taoaoxiang / draw_sphere

draw_sphere

Let's draw sphere in different ways!

• Loop #1 Devide-n-Conquer
  • Python
  • C++
• Recursion #1
  • Python
  • C++
• Loop #2 Geometry
  • Python
  • C++
• Loop #3 Brute-Force
  • Python
  • C++
  • CUDA
• Loop #4 Brute-Force All
  • Python
  • C++
  • CUDA
• [?] Random Forest Classification
  • [?] Failed
• [?] Random Forest Regression
  • [?] Failed
• Others

---

Loop #1 Devide-n-Conquer

Benchmark performance:

• Python: 0.087s
• C++: 0.682s

Total points:

• Python: 125834
• C++: 127560

Method:

1. Iterate Z-axis [0, r], val: c
2. Iterate X-axis [r, c], val: a
3. Iterate Y-axis [0, a], val: b
4. Draw a point (a, b, c) in the red area
5. Draw a point (b, a, c) in the green area
6. Draw a point (c, b, a) in the blue area
7. Iterate 8 octants (+,+,+) to (-,-,-)

Representation (VMD):

---

Recursion #1

Benchmark performance:

• Python: 2.988s
• C++: 14.581s

Total points:

• Python: 125858
• C++: 127752

Method:

1. Go over the first octant, find a qualified point (a, b, c)
2. Store the point (a, b, c)
3. Move the point, one direction at one time
   • (a +/- resolution, b, c)
   • (a, b +/- resolution, c)
   • (a, b, c +/- resolution)
4. Iterate 8 octants (+,+,+) to (-,-,-)

Note:
If the terminal complains about Segmentation fault (core dumped), modify ulimit in termial with the command below: ulimit -s unlimited

Representation (VMD):

---

Loop #2 Geometry

Benchmark performance:

• Python: 0.088s
• C++:

Total points:

• Python: 125858
• C++:

Method:

1. Iterate Z-axis [0, r], val: c
2. Iterate X-axis [r, c], val: a
3. Get the y-axis coordinate is ~= b, val: b
4. Draw a point (a, b, c) in the green area
5. Draw a point (c, b, a) in the green area
6. Iterate 8 octants (+,+,+) to (-,-,-)

Representation (VMD):

---

Loop #3 Brute-Force

Benchmark performance:

• Python:
• C++: 0.506s
• CUDA: 0.490s

Total points:

• Python:
• C++: 127856
• CUDA: 127840

Method:

1. Generate 1-D array, length: ((N+1) * (N+1) * (N+1))
2. Iterate the array, mark qualified points in the megenta region
3. Iterate 8 octants (+,+,+) to (-,-,-)

Representation (VMD):

---

Loop #3 Brute-Force All

Benchmark performance:

• Python:
• C++: 0.518s
• CUDA: 0.508s

Total points:

• Python:
• C++: 125930
• CUDA: 125914

Method:

1. Generate 1-D array, length: ((2N+1) * (2N+1) * (2N+1))
2. Iterate the array, mark qualified points in the yellow region

Representation (VMD):

---

Random Forest Regression

Benchmark performance:

• Python:

Total points:

• Python:

Method:

1. Training on the red dots
2. Test

Representation (VMD):