Skip to content

Hexbin client fan expansion: 42 floats/cell CPU+GPU where instancing needs 3 #175

Description

@Alek99

From a full data-transport audit at b0f8780.

The wire contract is already right (centers + one scalar per cell, python/xy/_payload.py:484-505). But _buildHexbinMark (js/src/50_chartview.ts:2567-2613) expands the hex fan 6x on the CPU: 6 columns x n*6 f32 allocated + 6 uploads, cval expanded x6 — 42 floats/cell in GPU memory where an instanced draw needs center (2) + cval (1) per instance with the constant ring as uniforms or a tiny shared vertex buffer (offsets added in the vertex shader, the way BAR_VS uses u_width). ~14x buffer-volume reduction; needs a small dedicated program instead of reusing MESH.

Metadata

Metadata

Assignees

No one assigned

    Labels

    needs investigateAudit finding awaiting validation/measurementperformanceReduces copies, bytes, or CPU on the data path

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions