lvgl-benchmark

Wraps the upstream lv_demo_benchmark() — LVGL's canonical performance harness. Walks a fixed scene set (gradients, alpha blends, text rendering, image rotation, …) and reports per-scene FPS.

Source: examples/lvgl-benchmark/.

Why run it

Compare the same demo across SoMs to see what the capabilities: block in metadata/e1m_modules/E1M-*.yaml is actually buying you at runtime:

• E1M-AEN701 (E7) has GPU2D + DAVE2D + DMA2D populated. The §D.lib.loader emits CONFIG_ALP_LVGL_GPU2D=y + CONFIG_ALP_LVGL_DMA2D=y, and LVGL hardware-accelerates the blits.
• E1M-AEN301 (E3) has no GPU2D. Loader emits CONFIG_ALP_LVGL_DMA2D=y only; blits fall back to CPU. The FPS drop is the visible measure of what GPU2D buys you.

This is the requires_cap capability-keyed binding system in action — same source, different CONFIG_* flags per SKU automatically.

board.yaml

schema_version: 2

som:
  sku: E1M-AEN701

carrier:
  name: E1M-EVK

cores:
  m55_hp:
    os: zephyr
    app: ./src
    peripherals: [spi, gpio]
    libraries: [lvgl]

chips:
  - st7789

diagnostics:
  log_level: info

See also

• lvgl-widgets-demo · lvgl-music-player
• Architecture: capability validation
• Examples overview