frozenleaves/oneDNN
1
0
Fork 0
mirror of https://github.com/uxlfoundation/oneDNN.git synced 2025-10-20 18:43:49 +08:00
Code Issues Packages Projects Releases Wiki Activity

build: cmake: unsupported compilers cleanup

Browse Source
This commit is contained in:
Palicki, Stefan
2025-01-02 15:35:03 -08:00
committed by Stefan Palicki
parent c7827a8f3a
commit 664b22abd5
33 changed files with 171 additions and 277 deletions
Download Patch File Download Diff File Expand all files Collapse all files

19
README.md
View File

@ -295,9 +295,9 @@ Runtime-specific dependencies:
### Validated Configurations
CPU engine was validated on RedHat\* Enterprise Linux 8 with
* GNU Compiler Collection 5.4, 6.1, 7.2, 8.1, 9.1, 11.1, 11.3
* Clang\* 7.1, 8.0, 9.0, 14.0.6
x86-64 CPU engine was validated on RedHat\* Enterprise Linux 8 with
* GNU Compiler Collection 8.5, 9.5, 11.1, 11.3
* Clang\* 11.0, 14.0.6
* [Intel oneAPI DPC++/C++ Compiler] 2024.0
on Windows Server\* 2019 with
@ -307,16 +307,19 @@ on Windows Server\* 2019 with
on macOS 11 (Big Sur) with
* Apple LLVM version 13.0
on Ubuntu 20.04 AArch64 with
* GNU Compiler Collection 7.0, 8.0, 9.0, 10.0
* Clang\* 9.0, 17.0
AArch64 CPU engine was validated on Ubuntu 22.04 with
* GNU Compiler Collection 10.0, 13.0
* Clang\* 17.0
* [Arm Compiler for Linux] 24.04
* [Arm Compute Library (ACL)] built for armv8-a arch, latest stable version
available at the time of release
on macOS 14 (Sonoma) with
* Apple LLVM version 15.0
GPU engine was validated on Ubuntu\* 22.04 with
* GNU Compiler Collection 7.2, 8.1, and 9.1
* Clang 7.1, 8.0, 9.0
* GNU Compiler Collection 8.5, and 9.5
* Clang 11.0
* [Intel oneAPI DPC++/C++ Compiler] 2024.0
* [Intel Software for General Purpose GPU capabilities] latest stable version
available at the time of release

9
cmake/OpenMP.cmake
View File

@ -1,5 +1,5 @@
#===============================================================================
# Copyright 2017-2024 Intel Corporation
# Copyright 2017-2025 Intel Corporation
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
@ -52,13 +52,6 @@ if(NOT OpenMP_CXX_FOUND AND MSVC AND CMAKE_CXX_COMPILER_ID MATCHES "(Clang|Intel
# The ICX driver doesn't link OpenMP library even if `/Qopenmp`
# was specified.
set(OpenMP_FLAGS "/Qopenmp -Xclang --dependent-lib=libiomp5md")
else()
if(CMAKE_CXX_COMPILER_VERSION VERSION_LESS "10.0")
# version < 10 can't pass cl-style `/openmp` flag
set(OpenMP_FLAGS "-Xclang -fopenmp")
# ... and requires explicit linking against omp library
set(OpenMP_CXX_LIBRARIES "libomp.lib")
endif()
endif()
set(OpenMP_C_FLAGS ${OpenMP_FLAGS})
set(OpenMP_CXX_FLAGS ${OpenMP_FLAGS})

16
cmake/SDL.cmake
View File

@ -1,5 +1,5 @@
#===============================================================================
# Copyright 2017-2024 Intel Corporation
# Copyright 2017-2025 Intel Corporation
# Copyright 2021 FUJITSU LIMITED
#
# Licensed under the Apache License, Version 2.0 (the "License");
@ -30,14 +30,10 @@ macro(sdl_unix_common_ccxx_flags var)
append(${var} "-fPIC -Wformat -Wformat-security")
endmacro()
macro(sdl_gnu_common_ccxx_flags var gnu_version)
if(${gnu_version} VERSION_LESS 4.9)
append(${var} "-fstack-protector-all")
else()
append(${var} "-fstack-protector-strong")
if(NOT (${gnu_version} VERSION_LESS 8.0) AND (DNNL_TARGET_ARCH STREQUAL "X64"))
append(${var} "-fcf-protection=full")
endif()
macro(sdl_gnu_common_ccxx_flags var)
append(${var} "-fstack-protector-strong")
if(DNNL_TARGET_ARCH STREQUAL "X64")
append(${var} "-fcf-protection=full")
endif()
endmacro()
@ -65,7 +61,7 @@ if(UNIX)
append(ONEDNN_SDL_COMPILER_FLAGS "-D_FORTIFY_SOURCE=2")
endif()
if("${CMAKE_CXX_COMPILER_ID}" STREQUAL "GNU")
sdl_gnu_common_ccxx_flags(ONEDNN_SDL_COMPILER_FLAGS CMAKE_CXX_COMPILER_VERSION)
sdl_gnu_common_ccxx_flags(ONEDNN_SDL_COMPILER_FLAGS)
sdl_gnu_src_ccxx_flags(CMAKE_SRC_CCXX_FLAGS)
sdl_gnu_example_ccxx_flags(CMAKE_EXAMPLE_CCXX_FLAGS)
elseif(CMAKE_CXX_COMPILER_ID MATCHES "(Apple)?[Cc]lang")

8
cmake/coverage.cmake
View File

@ -1,5 +1,5 @@
#===============================================================================
# Copyright 2019-2024 Intel Corporation
# Copyright 2019-2025 Intel Corporation
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
@ -36,11 +36,7 @@ if("${DNNL_CODE_COVERAGE}" STREQUAL "GCOV")
message(FATAL_ERROR "GCOV not found in path")
endif()
if(CMAKE_CXX_COMPILER_ID MATCHES "(Apple)?[Cc]lang")
if("${CMAKE_CXX_COMPILER_VERSION}" VERSION_LESS 3)
message(FATAL_ERROR "Clang version must be 3.0.0 or greater! Aborting...")
endif()
elseif(NOT "${CMAKE_CXX_COMPILER_ID}" STREQUAL "GNU")
if(NOT CMAKE_CXX_COMPILER_ID MATCHES "(Apple)?[Cc]lang|GNU")
message(FATAL_ERROR "Unsupported compiler: ${CMAKE_CXX_COMPILER_ID}")
endif()

4
cmake/host_compiler.cmake
View File

@ -1,5 +1,5 @@
#===============================================================================
# Copyright 2021-2024 Intel Corporation
# Copyright 2021-2025 Intel Corporation
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
@ -78,7 +78,7 @@ if(DPCPP_HOST_COMPILER_KIND MATCHES "^(GNU|CLANG)$")
if(DPCPP_HOST_COMPILER_KIND STREQUAL "GNU")
platform_gnu_nowarn_ccxx_flags(DPCPP_CXX_NOWARN_FLAGS ${DPCPP_HOST_COMPILER_MAJOR_VER}.${DPCPP_HOST_COMPILER_MINOR_VER})
sdl_gnu_common_ccxx_flags(DPCPP_HOST_COMPILER_OPTS DPCPP_HOST_COMPILER_VER)
sdl_gnu_common_ccxx_flags(DPCPP_HOST_COMPILER_OPTS)
sdl_gnu_src_ccxx_flags(DPCPP_SRC_CXX_FLAGS)
sdl_gnu_example_ccxx_flags(DPCPP_EXAMPLE_CXX_FLAGS)

14
cmake/host_compiler_id.cmake
View File

@ -1,5 +1,5 @@
#===============================================================================
# Copyright 2024 Intel Corporation
# Copyright 2024-2025 Intel Corporation
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
@ -98,13 +98,13 @@ message(STATUS "Host compiler version: ${DPCPP_HOST_COMPILER_MAJOR_VER}.${DPCPP_
# Check the version of the provided host compiler.
if(DPCPP_HOST_COMPILER_KIND STREQUAL "GNU")
if((DPCPP_HOST_COMPILER_MAJOR_VER LESS 7) OR (DPCPP_HOST_COMPILER_MAJOR_VER EQUAL 7 AND DPCPP_HOST_COMPILER_MINOR_VER LESS 4))
message(FATAL_ERROR "The minimum version of ${DPCPP_HOST_COMPILER_KIND} host compiler is 7.4.")
endif()
endif()
if(DPCPP_HOST_COMPILER_KIND STREQUAL "CLANG")
if(DPCPP_HOST_COMPILER_MAJOR_VER LESS 8)
message(FATAL_ERROR "The minimum version of ${DPCPP_HOST_COMPILER_KIND} host compiler is 8.0.")
endif()
endif()
if(DPCPP_HOST_COMPILER_KIND STREQUAL "CLANG")
if(DPCPP_HOST_COMPILER_MAJOR_VER LESS 11)
message(FATAL_ERROR "The minimum version of ${DPCPP_HOST_COMPILER_KIND} host compiler is 11.0.")
endif()
endif()

11
cmake/platform.cmake
View File

@ -362,12 +362,6 @@ elseif(UNIX OR MINGW)
append(CMAKE_CCXX_FLAGS "-Wno-ignored-attributes")
endif()
# XXX: Suppress an erroneous warning of nested lambda visibility
# exceeding that of the containing class (GCC Bugzilla - Bug 80947).
if (CMAKE_CXX_COMPILER_VERSION VERSION_LESS 8 AND CMAKE_CXX_COMPILER_VERSION VERSION_GREATER 6.0)
append(CMAKE_CCXX_FLAGS "-Wno-attributes")
endif()
if(DNNL_TARGET_ARCH STREQUAL "AARCH64")
if (NOT CMAKE_BUILD_TYPE STREQUAL "Debug")
set(DEF_ARCH_OPT_FLAGS "-O3")
@ -486,10 +480,5 @@ if (DNNL_TARGET_ARCH STREQUAL "RV64")
message(STATUS "DNNL_RISCV_USE_RVV_INTRINSICS: ${DNNL_RISCV_USE_RVV_INTRINSICS}")
endif()
# Old compiler versions do not support warnings available on newer compilers.
if(CMAKE_CXX_COMPILER_ID MATCHES "(Apple)?[Cc]lang" AND CMAKE_CXX_COMPILER_VERSION VERSION_LESS 8.0.0)
append(CMAKE_CCXX_FLAGS "-Wno-unknown-warning-option")
endif()
append(CMAKE_C_FLAGS "${CMAKE_CCXX_FLAGS}")
append(CMAKE_CXX_FLAGS "${CMAKE_CCXX_FLAGS}")

40
src/common/compiler_workarounds.hpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2020-2024 Intel Corporation
* Copyright 2020-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -17,11 +17,6 @@
#ifndef COMPILER_WORKAROUNDS_HPP
#define COMPILER_WORKAROUNDS_HPP
#if (defined __GNUC__) && (!defined(__INTEL_COMPILER)) \
&& (!defined(__INTEL_LLVM_COMPILER)) && (!defined(__clang__major__))
#define NEED_GCC_WA_CHECK 1
#endif
// Workaround 01: clang.
//
// Clang has an issue [1] with `#pragma omp simd` that might lead to segfault.
@ -32,7 +27,7 @@
// vectorization for clang altogether for now.
//
// [1] https://bugs.llvm.org/show_bug.cgi?id=48104
#if (defined __clang_major__) && (__clang_major__ >= 6)
#if (defined __clang_major__) && (__clang_major__ < 13)
#define CLANG_WA_01_SAFE_TO_USE_OMP_SIMD 0
#else
#define CLANG_WA_01_SAFE_TO_USE_OMP_SIMD 1
@ -40,40 +35,15 @@
// Workaround 02: clang.
//
// Clang 6+ generates incorrect code with OMP_SIMD in some particular cases.
// Clang generates incorrect code with OMP_SIMD in some particular cases.
// Unlike CLANG_WA_01_SAFE_TO_USE_OMP_SIMD, the issue happens even with -O3.
#if (defined __clang_major__) && (__clang_major__ >= 6)
#if (defined __clang_major__) && (__clang_major__ < 13)
#define CLANG_WA_02_SAFE_TO_USE_OMP_SIMD 0
#else
#define CLANG_WA_02_SAFE_TO_USE_OMP_SIMD 1
#endif
// Workaround 03: GCC
//
// For very large functions with too much control flow (i.e. if, switch, goto
// statements), GCC 7 may struggle to perform optimizations based on tree
// dominator (i.e. -ftree-dominator-opts, which is enabled with O1), thereby
// producing an internal compiler error (ICE). Specifically, it seems that the
// jump threading optimization is the culprit, which cannot be disabled on its
// own. There is no reliable way to reproduce the ICE, therefore it is not clear
// which __GCC_MINOR__ version fixes issue.
#if (defined NEED_GCC_WA_CHECK) && (__GNUC__ == 7)
#define GCC_WA_NO_TREE_DOMINATOR_OPTS 1
#else
#define GCC_WA_NO_TREE_DOMINATOR_OPTS 0
#endif
// Workaround 05: GCC
//
// NOTE: inside lambda, type cast variables captured by reference using
// either c-like "(type)var" or functional "type(var)" notation in order
// to avoid gcc7 bug with c++14 standard
// (https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83204).
#if (defined NEED_GCC_WA_CHECK) && (__GNUC__ <= 7)
#define GCC_WA_LAMBDA_C_CAST
#endif
// Workaround 05: c++17 vs c++20
// Workaround 03: MSVC c++17 vs c++20
//
// C++17/20 are contradictory wrt capturing this and using default '=' capture.
// - C++17 and before have to return a warning for the [=, this] capture as

14
src/common/dnnl_thread.hpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2017-2023 Intel Corporation
* Copyright 2017-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -192,18 +192,6 @@ inline int dnnl_get_current_num_threads() {
#define PRAGMA_OMP_SIMD(...) PRAGMA_MACRO(CHAIN2(omp, simd __VA_ARGS__))
#endif // defined(_MSC_VER) && !defined(__clang__) && !defined(__INTEL_COMPILER)
// process simdlen; it is supported for Clang >= 3.9; ICC >= 17.0; GCC >= 6.1
// No support on Windows.
#if (defined(__clang_major__) \
&& (__clang_major__ < 3 \
|| (__clang_major__ == 3 && __clang_minor__ < 9))) \
|| (defined(__INTEL_COMPILER) && __INTEL_COMPILER < 1700) \
|| (!defined(__INTEL_COMPILER) && !defined(__clang__) \
&& (defined(_MSC_VER) || __GNUC__ < 6 \
|| (__GNUC__ == 6 && __GNUC_MINOR__ < 1)))
#define simdlen(x)
#endif // long simdlen if
namespace dnnl {
namespace impl {

7
src/common/primitive_attr_quant.hpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2024 Intel Corporation
* Copyright 2024-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -42,10 +42,7 @@ struct runtime_scales_t;
const runtime_scales_t &default_runtime_scale();
struct runtime_scales_t : public c_compatible {
// Clang-3.8.1 raises an error for a default initialization of a const
// object. Const runtime_scales_t object is used as default_scales.
// runtime_scales_t() = default;
runtime_scales_t() {}
runtime_scales_t() = default;
runtime_scales_t &operator=(const runtime_scales_t &rhs) {
mask_ = rhs.mask_;

6
src/cpu/matmul/gemm_bf16_matmul.cpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2019-2024 Intel Corporation
* Copyright 2019-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -271,11 +271,7 @@ status_t gemm_bf16_matmul_t<dst_type>::execute_ref(
const dim_t acc_stride = gemm_based::get_scratchpad_block_elements(
batch, M, N, use_single_gemm_call, nthr);
#ifdef GCC_WA_LAMBDA_C_CAST
parallel(nthr, [= WA_THIS_COPY_CAPTURE, &st](int ithr, int nthr) {
#else
parallel(nthr, [&](int ithr, int nthr) {
#endif
size_t t_work_start {0}, t_work_end {0};
balance211(work_amount, nthr, ithr, t_work_start, t_work_end);

6
src/cpu/matmul/gemm_x8s8s32x_matmul.cpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2019-2024 Intel Corporation
* Copyright 2019-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -297,11 +297,7 @@ status_t gemm_x8s8s32x_matmul_t::execute_ref(const exec_ctx_t &ctx) const {
bool postops_in_matmul = need_post_processing(pd(), dst_zero_point_f32);
assert(IMPLICATION(postops_in_matmul, params.has_pp_kernel_));
#ifdef GCC_WA_LAMBDA_C_CAST
parallel(nthr, [= WA_THIS_COPY_CAPTURE, &st](int ithr, int nthr) {
#else
parallel(nthr, [&](int ithr, int nthr) {
#endif
size_t t_work_start {0}, t_work_end {0};
balance211(work_amount, nthr, ithr, t_work_start, t_work_end);

4
src/cpu/x64/gemm/amx/jit_avx512_core_amx_copy_kern.hpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2020-2021 Intel Corporation
* Copyright 2020-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -66,7 +66,7 @@ protected:
void copy_n(int unroll_n, Xbyak::Label &epilogue);
void copy_ns(int unroll_n, Xbyak::Label &epilogue);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
private:
static const int offset_a_ = 0, offset_b_ = 0;

4
src/cpu/x64/gemm/amx/jit_avx512_core_amx_gemm_kern.hpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2020-2021 Intel Corporation
* Copyright 2020-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -31,7 +31,7 @@ public:
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_amx_gemm_kern);
protected:
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
const int typea;
const int typeb;
const int typec;

18
src/cpu/x64/gemm/bf16/common_s16.hpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2019-2024 Intel Corporation
* Copyright 2019-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -26,7 +26,7 @@ namespace x64 {
class jit_avx512_core_s16_48x8_copy_an_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_s16_48x8_copy_an_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx512_core_s16_48x8_copy_an_kern();
@ -34,7 +34,7 @@ public:
class jit_avx512_core_s16_48x8_copy_at_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_s16_48x8_copy_at_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx512_core_s16_48x8_copy_at_kern();
@ -42,7 +42,7 @@ public:
class jit_avx512_core_s16_48x8_copy_bn_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_s16_48x8_copy_bn_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx512_core_s16_48x8_copy_bn_kern();
@ -50,7 +50,7 @@ public:
class jit_avx512_core_s16_48x8_copy_bt_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_s16_48x8_copy_bt_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx512_core_s16_48x8_copy_bt_kern();
@ -58,7 +58,7 @@ public:
class jit_avx512_core_s16_24x8_copy_an_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_s16_24x8_copy_an_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx512_core_s16_24x8_copy_an_kern();
@ -66,7 +66,7 @@ public:
class jit_avx512_core_s16_24x8_copy_at_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_s16_24x8_copy_at_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx512_core_s16_24x8_copy_at_kern();
@ -74,7 +74,7 @@ public:
class jit_avx512_core_s16_24x8_copy_bn_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_s16_24x8_copy_bn_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx512_core_s16_24x8_copy_bn_kern();
@ -82,7 +82,7 @@ public:
class jit_avx512_core_s16_24x8_copy_bt_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_s16_24x8_copy_bt_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx512_core_s16_24x8_copy_bt_kern();

4
src/cpu/x64/gemm/bf16/jit_avx512_core_gemm_bf16bf16f32_kern.hpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2019-2021 Intel Corporation
* Copyright 2019-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -58,7 +58,7 @@ protected:
void innerloop(int unroll_m, int unroll_n);
void outerloop(int unroll_x, int unroll_y, Xbyak::Label *&outerloop_label);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
private:
static const int UNROLL_N_ = 8;

4
src/cpu/x64/gemm/bf16/jit_avx512_core_gemv_bf16bf16f32_kern.hpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2020-2021 Intel Corporation
* Copyright 2020-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -52,7 +52,7 @@ protected:
void outerloop(int unroll_y, Xbyak::Label *&cur_outerloop_label,
Xbyak::Label *&outerloop_end_label);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
private:
static const int UNROLL_M_ = 64;

59
src/cpu/x64/gemm/f32/common_f32.hpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2019-2024 Intel Corporation
* Copyright 2019-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -26,7 +26,7 @@ namespace x64 {
class jit_avx512_core_f32_copy_an_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_f32_copy_an_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx512_core_f32_copy_an_kern();
@ -34,11 +34,10 @@ public:
class jit_avx512_core_f32_copy_at_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_f32_copy_at_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
void generate_part1(const Xbyak::Label &, const Xbyak::Label &,
const Xbyak::Label &, const Xbyak::Label &) ATTRIBUTE_OPTIMIZE;
void generate_part2(Xbyak::Label, Xbyak::Label, Xbyak::Label,
Xbyak::Label) ATTRIBUTE_OPTIMIZE;
const Xbyak::Label &, const Xbyak::Label &);
void generate_part2(Xbyak::Label, Xbyak::Label, Xbyak::Label, Xbyak::Label);
public:
jit_avx512_core_f32_copy_at_kern();
@ -46,7 +45,7 @@ public:
class jit_avx512_core_f32_copy_bn_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_f32_copy_bn_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx512_core_f32_copy_bn_kern();
@ -54,7 +53,7 @@ public:
class jit_avx512_core_f32_copy_bt_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_f32_copy_bt_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx512_core_f32_copy_bt_kern();
@ -62,7 +61,7 @@ public:
class jit_avx2_f32_copy_an_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_f32_copy_an_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_f32_copy_an_kern();
@ -70,7 +69,7 @@ public:
class jit_avx2_f32_copy_at_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_f32_copy_at_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_f32_copy_at_kern();
@ -78,7 +77,7 @@ public:
class jit_avx2_f32_copy_bn_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_f32_copy_bn_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_f32_copy_bn_kern();
@ -86,7 +85,7 @@ public:
class jit_avx2_f32_copy_bt_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_f32_copy_bt_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_f32_copy_bt_kern();
@ -94,7 +93,7 @@ public:
class jit_avx_f32_copy_an_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_f32_copy_an_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_f32_copy_an_kern();
@ -102,7 +101,7 @@ public:
class jit_avx_f32_copy_at_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_f32_copy_at_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_f32_copy_at_kern();
@ -110,7 +109,7 @@ public:
class jit_avx_f32_copy_bn_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_f32_copy_bn_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_f32_copy_bn_kern();
@ -118,7 +117,7 @@ public:
class jit_avx_f32_copy_bt_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_f32_copy_bt_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_f32_copy_bt_kern();
@ -126,11 +125,10 @@ public:
class jit_avx_kernel_b0_sgemm_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_kernel_b0_sgemm_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
void generate_part1(const Xbyak::Label &, const Xbyak::Label &,
const Xbyak::Label &, const Xbyak::Label &) ATTRIBUTE_OPTIMIZE;
void generate_part2(Xbyak::Label, Xbyak::Label, Xbyak::Label,
Xbyak::Label) ATTRIBUTE_OPTIMIZE;
const Xbyak::Label &, const Xbyak::Label &);
void generate_part2(Xbyak::Label, Xbyak::Label, Xbyak::Label, Xbyak::Label);
public:
jit_avx_kernel_b0_sgemm_kern();
@ -138,11 +136,10 @@ public:
class jit_avx_kernel_sgemm_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_kernel_sgemm_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate_part1(const Xbyak::Label &, const Xbyak::Label &,
const Xbyak::Label &) ATTRIBUTE_OPTIMIZE;
void generate_part2(
Xbyak::Label &, Xbyak::Label &, Xbyak::Label &) ATTRIBUTE_OPTIMIZE;
void generate() override;
void generate_part1(
const Xbyak::Label &, const Xbyak::Label &, const Xbyak::Label &);
void generate_part2(Xbyak::Label &, Xbyak::Label &, Xbyak::Label &);
public:
jit_avx_kernel_sgemm_kern();
@ -150,7 +147,7 @@ public:
class jit_sse41_f32_copy_an_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_f32_copy_an_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_f32_copy_an_kern();
@ -158,7 +155,7 @@ public:
class jit_sse41_f32_copy_at_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_f32_copy_at_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_f32_copy_at_kern();
@ -166,7 +163,7 @@ public:
class jit_sse41_f32_copy_bn_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_f32_copy_bn_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_f32_copy_bn_kern();
@ -174,7 +171,7 @@ public:
class jit_sse41_f32_copy_bt_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_f32_copy_bt_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_f32_copy_bt_kern();
@ -182,7 +179,7 @@ public:
class jit_sse41_kernel_b0_sgemm_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_kernel_b0_sgemm_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_kernel_b0_sgemm_kern();
@ -190,7 +187,7 @@ public:
class jit_sse41_kernel_sgemm_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_kernel_sgemm_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_kernel_sgemm_kern();

4
src/cpu/x64/gemm/f32/jit_avx2_kernel_sgemm_kern.hpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2019-2024 Intel Corporation
* Copyright 2019-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -79,7 +79,7 @@ class jit_avx2_kernel_sgemm_kern : public jit_generator {
void prefetchA_beforeFMA(int um, int un, int k_idx, int n_idx, int m_idx);
void prefetchC_afterBload(int um, int un, int k_idx, int n_idx);
void prefetchC_beforeKloop(int um);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
template <typename T_reg, typename T_desta, typename T_srca>
void loadA_betweenFMAs(int um, int un, int k_idx, int n_idx, int m_idx,

4
src/cpu/x64/gemm/f32/jit_avx512_common_gemm_f32.cpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2017-2024 Intel Corporation
* Copyright 2017-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -70,7 +70,7 @@ struct xbyak_gemm_t : public jit_generator {
, hasBias(hasBias)
, STACK_K_CAPACITY((STACK_CAPACITY - 256) / (SIZE * UNROLL_M)) {}
void generate() override ATTRIBUTE_OPTIMIZE {
void generate() override {
using namespace Xbyak;
bool isBeta0 = (beta == 0.0);
bool isBetaN = (!isBeta0 && beta != 1.0);

4
src/cpu/x64/gemm/f32/jit_avx512_core_gemm_smalln_tn_f32_kern.cpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2020-2024 Intel Corporation
* Copyright 2020-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -49,7 +49,7 @@ struct xbyak_gemm_smalln_tn_t : public jit_generator {
xbyak_gemm_smalln_tn_t(int N, float beta, float alpha)
: jit_generator(jit_name()), N(N), beta(beta), alpha(alpha) {}
void generate() override ATTRIBUTE_OPTIMIZE {
void generate() override {
using namespace Xbyak;
/**
* numN = 1 : 16 rows of A, 1x16 accumulators

4
src/cpu/x64/gemm/f32/jit_avx_gemm_f32.cpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2016-2024 Intel Corporation
* Copyright 2016-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -1966,7 +1966,7 @@ struct xbyak_gemm_t : public jit_generator {
if (hasBias) { add(BIAS, unroll_m * SIZE); }
}
void generate() override ATTRIBUTE_OPTIMIZE {
void generate() override {
assert(IMPLICATION(!is_avx2, mayiuse(avx)));
preamble();

4
src/cpu/x64/gemm/f32/jit_avx_gemv_t_f32_kern.hpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2019-2021 Intel Corporation
* Copyright 2019-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -40,7 +40,7 @@ protected:
void innerloop(int unroll_m, int unroll_n);
void outerloop(int unroll_x, int unroll_y, Xbyak::Label *&outerloop_label);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
private:
static const int M_UNROLL_ = 16;

4
src/cpu/x64/gemm/f32/jit_sse41_gemv_n_f32_kern.hpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2021 Intel Corporation
* Copyright 2021-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -61,7 +61,7 @@ protected:
Xbyak::Label *&cur_outerloop_label,
Xbyak::Label *&outerloop_end_label);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
private:
static const int max_um_vecs_ = 16;

4
src/cpu/x64/gemm/f32/jit_sse41_gemv_t_f32_kern.hpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2019-2021 Intel Corporation
* Copyright 2019-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -38,7 +38,7 @@ protected:
void innerloop(int unroll_m, int unroll_n);
void outerloop(int unroll_x, int unroll_y, Xbyak::Label *&outerloop_label);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
private:
static const int M_UNROLL_ = 8;

114
src/cpu/x64/gemm/s8x8s32/common_u8.hpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2019-2024 Intel Corporation
* Copyright 2019-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -32,7 +32,7 @@ namespace x64 {
class jit_avx512_core_u8_copy_an_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_u8_copy_an_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx512_core_u8_copy_an_kern();
@ -40,7 +40,7 @@ public:
class jit_avx512_core_u8_copy_at_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_u8_copy_at_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx512_core_u8_copy_at_kern();
@ -48,7 +48,7 @@ public:
class jit_avx512_core_u8_copy_bn_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_u8_copy_bn_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
bool s8_case;
public:
@ -57,7 +57,7 @@ public:
class jit_avx512_core_u8_copy_bt_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_u8_copy_bt_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
bool s8_case;
public:
@ -66,7 +66,7 @@ public:
class jit_avx512_core_u8_copy_sum_an_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_u8_copy_sum_an_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx512_core_u8_copy_sum_an_kern();
@ -74,7 +74,7 @@ public:
class jit_avx512_core_u8_copy_sum_at_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_u8_copy_sum_at_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx512_core_u8_copy_sum_at_kern();
@ -82,7 +82,7 @@ public:
class jit_avx512_core_u8_copy_sum_bn_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_u8_copy_sum_bn_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
bool s8_case;
public:
@ -91,7 +91,7 @@ public:
class jit_avx512_core_u8_copy_sum_bt_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx512_core_u8_copy_sum_bt_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
bool s8_case;
public:
@ -100,7 +100,7 @@ public:
class jit_avx2_vnni_u8_copy_an_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_vnni_u8_copy_an_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_vnni_u8_copy_an_kern();
@ -108,7 +108,7 @@ public:
class jit_avx2_vnni_u8_copy_at_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_vnni_u8_copy_at_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_vnni_u8_copy_at_kern();
@ -116,7 +116,7 @@ public:
class jit_avx2_vnni_u8_copy_bn_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_vnni_u8_copy_bn_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_vnni_u8_copy_bn_kern();
@ -124,7 +124,7 @@ public:
class jit_avx2_vnni_u8_copy_bt_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_vnni_u8_copy_bt_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_vnni_u8_copy_bt_kern();
@ -132,7 +132,7 @@ public:
class jit_avx2_vnni_u8_copy_sum_an_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_vnni_u8_copy_sum_an_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_vnni_u8_copy_sum_an_kern();
@ -140,7 +140,7 @@ public:
class jit_avx2_vnni_u8_copy_sum_at_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_vnni_u8_copy_sum_at_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_vnni_u8_copy_sum_at_kern();
@ -148,7 +148,7 @@ public:
class jit_avx2_vnni_u8_copy_sum_bn_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_vnni_u8_copy_sum_bn_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_vnni_u8_copy_sum_bn_kern();
@ -156,7 +156,7 @@ public:
class jit_avx2_vnni_u8_copy_sum_bt_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_vnni_u8_copy_sum_bt_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_vnni_u8_copy_sum_bt_kern();
@ -164,7 +164,7 @@ public:
class jit_avx2_u8_copy_an_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_u8_copy_an_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_u8_copy_an_kern();
@ -172,7 +172,7 @@ public:
class jit_avx2_u8_copy_at_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_u8_copy_at_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_u8_copy_at_kern();
@ -180,7 +180,7 @@ public:
class jit_avx2_u8_copy_bn_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_u8_copy_bn_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_u8_copy_bn_kern();
@ -188,7 +188,7 @@ public:
class jit_avx2_u8_copy_bt_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_u8_copy_bt_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_u8_copy_bt_kern();
@ -196,7 +196,7 @@ public:
class jit_avx2_u8_copy_sum_an_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_u8_copy_sum_an_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_u8_copy_sum_an_kern();
@ -204,7 +204,7 @@ public:
class jit_avx2_u8_copy_sum_at_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_u8_copy_sum_at_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_u8_copy_sum_at_kern();
@ -212,7 +212,7 @@ public:
class jit_avx2_u8_copy_sum_bn_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_u8_copy_sum_bn_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_u8_copy_sum_bn_kern();
@ -220,7 +220,7 @@ public:
class jit_avx2_u8_copy_sum_bt_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx2_u8_copy_sum_bt_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx2_u8_copy_sum_bt_kern();
@ -228,7 +228,7 @@ public:
class jit_avx_u8_copy_an_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_u8_copy_an_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_u8_copy_an_kern();
@ -236,7 +236,7 @@ public:
class jit_avx_u8_copy_at_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_u8_copy_at_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_u8_copy_at_kern();
@ -244,7 +244,7 @@ public:
class jit_avx_u8_copy_bn_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_u8_copy_bn_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_u8_copy_bn_kern();
@ -252,7 +252,7 @@ public:
class jit_avx_u8_copy_bt_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_u8_copy_bt_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_u8_copy_bt_kern();
@ -260,7 +260,7 @@ public:
class jit_avx_u8_copy_sum_an_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_u8_copy_sum_an_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_u8_copy_sum_an_kern();
@ -268,7 +268,7 @@ public:
class jit_avx_u8_copy_sum_at_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_u8_copy_sum_at_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_u8_copy_sum_at_kern();
@ -276,7 +276,7 @@ public:
class jit_avx_u8_copy_sum_bn_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_u8_copy_sum_bn_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_u8_copy_sum_bn_kern();
@ -284,7 +284,7 @@ public:
class jit_avx_u8_copy_sum_bt_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_u8_copy_sum_bt_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_u8_copy_sum_bt_kern();
@ -292,7 +292,7 @@ public:
class jit_avx_kernel_b0_gemm_s8u8s32_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_kernel_b0_gemm_s8u8s32_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_kernel_b0_gemm_s8u8s32_kern();
@ -300,7 +300,7 @@ public:
class jit_avx_kernel_b0_b_gemm_s8u8s32_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_kernel_b0_b_gemm_s8u8s32_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_kernel_b0_b_gemm_s8u8s32_kern();
@ -308,7 +308,7 @@ public:
class jit_avx_kernel_b0_r_gemm_s8u8s32_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_kernel_b0_r_gemm_s8u8s32_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_kernel_b0_r_gemm_s8u8s32_kern();
@ -316,7 +316,7 @@ public:
class jit_avx_kernel_b0_c_gemm_s8u8s32_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_kernel_b0_c_gemm_s8u8s32_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_kernel_b0_c_gemm_s8u8s32_kern();
@ -324,7 +324,7 @@ public:
class jit_avx_kernel_gemm_s8u8s32_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_kernel_gemm_s8u8s32_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_kernel_gemm_s8u8s32_kern();
@ -332,7 +332,7 @@ public:
class jit_avx_kernel_b_gemm_s8u8s32_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_kernel_b_gemm_s8u8s32_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_kernel_b_gemm_s8u8s32_kern();
@ -340,7 +340,7 @@ public:
class jit_avx_kernel_r_gemm_s8u8s32_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_kernel_r_gemm_s8u8s32_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_kernel_r_gemm_s8u8s32_kern();
@ -348,7 +348,7 @@ public:
class jit_avx_kernel_c_gemm_s8u8s32_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_avx_kernel_c_gemm_s8u8s32_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_avx_kernel_c_gemm_s8u8s32_kern();
@ -356,7 +356,7 @@ public:
class jit_sse41_u8_copy_an_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_u8_copy_an_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_u8_copy_an_kern();
@ -364,7 +364,7 @@ public:
class jit_sse41_u8_copy_at_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_u8_copy_at_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_u8_copy_at_kern();
@ -372,7 +372,7 @@ public:
class jit_sse41_u8_copy_bn_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_u8_copy_bn_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_u8_copy_bn_kern();
@ -380,7 +380,7 @@ public:
class jit_sse41_u8_copy_bt_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_u8_copy_bt_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_u8_copy_bt_kern();
@ -388,7 +388,7 @@ public:
class jit_sse41_u8_copy_sum_an_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_u8_copy_sum_an_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_u8_copy_sum_an_kern();
@ -396,7 +396,7 @@ public:
class jit_sse41_u8_copy_sum_at_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_u8_copy_sum_at_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_u8_copy_sum_at_kern();
@ -404,7 +404,7 @@ public:
class jit_sse41_u8_copy_sum_bn_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_u8_copy_sum_bn_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_u8_copy_sum_bn_kern();
@ -412,7 +412,7 @@ public:
class jit_sse41_u8_copy_sum_bt_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_u8_copy_sum_bt_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_u8_copy_sum_bt_kern();
@ -420,7 +420,7 @@ public:
class jit_sse41_kernel_b0_gemm_s8u8s32_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_kernel_b0_gemm_s8u8s32_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_kernel_b0_gemm_s8u8s32_kern();
@ -428,7 +428,7 @@ public:
class jit_sse41_kernel_b0_b_gemm_s8u8s32_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_kernel_b0_b_gemm_s8u8s32_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_kernel_b0_b_gemm_s8u8s32_kern();
@ -436,7 +436,7 @@ public:
class jit_sse41_kernel_b0_r_gemm_s8u8s32_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_kernel_b0_r_gemm_s8u8s32_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_kernel_b0_r_gemm_s8u8s32_kern();
@ -444,7 +444,7 @@ public:
class jit_sse41_kernel_b0_c_gemm_s8u8s32_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_kernel_b0_c_gemm_s8u8s32_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_kernel_b0_c_gemm_s8u8s32_kern();
@ -452,7 +452,7 @@ public:
class jit_sse41_kernel_gemm_s8u8s32_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_kernel_gemm_s8u8s32_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_kernel_gemm_s8u8s32_kern();
@ -460,7 +460,7 @@ public:
class jit_sse41_kernel_b_gemm_s8u8s32_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_kernel_b_gemm_s8u8s32_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_kernel_b_gemm_s8u8s32_kern();
@ -468,7 +468,7 @@ public:
class jit_sse41_kernel_r_gemm_s8u8s32_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_kernel_r_gemm_s8u8s32_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_kernel_r_gemm_s8u8s32_kern();
@ -476,7 +476,7 @@ public:
class jit_sse41_kernel_c_gemm_s8u8s32_kern : public jit_generator {
DECLARE_CPU_JIT_AUX_FUNCTIONS(jit_sse41_kernel_c_gemm_s8u8s32_kern);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
public:
jit_sse41_kernel_c_gemm_s8u8s32_kern();

4
src/cpu/x64/gemm/s8x8s32/jit_avx2_gemm_s8u8s32_kern.hpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2018-2021 Intel Corporation
* Copyright 2018-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -51,7 +51,7 @@ protected:
void innerloop(int unroll_m, int unroll_n);
void outerloop(int unroll_x, int unroll_y, Xbyak::Label *&outerloop_label);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
private:
static const int IGEMM_UNROLL_N_ = 4;

4
src/cpu/x64/gemm/s8x8s32/jit_avx512_core_gemm_s8u8s32_kern.hpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2018-2021 Intel Corporation
* Copyright 2018-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -50,7 +50,7 @@ protected:
void innerloop(int unroll_m, int unroll_n);
void outerloop(int unroll_x, int unroll_y, Xbyak::Label *&outerloop_label);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
private:
static const int IGEMM_UNROLL_M_ = 48;

4
src/cpu/x64/gemm/s8x8s32/jit_avx512_core_kernel_gemv_s8x8s32_kern.hpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2019-2024 Intel Corporation
* Copyright 2019-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -44,7 +44,7 @@ class jit_avx512_core_gemv_s8x8s32_kern : public jit_generator {
void shuffle_and_add(
Xbyak::Zmm, Xbyak::Zmm, Xbyak::Zmm, Xbyak::Zmm, Xbyak::Zmm);
void update_c(int, Xbyak::Reg64, int, Xbyak::Opmask);
void generate() override ATTRIBUTE_OPTIMIZE;
void generate() override;
cpu_isa_t isa = isa_undef;
ver_t ver = ver_t::undef;

6
src/cpu/x64/jit_generator.hpp
View File

@ -39,12 +39,6 @@
#define OFFSET_SHADOWSPACE 0x28
#endif
#if GCC_WA_NO_TREE_DOMINATOR_OPTS
#define ATTRIBUTE_OPTIMIZE __attribute__((optimize("no-tree-dominator-opts")))
#else
#define ATTRIBUTE_OPTIMIZE
#endif
#define DECLARE_CPU_JIT_AUX_FUNCTIONS(gen_name) \
const char *name() const override { return STRINGIFY(gen_name); } \
const char *source_file() const override { return __FILE__; } \

13
src/cpu/x64/jit_uni_xf16_sum.cpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2019-2023 Intel Corporation
* Copyright 2019-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -436,13 +436,6 @@ status_t jit_xf16_sum_t<src_data_type, dst_data_type, isa>::execute(
const dim_t num_blocks = nelems / num_elems_in_block;
const dim_t tail = nelems % num_elems_in_block;
#if defined(__GNUC__) && __GNUC__ == 4 && __GNUC_MINOR__ == 8 \
&& __GNUC_PATCHLEVEL__ == 3
// GCC issues a false positive warning 'array subscript is above array bounds'
// with gcc 4.8.3 + -march=native option, so disable it for now
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Warray-bounds"
#endif
parallel(0, [&](const int ithr, const int nthr) {
dim_t start {0}, end {0};
balance211(num_blocks, nthr, ithr, start, end);
@ -477,10 +470,6 @@ status_t jit_xf16_sum_t<src_data_type, dst_data_type, isa>::execute(
(*kernel_)(&arg);
}
});
#if defined(__GNUC__) && __GNUC__ == 4 && __GNUC_MINOR__ == 8 \
&& __GNUC_PATCHLEVEL__ == 3
#pragma GCC diagnostic pop
#endif
return status::success;
}

12
src/gpu/intel/jit/codegen/bank_conflict_allocation.cpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2021-2024 Intel Corporation
* Copyright 2021-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -25,12 +25,6 @@
#include <vector>
#include <initializer_list>
#if defined(__GNUC__) && __GNUC__ == 7
// GCC 7.x issues a false positive warning 'array subscript is above array bounds'
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Warray-bounds"
#endif
namespace dnnl {
namespace impl {
namespace gpu {
@ -849,7 +843,3 @@ bank_conflict_allocation_t bank_conflict_allocation_t::create(
} // namespace gpu
} // namespace impl
} // namespace dnnl
#if defined(__GNUC__) && __GNUC__ == 7
#pragma GCC diagnostic pop
#endif

16
src/graph/backend/graph_compiler/core/src/runtime/microkernel/cpu/brgemm_onednn.cpp
View File

@ -1,5 +1,5 @@
/*******************************************************************************
* Copyright 2020-2024 Intel Corporation
* Copyright 2020-2025 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -825,7 +825,7 @@ SC_API void dnnl_brgemm_call(brgemm_kernel_info *brg_desc, const void *A,
brgemm_batch_element_t *batch = (brgemm_batch_element_t *)_malloca(
num * sizeof(brgemm_batch_element_t));
#else
#if SC_IS_DPCPP() || (defined(CLANGVERSION) && CLANGVERSION <= 3)
#if SC_IS_DPCPP()
std::unique_ptr<brgemm_batch_element_t[]> batch_v(
new brgemm_batch_element_t[num]);
brgemm_batch_element_t *batch = batch_v.get();
@ -876,7 +876,7 @@ SC_API void dnnl_brgemm_call_postops(brgemm_kernel_info *brg_desc,
brgemm_batch_element_t *batch = (brgemm_batch_element_t *)_malloca(
num * sizeof(brgemm_batch_element_t));
#else
#if SC_IS_DPCPP() || (defined(CLANGVERSION) && CLANGVERSION <= 3)
#if SC_IS_DPCPP()
std::unique_ptr<brgemm_batch_element_t[]> batch_v(
new brgemm_batch_element_t[num]);
brgemm_batch_element_t *batch = batch_v.get();
@ -930,7 +930,7 @@ SC_API void dnnl_brgemm_list_call(brgemm_kernel_info *brg_desc,
brgemm_batch_element_t *batch = (brgemm_batch_element_t *)_malloca(
batch_num * sizeof(brgemm_batch_element_t));
#else
#if SC_IS_DPCPP() || (defined(CLANGVERSION) && CLANGVERSION <= 3)
#if SC_IS_DPCPP()
std::unique_ptr<brgemm_batch_element_t[]> batch_v(
new brgemm_batch_element_t[batch_num]);
brgemm_batch_element_t *batch = batch_v.get();
@ -995,7 +995,7 @@ SC_API void dnnl_brgemm_list_call_postops(brgemm_kernel_info *brg_desc,
brgemm_batch_element_t *batch = (brgemm_batch_element_t *)_malloca(
batch_num * sizeof(brgemm_batch_element_t));
#else
#if SC_IS_DPCPP() || (defined(CLANGVERSION) && CLANGVERSION <= 3)
#if SC_IS_DPCPP()
std::unique_ptr<brgemm_batch_element_t[]> batch_v(
new brgemm_batch_element_t[batch_num]);
brgemm_batch_element_t *batch = batch_v.get();
@ -1087,7 +1087,7 @@ SC_API int dnnl_brgemm_init_update(const void *A, const void *B, void *C,
brgemm_batch_element_t *batch = (brgemm_batch_element_t *)_malloca(
batch_num * sizeof(brgemm_batch_element_t));
#else
#if SC_IS_DPCPP() || (defined(CLANGVERSION) && CLANGVERSION <= 3)
#if SC_IS_DPCPP()
std::unique_ptr<brgemm_batch_element_t[]> batch_v(
new brgemm_batch_element_t[batch_num]);
brgemm_batch_element_t *batch = batch_v.get();
@ -1154,7 +1154,7 @@ SC_API int dnnl_brgemm_update(const void *A, const void *B, void *C, int num,
brgemm_batch_element_t *batch = (brgemm_batch_element_t *)_malloca(
batch_num * sizeof(brgemm_batch_element_t));
#else
#if SC_IS_DPCPP() || (defined(CLANGVERSION) && CLANGVERSION <= 3)
#if SC_IS_DPCPP()
std::unique_ptr<brgemm_batch_element_t[]> batch_v(
new brgemm_batch_element_t[batch_num]);
brgemm_batch_element_t *batch = batch_v.get();
@ -1214,7 +1214,7 @@ static int dnnl_brgemm_list_update_func(const void **A_list,
brgemm_batch_element_t *batch = (brgemm_batch_element_t *)_malloca(
batch_num * sizeof(brgemm_batch_element_t));
#else
#if SC_IS_DPCPP() || (defined(CLANGVERSION) && CLANGVERSION <= 3)
#if SC_IS_DPCPP()
std::unique_ptr<brgemm_batch_element_t[]> batch_v(
new brgemm_batch_element_t[batch_num]);
brgemm_batch_element_t *batch = batch_v.get();