Profiling a headless vulkan compute kernel

When I run ncu on a headless compute vulkan app, I get the message that no kernels were profiled.

I get the same with ncu-ui

$ GIB_APIS=vk /usr/local/NVIDIA-Nsight-Compute-2023.2/ncu ./testgib
numboxes in aab file: 69
Found 3 physical devices.
dGPU NVIDIA GeForce RTX 3060 Laptop GPU
iGPU AMD Radeon Graphics (RADV RENOIR)
 CPU llvmpipe (LLVM 15.0.7, 256 bits)
Using NVIDIA GeForce RTX 3060 Laptop GPU
Debug utils labels: supported
5 mem types (2 mem heaps)
  23574 MiB of non-local memory [ ]
   6144 MiB of local memory [ device-local ]
  23574 MiB of non-local memory [ host-visible host-coherent ]
  23574 MiB of non-local memory [ host-visible host-coherent host-cached ]
   6144 MiB of local memory [ device-local host-visible host-coherent ]
gib-vk: Read 2097152 hemi dirs.
gib-vk: hem data uploaded to GPU.
gib-vk: Read 2097152 omni dirs.
gib-vk: omn data uploaded to GPU.
gib-vk: hit buffers created on GPU.
gib-vk: col buffers created on GPU.
gib-vk: fac buffer created on GPU.
gib-vk: tex buffer created on GPU.
gib-vk: constant buffers created on GPU.
Created descriptor set layout.
Created pipeline for kernel 0
Descriptor sets updated for kernel 0.
Created descriptor set layout.
Created pipeline for kernel 1
Descriptor sets updated for kernel 1.
Created descriptor set layout.
Created pipeline for kernel 2
Descriptor sets updated for kernel 2.
primary hits:   2062964
secondary hits: 1815956
histogram: 1109286/1099927/283905/113051/137979/1134772
==WARNING== No kernels were profiled.
==WARNING== Profiling kernels launched by child processes requires the --target-processes all option.

Is ncu supposed to work with vulkan? Or will it only profile cuda kernels?