A 200x200 grid of 40,000 points whose height comes from a Texture Buffer
Object (TBO), sampled in the vertex shader with
texelFetch(yPosSampler, gl_VertexID) -- no second vertex attribute, no
interleaving, just per-vertex data read straight out of a buffer by its
texel index. XZ positions sit in an ordinary vertex buffer; Y lives in the
TBO and gets rebuilt and re-uploaded every 20ms
(sin(x + offset) + cos(x - offset), offset creeping forward each tick),
so the whole grid animates without the vertex buffer itself ever changing.
There's no ncca.ngl wrapper for TBOs, so this one drops to raw
OpenGL.GL calls (GL_TEXTURE_BUFFER, GL_R32F, glTexBuffer) rather
than going through ShaderLib/Texture.
w/s: wireframe / fill (no visible effect here --GL_POINTShas no edges to outline, ported anyway to match the source)f/n: fullscreen / windowed- Left-drag : orbit, Right-drag : pan, Wheel : zoom
Esc: quit
main_webgpu.py reinterprets the OpenGL Texture Buffer Object + texelFetch
mechanism: WebGPU has neither, so the per-vertex height values instead live
in a read-only storage buffer (var<storage, read> array<f32>), indexed by
@builtin(vertex_index) in the vertex shader -- the same "raw per-vertex
data fed to the shader outside normal vertex attributes" lesson, WebGPU's
native tool for it. One platform limitation to note: WebGPU's point-list
topology has no point-size control, so points render at 1 pixel here versus
the OpenGL version's glPointSize(4).