Skip to main content

Table 4 Execution times for the CPU, GPU, area-efficient microarchitecture and many-object microarchitecture

From: A hybrid fixed-function and microprocessor solution for high-throughput broad-phase collision detection

   

Area-efficient

Many-object

   

microarchitecture

microarchitecture

Objects

CPU

GPU

None

1024

None

1024

2048

4096

8192

16,384

100

0.2213

0.9732

0.0005

 

0.0008

     

200

0.4478

0.9982

0.0016

 

0.0029

     

300

0.7990

1.0347

0.0033

 

0.0062

     

400

0.9994

1.0441

0.0057

 

0.0107

     

500

1.2966

1.0302

0.0087

 

0.0165

     

600

1.5770

1.0727

0.0124

 

0.0236

     

700

1.9871

1.0796

0.0168

 

0.0319

     

800

2.5760

1.0889

0.0217

 

0.0414

     

900

2.9700

1.1072

0.0273

 

0.0522

     

1000

3.1694

1.0945

0.0336

 

0.0643

     

2000

6.0023

1.2040

 

0.0423

0.2535

0.0804

    

4000

14.2332

1.4039

 

0.0913

1.0070

0.1727

0.4036

   

8000

30.7177

1.9585

 

0.1665

4.0140

0.3153

0.6270

1.7638

  

16,000

69.8204

3.6801

 

0.3074

16.0280

0.5821

1.2199

2.5936

8. 2487

 

32,000

159.5770

8.3066

 

0.5676

 

1.0749

2.2525

4.7891

15.2312

36.9132

64,000

375.3380

22.8188

 

1.0481

 

1.9848

4.1592

8.8431

28.1244

68.1603

128,000

834.9720

67.5312

 

1.9353

 

3.6649

7.6800

16.3288

51.9317

125.8579

256,000

1940.2600

187.3310

 

3.5735

 

6.7672

14.1811

30.1512

95.8919

232.3967

512,000

4423.3300

605.7710

 

6.5985

 

12.4957

26.1853

55.6742

177.0645

429.1206

1,024,000

9897.0100

1958.8776

 

12.1841

 

23.0732

48.3512

102.8025

326.9497

792.3714

  1. All execution times are in milliseconds. ‘CPU’ is Bullet’s DBVT, ‘GPU’ is Bullet’s GPU sweep and prune, and the numeric table headers indicate the spatial-partitioning cell size used with the microarchitectures. Bold results highlight the optimal time for each microarchitecture