Add a pyopencl demo

arcade/examples/pyopencl_test.py

@@ -0,0 +1,58 @@
+"""
+Using pyopencl enables high-performance large-scale computations, 
+such as frequent destruction and creation of interfaces, and large-scale position calculations.
+More information about pyopencl:
+https://mathema.tician.de/software/pyopencl/
+The following example demonstrates how to quickly and easily use pyopencl within arcade to run kernel code.
+"""
+import arcade
+import pyopencl as cl
+import numpy as np
+
+WINDOW_WIDTH = 800
+WINDOW_HEIGHT = 600
+PARTICLE_COUNT = 1000
+
+# Initialize OpenCL
+nvidia_platform = [platform for platform in cl.get_platforms() if platform.name == "NVIDIA CUDA"][0]
+nvidia_devices = nvidia_platform.get_devices()
+nvidia_context = cl.Context(devices=nvidia_devices)
+queue = cl.CommandQueue(nvidia_context)
+
+# OpenCL kernel: simple position update
+kernel_code = """
+__kernel void update_particles(__global float2 *positions, float dt) {
+    int gid = get_global_id(0);
+    float2 pos = positions[gid];
+    pos.y -= dt * 100.0f;  // Falling down
+    if (pos.y < 0.0f) pos.y = 600.0f;
+    positions[gid] = pos;
+}
+"""
+program = cl.Program(nvidia_context, kernel_code).build()
+
+# Initialize particle data
+positions_np = np.random.rand(PARTICLE_COUNT, 2).astype(np.float32)
+positions_np[:, 0] *= WINDOW_WIDTH
+positions_np[:, 1] *= WINDOW_HEIGHT
+positions_cl = cl.Buffer(nvidia_context, cl.mem_flags.READ_WRITE | cl.mem_flags.COPY_HOST_PTR, hostbuf=positions_np)
+
+class ParticleWindow(arcade.Window):
+    def __init__(self):
+        super().__init__(WINDOW_WIDTH, WINDOW_HEIGHT, "Arcade + OpenCL Particle Example")
+        self.particles = []
+
+    def on_draw(self):
+        self.clear()
+        for pos in positions_np:
+            arcade.draw_circle_filled(pos[0], pos[1], 2, arcade.color.WHITE)
+
+    def on_update(self, delta_time: float):
+        # Run the OpenCL kernel to update particles
+        global_size = (PARTICLE_COUNT,)
+        program.update_particles(queue, global_size, None, positions_cl, np.float32(delta_time))
+        cl.enqueue_copy(queue, positions_np, positions_cl)
+
+if __name__ == "__main__":
+    window = ParticleWindow()
+    arcade.run()