cmake: skip cvector-generator and export-lora when CPU backend is disabled

[arichiardi]

Overview

Both cvector-generator and export-lora link against CPU backend symbols (ggml_backend_cpu_init,
ggml_get_f32_nd, etc.) unconditionally. When building with -DGGML_CPU=OFF, these tools fail to
link, causing cmake --install to error on a missing binary:

CMake Error at build/tools/cvector-generator/cmake_install.cmake:52 (file):
  file INSTALL cannot find
  "<repo-path>/llama.cpp/build/bin/llama-cvector-generator": No such
  file or directory.
Call Stack (most recent call first):
  build/tools/cmake_install.cmake:122 (include)
  build/cmake_install.cmake:67 (include)

Both tools are guarded by NOT GGML_BACKEND_DL but not by GGML_CPU. Added the missing check.

Requirements

• I have read and agree with the contributing guidelines
• AI usage disclosure: YES - AI assisted in identifying the issue and proposing the fix

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Tightens the build conditions for certain tools/examples so they are only included when the CPU backend is enabled (in addition to requiring non-dynamic backend loading).

Changes:

• Restrict cvector-generator and export-lora subdirectories to build only when GGML_CPU is enabled and GGML_BACKEND_DL is disabled.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[CISC]

Bad merge.

[arichiardi]

Sorry let me rebase - fixed ✔️