@fugood/llama.node 0.6.3 → 1.0.0-beta.1

package/src/llama.cpp/ggml/src/ggml-sycl/CMakeLists.txt

@@ -1,189 +0,0 @@
-message(STATUS  "GGML_SYCL_TARGET=${GGML_SYCL_TARGET}")
-
-if (NOT GGML_SYCL_TARGET MATCHES "^(INTEL|NVIDIA|AMD)$")
-    message(FATAL_ERROR "Invalid backend chosen, supported options are INTEL, NVIDIA, or AMD")
-endif()
-
-check_cxx_compiler_flag("-fsycl" SUPPORTS_SYCL)
-
-if (DEFINED ENV{ONEAPI_ROOT})
-    message(STATUS "Using oneAPI Release SYCL compiler (icpx).")
-elseif(SUPPORTS_SYCL)
-    message(WARNING "Using open-source SYCL compiler (clang++). Didn't detect ENV {ONEAPI_ROOT}.
-        If you expected the oneAPI Release compiler, please install oneAPI & source it, like:
-        source /opt/intel/oneapi/setvars.sh")
-else()
-    message(FATAL_ERROR, "C++ compiler lacks SYCL support.")
-endif()
-message(STATUS "SYCL found")
-#todo: AOT
-
-ggml_add_backend_library(ggml-sycl
-                         ggml-sycl.cpp
-                         ../../include/ggml-sycl.h
-                        )
-
-file(GLOB   GGML_HEADERS_SYCL "*.hpp")
-file(GLOB   GGML_SOURCES_SYCL "*.cpp")
-target_sources(ggml-sycl PRIVATE ${GGML_HEADERS_SYCL} ${GGML_SOURCES_SYCL})
-
-if (WIN32)
-    # To generate a Visual Studio solution, using Intel C++ Compiler for ggml-sycl is mandatory
-    if( ${CMAKE_GENERATOR} MATCHES "Visual Studio" AND NOT (${CMAKE_GENERATOR_TOOLSET} MATCHES "Intel C"))
-        set_target_properties(ggml-sycl PROPERTIES VS_PLATFORM_TOOLSET "Intel C++ Compiler 2025")
-        set(CMAKE_CXX_COMPILER "icx")
-        set(CMAKE_CXX_COMPILER_ID "IntelLLVM")
-    endif()
-endif()
-
-find_package(IntelSYCL)
-if (IntelSYCL_FOUND)
-    # Use oneAPI CMake when possible
-    target_link_libraries(ggml-sycl PRIVATE IntelSYCL::SYCL_CXX)
-else()
-    # Fallback to the simplest way of enabling SYCL when using intel/llvm nightly for instance
-    target_compile_options(ggml-sycl PRIVATE "-fsycl")
-    target_link_options(ggml-sycl PRIVATE "-fsycl")
-endif()
-
-target_compile_options(ggml-sycl PRIVATE "-Wno-narrowing")
-
-# Link against oneDNN
-set(GGML_SYCL_DNNL 0)
-if(GGML_SYCL_DNN)
-    find_package(DNNL)
-    if(DNNL_FOUND)
-        if (NOT DEFINED DNNL_GPU_VENDOR)
-            # default to intel target
-            set(DNNL_GPU_VENDOR "INTEL")
-            if(NOT "${GGML_SYCL_TARGET}" STREQUAL "INTEL")
-                message(WARNING "oneDNN builds bundled with oneapi release only support INTEL target")
-            endif()
-        endif()
-
-        # Verify oneDNN was compiled for the same target as llama
-        if("${GGML_SYCL_TARGET}" STREQUAL "${DNNL_GPU_VENDOR}")
-            target_link_libraries(ggml-sycl PRIVATE DNNL::dnnl)
-            set(GGML_SYCL_DNNL 1)
-            get_target_property(CONFIGS DNNL::dnnl IMPORTED_CONFIGURATIONS)
-            foreach(CONFIG ${CONFIGS})
-                get_target_property(DNNL_LIB DNNL::dnnl IMPORTED_LOCATION_${CONFIG})
-                message(STATUS "Found oneDNN: ${DNNL_LIB}")
-            endforeach()
-        else()
-            message(WARNING
-                "oneDNN must be compiled for the same target as llama.cpp.
-                 llama.cpp: ${GGML_SYCL_TARGET}, oneDNN: ${DNNL_GPU_VENDOR}.
-                 Disabling oneDNN support.")
-        endif()
-    else()
-        message(STATUS "oneDNN not found, disabling oneDNN support")
-    endif()
-else()
-    message(STATUS "oneDNN support disabled by the user")
-endif()
-target_compile_definitions(ggml-sycl PRIVATE GGML_SYCL_DNNL=${GGML_SYCL_DNNL})
-
-if (GGML_SYCL_F16)
-    if (GGML_SYCL_TARGET STREQUAL "AMD")
-        message(WARNING "AMD target does not entirely support FP16 in the SYCL backend.")
-    endif()
-    add_compile_definitions(GGML_SYCL_F16)
-endif()
-
-if (GGML_SYCL_TARGET STREQUAL "NVIDIA")
-    add_compile_definitions(GGML_SYCL_WARP_SIZE=32)
-elseif (GGML_SYCL_TARGET STREQUAL "AMD")
-    # INFO: Allowed Sub_group_sizes are not consistent through all
-    # hip targets. For example, 64 is used for certain models, but the backend
-    # does not support it.
-    # Target archs tested working: gfx1030, gfx1031, (Only tested sub_group_size = 32)
-    add_compile_definitions(GGML_SYCL_WARP_SIZE=32)
-else()
-    add_compile_definitions(GGML_SYCL_WARP_SIZE=16)
-endif()
-
-if (GGML_SYCL_GRAPH)
-    target_compile_definitions(ggml-sycl PRIVATE GGML_SYCL_GRAPH)
-endif()
-
-# Link against Intel oneMKL or oneMath
-if (GGML_SYCL_TARGET STREQUAL "INTEL")
-    # Intel devices use Intel oneMKL directly instead of oneMath to avoid the limitation of linking Intel oneMKL statically
-    # See https://github.com/uxlfoundation/oneMath/issues/654
-    if (CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
-        set(SYCL_COMPILER ON)
-    endif()
-    find_package(MKL REQUIRED)
-    target_link_libraries(ggml-sycl PRIVATE MKL::MKL_SYCL::BLAS)
-    target_compile_definitions(ggml-sycl PRIVATE GGML_SYCL_USE_INTEL_ONEMKL)
-else()
-    find_package(oneMath QUIET)
-    if (NOT oneMath_FOUND)
-        message(STATUS "oneMath not found: oneMath will be automatically downloaded")
-        # Use FetchContent to automatically pull and build oneMath
-        include(FetchContent)
-        set(BUILD_FUNCTIONAL_TESTS False)
-        set(BUILD_EXAMPLES False)
-        set(TARGET_DOMAINS blas)
-        if (GGML_SYCL_TARGET STREQUAL "NVIDIA")
-            set(ENABLE_MKLCPU_BACKEND False)
-            set(ENABLE_MKLGPU_BACKEND False)
-            set(ENABLE_CUBLAS_BACKEND True)
-        elseif (GGML_SYCL_TARGET STREQUAL "AMD")
-            set(ENABLE_MKLCPU_BACKEND False)
-            set(ENABLE_MKLGPU_BACKEND False)
-            set(ENABLE_ROCBLAS_BACKEND True)
-            # Ensure setting a string variable here is not overriden by oneMath CACHE variables
-            cmake_policy(SET CMP0126 NEW)
-            # Setting the device architecture is only needed and useful for AMD devices in oneMath
-            set(HIP_TARGETS ${GGML_SYCL_DEVICE_ARCH} CACHE STRING "oneMath HIP target" FORCE)
-        endif()
-        FetchContent_Declare(
-            ONEMATH
-            GIT_REPOSITORY https://github.com/uxlfoundation/oneMath.git
-            GIT_TAG c255b1b4c41e2ee3059455c1f96a965d6a62568a
-        )
-        FetchContent_MakeAvailable(ONEMATH)
-        # Create alias to match with find_package targets name
-        function(onemath_alias target)
-            if (TARGET ${target}_obj)
-                # Silence verbose warnings from external libraries
-                target_compile_options(${target}_obj PRIVATE -w)
-            endif()
-            if (TARGET ${target})
-                add_library(ONEMATH::${target} ALIAS ${target})
-            endif()
-        endfunction()
-        onemath_alias(onemath)
-        onemath_alias(onemath_blas_mklcpu)
-        onemath_alias(onemath_blas_mklgpu)
-        onemath_alias(onemath_blas_cublas)
-        onemath_alias(onemath_blas_rocblas)
-    endif()
-
-    # Below oneMath compile-time dispatching is used for better performance
-    if (GGML_SYCL_TARGET STREQUAL "NVIDIA")
-        target_link_libraries(ggml-sycl PRIVATE ONEMATH::onemath_blas_cublas)
-        target_compile_options(ggml-sycl PRIVATE "-fsycl-targets=nvptx64-nvidia-cuda")
-        target_link_options(ggml-sycl PRIVATE "-fsycl-targets=nvptx64-nvidia-cuda")
-        target_compile_definitions(ggml-sycl PRIVATE GGML_SYCL_NVIDIA)
-    elseif (GGML_SYCL_TARGET STREQUAL "AMD")
-        if (NOT GGML_SYCL_DEVICE_ARCH)
-            message(ERROR "Can't enable SYCL hip backend, GGML_SYCL_DEVICE_ARCH has not been set.")
-        endif()
-        target_link_libraries(ggml-sycl PRIVATE ONEMATH::onemath_blas_rocblas)
-        target_compile_options(ggml-sycl PRIVATE "-fsycl-targets=amdgcn-amd-amdhsa")
-        target_link_options(ggml-sycl PRIVATE "-fsycl-targets=amdgcn-amd-amdhsa")
-        target_compile_definitions(ggml-sycl PRIVATE GGML_SYCL_AMD)
-    else()
-        # Fallback to oneMath runtime dispatcher
-        target_link_libraries(ggml-sycl PRIVATE ONEMATH::onemath)
-        target_compile_definitions(ggml-sycl PRIVATE GGML_SYCL_GENERIC)
-    endif()
-endif()
-
-if (GGML_SYCL_DEVICE_ARCH)
-    target_compile_options(ggml-sycl PRIVATE -Xsycl-target-backend --offload-arch=${GGML_SYCL_DEVICE_ARCH})
-    target_link_options(ggml-sycl PRIVATE -Xsycl-target-backend --offload-arch=${GGML_SYCL_DEVICE_ARCH})
-endif()

package/src/llama.cpp/ggml/src/ggml-sycl/backend.hpp

@@ -1,37 +0,0 @@
-//
-// MIT license
-// Copyright (C) 2024 Intel Corporation
-// SPDX-License-Identifier: MIT
-//
-
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-
-#ifndef GGML_SYCL_BACKEND_HPP
-#define GGML_SYCL_BACKEND_HPP
-
-#include "binbcast.hpp"
-#include "common.hpp"
-#include "concat.hpp"
-#include "conv.hpp"
-#include "convert.hpp"
-#include "cpy.hpp"
-#include "dequantize.hpp"
-#include "dmmv.hpp"
-#include "element_wise.hpp"
-#include "gla.hpp"
-#include "im2col.hpp"
-#include "mmq.hpp"
-#include "mmvq.hpp"
-#include "norm.hpp"
-#include "outprod.hpp"
-#include "quants.hpp"
-#include "rope.hpp"
-#include "softmax.hpp"
-#include "tsembd.hpp"
-#include "wkv.hpp"
-
-#endif  // GGML_SYCL_BACKEND_HPP

package/src/llama.cpp/ggml/src/ggml-sycl/binbcast.cpp

@@ -1,239 +0,0 @@
-#include "binbcast.hpp"
-
-#include <array>
-#include <cstddef>
-#include <cstdint>
-#include <sycl/sycl.hpp>
-
-#include "dpct/helper.hpp"
-#include "ggml.h"
-
-template <float (*bin_op)(const float, const float), typename src0_t, typename src1_t, typename dst_t>
-static __dpct_inline__ void k_bin_bcast_contiguous(const src0_t * __restrict__ src0, const src1_t * __restrict__ src1,
-                                                   dst_t * dst, std::size_t num_elements, const sycl::nd_item<1> & it) {
-    auto element_id   = it.get_global_id(0);
-    auto global_range = it.get_global_range(0);
-    for (; element_id < num_elements; element_id += global_range) {
-        auto  src0_float_val = sycl::vec(src0[element_id]).template convert<float, sycl::rounding_mode::rte>();
-        auto  src1_float_val = sycl::vec(src1[element_id]).template convert<float, sycl::rounding_mode::rte>();
-        float dst_val        = bin_op(src0_float_val[0], src1_float_val[0]);
-        auto  val_to_store   = sycl::vec(dst_val).template convert<dst_t, sycl::rounding_mode::rte>();
-        dst[element_id]      = val_to_store;
-    }
-}
-
-template <float (*bin_op)(const float, const float), typename src0_t, typename src1_t, typename dst_t>
-static __dpct_inline__ void k_bin_bcast(const src0_t * __restrict__ src0, const src1_t * __restrict__ src1, dst_t * dst,
-                                        int ne0, int ne1, int ne2, int ne3, int ne10, int ne11, int ne12, int ne13,
-                                        int s0, int s1, int s2, int s3, int s00, int s01, int s02, int s03, int s10,
-                                        int s11, int s12, int s13, std::size_t num_dst_elements,
-                                        const sycl::nd_item<1> & item_ct1) {
-    auto calculate_logical_index =
-        [](const std::array<int, 4> & dims, std::size_t element_id) __attribute__((always_inline))->std::array<int, 4> {
-        std::array<int, 4> logical_index;
-#pragma unroll(4)
-        for (int i = 3; i >= 0; i--) {
-            logical_index[i] = element_id % dims[i];
-            element_id /= dims[i];
-        }
-        return logical_index;
-    };
-
-    auto calculate_index = [](const std::array<int, 4> & dims, const std::array<int, 4> & strides,
-                              const std::array<int, 4> & indices) __attribute__((always_inline))
-                               ->std::size_t {
-        std::size_t index = 0;
-#pragma unroll(4)
-        for (int i = 0; i < 4; i++) {
-            auto index_i = indices[i];
-            if (indices[i] >= dims[i]) {
-                index_i = indices[i] % dims[i];
-            }
-            index += strides[i] * index_i;
-        }
-        return index;
-    };
-
-    auto element_id = item_ct1.get_global_id(0);
-    for (; element_id < num_dst_elements; element_id += item_ct1.get_global_range(0)) {
-        auto  logical_index  = calculate_logical_index({ ne3, ne2, ne1, ne0 }, element_id);
-        auto  src_0_index    = calculate_index({ ne3, ne2, ne1, ne0 }, { s03, s02, s01, s00 }, logical_index);
-        auto  src_1_index    = calculate_index({ ne13, ne12, ne11, ne10 }, { s13, s12, s11, s10 }, logical_index);
-        auto  dst_index      = calculate_index({ ne3, ne2, ne1, ne0 }, { s3, s2, s1, s0 }, logical_index);
-        auto  src0_float_val = sycl::vec(src0[src_0_index]).template convert<float, sycl::rounding_mode::rte>();
-        auto  src1_float_val = sycl::vec(src1[src_1_index]).template convert<float, sycl::rounding_mode::rte>();
-        float dst_val        = bin_op(src0_float_val[0], src1_float_val[0]);
-        auto  val_to_store   = sycl::vec(dst_val).template convert<dst_t, sycl::rounding_mode::rte>();
-        dst[dst_index]       = val_to_store;
-    }
-}
-
-template <float (*bin_op)(const float, const float)> struct bin_bcast_sycl {
-    template <typename src0_t, typename src1_t, typename dst_t>
-    void operator()(const src0_t * src0_dd, const src1_t * src1_dd, dst_t * dst_dd, const int64_t ne00,
-                    const int64_t ne01, const int64_t ne02, const int64_t ne03, const int64_t ne10, const int64_t ne11,
-                    const int64_t ne12, const int64_t ne13, const int64_t ne0, const int64_t ne1, const int64_t ne2,
-                    const int64_t ne3, const size_t nb00, const size_t nb01, const size_t nb02, const size_t nb03,
-                    const size_t nb10, const size_t nb11, const size_t nb12, const size_t nb13, const size_t nb0,
-                    const size_t nb1, const size_t nb2, const size_t nb3, const bool src0_is_contiguous,
-                    const bool src1_is_contiguous, const bool dst_is_contiguous, queue_ptr stream) {
-        auto check_bcast_required = [](const std::array<int64_t, 4> & src_dims,
-                                       const std::array<int64_t, 4> & dst_dims) -> bool {
-            for (int i = 0; i < 4; i++) {
-                if (dst_dims[i] > src_dims[i]) {
-                    return true;
-                }
-            }
-            return false;
-        };
-
-        dpct::has_capability_or_fail(stream->get_device(), { sycl::aspect::fp16 });
-
-        GGML_ASSERT(nb0 % sizeof(dst_t) == 0);
-        GGML_ASSERT(nb1 % sizeof(dst_t) == 0);
-        GGML_ASSERT(nb2 % sizeof(dst_t) == 0);
-        GGML_ASSERT(nb3 % sizeof(dst_t) == 0);
-
-        GGML_ASSERT(nb00 % sizeof(src0_t) == 0);
-        GGML_ASSERT(nb01 % sizeof(src0_t) == 0);
-        GGML_ASSERT(nb02 % sizeof(src0_t) == 0);
-        GGML_ASSERT(nb03 % sizeof(src0_t) == 0);
-
-        GGML_ASSERT(nb10 % sizeof(src1_t) == 0);
-        GGML_ASSERT(nb11 % sizeof(src1_t) == 0);
-        GGML_ASSERT(nb12 % sizeof(src1_t) == 0);
-        GGML_ASSERT(nb13 % sizeof(src1_t) == 0);
-
-        // dst strides in number of elements
-        size_t s0 = nb0 / sizeof(dst_t);
-        size_t s1 = nb1 / sizeof(dst_t);
-        size_t s2 = nb2 / sizeof(dst_t);
-        size_t s3 = nb3 / sizeof(dst_t);
-
-        // src1 strides in number of elements
-        size_t s10 = nb10 / sizeof(src0_t);
-        size_t s11 = nb11 / sizeof(src1_t);
-        size_t s12 = nb12 / sizeof(src1_t);
-        size_t s13 = nb13 / sizeof(src1_t);
-
-        // src0 strides in number of elements
-        size_t s00 = nb00 / sizeof(src0_t);
-        size_t s01 = nb01 / sizeof(src0_t);
-        size_t s02 = nb02 / sizeof(src0_t);
-        size_t s03 = nb03 / sizeof(src0_t);
-
-        std::size_t num_dst_elements = static_cast<std::size_t>(ne0) * static_cast<std::size_t>(ne1) *
-                                       static_cast<std::size_t>(ne2) * static_cast<std::size_t>(ne3);
-        std::size_t local_range  = 256;
-        std::size_t global_range = ceil_div(num_dst_elements, local_range) * local_range;
-
-        bool needs_broadcasting = check_bcast_required({ ne00, ne01, ne02, ne03 }, { ne0, ne1, ne2, ne3 }) ||
-                                  check_bcast_required({ ne10, ne11, ne12, ne13 }, { ne0, ne1, ne2, ne3 });
-        bool all_contiguous = src0_is_contiguous && src1_is_contiguous && dst_is_contiguous;
-
-        if (! needs_broadcasting && all_contiguous) {
-            stream->submit([&](sycl::handler & cgh) {
-                cgh.parallel_for(sycl::nd_range<1>({ global_range }, { local_range }), [=](sycl::nd_item<1> it) {
-                    k_bin_bcast_contiguous<bin_op>(src0_dd, src1_dd, dst_dd, num_dst_elements, it);
-                });
-            });
-        } else {
-            stream->submit([&](sycl::handler & cgh) {
-                cgh.parallel_for(sycl::nd_range<1>({ global_range }, { local_range }), [=](sycl::nd_item<1> it) {
-                    k_bin_bcast<bin_op>(src0_dd, src1_dd, dst_dd, ne0, ne1, ne2, ne3, ne10, ne11, ne12, ne13, s0, s1,
-                                        s2, s3, s00, s01, s02, s03, s10, s11, s12, s13, num_dst_elements, it);
-                });
-            });
-        }
-    }
-};
-
-template <class op>
-inline void ggml_sycl_op_bin_bcast(ggml_backend_sycl_context & ctx, const ggml_tensor * src0, const ggml_tensor * src1,
-                                   ggml_tensor * dst) {
-    dpct::queue_ptr main_stream = ctx.stream();
-    GGML_TENSOR_BINARY_OP_LOCALS
-
-    if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F32) {
-        op()((const float *) src0->data, (const float *) src1->data, (float *) dst->data, ne00, ne01, ne02, ne03, ne10,
-             ne11, ne12, ne13, ne0, ne1, ne2, ne3, nb00, nb01, nb02, nb03, nb10, nb11, nb12, nb13, nb0, nb1, nb2, nb3,
-             ggml_is_contiguous(src0), ggml_is_contiguous(src1), ggml_is_contiguous(dst), main_stream);
-    } else if (src0->type == GGML_TYPE_F16 && src1->type == GGML_TYPE_F16 && dst->type == GGML_TYPE_F16) {
-        op()((const sycl::half *) src0->data, (const sycl::half *) src1->data, (sycl::half *) dst->data, ne00, ne01,
-             ne02, ne03, ne10, ne11, ne12, ne13, ne0, ne1, ne2, ne3, nb00, nb01, nb02, nb03, nb10, nb11, nb12, nb13,
-             nb0, nb1, nb2, nb3, ggml_is_contiguous(src0), ggml_is_contiguous(src1), ggml_is_contiguous(dst),
-             main_stream);
-    } else if (src0->type == GGML_TYPE_F16 && src1->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F16) {
-        op()((const sycl::half *) src0->data, (const float *) src1->data, (sycl::half *) dst->data, ne00, ne01, ne02,
-             ne03, ne10, ne11, ne12, ne13, ne0, ne1, ne2, ne3, nb00, nb01, nb02, nb03, nb10, nb11, nb12, nb13, nb0, nb1,
-             nb2, nb3, ggml_is_contiguous(src0), ggml_is_contiguous(src1), ggml_is_contiguous(dst), main_stream);
-    } else if (src0->type == GGML_TYPE_I32 && src1->type == GGML_TYPE_I32 && dst->type == GGML_TYPE_I32) {
-        op()((const int32_t *) src0->data, (const int32_t *) src1->data, (int32_t *) dst->data, ne00, ne01, ne02, ne03,
-             ne10, ne11, ne12, ne13, ne0, ne1, ne2, ne3, nb00, nb01, nb02, nb03, nb10, nb11, nb12, nb13, nb0, nb1, nb2,
-             nb3, ggml_is_contiguous(src0), ggml_is_contiguous(src1), ggml_is_contiguous(dst), main_stream);
-    } else if (src0->type == GGML_TYPE_I16 && src1->type == GGML_TYPE_I16 && dst->type == GGML_TYPE_I16) {
-        op()((const int16_t *) src0->data, (const int16_t *) src1->data, (int16_t *) dst->data, ne00, ne01, ne02, ne03,
-             ne10, ne11, ne12, ne13, ne0, ne1, ne2, ne3, nb00, nb01, nb02, nb03, nb10, nb11, nb12, nb13, nb0, nb1, nb2,
-             nb3, ggml_is_contiguous(src0), ggml_is_contiguous(src1), ggml_is_contiguous(dst), main_stream);
-    } else {
-        fprintf(stderr, "%s: unsupported types: dst: %s, src0: %s, src1: %s\n", __func__, ggml_type_name(dst->type),
-                ggml_type_name(src0->type), ggml_type_name(src1->type));
-        GGML_ABORT("fatal error");
-    }
-}
-
-inline void ggml_sycl_op_add(ggml_backend_sycl_context & ctx, ggml_tensor *dst) {
-
-    ggml_sycl_op_bin_bcast<bin_bcast_sycl<op_add>>(ctx, dst->src[0], dst->src[1], dst);
-}
-
-inline void ggml_sycl_op_sub(ggml_backend_sycl_context & ctx, ggml_tensor *dst) {
-
-    ggml_sycl_op_bin_bcast<bin_bcast_sycl<op_sub>>(ctx, dst->src[0], dst->src[1], dst);
-}
-
-inline void ggml_sycl_op_mul(ggml_backend_sycl_context & ctx, ggml_tensor *dst) {
-
-    ggml_sycl_op_bin_bcast<bin_bcast_sycl<op_mul>>(ctx, dst->src[0], dst->src[1], dst);
-}
-
-inline void ggml_sycl_op_div(ggml_backend_sycl_context & ctx, ggml_tensor *dst) {
-
-    ggml_sycl_op_bin_bcast<bin_bcast_sycl<op_div>>(ctx, dst->src[0], dst->src[1], dst);
-}
-
-inline void ggml_sycl_op_repeat(ggml_backend_sycl_context & ctx, ggml_tensor *dst) {
-    ggml_sycl_op_bin_bcast<bin_bcast_sycl<op_repeat>>(ctx, dst, dst->src[0], dst);
-}
-
-
-void ggml_sycl_add(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_add(ctx, dst);
-    GGML_SYCL_DEBUG("call %s done\n", __func__);
-}
-
-void ggml_sycl_sub(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_sub(ctx, dst);
-    GGML_SYCL_DEBUG("call %s done\n", __func__);
-}
-
-void ggml_sycl_mul(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_mul(ctx, dst);
-    GGML_SYCL_DEBUG("call %s done\n", __func__);
-}
-
-void ggml_sycl_div(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_div(ctx, dst);
-    GGML_SYCL_DEBUG("call %s done\n", __func__);
-}
-
-void ggml_sycl_repeat(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    GGML_SYCL_DEBUG("call %s\n", __func__);
-    ggml_sycl_op_repeat(ctx, dst);
-    GGML_SYCL_DEBUG("call %s done\n", __func__);
-}
-

package/src/llama.cpp/ggml/src/ggml-sycl/binbcast.hpp

@@ -1,39 +0,0 @@
-#ifndef GGML_SYCL_BINBCAST_HPP
-#define GGML_SYCL_BINBCAST_HPP
-#include "common.hpp"
-
-
-static __dpct_inline__ float op_repeat(const float a, const float b) {
-    return b;
-    GGML_UNUSED(a);
-}
-
-static __dpct_inline__ float op_add(const float a, const float b) {
-    return a + b;
-}
-
-static __dpct_inline__ float op_sub(const float a, const float b) {
-    return a - b;
-}
-
-static __dpct_inline__ float op_mul(const float a, const float b) {
-    return a * b;
-}
-
-static __dpct_inline__ float op_div(const float a, const float b) {
-    return a / b;
-}
-
-void ggml_sycl_add(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-
-void ggml_sycl_sub(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-
-void ggml_sycl_mul(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-
-void ggml_sycl_div(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-
-void ggml_sycl_repeat(ggml_backend_sycl_context & ctx, ggml_tensor * dst);
-
-
-#endif //GGML_SYCL_BINBCAST_HPP
-

package/src/llama.cpp/ggml/src/ggml-sycl/common.cpp

@@ -1,83 +0,0 @@
-//
-// MIT license
-// Copyright (C) 2024 Intel Corporation
-// SPDX-License-Identifier: MIT
-//
-
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-
-#include "common.hpp"
-
-#include "ggml-backend-impl.h"
-#include "ggml-impl.h"
-
-int get_current_device_id() {
-  return dpct::dev_mgr::instance().current_device_id();
-}
-
-void* ggml_sycl_host_malloc(size_t size) try {
-  if (getenv("GGML_SYCL_NO_PINNED") != nullptr) {
-    return nullptr;
-  }
-
-  void* ptr = nullptr;
-  // allow to use dpct::get_in_order_queue() for host malloc
-  dpct::err0 err = CHECK_TRY_ERROR(
-      ptr = (void*)sycl::malloc_host(size, dpct::get_in_order_queue()));
-
-  if (err != 0) {
-    // clear the error
-    GGML_LOG_ERROR("WARNING: failed to allocate %.2f MB of pinned memory: %s\n", size / 1024.0 / 1024.0,    "syclGetErrorString is not supported");
-    return nullptr;
-  }
-
-  return ptr;
-} catch (sycl::exception const& exc) {
-  std::cerr << exc.what() << "Exception caught at file:" << __FILE__
-            << ", line:" << __LINE__ << std::endl;
-  std::exit(1);
-}
-
-void ggml_sycl_host_free(void* ptr) try {
-  // allow to use dpct::get_in_order_queue() for host malloc
-  SYCL_CHECK(CHECK_TRY_ERROR(sycl::free(ptr, dpct::get_in_order_queue())));
-} catch (sycl::exception const& exc) {
-  std::cerr << exc.what() << "Exception caught at file:" << __FILE__
-            << ", line:" << __LINE__ << std::endl;
-  std::exit(1);
-}
-
-bool gpu_has_xmx(sycl::device &dev) {
-    return dev.has(sycl::aspect::ext_intel_matrix);
-}
-
-int64_t downsample_sycl_global_range(int64_t accumulate_block_num, int64_t block_size) {
-  const int64_t max_range = std::numeric_limits<int>::max();
-  int64_t sycl_down_blk_size = block_size;
-  int64_t global_range = accumulate_block_num * sycl_down_blk_size;
-  while(global_range > max_range) {
-      sycl_down_blk_size /= 2;
-      global_range = accumulate_block_num * sycl_down_blk_size;
-  }
-  return sycl_down_blk_size;
-}
-
-void release_extra_gpu(ggml_tensor_extra_gpu * extra, std::vector<queue_ptr> streams) {
-    for (int i = 0; i < ggml_sycl_info().device_count; ++i) {
-        for (int64_t is = 0; is < GGML_SYCL_MAX_STREAMS; ++is) {
-            if (extra->events[i][is] != nullptr) {
-                SYCL_CHECK(CHECK_TRY_ERROR(dpct::destroy_event(extra->events[i][is])));
-            }
-        }
-        if (extra->data_device[i] != nullptr && streams.size()>0) {
-            ggml_sycl_set_device(i);
-            SYCL_CHECK(
-                CHECK_TRY_ERROR(sycl::free(extra->data_device[i], *(streams[i]))));
-        }
-    }
-    delete extra;
-}