xref: /foundation/ai/mindspore/patches/0023-support-x86-emulator-build.patch

From a80a695b3f6ce3118cc01c31359762cfac35f02d Mon Sep 17 00:00:00 2001
From: zhangyanhui <zhangyanhui17@huawei.com>
Date: Thu, 13 Jun 2024 14:23:56 +0800
Subject: [PATCH] 0023-support-x86-emulator-build

---
 .../plugin/device/cpu/kernel/nnacl/BUILD.gn   | 92 +++++++++++++++----
 mindspore/lite/BUILD.gn                       | 60 ++++++------
 mindspore/lite/src/common/thread_utils.cc     |  2 +-
 mindspore/lite/src/litert/kernel/cpu/BUILD.gn | 28 +++++-
 .../cpu/fp32/convolution_delegate_fp32.cc     |  2 +
 ...volution_depthwise_slidewindow_x86_fp32.cc |  4 +-
 ...nvolution_depthwise_slidewindow_x86_fp32.h |  2 +-
 7 files changed, 138 insertions(+), 52 deletions(-)

diff --git a/mindspore/ccsrc/plugin/device/cpu/kernel/nnacl/BUILD.gn b/mindspore/ccsrc/plugin/device/cpu/kernel/nnacl/BUILD.gn
index d27817be..387a675a 100644
--- a/mindspore/ccsrc/plugin/device/cpu/kernel/nnacl/BUILD.gn
+++ b/mindspore/ccsrc/plugin/device/cpu/kernel/nnacl/BUILD.gn
@@ -46,7 +46,6 @@ config("nnacl_public_config") {
     }
   } else if (target_cpu == "x86_64") {
     cflags_c += [
-      "-mavx512f",
       "-mavx",
       "-mavx2",
       "-mfma",
@@ -56,8 +55,16 @@ config("nnacl_public_config") {
     defines += [
       "ENABLE_SSE",
       "ENABLE_AVX",
-      "ENABLE_AVX512",
     ]
+    # emulator not support avx512
+    if (!is_emulator) {
+      cflags_c += [
+        "-mavx512f",
+      ]
+      defines += [
+        "ENABLE_AVX512",
+      ]
+    }
   }
 }
 
@@ -102,7 +109,6 @@ c_kernel_sources = [
   "kernel/convolution_depthwise_sw.c",
   "kernel/convolution_im2col_arm32.c",
   "kernel/convolution_im2col_arm64.c",
-  "kernel/convolution_im2col_avx512.c",
   "kernel/convolution_im2col_avx.c",
   "kernel/convolution_im2col_base.c",
   "kernel/convolution_im2col.c",
@@ -136,7 +142,6 @@ c_kernel_sources = [
   "kernel/log_softmax.c",
   "kernel/matmul_arm32.c",
   "kernel/matmul_arm64.c",
-  "kernel/matmul_avx512.c",
   "kernel/matmul_avx.c",
   "kernel/matmul_base.c",
   "kernel/matmul.c",
@@ -169,10 +174,6 @@ c_kernel_sources = [
   "kernel/zeros_like.c",
 ]
 
-# list of ${NNACL_DIR}/experimental/*.c
-experimental_kernel_sources = [
-]
-
 # list of ${NNACL_DIR}/base/*.c
 base_kernel_sources = [
   "base/arithmetic_base.c",
@@ -221,7 +222,6 @@ fp32_kernel_sources = [
   "fp32/conv_common_fp32.c",
   "fp32/conv_depthwise_avx_fp32.c",
   "fp32/conv_depthwise_fp32.c",
-  "fp32/conv_im2col_avx512_fp32.c",
   "fp32/conv_im2col_fp32.c",
   "fp32/conv_sw_arm64_fp32.c",
   "fp32/conv_sw_avx_fp32.c",
@@ -246,8 +246,6 @@ fp32_kernel_sources = [
   "fp32/local_response_norm_fp32.c",
   "fp32/log_softmax_fp32.c",
   "fp32/lstm_fp32.c",
-  "fp32/matmul_avx512_fp32.c",
-  "fp32/matmul_avx512_mask_fp32.c",
   "fp32/matmul_avx_fp32.c",
   "fp32/matmul_fp32.c",
   "fp32/mul_fp32.c",
@@ -784,6 +782,13 @@ sse_avx_avx512_sources = [
   "assembly/avx/MatmulAvx.S",
 ]
 
+# only x86_64 real machine support avx512
+if (target_cpu == "x86_64" && !is_emulator) {
+  sse_avx_avx512_sources += [
+    "assembly/avx512/ConvDwFp32RowAVX512.S",
+  ]
+}
+
 gemm_avx512_kernel_sources = [
   "experimental/HPC-generator/gemm_avx512/nnacl_gemm_avx512_10x16_kernel_nhwc_fp32.c",
   "experimental/HPC-generator/gemm_avx512/nnacl_gemm_avx512_10x32_kernel_nhwc_fp32.c",
@@ -834,16 +839,64 @@ gemm_avx512_kernel_sources = [
   "experimental/HPC-generator/gemm_avx512/nnacl_gemm_avx512_9x32_kernel_nhwc_fp32.c",
 ]
 
+gemm_mask_avx512_kernel_sources = [
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_10x16_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_10x32_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_11x16_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_11x32_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_12x16_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_12x32_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_1x16_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_1x32_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_1x48_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_1x64_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_1x80_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_1x96_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_2x16_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_2x32_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_2x48_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_2x64_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_2x80_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_2x96_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_3x16_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_3x32_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_3x48_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_3x64_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_3x80_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_3x96_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_4x16_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_4x32_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_4x48_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_4x64_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_4x80_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_4x96_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_5x16_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_5x32_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_5x48_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_5x64_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_5x80_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_6x16_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_6x32_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_6x48_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_6x64_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_7x16_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_7x32_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_7x48_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_8x16_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_8x32_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_8x48_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_9x16_mask_kernel_nhwc_fp32.c",
+  "experimental/HPC-generator/gemm_mask_avx512/nnacl_gemm_mask_avx512_9x32_mask_kernel_nhwc_fp32.c",
+]
+
 fp32_kernel_sources -= no_fast_math_fp32_kernel_sources
 fp32_kernel_sources -= avx_fp32_kernel_sources
-fp32_kernel_sources -= avx512_fp32_kernel_sources
 fp32_kernel_sources -= arm64_fp32_kernel_sources
 
 # source files on all target
 nnacl_sources = common_sources
 nnacl_sources += base_kernel_sources
 nnacl_sources += c_kernel_sources
-nnacl_sources += experimental_kernel_sources
 nnacl_sources += fp32_kernel_sources
 nnacl_sources += fp32_sparse_kernel_sources
 nnacl_sources += fp32_grad_kernel_sources
@@ -854,7 +907,6 @@ nnacl_sources += infer_control_sources
 
 # source files on arm32
 arm_only_sources = arm32_assembly_sources
-#arm_only_sources += arm32_fp16_assembly_sources
 not_needed(arm32_fp16_assembly_sources)
 
 # source files on arm64
@@ -868,8 +920,16 @@ arm64_only_sources += arm64_fp32_kernel_sources
 # sources files on x86_64
 x86_64_only_sources = sse_avx_avx512_sources
 x86_64_only_sources += avx_fp32_kernel_sources
-x86_64_only_sources += avx512_fp32_kernel_sources
-x86_64_only_sources += gemm_avx512_kernel_sources
+# emulator not support avx512
+if (is_emulator) {
+  not_needed(avx512_fp32_kernel_sources)
+  not_needed(gemm_avx512_kernel_sources)
+  not_needed(gemm_mask_avx512_kernel_sources)
+} else {
+  x86_64_only_sources += avx512_fp32_kernel_sources
+  x86_64_only_sources += gemm_avx512_kernel_sources
+  x86_64_only_sources += gemm_mask_avx512_kernel_sources
+}
 
 if (target_cpu == "arm") {
   nnacl_sources += arm_only_sources
diff --git a/mindspore/lite/BUILD.gn b/mindspore/lite/BUILD.gn
index 467cdb6a..124c84c9 100644
--- a/mindspore/lite/BUILD.gn
+++ b/mindspore/lite/BUILD.gn
@@ -118,12 +118,6 @@ control_flow_kernel_sources = [
   "src/control_flow/kernel/identity_kernel.cc",
 ]
 
-experimental_sources = [
-]
-
-string_kernel_source = [
-]
-
 auto_parallel_source = [
   "src/litert/sub_graph_split.cc"
 ]
@@ -186,19 +180,11 @@ lite_mindrt_sources = [
 all_lite_sources += cxx_api_sources
 all_lite_sources += api_source
 all_lite_sources += control_flow_kernel_sources
-all_lite_sources += experimental_sources
-all_lite_sources += string_kernel_source
 all_lite_sources += auto_parallel_source
 all_lite_sources += custom_registry_sources
 all_lite_sources += weight_decode_source
 all_lite_sources += lite_mindrt_sources
 
-ops_base_sources = [
-#  "src/common/ops/anf_utils.cc", # disable runtiem convert
-#  "src/common/ops/ops_def.cc", # disable kernel executor
-#  "src/common/ops/ops_utils.cc" # disable kernel executor
-]
-
 basic_populate_sources = [
   "src/common/ops/populate/activation_grad_populate.cc",
   "src/common/ops/populate/activation_populate.cc",
@@ -346,8 +332,7 @@ control_populate_sources = [
   "src/common/ops/populate/control/tensorliststack_populate.cc",
 ]
 
-all_ops_sources = ops_base_sources
-all_ops_sources += basic_populate_sources
+all_ops_sources = basic_populate_sources
 all_ops_sources += string_populate_sources
 all_ops_sources += control_populate_sources
 
@@ -360,6 +345,12 @@ missing_sources = [
 
 all_sources += missing_sources
 
+SUPPORT_NNRT = false
+# currently, only arm/arm64 real machine support nnrt
+if ((target_cpu == "arm" || target_cpu == "arm64") && !is_emulator) {
+  SUPPORT_NNRT = true
+}
+
 ohos_shared_library("mindspore_lib") {
   deps = [
     "../ccsrc/plugin/device/cpu/kernel/nnacl/:nnacl_obj",
@@ -387,7 +378,6 @@ ohos_shared_library("mindspore_lib") {
     "../ccsrc/",
     "src/litert/kernel/cpu/",
     "../core/mindrt/src/",
-    "//foundation/ai/neural_network_runtime/",
   ]
 
   defines = [
@@ -418,6 +408,17 @@ ohos_shared_library("mindspore_lib") {
       "CL_HPP_TARGET_OPENCL_VERSION=120",
       "CL_HPP_MINIMUM_OPENCL_VERSION=120",
     ]
+  } else if (target_cpu == "x86_64") {
+    defines += [
+      "ENABLE_SSE",
+      "ENABLE_AVX",
+    ]
+    # emulator not support avx512
+    if (!is_emulator) {
+      defines += [
+        "ENABLE_AVX512",
+      ]
+    }
   }
 
   configs = [
@@ -434,10 +435,10 @@ ohos_shared_library("mindspore_lib") {
   output_name = "libmindspore-lite"
   output_extension = "so"
   innerapi_tags = [ "platformsdk" ]
-  SUPPORT_NNRT = true
   if (SUPPORT_NNRT) {
     if (mindspore_feature_nnrt_metagraph) {
       defines += [ "SUPPORT_NNRT_METAGRAPH" ]
+      sources += [ "src/litert/delegate/nnrt/hiai_foundation_wrapper.cc", ]
       print("enabled feature: mindspore_feature_nnrt_metagraph")
     }
     sources += [
@@ -445,7 +446,6 @@ ohos_shared_library("mindspore_lib") {
       "src/litert/delegate/nnrt/nnrt_delegate.cc",
       "src/litert/delegate/nnrt/nnrt_model_kernel.cc",
       "src/litert/delegate/nnrt/nnrt_allocator.cc",
-      "src/litert/delegate/nnrt/hiai_foundation_wrapper.cc",
       "src/litert/delegate/nnrt/extension_options_parser.cc",
     ]
     include_dirs += [
@@ -453,6 +453,7 @@ ohos_shared_library("mindspore_lib") {
       "../../mindspore/core/ir",
       "mindir/include",
       "mindir/inner_headers",
+      "//foundation/ai/neural_network_runtime/",
     ]
 
     external_deps += [ "neural_network_runtime:nnrt_target" ]
@@ -499,11 +500,9 @@ ohos_shared_library("mindspore_ndk") {
     "../../third_party/",
     "./schema/",
     "../ccsrc/",
-    "//foundation/ai/neural_network_runtime/",
   ]
 
   defines = [
-    "SUPPORT_NNRT",
     "MS_COMPILE_OHOS",
     "PRIMITIVE_WRITEABLE",
     "RUNTIME_PASS_CLIP",
@@ -512,9 +511,18 @@ ohos_shared_library("mindspore_ndk") {
     "ENABLE_HI_APP_EVENT",
   ]
 
-  if (mindspore_feature_nnrt_metagraph) {
-    defines += [ "SUPPORT_NNRT_METAGRAPH" ]
-    print("enabled feature: mindspore_feature_nnrt_metagraph")
+  if (SUPPORT_NNRT) {
+    include_dirs += [
+      "//foundation/ai/neural_network_runtime/",
+    ]
+    defines += [
+      "SUPPORT_NNRT",
+    ]
+    if (mindspore_feature_nnrt_metagraph) {
+      defines += [ "SUPPORT_NNRT_METAGRAPH" ]
+      print("enabled feature: mindspore_feature_nnrt_metagraph")
+    }
+    external_deps = [ "neural_network_runtime:nnrt_target" ]
   }
 
   configs = [
@@ -523,8 +531,6 @@ ohos_shared_library("mindspore_ndk") {
     ":secure_option",
   ]
 
-  external_deps = [ "neural_network_runtime:nnrt_target" ]
-
   remove_configs = [ "//build/config/compiler:no_rtti" ]
 
   output_name = "libmindspore_lite_ndk"
@@ -749,4 +755,4 @@ config("secure_option") {
 
 config("train_kernel_option") {
   cflags_cc = [ "-fno-finite-math-only" ]
-}
+}
\ No newline at end of file
diff --git a/mindspore/lite/src/common/thread_utils.cc b/mindspore/lite/src/common/thread_utils.cc
index 28c8e1cd..28c7acab 100644
--- a/mindspore/lite/src/common/thread_utils.cc
+++ b/mindspore/lite/src/common/thread_utils.cc
@@ -17,7 +17,7 @@
 #if defined(__linux__) && !defined(ENABLE_ARM)
 #include "src/common/thread_utils.h"
 #include <sys/stat.h>
-#include <wait.h>
+#include <sys/wait.h>
 #include "src/common/log_adapter.h"
 
 namespace mindspore {
diff --git a/mindspore/lite/src/litert/kernel/cpu/BUILD.gn b/mindspore/lite/src/litert/kernel/cpu/BUILD.gn
index 297fc6f6..d51b9f4a 100644
--- a/mindspore/lite/src/litert/kernel/cpu/BUILD.gn
+++ b/mindspore/lite/src/litert/kernel/cpu/BUILD.gn
@@ -52,7 +52,6 @@ cpu_kernel_sources = [
     "fp32/convolution_fp32.cc",
     "fp32/convolution_im2col_arm32_fp32.cc",
     "fp32/convolution_im2col_arm64_fp32.cc",
-    "fp32/convolution_im2col_avx512_fp32.cc",
     "fp32/convolution_im2col_avx_fp32.cc",
     "fp32/convolution_im2col_base_fp32.cc",
     "fp32/convolution_im2col_fp32.cc",
@@ -90,7 +89,6 @@ cpu_kernel_sources = [
     "fp32/lstm_non_mindir_fp32.cc",
     "fp32/matmul_fp32_arm32.cc",
     "fp32/matmul_fp32_arm64.cc",
-    "fp32/matmul_fp32_avx512.cc",
     "fp32/matmul_fp32_avx.cc",
     "fp32/matmul_fp32_base.cc",
     "fp32/matmul_fp32.cc",
@@ -125,7 +123,7 @@ cpu_kernel_sources = [
 ]
 
 if ((target_cpu != "arm") && (target_cpu != "arm64")) {
-    cpu_kernel_sources += [ "src/runtime/kernel/cpu/fp32/cast_for_x86_fp16.cc" ]
+    cpu_kernel_sources += [ "fp32/cast_for_x86_fp16.cc" ]
 }
 
 arm64_cpu_kernel_sources = [
@@ -148,8 +146,6 @@ sse_avx_avx512_kernel_sources = [
   "fp32/convolution_im2col_avx_fp32.cc",
   "fp32/matmul_fp32_avx.cc",
   "fp32/convolution_winograd_avx_fp32.cc",
-  "fp32/convolution_im2col_avx512_fp32.cc",
-  "fp32/matmul_fp32_avx512.cc",
 ]
 
 fp16_kernel_sources = [
@@ -272,6 +268,18 @@ control_kernel_sources = [
     "control/tensorlist_stack.cc",
 ]
 
+# emulator not support avx512
+if (!is_emulator) {
+  cpu_kernel_sources += [
+    "fp32/convolution_im2col_avx512_fp32.cc",
+    "fp32/matmul_fp32_avx512.cc",
+  ]
+  sse_avx_avx512_kernel_sources += [
+    "fp32/convolution_im2col_avx512_fp32.cc",
+    "fp32/matmul_fp32_avx512.cc",
+  ]
+}
+
 all_cpu_kernel_sources = cpu_kernel_sources
 all_cpu_kernel_sources += int8_kernel_sources
 all_cpu_kernel_sources += string_kernel_sources
@@ -348,6 +356,16 @@ ohos_source_set("cpu_kernel_obj") {
       "CL_HPP_TARGET_OPENCL_VERSION=120",
       "CL_HPP_MINIMUM_OPENCL_VERSION=120",
     ]
+  } else if (target_cpu == "x86_64") {
+    defines += [
+      "ENABLE_SSE",
+      "ENABLE_AVX",
+    ]
+    if (!is_emulator) {
+      defines += [
+        "ENABLE_AVX512",
+      ]
+    }
   }
 
   cflags_cc = [
diff --git a/mindspore/lite/src/litert/kernel/cpu/fp32/convolution_delegate_fp32.cc b/mindspore/lite/src/litert/kernel/cpu/fp32/convolution_delegate_fp32.cc
index f907bbbf..ac693c44 100644
--- a/mindspore/lite/src/litert/kernel/cpu/fp32/convolution_delegate_fp32.cc
+++ b/mindspore/lite/src/litert/kernel/cpu/fp32/convolution_delegate_fp32.cc
@@ -49,7 +49,9 @@ using mindspore::schema::PrimitiveType_Conv2DFusion;
 
 namespace mindspore::kernel {
 namespace {
+#ifndef ENABLE_AVX
 constexpr int kMaxDwConvSWSize = 32;
+#endif
 }  // namespace
 
 float *ConvolutionDelegateCPUKernel::CopyData(const lite::Tensor *tensor) {
diff --git a/mindspore/lite/src/litert/kernel/cpu/fp32/convolution_depthwise_slidewindow_x86_fp32.cc b/mindspore/lite/src/litert/kernel/cpu/fp32/convolution_depthwise_slidewindow_x86_fp32.cc
index 568b9463..d35669ce 100644
--- a/mindspore/lite/src/litert/kernel/cpu/fp32/convolution_depthwise_slidewindow_x86_fp32.cc
+++ b/mindspore/lite/src/litert/kernel/cpu/fp32/convolution_depthwise_slidewindow_x86_fp32.cc
@@ -106,7 +106,7 @@ int ConvolutionDepthwiseSWCPUKernelX86::ReSize() {
   return RET_OK;
 }
 
-int ConvolutionDepthwiseSWCPUKernelX86::Execute(int task_id) {
+int ConvolutionDepthwiseSWCPUKernelX86::DoExecute(int task_id) {
   DepthwiseSWAvxFp32(packed_output_, packed_input_, reinterpret_cast<float *>(packed_weight_),
                      reinterpret_cast<float *>(bias_data_), conv_param_, sliding_, task_id);
   return RET_OK;
@@ -114,7 +114,7 @@ int ConvolutionDepthwiseSWCPUKernelX86::Execute(int task_id) {
 
 int ConvDwSWAvxRun(void *cdata, int task_id, float lhs_scale, float rhs_scale) {
   auto conv_dw = reinterpret_cast<ConvolutionDepthwiseSWCPUKernelX86 *>(cdata);
-  auto ret = conv_dw->Execute(task_id);
+  auto ret = conv_dw->DoExecute(task_id);
   if (ret != RET_OK) {
     MS_LOG(ERROR) << "ConvolutionDepthwiseSWRun in x86 error task_id[" << task_id << "] error_code[" << ret << "]";
     return RET_ERROR;
diff --git a/mindspore/lite/src/litert/kernel/cpu/fp32/convolution_depthwise_slidewindow_x86_fp32.h b/mindspore/lite/src/litert/kernel/cpu/fp32/convolution_depthwise_slidewindow_x86_fp32.h
index e959fe45..928321e5 100644
--- a/mindspore/lite/src/litert/kernel/cpu/fp32/convolution_depthwise_slidewindow_x86_fp32.h
+++ b/mindspore/lite/src/litert/kernel/cpu/fp32/convolution_depthwise_slidewindow_x86_fp32.h
@@ -35,7 +35,7 @@ class ConvolutionDepthwiseSWCPUKernelX86 : public ConvolutionBaseCPUKernel {
   int ReSize() override;
   int Run() override;
 
-  int Execute(int task_id);
+  int DoExecute(int task_id);
 
  private:
   void FreePackedInputOutput();
-- 
2.25.1