Speeding up bullet simulations

[alda30]

Hi,

The way that I use bullet3 is to write my code in c++ in a cpp file and use makefile to execute bullet and linking the libraries.
my makefile contains:

#Compiler setting
CC=				g++
CFLAGS=			-c -Wall
GLFLAGS=		-lGL -lGLU -lglut
SDLFLAGS=		-lSDL -lSL_image
TFLAGS=			-std=c++0x 
#File name setting
CODE=			helloWorld.cpp
#Bullet3 libraries and headers addresses
B3SOURCE=		~/Bullet/bullet3/src/
B3WORLDIMPORTER=~/Bullet/bullet3/bin/libBulletWorldImporter_gmake_x64_release.a
B3SB=			~/Bullet/bullet3/bin/libBulletSoftBody_gmake_x64_release.a
B3INVDUTIL=		~/Bullet/bullet3/bin/libBulletInverseDynamicsUtils_gmake_x64_release.a
B3INVDD=		~/Bullet/bullet3/bin/libBulletInverseDynamics_gmake_x64_release.a
B3D=			~/Bullet/bullet3/bin/libBulletDynamics_gmake_x64_release.a
B3C=			~/Bullet/bullet3/bin/libBulletCollision_gmake_x64_release.a
B3LM=			~/Bullet/bullet3/bin/libLinearMath_gmake_x64_release.a
B3FILELOADER=	~/Bullet/bullet3/bin/libBulletFileLoader_gmake_x64_release.a
B3BC=			~/Bullet/bullet3/bin/libBullet3Common_gmake_x64_release.a
B3OPENCL=		~/Bullet/bullet3/bin/libBullet3OpenCL_clew_gmake_x64_release.a
B3CLD=			~/Bullet/bullet3/bin/libBullet3Dynamics_gmake_x64_release.a
B3CLC=			~/Bullet/bullet3/bin/libBullet3Collision_gmake_x64_release.a
B3CLG=			~/Bullet/bullet3/bin/libBullet3Geometry_gmake_x64_release.a
#Output setting
OUTPUT=			output.out
EXECUTE=		./$(OUTPUT)
TEXTOUTPUT=		output.txt

compile: $(CODE)
	@  $(CC) $(CODE) -I $(B3SOURCE) $(B3WORLDIMPORTER) $(B3SB) $(B3INVDUTIL) $(B3INVDD) $(B3D) $(B3C) $(B3FILELOADER) $(B3LM) $(B3BC) $(B3OPENCL) $(B3CLD) $(B3CLC) $(B3CLG) $(TFLAGS) -o $(OUTPUT)

run: $(OUTPUT)
	@  $(EXECUTE)

My simulation scene contains thousands of rigidbodies. But the problem is that during the running the simulation, when I try "gmonitor" to monitor the GPU usage, I see almost no GPU percentage. However, when I run Bullet example browser, and during the simulation, do a "gmonitor", my GPU occupation rate is up to 30-40%.

I guess the linking in makefile is not correct. Can someone please help me in speeding up my simulation by using my GPU?

Also, as far as I know there is no mult-threading support for at CPU level right?

Regards,
Alex Davdison

[alda30]

Any help?