提交 b0f9b5dc 编写于 作者: W wsc

Clean building warnings for arm64 platform.

上级 226f019e
......@@ -147,7 +147,7 @@ class LogWriter {
LogWriter(const LocationInfo &location, MsLogLevel log_level, SubModuleId submodule,
ExceptionType excp_type = NoExceptionType)
: location_(location), log_level_(log_level), submodule_(submodule), exception_type_(excp_type) {}
: location_(location), log_level_(log_level), exception_type_(excp_type) {}
~LogWriter() = default;
void operator<(const LogStream &stream) const noexcept __attribute__((visibility("default")));
......@@ -161,7 +161,6 @@ class LogWriter {
LocationInfo location_;
MsLogLevel log_level_;
SubModuleId submodule_;
ExceptionType exception_type_;
inline static ExceptionHandler exception_handler_ = nullptr;
......
......@@ -65,22 +65,21 @@ set(CMAKE_VERBOSE_MAKEFILE on)
add_compile_definitions(USE_ANDROID_LOG)
add_compile_definitions(NO_DLIB)
add_compile_options(-fPIC)
if (NOT PLATFORM_ARM64 AND NOT PLATFORM_ARM32)
if ("${CMAKE_BUILD_TYPE}" STREQUAL "Debug")
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -DDebug -g")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -DDebug -g")
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fvisibility=default")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fvisibility=default")
else ()
## enable for binscope for release
set(CMAKE_C_FLAGS "-fPIC -fPIE -D_FORTIFY_SOURCE=2 -O2 -Wall -Werror -fstack-protector-strong -Wno-attributes -Wno-deprecated-declarations ${CMAKE_C_FLAGS}")
set(CMAKE_CXX_FLAGS "-fPIC -fPIE -D_FORTIFY_SOURCE=2 -O2 -Wall -Werror -fstack-protector-strong -Wno-attributes -Wno-deprecated-declarations ${CMAKE_CXX_FLAGS}")
if (NOT WIN32)
set(CMAKE_SHARED_LINKER_FLAGS "-Wl,-z,relro,-z,now -Wl,-z,noexecstack ${CMAKE_SHARED_LINKER_FLAGS}")
set(CMAKE_EXE_LINKER_FLAGS "-Wl,-z,relro,-z,now -Wl,-z,noexecstack ${CMAKE_EXE_LINKER_FLAGS}")
endif()
string(REPLACE " -g " " " CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
endif ()
if ("${CMAKE_BUILD_TYPE}" STREQUAL "Debug")
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -DDebug -g")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -DDebug -g")
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -fvisibility=default")
set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -fvisibility=default")
else ()
## enable for binscope for release
set(CMAKE_C_FLAGS "-fPIC -fPIE -D_FORTIFY_SOURCE=2 -O2 -Wall -Werror -fstack-protector-strong -Wno-attributes -Wno-deprecated-declarations -Wno-missing-braces -Wno-overloaded-virtual ${CMAKE_C_FLAGS}")
set(CMAKE_CXX_FLAGS "-fPIC -fPIE -D_FORTIFY_SOURCE=2 -O2 -Wall -Werror -fstack-protector-strong -Wno-attributes -Wno-deprecated-declarations -Wno-missing-braces -Wno-overloaded-virtual ${CMAKE_CXX_FLAGS}")
if (NOT WIN32)
set(CMAKE_SHARED_LINKER_FLAGS "-Wl,-z,relro,-z,now -Wl,-z,noexecstack ${CMAKE_SHARED_LINKER_FLAGS}")
set(CMAKE_EXE_LINKER_FLAGS "-Wl,-z,relro,-z,now -Wl,-z,noexecstack ${CMAKE_EXE_LINKER_FLAGS}")
endif()
string(REPLACE " -g " " " CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS}")
endif ()
if (BUILD_DEVICE)
......
......@@ -51,6 +51,8 @@ void TileOneDimension(float *inData, float *outData, int dim, size_t ndim, int *
int *outStrides, int *multiple);
void ComputeStrides(int *shape, int *strides, int ndim);
void CalcMultiplesAndStrides(ArithmeticParameter *param);
void TileDimensions(float *data0, float *data1, float *tile_data0, float *tile_data1, ArithmeticParameter *param);
void TileDimensionsUint8(uint8_t *data0, uint8_t *data1, uint8_t *tile_data0, uint8_t *tile_data1,
ArithmeticParameter *param);
......
......@@ -395,7 +395,6 @@ void Conv3x3Fp16(float16_t *input_data, float16_t *transed_weight, const float16
int input_batch = conv_param->input_batch_;
for (int batch = 0; batch < input_batch; batch++) {
int in_batch_offset = batch * ic4 * C4NUM * conv_param->input_h_ * conv_param->input_w_;
int tmp_out_batch_offset = batch * oc8 * C8NUM * out_w_block * out_h_block * output_unit * output_unit;
for (int thread_id = task_id; thread_id < output_tile_count; thread_id += thread_count) {
int start_index = thread_id * tile_num;
......
......@@ -55,7 +55,6 @@ void Im2ColPackUnitFp16(float16_t *input_data, ConvParameter *conv_param, float1
int in_w = conv_param->input_w_;
int out_w = conv_param->output_w_;
int channel_block = UP_DIV(in_channel, 4);
int kernel_plane = kernel_h * kernel_w;
for (int i = 0; i < real_cal_num; i++) {
int block_start = block_index + i;
......
......@@ -607,7 +607,7 @@ void WinogradInputTransformFp16(const float16_t *input_data, float16_t *trans_in
for (int j = 0; j < (interval_x_e - interval_x_s); j++) {
int src_x_offset = src_y_offset + j * ic8 * C8NUM;
int dst_x_offset = dst_y_offset + j * C8NUM;
float16_t *src_addr = input_data + src_x_offset;
const float16_t *src_addr = input_data + src_x_offset;
float16_t *dst_addr = tmp_data + dst_x_offset;
#ifdef ENABLE_NEON
vst1q_f16(dst_addr, vld1q_f16(src_addr));
......
......@@ -28,7 +28,7 @@ void IndirectGemmInt8(int8_t *dst, int32_t *tmp_dst, const int8_t *src, const in
int32_t out_zp = conv_param->conv_quant_arg_.output_quant_args_[0].zp_;
int32_t act_min = conv_param->conv_quant_arg_.out_act_min_[0];
int32_t act_max = conv_param->conv_quant_arg_.out_act_max_[0];
int oc4 = UP_DIV(output_channel, C4NUM);
#ifdef ENABLE_ARM64
size_t asymmetric = conv_param->conv_quant_arg_.asymmetric_ & FILTER_ASYMMETRIC;
size_t per_channel = conv_param->conv_quant_arg_.per_channel_ & FILTER_PER_CHANNEL;
......@@ -36,6 +36,7 @@ void IndirectGemmInt8(int8_t *dst, int32_t *tmp_dst, const int8_t *src, const in
output_channel * sizeof(int8_t), input_sum, act_min, act_max, out_zp, out_multiplier,
shift_before, shift_after, asymmetric, per_channel);
#else
int oc4 = UP_DIV(output_channel, C4NUM);
int tile_num = conv_param->tile_num_;
int plane_c4 = UP_DIV(kernel_plane, C4NUM);
for (int oc = 0; oc < output_channel; oc++) {
......
......@@ -63,16 +63,17 @@ void RowMajor2Row16x4MajorInt8(void *src_ptr, void *dst_ptr, int row, int col) {
for (int ri = 0; ri < row_4div; ri += C4NUM) {
for (int ci = 0; ci < col_16div; ci += C16NUM) {
#ifdef ENABLE_ARM64
size_t col_offset = col;
int8_t *src_c = src_r + ci;
int8_t *dst_c = dst_r + ci * C4NUM;
asm volatile(
"mov x10, %[src_c] \n"
"mov x11, %[dst_c] \n"
"ld1 {v0.16b}, [x10], %[col]\n"
"ld1 {v1.16b}, [x10], %[col]\n"
"ld1 {v2.16b}, [x10], %[col]\n"
"ld1 {v3.16b}, [x10], %[col]\n"
"ld1 {v0.16b}, [x10], %[col_offset]\n"
"ld1 {v1.16b}, [x10], %[col_offset]\n"
"ld1 {v2.16b}, [x10], %[col_offset]\n"
"ld1 {v3.16b}, [x10], %[col_offset]\n"
"st1 {v0.16b}, [x11], #16\n"
"st1 {v1.16b}, [x11], #16\n"
......@@ -80,7 +81,7 @@ void RowMajor2Row16x4MajorInt8(void *src_ptr, void *dst_ptr, int row, int col) {
"st1 {v3.16b}, [x11], #16\n"
:
: [ dst_c ] "r"(dst_c), [ src_c ] "r"(src_c), [ col ] "r"(col)
: [ dst_c ] "r"(dst_c), [ src_c ] "r"(src_c), [ col_offset ] "r"(col_offset)
: "x10", "x11", "v0", "v1", "v2", "v3");
#else
MatrixPack4x16UnitInt8(src_r + ci, dst_r + ci * C4NUM, C4NUM, C16NUM, col);
......
......@@ -1225,9 +1225,9 @@ void Conv3x3Uint8OutputUnit(const int32_t *gemm_out, const int32_t *bias_data, i
ls = vld1q_s32(left_shift);
rs = vld1q_s32(right_shift);
} else {
out_multiplier = vdupq_n_s32(quant_multiplier);
ls = vdupq_n_s32(left_shift);
rs = vdupq_n_s32(right_shift);
out_multiplier = vdupq_n_s32(quant_multiplier[0]);
ls = vdupq_n_s32(left_shift[0]);
rs = vdupq_n_s32(right_shift[0]);
}
int32x4_t out_zp = vdupq_n_s32(output_zp);
int32x4_t output_min = vdupq_n_s32(out_min);
......
......@@ -43,7 +43,7 @@ std::vector<size_t> GetGraphInputNodes(const schema::MetaGraph *meta_graph) {
}
}
}
return std::move(ret);
return ret;
}
std::vector<size_t> GetGraphOutputNodes(const schema::MetaGraph *meta_graph) {
......@@ -64,7 +64,7 @@ std::vector<size_t> GetGraphOutputNodes(const schema::MetaGraph *meta_graph) {
}
}
}
return std::move(ret);
return ret;
}
// NODE_ID OpNode::ID() { return id; }
......
......@@ -54,7 +54,10 @@ int Resize::InferShape(std::vector<lite::tensor::Tensor *> inputs_, std::vector<
if (input == nullptr) {
return 1;
}
MS_ASSERT(input->shape().size() == kInputRank);
if (input->shape().size() != kInputRank) {
MS_LOG(ERROR) << "Size of input shape is wrong.";
return RET_ERROR;
}
auto output = outputs_.front();
if (output == nullptr) {
......
......@@ -40,8 +40,14 @@ int PriorBoxCPUKernel::Init() {
return RET_NULL_PTR;
}
MS_ASSERT(in_tensors_.size() == kInputNum);
MS_ASSERT(out_tensors_.size() == kOutputNum);
if (in_tensors_.size() != kInputNum) {
MS_LOG(ERROR) << "Size of input tensors is wrong.";
return RET_ERROR;
}
if (in_tensors_.size() != kOutputNum) {
MS_LOG(ERROR) << "Size of input tensors is wrong.";
return RET_ERROR;
}
if (!InferShapeDone()) {
return RET_OK;
......
......@@ -46,7 +46,6 @@ class ArithmeticFP16CPUKernel : public LiteKernel {
private:
void FreeTmpBuffer();
int break_pos_;
int outside_;
int out_thread_stride_;
int out_count_;
float16_t *tile_data0_ = nullptr;
......
......@@ -44,7 +44,6 @@ class ReduceFp16CPUKernel : public ReduceBaseCPUKernel {
private:
Reducer reducer_ = nullptr;
std::vector<float16_t *> data_buffers_;
const float *src_data_ = nullptr;
float *dst_data_ = nullptr;
float16_t *fp16_input_ = nullptr;
const float16_t *fp16_src_data_ = nullptr;
......
......@@ -111,8 +111,8 @@ int SplitFp16CPUKernel::Run() {
context_->allocator->Free(output_ptr_[i]);
output_ptr_[i] = nullptr;
}
return RET_OK;
}
return RET_OK;
}
kernel::LiteKernel *CpuSplitFp16KernelCreator(const std::vector<lite::tensor::Tensor *> &inputs,
......
......@@ -30,10 +30,6 @@ using mindspore::lite::RET_OP_EXECUTE_FAILURE;
using mindspore::schema::PrimitiveType_Transpose;
namespace mindspore::kernel {
namespace {
constexpr int kTransposeInputNum = 1;
constexpr int kTransposeOutputNum = 1;
} // namespace
int TransposeFp16CPUKernel::Init() {
TransposeParameter *param = reinterpret_cast<TransposeParameter *>(this->op_parameter_);
num_unit_ = static_cast<int>(in_tensors_[kInputIndex]->shape().at(param->perm_[kNHWC_H]));
......
......@@ -46,7 +46,7 @@ class ArithmeticSelfCPUKernel : public LiteKernel {
explicit ArithmeticSelfCPUKernel(OpParameter *parameter, const std::vector<lite::tensor::Tensor *> &inputs,
const std::vector<lite::tensor::Tensor *> &outputs, const lite::Context *ctx,
const mindspore::lite::PrimitiveC *primitive)
: LiteKernel(parameter, inputs, outputs, ctx, primitive), ctx_(ctx), thread_count_(ctx->thread_num_) {
: LiteKernel(parameter, inputs, outputs, ctx, primitive), thread_count_(ctx->thread_num_) {
switch (parameter->type_) {
case PrimitiveType_Abs:
arithmeticSelf_run_ = ElementAbs;
......@@ -102,7 +102,6 @@ class ArithmeticSelfCPUKernel : public LiteKernel {
size_t data_size_;
ArithmeticSelfParameter *arithmeticSelfParameter_;
ArithmeticSelfRun arithmeticSelf_run_;
const Context *ctx_;
int thread_count_;
float *in_ptr_;
float *out_ptr_;
......
......@@ -28,12 +28,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_ConstantOfShape;
namespace mindspore::kernel {
namespace {
constexpr int kInputNum = 1;
constexpr int kOutputNum = 1;
} // namespace
int ConstantOfShapeCPUKernel::Init() { return RET_OK; }
int ConstantOfShapeCPUKernel::ReSize() { return RET_OK; }
......
......@@ -32,7 +32,7 @@ class ExpandDimsCPUKernel : public LiteKernel {
ExpandDimsCPUKernel(OpParameter *parameter, const std::vector<lite::tensor::Tensor *> &inputs,
const std::vector<lite::tensor::Tensor *> &outputs, const lite::Context *ctx,
const mindspore::lite::PrimitiveC *primitive)
: LiteKernel(parameter, inputs, outputs, ctx, primitive), ctx_(ctx), thread_count_(ctx->thread_num_) {}
: LiteKernel(parameter, inputs, outputs, ctx, primitive), thread_count_(ctx->thread_num_) {}
~ExpandDimsCPUKernel() override = default;
int Init() override;
......@@ -46,7 +46,6 @@ class ExpandDimsCPUKernel : public LiteKernel {
size_t data_size_;
float *in_ptr_;
float *out_ptr_;
const Context *ctx_;
int thread_count_;
};
} // namespace mindspore::kernel
......
......@@ -28,12 +28,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_Fill;
namespace mindspore::kernel {
namespace {
constexpr int kInputNum = 1;
constexpr int kOutputNum = 1;
} // namespace
int FillCPUKernel::Init() {
if (!InferShapeDone()) {
return RET_OK;
......
......@@ -30,7 +30,7 @@ class FillCPUKernel : public LiteKernel {
FillCPUKernel(OpParameter *parameter, const std::vector<lite::tensor::Tensor *> &inputs,
const std::vector<lite::tensor::Tensor *> &outputs, const lite::Context *ctx,
const mindspore::lite::PrimitiveC *primitive)
: LiteKernel(parameter, inputs, outputs, ctx, primitive), ctx_(ctx), thread_count_(ctx->thread_num_) {}
: LiteKernel(parameter, inputs, outputs, ctx, primitive), thread_count_(ctx->thread_num_) {}
~FillCPUKernel() override = default;
int Init() override;
......@@ -44,7 +44,6 @@ class FillCPUKernel : public LiteKernel {
int data_size_;
float src_data_;
float *out_ptr_;
const Context *ctx_;
int thread_count_;
};
} // namespace mindspore::kernel
......
......@@ -32,7 +32,7 @@ class GatherNdCPUKernel : public LiteKernel {
GatherNdCPUKernel(OpParameter *parameter, const std::vector<lite::tensor::Tensor *> &inputs,
const std::vector<lite::tensor::Tensor *> &outputs, const Context *ctx,
const mindspore::lite::PrimitiveC *primitive)
: LiteKernel(parameter, inputs, outputs, ctx, primitive), ctx_(ctx), thread_count_(ctx->thread_num_) {}
: LiteKernel(parameter, inputs, outputs, ctx, primitive), thread_count_(ctx->thread_num_) {}
~GatherNdCPUKernel() override;
int Init() override;
......@@ -48,7 +48,6 @@ class GatherNdCPUKernel : public LiteKernel {
int *in_offset_ = nullptr;
float *in_ptr_;
float *out_ptr_;
const Context *ctx_;
int thread_count_;
};
} // namespace mindspore::kernel
......
......@@ -30,7 +30,6 @@ class PowerCPUKernel : public PowerBaseCPUKernel {
const std::vector<lite::tensor::Tensor *> &outputs, const lite::Context *ctx,
const mindspore::lite::PrimitiveC *primitive)
: PowerBaseCPUKernel(param, inputs, outputs, ctx, primitive),
ctx_(ctx),
thread_count_(ctx->thread_num_),
power_(reinterpret_cast<PowerParameter *>(op_parameter_)->power_),
scale_(reinterpret_cast<PowerParameter *>(op_parameter_)->scale_),
......@@ -43,7 +42,6 @@ class PowerCPUKernel : public PowerBaseCPUKernel {
int RunImpl(int task_id);
private:
const lite::Context *ctx_;
int thread_count_;
float power_;
float scale_;
......
......@@ -27,12 +27,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_Range;
namespace mindspore::kernel {
namespace {
constexpr int kInputNum = 0;
constexpr int kOutputNum = 1;
} // namespace
int RangeCPUKernel::Init() { return RET_OK; }
int RangeCPUKernel::ReSize() { return RET_OK; }
......
......@@ -27,12 +27,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_Rank;
namespace mindspore::kernel {
namespace {
constexpr int kInputNum = 1;
constexpr int kOutputNum = 1;
} // namespace
int RankCPUKernel::Init() { return RET_OK; }
int RankCPUKernel::ReSize() { return RET_OK; }
......
......@@ -31,7 +31,7 @@ class ReverseCPUKernel : public LiteKernel {
ReverseCPUKernel(OpParameter *parameter, const std::vector<lite::tensor::Tensor *> &inputs,
const std::vector<lite::tensor::Tensor *> &outputs, const lite::Context *ctx,
const mindspore::lite::PrimitiveC *primitive)
: LiteKernel(parameter, inputs, outputs, ctx, primitive), ctx_(ctx), thread_count_(ctx->thread_num_) {}
: LiteKernel(parameter, inputs, outputs, ctx, primitive), thread_count_(ctx->thread_num_) {}
~ReverseCPUKernel() {
if (tmp_ != nullptr) {
free(tmp_);
......@@ -52,7 +52,6 @@ class ReverseCPUKernel : public LiteKernel {
int strides_[REVERSE_STRIDE_MAX_SIZE];
int inCount_[REVERSE_STRIDE_MAX_SIZE];
int outCount_[REVERSE_STRIDE_MAX_SIZE];
const Context *ctx_;
int thread_count_;
int *tmp_ = nullptr;
float *in_ptr_;
......
......@@ -30,8 +30,6 @@ using mindspore::schema::PrimitiveType_ScatterND;
namespace mindspore::kernel {
namespace {
constexpr int kScatterNDInputNum = 3;
constexpr int kScatterNDOutputNum = 1;
constexpr int kScatterShapeIndex = 0;
constexpr int kScatterIndicesIndex = 1;
constexpr int kScatterUpdateIndex = 2;
......
......@@ -26,10 +26,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_Shape;
namespace mindspore::kernel {
namespace {
constexpr int kShapeInputNum = 1;
constexpr int kShapeOutputNum = 1;
} // namespace
int ShapeCPUKernel::Init() { return RET_OK; }
int ShapeCPUKernel::ReSize() { return RET_OK; }
......
......@@ -27,11 +27,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_Squeeze;
namespace mindspore::kernel {
namespace {
constexpr int kSqueezeInputNum = 1;
constexpr int kSqueezeOutputNum = 1;
} // namespace
int SqueezeCPUKernel::Init() { return RET_OK; }
int SqueezeCPUKernel::ReSize() { return RET_OK; }
......
......@@ -29,10 +29,6 @@ using mindspore::lite::RET_OP_EXECUTE_FAILURE;
using mindspore::schema::PrimitiveType_Transpose;
namespace mindspore::kernel {
namespace {
constexpr int kTransposeInputNum = 1;
constexpr int kTransposeOutputNum = 1;
} // namespace
int TransposeCPUKernel::Init() {
TransposeParameter *param = reinterpret_cast<TransposeParameter *>(this->op_parameter_);
num_unit_ = static_cast<int>(in_tensors_[kInputIndex]->shape().at(param->perm_[kNHWC_H]));
......
......@@ -27,9 +27,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_ZerosLike;
namespace mindspore::kernel {
constexpr int kInputNum = 1;
constexpr int kOutputNum = 1;
int ZerosLikeCPUKernel::Init() { return RET_OK; }
int ZerosLikeCPUKernel::Run() {
......
......@@ -92,7 +92,7 @@ int QuantizedAddCPUKernel::Run() {
input0_data_ = static_cast<int8_t *>(ctx_->allocator->Malloc(out_tensors_.at(0)->Size()));
input1_data_ = static_cast<int8_t *>(ctx_->allocator->Malloc(out_tensors_.at(0)->Size()));
ArithmeticParameter tile_para = {0};
ArithmeticParameter tile_para;
tile_para.ndim_ = out_tensors_.at(0)->shape().size();
for (size_t i = 0; i < tile_para.ndim_; i++) {
tile_para.in_shape0_[i] = in_tensors_.at(0)->DimensionSize(i);
......
......@@ -45,7 +45,7 @@ class ArithmeticSelfInt8CPUKernel : public LiteKernel {
explicit ArithmeticSelfInt8CPUKernel(OpParameter *parameter, const std::vector<lite::tensor::Tensor *> &inputs,
const std::vector<lite::tensor::Tensor *> &outputs, const Context *ctx,
const mindspore::lite::PrimitiveC *primitive)
: LiteKernel(parameter, inputs, outputs, ctx, primitive), ctx_(ctx), thread_count_(ctx->thread_num_) {
: LiteKernel(parameter, inputs, outputs, ctx, primitive), thread_count_(ctx->thread_num_) {
switch (parameter->type_) {
case PrimitiveType_Round:
arithmeticSelf_run_ = Int8ElementRound;
......@@ -98,7 +98,6 @@ class ArithmeticSelfInt8CPUKernel : public LiteKernel {
size_t data_size_;
ArithmeticSelfParameter *para_;
ArithmeticSelfInt8Run arithmeticSelf_run_;
const Context *ctx_;
int thread_count_;
int8_t *in_ptr_;
int8_t *out_ptr_;
......
......@@ -104,7 +104,7 @@ int DivInt8CPUKernel::Run() {
}
if (broadcast_) {
ArithmeticParameter tile_para = {0};
ArithmeticParameter tile_para;
tile_para.ndim_ = out_tensors_.at(0)->shape().size();
for (size_t i = 0; i < tile_para.ndim_; i++) {
tile_para.in_shape0_[i] = in_tensors_.at(0)->DimensionSize(i);
......
......@@ -77,7 +77,7 @@ int MulInt8CPUKernel::Run() {
input0_data_ = static_cast<int8_t *>(ctx_->allocator->Malloc(out_tensors_.at(0)->Size()));
input1_data_ = static_cast<int8_t *>(ctx_->allocator->Malloc(out_tensors_.at(0)->Size()));
ArithmeticParameter tile_para = {0};
ArithmeticParameter tile_para;
tile_para.ndim_ = out_tensors_.at(0)->shape().size();
for (size_t i = 0; i < tile_para.ndim_; i++) {
tile_para.in_shape0_[i] = in_tensors_.at(0)->DimensionSize(i);
......
......@@ -30,12 +30,6 @@ using mindspore::lite::RET_NULL_PTR;
using mindspore::lite::RET_OK;
namespace mindspore::kernel {
namespace {
constexpr int kInputNum = 1;
constexpr int kOutputNum = 1;
constexpr size_t kRank = 4;
} // namespace
int ResizeInt8CPUKernel::Init() {
auto ret = ResizeBaseCPUKernel::Init();
if (ret != RET_OK) {
......
......@@ -128,7 +128,7 @@ int SubInt8CPUKernel::Run() {
}
if (broadcast_) {
ArithmeticParameter tile_para = {0};
ArithmeticParameter tile_para;
tile_para.ndim_ = out_tensors_.at(0)->shape().size();
for (size_t i = 0; i < tile_para.ndim_; i++) {
tile_para.in_shape0_[i] = in_tensors_.at(0)->DimensionSize(i);
......
......@@ -30,7 +30,7 @@ class Unsqueezeint8CPUKernel : public LiteKernel {
Unsqueezeint8CPUKernel(OpParameter *parameter, const std::vector<lite::tensor::Tensor *> &inputs,
const std::vector<lite::tensor::Tensor *> &outputs, const Context *ctx,
const mindspore::lite::PrimitiveC *primitive)
: LiteKernel(parameter, inputs, outputs, ctx, primitive), ctx_(ctx), thread_count_(ctx->thread_num_) {
: LiteKernel(parameter, inputs, outputs, ctx, primitive), thread_count_(ctx->thread_num_) {
Unsq_para_ = reinterpret_cast<UnSqueezeParameter *>(op_parameter_);
Unsq_para_->thread_count_ = op_parameter_->thread_num_;
}
......@@ -42,14 +42,12 @@ class Unsqueezeint8CPUKernel : public LiteKernel {
int DoUnsqueeze(int task_id);
private:
UnSqueezeQuantArg *quant_Unsqueeze_parm_;
UnSqueezeParameter *Unsq_para_;
int thread_sz_count_;
int thread_sz_stride_;
int data_size_;
float *in_ptr_;
float *out_ptr_;
const Context *ctx_;
int thread_count_;
};
} // namespace mindspore::kernel
......
......@@ -57,7 +57,6 @@ class SubGraphOpenCLKernel : public SubGraphKernel {
std::vector<std::vector<kernel::LiteKernel *>> *out_kernels, bool is_from);
private:
SubGraphOpenCLParameter *subgraph_ocl_parameter_;
lite::opencl::OpenCLAllocator *allocator_;
std::vector<lite::tensor::Tensor *> in_convert_tensors_;
std::vector<lite::tensor::Tensor *> out_convert_tensors_;
......
......@@ -17,5 +17,10 @@ else()
target_link_libraries(timeprofile mindspore-lite pthread)
endif()
install(TARGETS timeprofile
RUNTIME DESTINATION ${MAIN_DIR}/time_profile COMPONENT ${COMPONENT_NAME})
\ No newline at end of file
if (PLATFORM_ARM32 OR PLATFORM_ARM64)
install(TARGETS timeprofile
RUNTIME DESTINATION ${MAIN_DIR}/time_profile COMPONENT ${COMPONENT_NAME})
else()
install(TARGETS timeprofile
RUNTIME DESTINATION ${MAIN_DIR}/time_profile COMPONENT ${RUN_X86_COMPONENT_NAME})
endif()
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册