Building whisper.cpp on Nvidia Jetson AGX Orin - Seeking Updated Build Instructions

[stenhelge]

We are currently running whisper.cpp version 1.5.1 on a Nvidia Jetson AGX Orin. Our current build process involves:

1. Modifying the Makefile by commenting out the following section:

ifneq ($(filter aarch64%,$(UNAME_M)),)
CFLAGS += -mcpu=native
CXXFLAGS += -mcpu=native
endif

2. Building with GPU support using:

WHISPER_CUBLAS=1 make -j

This setup is working successfully, and we perform transcriptions daily without issues.
However, since version 1.5.1, the project has undergone significant changes, including new build procedures. These updated build procedures are not compatible with the Nvidia Jetson platform.
Has anyone successfully built a newer version of whisper.cpp on the Nvidia Jetson AGX Orin platform? If so, could you share your build configuration and steps?
Thank you in advance for your assistance.

[Tharit]

For what its worth, it works for me just fine on an Jetson Orin NX running Jetpack 6.0, as per the documentation.
That is, just with

cmake -B build -DGGML_CUDA=1