Remove INTEL_CPU

Author: Rbiessy

docs/backend/SYCL.md

@@ -630,7 +630,7 @@ use 1 SYCL GPUs: [0] with Max compute units:512
 | Name               | Value                                 | Function                                    |
 |--------------------|---------------------------------------|---------------------------------------------|
 | GGML_SYCL          | ON (mandatory)                        | Enable build with SYCL code path.<br>FP32 path - recommended for better perforemance than FP16 on quantized model|
-| GGML_SYCL_TARGET   | INTEL *(default)* \| INTEL_CPU \| INTEL_GPU \| NVIDIA \| AMD    | Set the SYCL target device type.            |
+| GGML_SYCL_TARGET   | INTEL *(default)* \| NVIDIA \| AMD    | Set the SYCL target device type.            |
 | GGML_SYCL_DEVICE_ARCH | Optional (except for AMD)          | Set the SYCL device architecture, optional except for AMD. Setting the device architecture can improve the performance. See the table [--offload-arch](https://github.com/intel/llvm/blob/sycl/sycl/doc/design/OffloadDesign.md#--offload-arch) for a list of valid architectures. |
 | GGML_SYCL_F16      | OFF *(default)* \|ON *(optional)*     | Enable FP16 build with SYCL code path.      |
 | CMAKE_C_COMPILER   | `icx` *(Linux)*, `icx/cl` *(Windows)* | Set `icx` compiler for SYCL code path.      |

examples/sycl/README.md

@@ -14,7 +14,7 @@ List all SYCL devices with ID, compute capability, max work group size, ect.
 
 1. Build the llama.cpp for SYCL for the specified target *(using GGML_SYCL_TARGET)*.
 
-2. Enable oneAPI running environment *(if GGML_SYCL_TARGET is set to INTEL -default-, INTEL_CPU or INTEL_GPU)*
+2. Enable oneAPI running environment *(if GGML_SYCL_TARGET is set to INTEL -default-)*
 
 ```
 source /opt/intel/oneapi/setvars.sh

ggml/cmake/ggml-config.cmake.in

@@ -78,7 +78,7 @@ if (NOT GGML_SHARED_LIB)
 
     if (GGML_SYCL)
         find_package(DNNL)
-        if (${DNNL_FOUND} AND GGML_SYCL_TARGET MATCHES "INTEL.*")
+        if (${DNNL_FOUND} AND GGML_SYCL_TARGET STREQUAL "INTEL")
             list(APPEND GGML_SYCL_INTERFACE_LINK_LIBRARIES DNNL::dnnl)
         endif()
         if (WIN32)

ggml/src/ggml-sycl/CMakeLists.txt

@@ -1,6 +1,6 @@
 message(STATUS  "GGML_SYCL_TARGET=${GGML_SYCL_TARGET}")
 
-if (NOT GGML_SYCL_TARGET MATCHES "^(INTEL.*|NVIDIA|AMD)$")
+if (NOT GGML_SYCL_TARGET MATCHES "^(INTEL|NVIDIA|AMD)$")
     message(FATAL_ERROR "Invalid backend chosen, supported options are INTEL, NVIDIA, or AMD")
 endif()
 
@@ -49,13 +49,13 @@ target_sources(ggml-sycl PRIVATE ${GGML_HEADERS_SYCL} ${GGML_SOURCES_SYCL})
 find_package(DNNL)
 message("-- DNNL found:" ${DNNL_FOUND})
 
-if (GGML_SYCL_TARGET MATCHES "INTEL.*")
+if (GGML_SYCL_TARGET STREQUAL "INTEL")
     add_compile_definitions(GGML_SYCL_DNNL=${DNNL_FOUND})
 else()
     add_compile_definitions(GGML_SYCL_DNNL=0)
 endif()
 
-if (${DNNL_FOUND} AND GGML_SYCL_TARGET MATCHES "INTEL.*")
+if (${DNNL_FOUND} AND GGML_SYCL_TARGET STREQUAL "INTEL")
     target_link_libraries(ggml-sycl PRIVATE DNNL::dnnl)
 endif()
 
@@ -116,12 +116,9 @@ if (NOT oneMath_FOUND)
 endif()
 
 # Below oneMath compile-time dispatching is used for better performance
-if (GGML_SYCL_TARGET STREQUAL "INTEL_CPU")
-    target_link_libraries(ggml-sycl PRIVATE ONEMATH::onemath_blas_mklcpu)
-    target_compile_definitions(ggml-sycl PRIVATE GGML_SYCL_INTEL GGML_SYCL_INTEL_CPU)
-elseif (GGML_SYCL_TARGET STREQUAL "INTEL_GPU")
+if (GGML_SYCL_TARGET STREQUAL "INTEL")
     target_link_libraries(ggml-sycl PRIVATE ONEMATH::onemath_blas_mklgpu)
-    target_compile_definitions(ggml-sycl PRIVATE GGML_SYCL_INTEL GGML_SYCL_INTEL_GPU)
+    target_compile_definitions(ggml-sycl PRIVATE GGML_SYCL_INTEL)
 elseif (GGML_SYCL_TARGET STREQUAL "NVIDIA")
     target_link_libraries(ggml-sycl PRIVATE ONEMATH::onemath_blas_cublas)
     target_compile_options(ggml-sycl PRIVATE "-fsycl-targets=nvptx64-nvidia-cuda")

ggml/src/ggml-sycl/dpct/helper.hpp

@@ -98,9 +98,7 @@ inline auto get_onemath_backend(sycl::queue& queue)
 // If the backend is known at compile-time, use oneMath backend_selector to use
 // compile-time dispatching and avoid the need to dlopen libraries. Otherwise
 // fallback to runtime dispatching.
-#if defined(GGML_SYCL_INTEL_CPU)
-    return oneapi::math::backend_selector<oneapi::math::backend::mklcpu>{ queue };
-#elif defined(GGML_SYCL_INTEL_GPU)
+#if defined(GGML_SYCL_INTEL)
     return oneapi::math::backend_selector<oneapi::math::backend::mklgpu>{ queue };
 #elif defined(GGML_SYCL_NVIDIA)
     return oneapi::math::backend_selector<oneapi::math::backend::cublas>{ queue };