdaDt; btVector3 angImp0 = (invInertiaA* angular0)*rambdaDt;
float4 angImp1 = mtMul1(invInertiaB, angular1)*rambdaDt; btVector3 angImp1 = (invInertiaB* angular1)*rambdaDt;
#ifdef _WIN32 #ifdef _WIN32
btAssert(_finite(linImp0.x)); btAssert(_finite(linImp0.x()));
btAssert(_finite(linImp1.x)); btAssert(_finite(linImp1.x()));
#endif #endif
linVelA += linImp0; linVelA += linImp0;
angVelA += angImp0; angVelA += angImp0;
@@ -444,17 +421,18 @@ void solveContact(btGpuConstraint4& cs,
} }
{ // angular damping for point constraint { // angular damping for point constraint
float4 ab = normalize3( posB - posA ); btVector3 ab = ( posB - posA ).normalized();
float4 ac = normalize3( center - posA ); btVector3 ac = ( center - posA ).normalized();
if( dot3F4( ab, ac ) > 0.95f || (invMassA == 0.f || invMassB == 0.f)) if( btDot( ab, ac ) > 0.95f || (invMassA == 0.f || invMassB == 0.f))
{ {
float angNA = dot3F4( n, angVelA ); float angNA = btDot( n, angVelA );
float angNB = dot3F4( n, angVelB ); float angNB = btDot( n, angVelB );
angVelA -= (angNA*0.1f)*n; angVelA -= (angNA*0.1f)*n;
angVelB -= (angNB*0.1f)*n; angVelB -= (angNB*0.1f)*n;
} }
} }
} }
@@ -465,7 +443,7 @@ struct SolveTask// : public ThreadPool::Task
: m_bodies( bodies ), m_shapes( shapes ), m_constraints( constraints ), m_start( start ), m_nConstraints( nConstraints ), : m_bodies( bodies ), m_shapes( shapes ), m_constraints( constraints ), m_start( start ), m_nConstraints( nConstraints ),
m_solveFriction( true ){} m_solveFriction( true ){}
u16 getType(){ return 0; } unsigned short int getType(){ return 0; }
void run(int tIdx) void run(int tIdx)
{ {
@@ -486,8 +464,8 @@ struct SolveTask// : public ThreadPool::Task
float maxRambdaDt[4] = {FLT_MAX,FLT_MAX,FLT_MAX,FLT_MAX}; float maxRambdaDt[4] = {FLT_MAX,FLT_MAX,FLT_MAX,FLT_MAX};
float minRambdaDt[4] = {0.f,0.f,0.f,0.f}; float minRambdaDt[4] = {0.f,0.f,0.f,0.f};
solveContact<false>( m_constraints[i], (float4&)bodyA.m_pos, (float4&)bodyA.m_linVel, (float4&)bodyA.m_angVel, bodyA.m_invMass, (const Matrix3x3 &)m_shapes[aIdx].m_invInertiaWorld, solveContact<false>( m_constraints[i], (btVector3&)bodyA.m_pos, (btVector3&)bodyA.m_linVel, (btVector3&)bodyA.m_angVel, bodyA.m_invMass, (const btMatrix3x3 &)m_shapes[aIdx].m_invInertiaWorld,
(float4&)bodyB.m_pos, (float4&)bodyB.m_linVel, (float4&)bodyB.m_angVel, bodyB.m_invMass, (const Matrix3x3 &)m_shapes[bIdx].m_invInertiaWorld, (btVector3&)bodyB.m_pos, (btVector3&)bodyB.m_linVel, (btVector3&)bodyB.m_angVel, bodyB.m_invMass, (const btMatrix3x3 &)m_shapes[bIdx].m_invInertiaWorld,
maxRambdaDt, minRambdaDt ); maxRambdaDt, minRambdaDt );
} }
@@ -508,8 +486,8 @@ struct SolveTask// : public ThreadPool::Task
minRambdaDt[j] = -maxRambdaDt[j]; minRambdaDt[j] = -maxRambdaDt[j];
} }
solveFriction( m_constraints[i], (float4&)bodyA.m_pos, (float4&)bodyA.m_linVel, (float4&)bodyA.m_angVel, bodyA.m_invMass,(const Matrix3x3 &) m_shapes[aIdx].m_invInertiaWorld, solveFriction( m_constraints[i], (btVector3&)bodyA.m_pos, (btVector3&)bodyA.m_linVel, (btVector3&)bodyA.m_angVel, bodyA.m_invMass,(const btMatrix3x3 &) m_shapes[aIdx].m_invInertiaWorld,
(float4&)bodyB.m_pos, (float4&)bodyB.m_linVel, (float4&)bodyB.m_angVel, bodyB.m_invMass,(const Matrix3x3 &) m_shapes[bIdx].m_invInertiaWorld, (btVector3&)bodyB.m_pos, (btVector3&)bodyB.m_linVel, (btVector3&)bodyB.m_angVel, bodyB.m_invMass,(const btMatrix3x3 &) m_shapes[bIdx].m_invInertiaWorld,
maxRambdaDt, minRambdaDt ); maxRambdaDt, minRambdaDt );
} }

40
opencl/gpu_rigidbody/host/Solver.h
View File

@@ -19,24 +19,8 @@ subject to the following restrictions:
#include "../../parallel_primitives/host/btOpenCLArray.h" #include "../../parallel_primitives/host/btOpenCLArray.h"
#include "../host/btGpuConstraint4.h" #include "../host/btGpuConstraint4.h"
//#include <AdlPhysics/TypeDefinition.h>
//#include "AdlRigidBody.h"
#include "../../gpu_sat/host/btRigidBodyCL.h" #include "../../gpu_sat/host/btRigidBodyCL.h"
#include "../../gpu_sat/host/btContact4.h" #include "../../gpu_sat/host/btContact4.h"
#include "../Stubs/AdlMath.h"
#include "../Stubs/AdlMatrix3x3.h"
//#include "AdlPhysics/Batching/Batching.h>
#define MYF4 float4
#define MAKE_MYF4 make_float4
//#define MYF4 float4sse
//#define MAKE_MYF4 make_float4sse
#include "../host/btGpuConstraint4.h" #include "../host/btGpuConstraint4.h"
#include "../../parallel_primitives/host/btPrefixScanCL.h" #include "../../parallel_primitives/host/btPrefixScanCL.h"
@@ -46,6 +30,7 @@ subject to the following restrictions:
#include "../../basic_initialize/btOpenCLUtils.h" #include "../../basic_initialize/btOpenCLUtils.h"
#define BTNEXTMULTIPLEOF(num, alignment) (((num)/(alignment) + (((num)%(alignment)==0)?0:1))*(alignment))
class SolverBase class SolverBase
{ {
@@ -56,9 +41,9 @@ class SolverBase
{ {
ConstraintData(): m_b(0.f), m_appliedRambdaDt(0.f) {} ConstraintData(): m_b(0.f), m_appliedRambdaDt(0.f) {}
float4 m_linear; // have to be normalized btVector3 m_linear; // have to be normalized
float4 m_angular0; btVector3 m_angular0;
float4 m_angular1; btVector3 m_angular1;
float m_jacCoeffInv; float m_jacCoeffInv;
float m_b; float m_b;
float m_appliedRambdaDt; float m_appliedRambdaDt;
@@ -67,8 +52,8 @@ class SolverBase
unsigned int m_bodyBPtr; unsigned int m_bodyBPtr;
bool isInvalid() const { return ((unsigned int)m_bodyAPtr+(unsigned int)m_bodyBPtr) == 0; } bool isInvalid() const { return ((unsigned int)m_bodyAPtr+(unsigned int)m_bodyBPtr) == 0; }
float getFrictionCoeff() const { return m_linear.w; } float getFrictionCoeff() const { return m_linear[3]; }
void setFrictionCoeff(float coeff) { m_linear.w = coeff; } void setFrictionCoeff(float coeff) { m_linear[3] = coeff; }
}; };
struct ConstraintCfg struct ConstraintCfg
@@ -135,12 +120,6 @@ class Solver : public SolverBase
virtual ~Solver(); virtual ~Solver();
/* void reorderConvertToConstraints( const btOpenCLArray<btRigidBodyCL>* bodyBuf,
const btOpenCLArray<btInertiaCL>* shapeBuf,
btOpenCLArray<btContact4>* contactsIn, btOpenCLArray<btGpuConstraint4>* contactCOut, void* additionalData,
int nContacts, const ConstraintCfg& cfg );
*/
void solveContactConstraint( const btOpenCLArray<btRigidBodyCL>* bodyBuf, const btOpenCLArray<btInertiaCL>* inertiaBuf, void solveContactConstraint( const btOpenCLArray<btRigidBodyCL>* bodyBuf, const btOpenCLArray<btInertiaCL>* inertiaBuf,
btOpenCLArray<btGpuConstraint4>* constraint, void* additionalData, int n ,int maxNumBatches); btOpenCLArray<btGpuConstraint4>* constraint, void* additionalData, int n ,int maxNumBatches);
@@ -153,16 +132,11 @@ class Solver : public SolverBase
btOpenCLArray<btContact4>* contactsIn, btOpenCLArray<btGpuConstraint4>* contactCOut, void* additionalData, btOpenCLArray<btContact4>* contactsIn, btOpenCLArray<btGpuConstraint4>* contactCOut, void* additionalData,
int nContacts, const ConstraintCfg& cfg ); int nContacts, const ConstraintCfg& cfg );
/* void sortContacts( const btOpenCLArray<btRigidBodyCL>* bodyBuf,
btOpenCLArray<btContact4>* contactsIn, void* additionalData,
int nContacts, const ConstraintCfg& cfg );
*/
void batchContacts( btOpenCLArray<btContact4>* contacts, int nContacts, btOpenCLArray<unsigned int>* n, btOpenCLArray<unsigned int>* offsets, int staticIdx ); void batchContacts( btOpenCLArray<btContact4>* contacts, int nContacts, btOpenCLArray<unsigned int>* n, btOpenCLArray<unsigned int>* offsets, int staticIdx );
}; };
#undef MYF4
#undef MAKE_MYF4
#endif //__ADL_SOLVER_H #endif //__ADL_SOLVER_H

20
opencl/gpu_rigidbody/host/btGpuBatchingPgsSolver.cpp
View File

@@ -9,7 +9,7 @@
#include <string.h> #include <string.h>
#include "../../basic_initialize/btOpenCLUtils.h" #include "../../basic_initialize/btOpenCLUtils.h"
#include "../host/btConfig.h" #include "../host/btConfig.h"
#include "../Stubs/Solver.h" #include "Solver.h"
#define SOLVER_SETUP_KERNEL_PATH "opencl/gpu_rigidbody/kernels/solverSetup.cl" #define SOLVER_SETUP_KERNEL_PATH "opencl/gpu_rigidbody/kernels/solverSetup.cl"
@@ -25,7 +25,7 @@
#include "../kernels/batchingKernels.h" #include "../kernels/batchingKernels.h"
#define BTNEXTMULTIPLEOF(num, alignment) (((num)/(alignment) + (((num)%(alignment)==0)?0:1))*(alignment))
enum enum
{ {
BT_SOLVER_N_SPLIT = 16, BT_SOLVER_N_SPLIT = 16,
@@ -452,10 +452,10 @@ void btGpuBatchingPgsSolver::solveContacts(int numBodies, cl_mem bodyBuf, cl_mem
{ {
int sortSize = NEXTMULTIPLEOF( nContacts, sortAlignment ); int sortSize = BTNEXTMULTIPLEOF( nContacts, sortAlignment );
btOpenCLArray<u32>* countsNative = m_data->m_solverGPU->m_numConstraints; btOpenCLArray<unsigned int>* countsNative = m_data->m_solverGPU->m_numConstraints;
btOpenCLArray<u32>* offsetsNative = m_data->m_solverGPU->m_offsets; btOpenCLArray<unsigned int>* offsetsNative = m_data->m_solverGPU->m_offsets;
{ // 2. set cell idx { // 2. set cell idx
BT_PROFILE("GPU set cell idx"); BT_PROFILE("GPU set cell idx");
@@ -584,11 +584,11 @@ void btGpuBatchingPgsSolver::solveContacts(int numBodies, cl_mem bodyBuf, cl_mem
btOpenCLArray<btContact4>* contactsIn = m_data->m_pBufContactOutGPU; btOpenCLArray<btContact4>* contactsIn = m_data->m_pBufContactOutGPU;
contactsIn->copyToHost(cpuContacts); contactsIn->copyToHost(cpuContacts);
btOpenCLArray<u32>* countsNative = m_data->m_solverGPU->m_numConstraints; btOpenCLArray<unsigned int>* countsNative = m_data->m_solverGPU->m_numConstraints;
btOpenCLArray<u32>* offsetsNative = m_data->m_solverGPU->m_offsets; btOpenCLArray<unsigned int>* offsetsNative = m_data->m_solverGPU->m_offsets;
btAlignedObjectArray<u32> nNativeHost; btAlignedObjectArray<unsigned int> nNativeHost;
btAlignedObjectArray<u32> offsetsNativeHost; btAlignedObjectArray<unsigned int> offsetsNativeHost;
{ {
BT_PROFILE("countsNative/offsetsNative copyToHost"); BT_PROFILE("countsNative/offsetsNative copyToHost");
@@ -665,7 +665,7 @@ void btGpuBatchingPgsSolver::solveContacts(int numBodies, cl_mem bodyBuf, cl_mem
} }
else else
{ {
//m_data->m_solverGPU->solveContactConstraintHost(m_data->m_bodyBufferGPU, m_data->m_inertiaBufferGPU, m_data->m_contactCGPU,0, nContactOut ,maxNumBatches); m_data->m_solverGPU->solveContactConstraintHost(m_data->m_bodyBufferGPU, m_data->m_inertiaBufferGPU, m_data->m_contactCGPU,0, nContactOut ,maxNumBatches);
} }
clFinish(m_data->m_queue); clFinish(m_data->m_queue);

2
opencl/gpu_rigidbody/host/btGpuRigidBodyPipeline.cpp
View File

@@ -13,7 +13,7 @@
#include "../../gpu_sat/host/btRigidBodyCL.h" #include "../../gpu_sat/host/btRigidBodyCL.h"
#include "../../gpu_sat/host/btContact4.h" #include "../../gpu_sat/host/btContact4.h"
#include "btGpuBatchingPgsSolver.h" #include "btGpuBatchingPgsSolver.h"
#include "../Stubs/Solver.h" #include "Solver.h"
#include "btConfig.h" #include "btConfig.h"