On 16/05/2025 18:53, Tomeu Vizoso wrote:

See Chapter 36 "RKNN" from the RK3588 TRM (Part 1).

This is a derivative of NVIDIA's NVDLA, but with its own front-end processor.

The IP is divided in three cores, programmed independently. The first core though is special, requiring to be powered on before any of the others can be used.

The IOMMU of the first core is also special in that it has two subunits (read/write?) that need to be programmed in sync.

v2:

• Have one device for each NPU core (Sebastian Reichel)
• Have one device for each IOMMU (Sebastian Reichel)
• Correctly sort nodes (Diederik de Haas)
• Add rockchip,iommu compatible to IOMMU nodes (Sebastian Reichel)

v3:

• Adapt to a split of the register block in the DT bindings (Nicolas Frattaroli)

Signed-off-by: Tomeu Vizoso tomeu@tomeuvizoso.net

arch/arm64/boot/dts/rockchip/rk3588-base.dtsi | 85 +++++++++++++++++++++++++++ 1 file changed, 85 insertions(+)

diff --git a/arch/arm64/boot/dts/rockchip/rk3588-base.dtsi b/arch/arm64/boot/dts/rockchip/rk3588-base.dtsi index 1e18ad93ba0ebdad31642b88ff0f90ef4e8dc76f..7b961ab838212fad8e4a70390fdc917a828433a9 100644 --- a/arch/arm64/boot/dts/rockchip/rk3588-base.dtsi +++ b/arch/arm64/boot/dts/rockchip/rk3588-base.dtsi @@ -1136,6 +1136,91 @@ power-domain@RK3588_PD_SDMMC { }; };

• rknn_core_top: npu-core@fdab0000 {

npu@

• compatible = "rockchip,rk3588-rknn-core-top", "rockchip,rknn-core-top";
  

You never tested this. Test before sending instead of relying on us or after merging.

Best regards, Krzysztof