Bug: [SYCL] linker fails with undefined reference to symbol

[qnixsynapse]

What happened?

With the latest master branch, the linker fails with "undefined reference to symbol '_ZNK4sycl3_V16device8get_infoINS0_3ext5intel4info6device9device_idEEENS0_6detail19is_device_info_descIT_E11return_typeEv'"

Name and Version

Latest Master; commit hash: 822b632
intel-oneapi-basekit: 2024.1.0.596-3
Intel Compute Runtime: 24.31.30508.7-1

What operating system are you seeing the problem on?

Linux

Relevant log output

[100%] Linking CXX executable ../../bin/llama-server
cd /home/qnixsynapse/Public/Projects/llama.cpp/build/examples/server && /usr/bin/cmake -E cmake_link_script CMakeFiles/llama-server.dir/link.txt --verbose=1
/opt/intel/oneapi/compiler/2024.1/bin/icpx -O3 -DNDEBUG "CMakeFiles/llama-server.dir/server.cpp.o" -o ../../bin/llama-server  ../../common/libcommon.a ../../src/libllama.a ../../ggml/src/libggml.a /opt/intel/oneapi/compiler/2024.1/lib/libiomp5.so /usr/lib64/libpthread.a /opt/intel/oneapi/dnnl/2024.1/lib/libdnnl.so.3.4 -lOpenCL -lpthread /opt/intel/oneapi/tbb/2021.12/lib/intel64/gcc4.8/libtbb.so.12 -lOpenCL -lmkl_core -lpthread -lm -ldl -lmkl_sycl_blas -lmkl_intel_ilp64 -lmkl_tbb_thread -lm
/usr/bin/ld: ../../ggml/src/libggml.a(ggml-sycl.cpp.o): undefined reference to symbol '_ZNK4sycl3_V16device8get_infoINS0_3ext5intel4info6device9device_idEEENS0_6detail19is_device_info_descIT_E11return_typeEv'
/usr/bin/ld: /opt/intel/oneapi/compiler/2024.1/lib/libsycl.so.7: error adding symbols: DSO missing from command line
icpx: error: linker command failed with exit code 1 (use -v to see invocation)
make[3]: *** [examples/server/CMakeFiles/llama-server.dir/build.make:169: bin/llama-server] Error 1
make[3]: Leaving directory '/home/qnixsynapse/Public/Projects/llama.cpp/build'
make[2]: *** [CMakeFiles/Makefile2:3277: examples/server/CMakeFiles/llama-server.dir/all] Error 2
make[2]: Leaving directory '/home/qnixsynapse/Public/Projects/llama.cpp/build'
make[1]: *** [CMakeFiles/Makefile2:3284: examples/server/CMakeFiles/llama-server.dir/rule] Error 2
make[1]: Leaving directory '/home/qnixsynapse/Public/Projects/llama.cpp/build'
make: *** [Makefile:1297: llama-server] Error 2

[airMeng]

should be fixed in #9497, feel free to reopen

[qnixsynapse]

Thank you!

[qnixsynapse]

Builds correctly but crashed at runtime with this error: No kernel named _ZTSZL14scale_f32_syclPKfPffiPN4sycl3_V15queueEEUlNS3_7nd_itemILi3EEEE_ was found -46 (PI_ERROR_INVALID_KERNEL_NAME) Exception caught at file:ggml/src/ggml-sycl.cpp, line:2741"

Edit: Looks like it can't find any kernels: TANH(type=f32,ne_a=[128,2,2,2],v=0): No kernel named _ZTSZL13tanh_f32_syclPKfPfiPN4sycl3_V15queueEEUlNS3_7nd_itemILi3EEEE_ was found -46 (PI_ERROR_INVALID_KERNEL_NAME)

Edit: Had to turn on SHARED_LIBS to fix it.