之前在一些三维软件上实现过cuda的metaball3d，由于对Python和简单的运动有点不太熟悉，只能简单实现一个metaball2d，下一步打算试试marching squares和metaball3d
import taichi as ti
import random
ti.init(arch=ti.gpu)
num = 8
width = 1080
height = 720
speed = 2
isoV = 50
pixels = ti.var(dt=ti.f32, shape=(width, height))
vs = ti.Vector(2, dt=ti.f32, shape=num) # position
randomX = ti.var(ti.f32, num)
randomY = ti.var(ti.f32, num)
randomR = ti.var(ti.f32, num)
for i in range(num):
vs[i] = [random.randint(0, width), random.randint(0, height)]
randomX[i] = random.randint(-speed, speed)
randomY[i] = random.randint(-speed, speed)
randomR[i] = random.randint(100, 200)
def move():
for k in range(num):
vs[k].x += randomX[k]
vs[k].y += randomY[k]
if vs[k].x - randomR[k] < 0:
randomX[k] = abs(randomX[k])
if vs[k].x + randomR[k] > width:
randomX[k] = -abs(randomX[k])
if vs[k].y - randomR[k] < 0:
randomY[k] = abs(randomY[k])
if vs[k].y + randomR[k] > height:
randomY[k] = -abs(randomY[k])
@ti.kernel
def drawPixels(iso:ti.f32):
for i, j in pixels: # Parallized over all pixels
fxy = 0
for k in range(num):
dx = i + 0.5 - vs[k].x
dy = j + 0.5 - vs[k].y
d2 = dx * dx + dy * dy
ti.atomic_add(fxy,randomR[k] ** 2 / d2)
if fxy >= iso:
pixels[i,j] = 0.2
else:
pixels[i, j] = 1
gui = ti.GUI("Metaballs2d", (width, height)) # 设置gui
for i in range(1000000):
move()
drawPixels(isoV)
gui.set_image(pixels)
gui.show()
