p60975123
/
mindspore
forked from mindspore-Ecosystem/mindspore
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

!17774 clean code 

From: @zhaozhenlong
Reviewed-by: @jpc_chenjianping,@zhanghaibo5
Signed-off-by: @zhanghaibo5
This commit is contained in:
mindspore-ci-bot 2021-06-04 17:16:34 +08:00 committed by Gitee
parent 67ba618e95 689019813c
commit 3ec4530ede
79 changed files with 362 additions and 272 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
mindspore/lite/src/runtime/kernel/arm/fp32/gatherNd_fp32.cc
View File

@ -15,7 +15,7 @@
*/
#include "src/runtime/kernel/arm/fp32/gatherNd_fp32.h"
#include <string.h>
#include <cstring>
#include <limits>
#include <vector>
#include "schema/model_generated.h"

5
mindspore/lite/src/runtime/kernel/arm/fp32/l2_norm_fp32.cc
View File

@ -144,9 +144,10 @@ int L2NormCPUKernel::Run() {
auto input_shape = in_tensors().at(kInputIndex)->shape();
input_ptr_ = reinterpret_cast<float *>(in_tensors_.at(kInputIndex)->MutableData());
output_ptr_ = reinterpret_cast<float *>(out_tensors_.at(kOutputIndex)->MutableData());
int ret;
if (l2_norm_param_->axis_num_ == 0 || l2_norm_param_->axis_num_ == input_shape.size()) {
// all axis
auto ret = static_cast<const lite::InnerContext *>(this->context_)
ret = static_cast<const lite::InnerContext *>(this->context_)
->thread_pool_->ParallelLaunch(SquareSumRun, this, context_->thread_num_);
if (ret != RET_OK) {
MS_LOG(ERROR) << "L2Norm error: error_code[" << ret << "]";
@ -164,7 +165,7 @@ int L2NormCPUKernel::Run() {
return RET_ERROR;
}
} else if (l2_norm_param_->axis_num_ == 1 && l2_norm_param_->axis_[0] == static_cast<int>(input_shape.size()) - 1) {
auto ret = static_cast<const lite::InnerContext *>(this->context_)
ret = static_cast<const lite::InnerContext *>(this->context_)
->thread_pool_->ParallelLaunch(L2NormTrailingAxisRun, this, context_->thread_num_);
if (ret != RET_OK) {
MS_LOG(ERROR) << "L2Norm error: error_code[" << ret << "]";

1
mindspore/lite/src/runtime/kernel/arm/fp32/local_response_norm_fp32.cc
View File

@ -28,7 +28,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_LRN;
namespace mindspore::kernel {
int LocalResponseNormCPUKernel::Init() { return RET_OK; }
int LocalResponseNormCPUKernel::ReSize() { return RET_OK; }

2
mindspore/lite/src/runtime/kernel/arm/fp32/lstm_fp32.cc
View File

@ -15,7 +15,7 @@
*/
#include "src/runtime/kernel/arm/fp32/lstm_fp32.h"
#include <float.h>
#include <cfloat>
#include <vector>
#include "schema/model_generated.h"
#include "src/kernel_registry.h"

4
mindspore/lite/src/runtime/kernel/arm/fp32/matmul_fp32_base.cc
View File

@ -196,9 +196,9 @@ int MatmulFp32BaseCPUKernel::InitMatrixB(const float *src_ptr) {
memcpy(b_pack_ptr_, src_ptr, params_->batch * params_->col_ * params_->deep_ * sizeof(float));
} else {
for (int i = 0; i < params_->batch; i++) {
const float *src = src_ptr + i * params_->deep_ * params_->col_;
const float *src_data = src_ptr + i * params_->deep_ * params_->col_;
float *dst = b_pack_ptr_ + i * params_->deep_ * params_->col_;
RowMajor2ColMajor(src, dst, params_->deep_, params_->col_);
RowMajor2ColMajor(src_data, dst, params_->deep_, params_->col_);
}
}
return RET_OK;

9
mindspore/lite/src/runtime/kernel/arm/fp32/pad_fp32.cc
View File

@ -85,7 +85,7 @@ void PadCPUKernel::InitMirrorPadBlock() {
int cur_input = 1;
int cur_output = 1;
for (size_t i = 0; i < COMM_SHAPE_SIZE; ++i) {
if (1 < cur_input) {
if (cur_input > 1) {
input_separate_dims.emplace_back(cur_input);
output_separate_dims.emplace_back(cur_output);
separate_offset.emplace_back(0);
@ -355,20 +355,21 @@ void PadCPUKernel::CalculateStrides() {
}
int PadCPUKernel::HandleMirrorPad() {
int ret;
if (in_tensors_.size() == 1) {
auto input_shape = in_tensors_.at(0)->shape();
int rank = static_cast<int>(input_shape.size());
auto ret = ExtendShape(in_, COMM_SHAPE_SIZE, input_shape.data(), rank);
ret = ExtendShape(in_, COMM_SHAPE_SIZE, input_shape.data(), rank);
if (ret != RET_OK) {
return ret;
}
} else {
auto ret = CopyPaddingFromInput();
ret = CopyPaddingFromInput();
if (ret != RET_OK) {
return ret;
}
}
auto ret = CheckPaddings(pad_param_->paddings_, COMM_SHAPE_SIZE, in_, pad_param_->pad_mode_);
ret = CheckPaddings(pad_param_->paddings_, COMM_SHAPE_SIZE, in_, pad_param_->pad_mode_);
if (ret != RET_OK) {
return ret;
}

2
mindspore/lite/src/runtime/kernel/arm/fp32/pooling_fp32.cc
View File

@ -15,7 +15,7 @@
*/
#include "src/runtime/kernel/arm/fp32/pooling_fp32.h"
#include <float.h>
#include <cfloat>
#include "nnacl/fp32/pooling_fp32.h"
#include "src/kernel_registry.h"
#include "src/runtime/runtime_api.h"

2
mindspore/lite/src/runtime/kernel/arm/fp32/reverse_fp32.cc
View File

@ -14,7 +14,7 @@
* limitations under the License.
*/
#include "src/runtime/kernel/arm/fp32/reverse_fp32.h"
#include <string.h>
#include <cstring>
#include <vector>
#include "schema/model_generated.h"
#include "src/kernel_registry.h"

1
mindspore/lite/src/runtime/kernel/arm/fp32/roi_pooling_fp32.cc
View File

@ -29,7 +29,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_ROIPooling;
namespace mindspore::kernel {
int ROIPoolingCPUKernel::Init() {
if (!InferShapeDone()) {
return RET_OK;

2
mindspore/lite/src/runtime/kernel/arm/fp32/scale_fp32.cc
View File

@ -15,7 +15,7 @@
*/
#include "src/runtime/kernel/arm/fp32/scale_fp32.h"
#include <string.h>
#include <cstring>
#include <vector>
#include "schema/model_generated.h"
#include "src/kernel_registry.h"

3
mindspore/lite/src/runtime/kernel/arm/fp32/scatter_nd_fp32.cc
View File

@ -15,7 +15,7 @@
*/
#include "src/runtime/kernel/arm/fp32/scatter_nd_fp32.h"
#include <string.h>
#include <cstring>
#include <vector>
#include "schema/model_generated.h"
#include "src/kernel_registry.h"
@ -85,7 +85,6 @@ int ScatterNDCPUKernel::ReSize() {
return RET_ERROR;
}
}
// for (size_t i = 0; i < static_cast<size_t>(indice_unit_rank); i++) {}
// calculate unit_size_
unit_size_ = 1;

1
mindspore/lite/src/runtime/kernel/arm/fp32/skip_gram_fp32.cc
View File

@ -27,7 +27,6 @@ using mindspore::lite::StringPack;
using mindspore::schema::PrimitiveType_SkipGram;
namespace mindspore::kernel {
int SkipGramCPUKernel::Init() {
if (!InferShapeDone()) {
return RET_OK;

2
mindspore/lite/src/runtime/kernel/arm/fp32/softmax_fp32.cc
View File

@ -15,7 +15,7 @@
*/
#include "src/runtime/kernel/arm/fp32/softmax_fp32.h"
#include <string.h>
#include <cstring>
#include <vector>
#include "nnacl/fp32/softmax_fp32.h"
#include "schema/model_generated.h"

1
mindspore/lite/src/runtime/kernel/arm/fp32/space_to_depth_fp32.cc
View File

@ -32,7 +32,6 @@ using mindspore::lite::RET_PARAM_INVALID;
using mindspore::schema::PrimitiveType_SpaceToDepth;
namespace mindspore::kernel {
int SpaceToDepthCPUKernel::Init() {
SpaceToDepthParameter *param = reinterpret_cast<SpaceToDepthParameter *>(op_parameter_);
if (param->block_size_ <= 0) {

3
mindspore/lite/src/runtime/kernel/arm/fp32_grad/adam.cc
View File

@ -29,10 +29,9 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_Adam;
namespace mindspore::kernel {
int AdamCPUKernel::ReSize() { return RET_OK; }
static int DoAdam(float *m, float *v, float *gradient, float *weight, float beta1, float beta2, float beta1_power,
static int DoAdam(float *m, float *v, const float *gradient, float *weight, float beta1, float beta2, float beta1_power,
float beta2_power, float eps, float learning_rate, bool nesterov, int start, int end) {
if ((1.f - beta1_power) <= 0.0f) {
MS_LOG(ERROR) << "divisor cannot be 0 or below";

2
mindspore/lite/src/runtime/kernel/arm/fp32_grad/apply_momentum.cc
View File

@ -30,7 +30,7 @@ using mindspore::schema::PrimitiveType_ApplyMomentum;
namespace mindspore::kernel {
int ApplyMomentumCPUKernel::ReSize() { return RET_OK; }
static int DoApplyMomentum(float *weight, float *accumulate, float learning_rate, float *gradient, float moment,
static int DoApplyMomentum(float *weight, float *accumulate, float learning_rate, const float *gradient, float moment,
bool nesterov, int start, int end) {
if (nesterov) {
for (int i = start; i < end; i++) {

3
mindspore/lite/src/runtime/kernel/arm/fp32_grad/arithmetic_grad.cc
View File

@ -28,7 +28,6 @@ using mindspore::lite::RET_ERROR;
using mindspore::lite::RET_OK;
namespace mindspore::kernel {
int ArithmeticGradCPUKernel::Init() {
auto dx1 = out_tensors_[0];
auto dx2 = out_tensors_[1];
@ -240,7 +239,7 @@ kernel::InnerKernel *CpuArithmeticGradFp32KernelCreator(const std::vector<lite::
const std::vector<lite::Tensor *> &outputs,
OpParameter *opParameter, const lite::Context *ctx,
const kernel::KernelKey &desc) {
MS_ASSERT(nullptr != opParameter);
MS_ASSERT(opParameter != nullptr);
if (opParameter == nullptr) {
return nullptr;
}

1
mindspore/lite/src/runtime/kernel/arm/fp32_grad/assign.cc
View File

@ -28,7 +28,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_Assign;
namespace mindspore::kernel {
int AssignCPUKernel::ReSize() { return RET_OK; }
int AssignCPUKernel::Execute(int task_id) {

3
mindspore/lite/src/runtime/kernel/arm/fp32_grad/bias_grad.cc
View File

@ -28,7 +28,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_BiasAddGrad;
namespace mindspore::kernel {
int BiasGradCPUKernel::ReSize() {
auto dims = in_tensors_[0]->shape();
bias_param->ndim_ = dims.size();
@ -37,7 +36,7 @@ int BiasGradCPUKernel::ReSize() {
bias_param->out_shape_[i] = 1; // 1 dimension for N,H,W,
}
bias_param->out_shape_[bias_param->ndim_ - 1] = dims[bias_param->ndim_ - 1];
for (auto i = bias_param->ndim_; i < 4; i++) {
for (auto i = bias_param->ndim_; i < DIMENSION_4D; i++) {
bias_param->in_shape0_[i] = 0;
bias_param->out_shape_[i] = 0;
}

16
mindspore/lite/src/runtime/kernel/arm/fp32_grad/bn_grad.cc
View File

@ -15,7 +15,7 @@
*/
#include "src/runtime/kernel/arm/fp32_grad/bn_grad.h"
#include <math.h>
#include <cmath>
#include <algorithm>
#include <vector>
#include <string>
@ -35,10 +35,15 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_BatchNormGrad;
namespace mindspore::kernel {
namespace {
constexpr int kWsMultiplier = 2;
constexpr int kMaxTaskNum = 4;
} // namespace
int BNGradCPUKernel::ReSize() {
auto *input_x = in_tensors_.at(1);
int channels = input_x->shape().at(kNHWC_C);
ws_size_ = 2 * channels;
ws_size_ = kWsMultiplier * channels;
set_workspace_size(ws_size_ * sizeof(float));
return RET_OK;
}
@ -85,7 +90,7 @@ int BNGradCPUKernel::Execute(int task_id) {
count = (count < 0) ? 0 : count;
switch (stage) {
case 0: {
for (int job = task_id; job < 4; job += thread_num) {
for (int job = task_id; job < kMaxTaskNum; job += thread_num) {
switch (job) {
case 0:
var2Invar(save_var, input_var->ElementsNum(), bn_param->epsilon_);
@ -134,8 +139,9 @@ int BNGradRun(void *cdata, int task_id) {
int BNGradCPUKernel::Run() {
stage_ = 0;
thread_num_ = context_->thread_num_;
int error_code;
if (thread_num_ == 1) {
int error_code = static_cast<const lite::InnerContext *>(this->context_)
error_code = static_cast<const lite::InnerContext *>(this->context_)
->thread_pool_->ParallelLaunch(BNGradRun, this, thread_num_);
if (error_code != RET_OK) {
MS_LOG(ERROR) << "BN function error error_code[" << error_code << "]";
@ -145,7 +151,7 @@ int BNGradCPUKernel::Run() {
const std::vector<int> threads = {thread_num_, 1, thread_num_};
for (size_t stage = 0; stage < threads.size(); stage++) {
stage_ = static_cast<int>(stage);
int error_code = static_cast<const lite::InnerContext *>(this->context_)
error_code = static_cast<const lite::InnerContext *>(this->context_)
->thread_pool_->ParallelLaunch(BNGradRun, this, threads.at(stage));
if (error_code != RET_OK) {
MS_LOG(ERROR) << "BN function error error_code[" << error_code << "]";

36
mindspore/lite/src/runtime/kernel/arm/fp32_grad/convolution.cc
View File

@ -94,7 +94,7 @@ int ConvolutionTrainCPUKernel::Execute(int task_id) {
const int k_h = conv_param_->kernel_h_;
const int k_w = conv_param_->kernel_w_;
const int batch = conv_param_->output_batch_;
const int out_ch = conv_param_->output_channel_; // out_y->shape()[3];
const int out_ch = conv_param_->output_channel_;
const int groups = conv_param_->group_;
const int out_h = conv_param_->output_h_;
const int out_w = conv_param_->output_w_;
@ -103,18 +103,22 @@ int ConvolutionTrainCPUKernel::Execute(int task_id) {
const int k = k_h * k_w * in_ch / groups;
float *workspace_temp = static_cast<float *>(workspace());
float *mat_workspace = workspace_temp + ws_size_;
int real_chunk;
float *mat_a = nullptr;
float *im = nullptr;
const float *mat_b = nullptr;
float *mat_c = nullptr;
if (do_dw_) {
const int kernel_spatial = k_h * k_w;
for (int i = 0; i < batch; ++i) {
for (int ci = 0; ci < m; ci += chunk_) {
int real_chunk = MSMIN(m - ci, chunk_);
float *mat_a = workspace_temp;
float *im = x_addr + (i * in_ch * in_h * in_w);
real_chunk = MSMIN(m - ci, chunk_);
mat_a = workspace_temp;
im = x_addr + (i * in_ch * in_h * in_w);
RollingIm2ColPackDwUnitFp32(im, conv_param_, mat_a, real_chunk, ci);
for (int j = 0; j < groups; ++j) {
const float *mat_b = w_addr + j * nweights / groups;
float *mat_c = y_addr + (i * groups) * n * m + j * (out_ch / groups) + ci * out_ch;
mat_b = w_addr + j * nweights / groups;
mat_c = y_addr + (i * groups) * n * m + j * (out_ch / groups) + ci * out_ch;
GemmMatmul(0, 1, real_chunk, n, k, 1, mat_a + (j * kernel_spatial), k * groups, mat_b, k, 0, mat_c, out_ch,
mat_workspace);
}
@ -124,24 +128,24 @@ int ConvolutionTrainCPUKernel::Execute(int task_id) {
for (int i = 0; i < batch; ++i) {
for (int j = 0; j < groups; ++j) {
for (int ci = 0; ci < m; ci += chunk_) {
int real_chunk = MSMIN(m - ci, chunk_);
float *mat_a = workspace_temp;
const float *mat_b = w_addr + j * nweights / groups;
float *mat_c = y_addr + (i * groups) * n * m + j * (out_ch / groups) + ci * out_ch;
float *im = x_addr + i * in_ch * in_h * in_w + j * (in_ch / groups);
real_chunk = MSMIN(m - ci, chunk_);
mat_a = workspace_temp;
mat_b = w_addr + j * nweights / groups;
mat_c = y_addr + (i * groups) * n * m + j * (out_ch / groups) + ci * out_ch;
im = x_addr + i * in_ch * in_h * in_w + j * (in_ch / groups);
RollingIm2ColPackUnitFp32(im, conv_param_, mat_a, real_chunk, ci);
GemmMatmul(0, 1, real_chunk, n, k, 1, mat_a, k, mat_b, k, 0, mat_c, out_ch, mat_workspace);
}
}
}
} else {
const float *mat_b = w_addr;
mat_b = w_addr;
const size_t in_plane_size = in_ch * in_h * in_w;
for (int i = 0; i < batch; ++i) {
float *im = x_addr + i * in_plane_size;
im = x_addr + i * in_plane_size;
for (int ci = 0; ci < m; ci += chunk_) {
int real_chunk = MSMIN(m - ci, chunk_);
float *mat_c = y_addr + i * n * m + ci * out_ch;
real_chunk = MSMIN(m - ci, chunk_);
mat_c = y_addr + i * n * m + ci * out_ch;
int input_height = ci / out_w * conv_param_->stride_h_;
int input_width = ci % out_w * conv_param_->stride_w_;
int offset = (input_height * in_w + input_width) * in_ch;

34
mindspore/lite/src/runtime/kernel/arm/fp32_grad/convolution_grad_filter.cc
View File

@ -111,7 +111,11 @@ int ConvolutionGradFilterCPUKernel::Execute(int task_id) {
count = (count < 0) ? 0 : count;
int start = stride * task_id;
int end = start + count;
int real_chunk;
float *mat_a = nullptr;
float *mat_b = nullptr;
float *mat_c = nullptr;
float *im = nullptr;
if (do_dw_) {
#ifdef ENABLE_ARM
stride = UP_DIV(k_h * k_w, thread_num);
@ -128,13 +132,13 @@ int ConvolutionGradFilterCPUKernel::Execute(int task_id) {
const int kernel_spatial = k_h * k_w;
for (int i = 0; i < batch; ++i) {
for (int ci = 0; ci < m; ci += chunk_) {
int real_chunk = MSMIN(m - ci, chunk_);
float *mat_b = workspace_temp + task_id * ws_size_;
float *im = x_addr + (i * in_ch * in_h * in_w);
real_chunk = MSMIN(m - ci, chunk_);
mat_b = workspace_temp + task_id * ws_size_;
im = x_addr + (i * in_ch * in_h * in_w);
RollingIm2ColPackDwUnitFp32(im, conv_param, mat_b, real_chunk, ci);
for (int j = start; j < end; ++j) {
float *mat_a = dy_addr + (i * groups) * m * k + j * (out_ch / groups) + ci * out_ch;
float *mat_c = dw_addr + j * nweights / groups;
mat_a = dy_addr + (i * groups) * m * k + j * (out_ch / groups) + ci * out_ch;
mat_c = dw_addr + j * nweights / groups;
GemmMatmul(1, 0, k, n, real_chunk, 1, mat_a, out_ch, mat_b + (j * kernel_spatial), n * groups, 1, mat_c, n,
mat_workspace);
}
@ -145,11 +149,11 @@ int ConvolutionGradFilterCPUKernel::Execute(int task_id) {
for (int i = start; i < end; ++i) {
for (int ci = 0; ci < m; ci += chunk_) {
for (int j = 0; j < groups; ++j) {
int real_chunk = MSMIN(m - ci, chunk_);
float *mat_a = dy_addr + (i * groups) * m * k + j * (out_ch / groups) + ci * out_ch;
float *mat_b = workspace_temp + task_id * ws_size_;
float *mat_c = dw_addr + j * nweights / groups;
float *im = x_addr + (i * in_ch * in_h * in_w) + j * (in_ch / groups);
real_chunk = MSMIN(m - ci, chunk_);
mat_a = dy_addr + (i * groups) * m * k + j * (out_ch / groups) + ci * out_ch;
mat_b = workspace_temp + task_id * ws_size_;
mat_c = dw_addr + j * nweights / groups;
im = x_addr + (i * in_ch * in_h * in_w) + j * (in_ch / groups);
RollingIm2ColPackUnitFp32(im, conv_param, mat_b, real_chunk, ci);
GemmMatmul(1, 0, k, n, real_chunk, 1, mat_a, out_ch, mat_b, n, 0, mat_tmp, n, mat_workspace);
std::unique_lock<std::mutex> merge_lock(lock_);
@ -158,13 +162,13 @@ int ConvolutionGradFilterCPUKernel::Execute(int task_id) {
}
}
} else {
float *mat_c = dw_addr;
mat_c = dw_addr;
const size_t in_plane_size = in_ch * in_h * in_w;
for (int i = start; i < end; ++i) {
for (int ci = 0; ci < m; ci += chunk_) {
int real_chunk = MSMIN(m - ci, chunk_);
float *mat_a = dy_addr + i * m * k + ci * out_ch;
float *im = x_addr + i * in_plane_size;
real_chunk = MSMIN(m - ci, chunk_);
mat_a = dy_addr + i * m * k + ci * out_ch;
im = x_addr + i * in_plane_size;
int input_h = ci / out_w * conv_param->stride_h_;
int input_w = ci % out_w * conv_param->stride_w_;
int offset = (input_h * in_w + input_w) * in_ch;

4
mindspore/lite/src/runtime/kernel/arm/fp32_grad/convolution_grad_input.cc
View File

@ -77,8 +77,8 @@ int ConvolutionGradInputCPUKernel::Execute(int task_id) {
int in_ch = conv_param->input_channel_;
int in_h = conv_param->input_h_;
int in_w = conv_param->input_w_;
int k_h = conv_param->kernel_h_; // out_dw->shape()[1];
int k_w = conv_param->kernel_w_; // out_dw->shape()[2];
int k_h = conv_param->kernel_h_;
int k_w = conv_param->kernel_w_;
int batch = conv_param->output_batch_;
int out_ch = conv_param->output_channel_;
int groups = conv_param->group_;

1
mindspore/lite/src/runtime/kernel/arm/fp32_grad/dropout.cc
View File

@ -31,7 +31,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_Dropout;
namespace mindspore::kernel {
int DropoutCPUKernel::Init() {
auto param = reinterpret_cast<DropoutParameter *>(op_parameter_);
if (param == nullptr) {

2
mindspore/lite/src/runtime/kernel/arm/fp32_grad/dropout_grad.cc
View File

@ -31,7 +31,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_DropoutGrad;
namespace mindspore::kernel {
int DropoutGradCPUKernel::Init() {
auto param = reinterpret_cast<DropoutParameter *>(op_parameter_);
if (param == nullptr) {
@ -64,7 +63,6 @@ int DropoutGradCPUKernel::Execute(int task_id) {
auto length = in_tensors_.at(kInputIndex)->ElementsNum();
int stride = UP_DIV(length, thread_count_);
int count = MSMIN(stride, length - stride * task_id);
if (count > 0) {
int start = stride * task_id;
DropoutGrad(&(yt_ptr[start]), &(mask_ptr[start]), &(output_ptr[start]), count, scale_);

1
mindspore/lite/src/runtime/kernel/arm/fp32_grad/pooling_grad.cc
View File

@ -68,7 +68,6 @@ int PoolingGradCPUKernel::Execute(int task_id) {
auto output_ptr = reinterpret_cast<float *>(out_tensors_.at(0)->MutableData());
int stride = UP_DIV(pool_param->output_batch_, thread_num_);
int count = MSMIN(stride, pool_param->output_batch_ - stride * task_id);
if (count > 0) {
int in_batch_size = pool_param->input_h_ * pool_param->input_w_ * pool_param->input_channel_;
int out_batch_size = pool_param->output_h_ * pool_param->output_w_ * pool_param->input_channel_;

1
mindspore/lite/src/runtime/kernel/arm/fp32_grad/resize_grad.cc
View File

@ -29,7 +29,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_ResizeGrad;
namespace mindspore::kernel {
float Scaling(size_t in_size, size_t out_size, bool align_corners) {
return (align_corners && out_size > 1) ? (in_size - 1) / static_cast<float>(out_size - 1)
: in_size / static_cast<float>(out_size);

13
mindspore/lite/src/runtime/kernel/arm/fp32_grad/sgd.cc
View File

@ -29,10 +29,9 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_SGD;
namespace mindspore::kernel {
int SgdCPUKernel::ReSize() { return RET_OK; }
int DoSgd(float *weight, float *accumulate, float *gradient, float learning_rate, float dampening, float moment,
int DoSgd(float *weight, float *accumulate, const float *gradient, float learning_rate, float dampening, float moment,
bool nesterov, int start, int end) {
if (moment > 0.f) {
if (nesterov) {
@ -54,8 +53,8 @@ int DoSgd(float *weight, float *accumulate, float *gradient, float learning_rate
return RET_OK;
}
int DoSgdInit(float *weight, float *accumulate, float *gradient, float *stat, float learning_rate, float dampening,
float moment, bool nesterov, int start, int end) {
int DoSgdInit(float *weight, float *accumulate, float *gradient, float *stat, float learning_rate, float moment,
bool nesterov, int start, int end) {
std::copy(&(gradient[start]), &(gradient[end]), &(accumulate[start]));
if (nesterov) {
for (int i = start; i < end; ++i) {
@ -106,8 +105,7 @@ int SgdCPUKernel::ExecuteInit(int task_id) {
int end = start + count;
if (count > 0) {
DoSgdInit(weight, accumulate, gradient, stat, learning_rate, sgd_param_->dampening_, moment,
sgd_param_->use_nesterov_, start, end);
DoSgdInit(weight, accumulate, gradient, stat, learning_rate, moment, sgd_param_->use_nesterov_, start, end);
}
return RET_OK;
}
@ -192,8 +190,7 @@ int SgdCPUKernel::OptimizerStep() {
DoSgd(weight, accumulate, grad_sum_, learning_rate, sgd_param_->dampening_, moment, sgd_param_->use_nesterov_,
start, end);
} else {
DoSgdInit(weight, accumulate, grad_sum_, stat, learning_rate, sgd_param_->dampening_, moment,
sgd_param_->use_nesterov_, start, end);
DoSgdInit(weight, accumulate, grad_sum_, stat, learning_rate, moment, sgd_param_->use_nesterov_, start, end);
}
std::fill(grad_sum_, grad_sum_ + length, 0);
OptimizerKernel::OptimizerStep();

3
mindspore/lite/src/runtime/kernel/arm/fp32_grad/sigmoid_cross_entropy_with_logits.cc
View File

@ -15,7 +15,7 @@
*/
#include "src/runtime/kernel/arm/fp32_grad/sigmoid_cross_entropy_with_logits.h"
#include <math.h>
#include <cmath>
#include "src/kernel_registry.h"
#include "include/errorcode.h"
#include "src/runtime/runtime_api.h"
@ -26,7 +26,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_SigmoidCrossEntropyWithLogits;
namespace mindspore::kernel {
int SigmoidCrossEntropyWithLogitsCPUKernel::ReSize() { return RET_OK; }
int SigmoidCrossEntropyWithLogitsCPUKernel::Execute(int task_id) {

3
mindspore/lite/src/runtime/kernel/arm/fp32_grad/sigmoid_cross_entropy_with_logits_grad.cc
View File

@ -15,7 +15,7 @@
*/
#include "src/runtime/kernel/arm/fp32_grad/sigmoid_cross_entropy_with_logits_grad.h"
#include <math.h>
#include <cmath>
#include "src/kernel_registry.h"
#include "include/errorcode.h"
#include "src/runtime/runtime_api.h"
@ -26,7 +26,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_SigmoidCrossEntropyWithLogitsGrad;
namespace mindspore::kernel {
int SigmoidCrossEntropyWithLogitsGradCPUKernel::ReSize() { return RET_OK; }
int SigmoidCrossEntropyWithLogitsGradCPUKernel::Execute(int task_id) {

1
mindspore/lite/src/runtime/kernel/arm/fp32_grad/smooth_l1_loss.cc
View File

@ -25,7 +25,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_SmoothL1Loss;
namespace mindspore::kernel {
int SmoothL1LossCPUKernel::ReSize() { return RET_OK; }
int SmoothL1LossCPUKernel::Execute(int task_id) {

1
mindspore/lite/src/runtime/kernel/arm/fp32_grad/smooth_l1_loss_grad.cc
View File

@ -25,7 +25,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_SmoothL1LossGrad;
namespace mindspore::kernel {
int SmoothL1LossGradCPUKernel::ReSize() { return RET_OK; }
int SmoothL1LossGradCPUKernel::Execute(int task_id) {

3
mindspore/lite/src/runtime/kernel/arm/fp32_grad/softmax_cross_entropy_with_logits.cc
View File

@ -14,7 +14,7 @@
* limitations under the License.
*/
#include <math.h>
#include <cmath>
#include "src/kernel_registry.h"
#include "nnacl/softmax_parameter.h"
#include "nnacl/fp32/softmax_fp32.h"
@ -28,7 +28,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_SoftmaxCrossEntropyWithLogits;
namespace mindspore::kernel {
int SoftmaxCrossEntropyWithLogitsCPUKernel::Init() { return ReSize(); }
void SoftmaxCrossEntropyWithLogitsCPUKernel::ForwardPostExecute(const float *labels, const float *logits, float *grads,

2
mindspore/lite/src/runtime/kernel/arm/fp32_grad/softmax_grad.cc
View File

@ -15,7 +15,7 @@
*/
#include "src/runtime/kernel/arm/fp32_grad/softmax_grad.h"
#include <string.h>
#include <cstring>
#include <vector>
#include "nnacl/fp32_grad/softmax_grad.h"
#include "schema/model_generated.h"

7
mindspore/lite/src/runtime/kernel/arm/fp32_grad/sparse_softmax_cross_entropy_with_logits.cc
View File

@ -1,5 +1,5 @@
/**
* Copyright 2020 Huawei Technologies Co., Ltd
* Copyright 2020-2021 Huawei Technologies Co., Ltd
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -14,7 +14,7 @@
* limitations under the License.
*/
#include <math.h>
#include <cmath>
#include "src/kernel_registry.h"
#include "nnacl/softmax_parameter.h"
#include "nnacl/fp32/softmax_fp32.h"
@ -29,7 +29,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_SparseSoftmaxCrossEntropyWithLogits;
namespace mindspore::kernel {
int SparseSoftmaxCrossEntropyWithLogitsCPUKernel::ReSize() { return RET_OK; }
int SparseSoftmaxCrossEntropyWithLogitsCPUKernel::ForwardPostExecute(const int *labels, const float *losses,
@ -162,7 +161,7 @@ int SparseSoftmaxCrossEntropyWithLogitsCPUKernel::Init() {
param->number_of_classes_ = dims.at(1);
param->batch_size_ = dims.at(0);
for (unsigned int i = 0; i < dims.size(); i++) param->input_shape_[i] = dims.at(i);
if (2 != this->in_tensors_.size()) {
if (this->in_tensors_.size() != 2) {
MS_LOG(ERROR) << "sparse softmax entropy loss should have two inputs";
return RET_ERROR;
}

1
mindspore/lite/src/runtime/kernel/arm/fp32_grad/strided_slice_grad.cc
View File

@ -31,7 +31,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_StridedSliceGrad;
namespace mindspore::kernel {
int StridedSliceGradCPUKernel::Init() {
if (!InferShapeDone()) {
return RET_OK;

1
mindspore/lite/src/runtime/kernel/arm/fp32_grad/unsorted_segment_sum.cc
View File

@ -30,7 +30,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_UnsortedSegmentSum;
namespace mindspore::kernel {
int UnsortedSegmentSumCPUKernel::Init() {
if (!InferShapeDone()) {
return RET_OK;

6
mindspore/lite/src/runtime/kernel/arm/int8/add_int8.cc
View File

@ -27,6 +27,10 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_AddFusion;
namespace mindspore::kernel {
namespace {
constexpr int kBaseShift = 20;
} // namespace
QuantizedAddCPUKernel::~QuantizedAddCPUKernel() {
if (para_ != nullptr) {
free(para_);
@ -52,7 +56,7 @@ int QuantizedAddCPUKernel::Init() {
const double in1_scale = input1->quant_params().front().scale;
const double out_scale = output->quant_params().front().scale;
para_->left_shift_ = 20;
para_->left_shift_ = kBaseShift;
const double twice_max_input_scale = 2 * std::max(in0_scale, in1_scale);
const double in0_multiplier = in0_scale / twice_max_input_scale;
const double in1_multiplier = in1_scale / twice_max_input_scale;

5
mindspore/lite/src/runtime/kernel/arm/int8/arithmetic_int8.cc
View File

@ -107,6 +107,7 @@ int ArithmeticInt8CPUKernel::DoArithmetic(int thread_id) {
auto output_data = reinterpret_cast<uint8_t *>(out_tensors_[0]->MutableData());
auto element_num = out_tensors_[0]->ElementsNum();
auto param = reinterpret_cast<ArithmeticParameter *>(op_parameter_);
int error_code;
if (param->broadcasting_ && arithmetic_run_ != nullptr) {
MS_ASSERT(op_parameter_->thread_num_ != 0);
int stride = UP_DIV(element_num, op_parameter_->thread_num_);
@ -115,14 +116,14 @@ int ArithmeticInt8CPUKernel::DoArithmetic(int thread_id) {
return RET_OK;
}
int error_code = arithmetic_run_(tile_data0_ + stride * thread_id, tile_data1_ + stride * thread_id,
error_code = arithmetic_run_(tile_data0_ + stride * thread_id, tile_data1_ + stride * thread_id,
output_data + stride * thread_id, count, &quant_args_);
if (error_code != RET_OK) {
MS_LOG(ERROR) << "Arithmetic run fail! ret: " << error_code;
return error_code;
}
} else if (arithmetic_run_ != nullptr) {
int error_code = arithmetic_run_(input0_data, input1_data1, output_data, element_num, &quant_args_);
error_code = arithmetic_run_(input0_data, input1_data1, output_data, element_num, &quant_args_);
if (error_code != RET_OK) {
MS_LOG(ERROR) << "Arithmetic run fail!ret: " << error_code;
return error_code;

2
mindspore/lite/src/runtime/kernel/arm/int8/batch_to_space_int8.cc
View File

@ -63,7 +63,7 @@ int BatchToSpaceInt8CPUKernel::Init() {
}
int BatchToSpaceInt8CPUKernel::ReSize() {
MS_ASSERT(in_tensors_.at(0)->shape().size() == 4);
MS_ASSERT(in_tensors_.at(0)->shape().size() == DIMENSION_4D);
return RET_OK;
}

2
mindspore/lite/src/runtime/kernel/arm/int8/batchnorm_int8.cc
View File

@ -15,7 +15,7 @@
*/
#include "src/runtime/kernel/arm/int8/batchnorm_int8.h"
#include <math.h>
#include <cmath>
#include "schema/model_generated.h"
#include "src/kernel_registry.h"
#include "include/errorcode.h"

6
mindspore/lite/src/runtime/kernel/arm/int8/concat_int8.cc
View File

@ -40,9 +40,9 @@ int ConcatInt8CPUKernel::Init() {
}
for (size_t i = 0; i < input_num; i++) {
auto *input_tensor = in_tensors_.at(i);
auto quant_args = input_tensor->quant_params();
concat_param_->quant_arg_.in_args_[i].scale_ = quant_args.front().scale;
concat_param_->quant_arg_.in_args_[i].zp_ = quant_args.front().zeroPoint;
auto in_quant_args = input_tensor->quant_params();
concat_param_->quant_arg_.in_args_[i].scale_ = in_quant_args.front().scale;
concat_param_->quant_arg_.in_args_[i].zp_ = in_quant_args.front().zeroPoint;
}
auto output_tensor = out_tensors_.at(kOutputIndex);

4
mindspore/lite/src/runtime/kernel/arm/int8/convolution_1x1_int8.cc
View File

@ -238,7 +238,7 @@ int Convolution1x1Int8CPUKernel::InitWeightBias() {
return RET_ERROR;
}
memset(bias_data_, 0, size * sizeof(int32_t));
if (in_tensors_.size() == 3) {
if (in_tensors_.size() == kInputSize2) {
memcpy(bias_data_, in_tensors_.at(kBiasIndex)->data_c(), output_channel * sizeof(int32_t));
}
@ -270,7 +270,7 @@ int Convolution1x1Int8CPUKernel::InitWeightBiasArm32() {
return RET_ERROR;
}
memset(bias_data_, 0, col2 * sizeof(int32_t));
if (in_tensors_.size() == 3) {
if (in_tensors_.size() == kInputSize2) {
memcpy(bias_data_, in_tensors_.at(kBiasIndex)->data_c(), output_channel * sizeof(int32_t));
}

9
mindspore/lite/src/runtime/kernel/arm/int8/convolution_3x3_int8.cc
View File

@ -23,6 +23,9 @@ using mindspore::lite::RET_ERROR;
using mindspore::lite::RET_OK;
namespace mindspore::kernel {
namespace {
constexpr size_t kUnitBufferMultipler = 4 * 4;
} // namespace
int ProcessFilterUint8(int8_t *origin_weight, int16_t *dst_weight, ConvParameter *conv_param) {
auto input_channel = conv_param->input_channel_;
auto output_channel = conv_param->output_channel_;
@ -82,7 +85,7 @@ int Convolution3x3Int8CPUKernel::InitWeightBias() {
int iC8 = UP_DIV(input_channel, C8NUM);
int oC4 = UP_DIV(output_channel, C4NUM);
// init weight
size_t transformed_size = iC8 * C8NUM * oC4 * C4NUM * 16 * sizeof(int16_t);
size_t transformed_size = iC8 * C8NUM * oC4 * C4NUM * kUnitBufferMultipler * sizeof(int16_t);
transformed_filter_addr_ = reinterpret_cast<int16_t *>(malloc(transformed_size));
if (transformed_filter_addr_ == nullptr) {
MS_LOG(ERROR) << "malloc transformed_filter_addr_ failed.";
@ -136,14 +139,14 @@ int Convolution3x3Int8CPUKernel::InitTmpBuffer() {
return RET_ERROR;
}
size_t block_unit_buffer_size = thread_count_ * 4 * 4 * C8NUM * sizeof(int16_t);
size_t block_unit_buffer_size = thread_count_ * kUnitBufferMultipler * C8NUM * sizeof(int16_t);
block_unit_buffer_ = reinterpret_cast<int16_t *>(ctx_->allocator->Malloc(block_unit_buffer_size));
if (block_unit_buffer_ == nullptr) {
MS_LOG(ERROR) << "malloc block_unit_buffer_ failed.";
return RET_ERROR;
}
size_t tmp_dst_buffer_size = thread_count_ * TILE_NUM * 16 * oc4 * C4NUM * sizeof(int32_t);
size_t tmp_dst_buffer_size = thread_count_ * TILE_NUM * kUnitBufferMultipler * oc4 * C4NUM * sizeof(int32_t);
tmp_dst_buffer_ = reinterpret_cast<int32_t *>(ctx_->allocator->Malloc(tmp_dst_buffer_size));
if (tmp_dst_buffer_ == nullptr) {
MS_LOG(ERROR) << "malloc tmp_dst_buffer_ failed.";

14
mindspore/lite/src/runtime/kernel/arm/int8/convolution_depthwise_3x3_int8.cc
View File

@ -23,6 +23,10 @@ using mindspore::lite::RET_ERROR;
using mindspore::lite::RET_OK;
namespace mindspore::kernel {
namespace {
constexpr int kConvDepthwise3x3BufferSize = 64 * 10 * 10;
constexpr int kChannelUnit = 8;
} // namespace
ConvolutionDepthwise3x3Int8CPUKernel::~ConvolutionDepthwise3x3Int8CPUKernel() {
if (sliding_ != nullptr) {
delete sliding_;
@ -40,7 +44,7 @@ int ConvolutionDepthwise3x3Int8CPUKernel::InitWeightBias() {
auto weight_tensor = in_tensors_.at(kWeightIndex);
auto origin_weight = reinterpret_cast<int8_t *>(weight_tensor->MutableData());
int channel = weight_tensor->Batch();
if (channel % 8 != 0) {
if (channel % kChannelUnit != 0) {
MS_LOG(ERROR) << "ConvolutionDepthwise3x3Int8CPUKernel doesn't support channel " << channel;
return RET_ERROR;
}
@ -63,8 +67,8 @@ int ConvolutionDepthwise3x3Int8CPUKernel::InitWeightBias() {
if (filter_per_channel) {
for (int i = 0; i < weight_tensor->Height() * weight_tensor->Width(); i++) {
for (int c = 0; c < channel; c++) {
int weight_zp = conv_param_->conv_quant_arg_.filter_quant_args_[c].zp_;
packed_weight_[i * channel + c] = (int16_t)(tmp_weight[i * channel + c] - weight_zp);
int per_channel_weight_zp = conv_param_->conv_quant_arg_.filter_quant_args_[c].zp_;
packed_weight_[i * channel + c] = (int16_t)(tmp_weight[i * channel + c] - per_channel_weight_zp);
}
}
} else {
@ -119,7 +123,7 @@ int ConvolutionDepthwise3x3Int8CPUKernel::ReSize() {
}
int ConvolutionDepthwise3x3Int8CPUKernel::Execute(int task_id) {
auto buffer = buffer_ + 64 * 10 * 10 * task_id;
auto buffer = buffer_ + kConvDepthwise3x3BufferSize * task_id;
ConvDw3x3Int8(output_ptr_, buffer, input_ptr_, packed_weight_, reinterpret_cast<int32_t *>(bias_data_), conv_param_,
sliding_, task_id);
return RET_OK;
@ -136,7 +140,7 @@ int ConvDw3x3Int8Run(void *cdata, int task_id) {
}
int ConvolutionDepthwise3x3Int8CPUKernel::InitBuffer() {
int buffer_size = 64 * 10 * 10 * conv_param_->thread_num_;
int buffer_size = kConvDepthwise3x3BufferSize * conv_param_->thread_num_;
buffer_ = reinterpret_cast<int8_t *>(context_->allocator->Malloc(buffer_size * sizeof(int8_t)));
if (buffer_ == nullptr) {
MS_LOG(ERROR) << "Malloc buffer failed.";

4
mindspore/lite/src/runtime/kernel/arm/int8/convolution_depthwise_int8.cc
View File

@ -56,8 +56,8 @@ int ConvolutionDepthwiseInt8CPUKernel::InitWeightBias() {
if (filter_per_channel) {
for (int i = 0; i < weight_tensor->Height() * weight_tensor->Width(); i++) {
for (int c = 0; c < channel; c++) {
int weight_zp = conv_param_->conv_quant_arg_.filter_quant_args_[c].zp_;
packed_weight_[i * channel + c] = (int16_t)(tmp_weight[i * channel + c] - weight_zp);
int per_channel_weight_zp = conv_param_->conv_quant_arg_.filter_quant_args_[c].zp_;
packed_weight_[i * channel + c] = (int16_t)(tmp_weight[i * channel + c] - per_channel_weight_zp);
}
}
} else {

2
mindspore/lite/src/runtime/kernel/arm/int8/convolution_depthwise_slidewindow_int8.cc
View File

@ -38,7 +38,7 @@ ConvolutionDepthwiseSWInt8CPUKernel::~ConvolutionDepthwiseSWInt8CPUKernel() {
int ConvolutionDepthwiseSWInt8CPUKernel::InitWeightBias() {
// init weight, int8 -> int16
// o, h, w, i -> o/8, h, w, i, 8; o == group, i == 1
// o, h, w, i -> o/8, h, w, i, 8; o equals to group, i equals to 1
auto weight_tensor = in_tensors_.at(kWeightIndex);
auto origin_weight = reinterpret_cast<int8_t *>(weight_tensor->MutableData());
int OC8 = UP_DIV(weight_tensor->Batch(), C8NUM);

4
mindspore/lite/src/runtime/kernel/arm/int8/deconvolution_depthwise_int8.cc
View File

@ -37,7 +37,7 @@ DeconvolutionDepthwiseInt8CPUKernel::~DeconvolutionDepthwiseInt8CPUKernel() {
int DeconvolutionDepthwiseInt8CPUKernel::InitWeightBias() {
// init weight: int8 -> int16
// o, h, w, i -> o/8, h, w, i, 8; o == group, i == 1
// o, h, w, i -> o/8, h, w, i, 8; o equals to group, i equals to 1
auto weight_tensor = in_tensors_.at(kWeightIndex);
auto origin_weight = reinterpret_cast<int8_t *>(weight_tensor->MutableData());
int OC4 = UP_DIV(weight_tensor->Batch(), C4NUM);
@ -87,7 +87,7 @@ int DeconvolutionDepthwiseInt8CPUKernel::InitSlideParam() {
int DeconvolutionDepthwiseInt8CPUKernel::InitBuffer() {
int pack_input_size = conv_param_->input_batch_ * conv_param_->input_h_ * conv_param_->input_w_ * C4NUM *
UP_DIV(conv_param_->input_channel_, 4);
UP_DIV(conv_param_->input_channel_, C4NUM);
packed_input_ = reinterpret_cast<int16_t *>(context_->allocator->Malloc(pack_input_size * sizeof(int16_t)));
if (packed_input_ == nullptr) {
MS_LOG(ERROR) << "Malloc buffer failed.";

2
mindspore/lite/src/runtime/kernel/arm/int8/deconvolution_int8.cc
View File

@ -132,7 +132,7 @@ int DeConvInt8CPUKernel::InitBiasWeight() {
return RET_ERROR;
}
memset(bias_data_, 0, size);
if (in_tensors_.size() == 3) {
if (in_tensors_.size() == kInputSize2) {
memcpy(bias_data_, in_tensors_.at(0)->MutableData(), conv_param_->output_channel_ * sizeof(int32_t));
}

1
mindspore/lite/src/runtime/kernel/arm/int8/depth_to_space_int8.cc
View File

@ -81,5 +81,4 @@ int DepthToSpaceInt8CPUKernel::Run() {
}
REG_KERNEL(kCPU, kNumberTypeInt8, PrimitiveType_DepthToSpace, LiteKernelCreator<DepthToSpaceInt8CPUKernel>)
} // namespace mindspore::kernel

2
mindspore/lite/src/runtime/kernel/arm/int8/gatherNd_int8.cc
View File

@ -15,7 +15,7 @@
*/
#include "src/runtime/kernel/arm/int8/gatherNd_int8.h"
#include <string.h>
#include <cstring>
#include <limits>
#include <vector>
#include "schema/model_generated.h"

1
mindspore/lite/src/runtime/kernel/arm/int8/gather_int8.cc
View File

@ -30,7 +30,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_Gather;
namespace mindspore::kernel {
int GatherInt8CPUKernel::Init() {
axis_ = (reinterpret_cast<GatherParameter *>(op_parameter_))->axis_;
auto in_quant_args = in_tensors_.at(0)->quant_params();

4
mindspore/lite/src/runtime/kernel/arm/int8/matmul_base_int8.cc
View File

@ -262,8 +262,8 @@ int MatmulBaseInt8CPUKernel::InitTmpBuffer() {
}
int MatmulBaseInt8CPUKernel::InitBias() {
if (in_tensors_.size() == 3) {
auto bias_tensor = in_tensors_[2];
if (in_tensors_.size() == kInputSize2) {
auto bias_tensor = in_tensors_[kBiasIndex];
int max_bias_data = UP_ROUND(bias_tensor->ElementsNum(), C4NUM);
bias_ptr_ = reinterpret_cast<int *>(malloc(max_bias_data * sizeof(int)));
if (bias_ptr_ == nullptr) {

21
mindspore/lite/src/runtime/kernel/arm/int8/mul_int8.cc
View File

@ -70,12 +70,12 @@ int MulInt8CPUKernel::Init() {
}
void MulInt8CPUKernel::CheckSameShapeSize(std::vector<int> in_tensor0_shape, std::vector<int> in_tensor1_shape) {
bool condition1 = in_tensor0_shape[0] == in_tensor1_shape[0];
bool condition2 = in_tensor0_shape[1] == 1;
bool condition3 = in_tensor0_shape[2] == 1;
bool condition4 = in_tensor0_shape[3] == in_tensor1_shape[3];
bool condition5 = in_tensor1_shape[1] == 1;
bool condition6 = in_tensor1_shape[2] == 1;
bool condition1 = in_tensor0_shape[kNHWC_N] == in_tensor1_shape[kNHWC_N];
bool condition2 = in_tensor0_shape[kNHWC_H] == 1;
bool condition3 = in_tensor0_shape[kNHWC_W] == 1;
bool condition4 = in_tensor0_shape[kNHWC_C] == in_tensor1_shape[kNHWC_C];
bool condition5 = in_tensor1_shape[kNHWC_H] == 1;
bool condition6 = in_tensor1_shape[kNHWC_W] == 1;
if (condition1 && condition2 && condition3 && condition4) {
fast_hw_broadcast_ = true;
} else if (condition1 && condition4 && condition5 && condition6) {
@ -91,11 +91,11 @@ void MulInt8CPUKernel::CheckIfFastImpl() {
if (in_tensor0->shape().size() == COMM_SHAPE_SIZE && in_tensor1->shape().size() == COMM_SHAPE_SIZE) {
CheckSameShapeSize(in_tensor0->shape(), in_tensor1->shape());
} else if (in_tensor0->shape().size() == 1 && in_tensor1->shape().size() == COMM_SHAPE_SIZE) {
if (in_tensor0->ElementsNum() == in_tensor1->shape()[3]) {
if (in_tensor0->ElementsNum() == in_tensor1->shape()[kNHWC_C]) {
fast_hw_broadcast_ = true;
}
} else if (in_tensor0->shape().size() == COMM_SHAPE_SIZE && in_tensor1->shape().size() == 1) {
if (in_tensor1->ElementsNum() == in_tensor0->shape()[3]) {
if (in_tensor1->ElementsNum() == in_tensor0->shape()[kNHWC_C]) {
fast_hw_broadcast_ = true;
input1_hw_broadcast_ = true;
}
@ -162,6 +162,7 @@ int MulInt8CPUKernel::Run() {
elements_num_ = out_tensors_.at(0)->ElementsNum();
count_unit_ = thread_count_ > 1 ? UP_DIV(elements_num_, thread_count_) : elements_num_;
int ret = RET_ERROR;
if (in_tensors_.at(0)->ElementsNum() != in_tensors_.at(1)->ElementsNum()) {
input0_data_ = static_cast<int8_t *>(ctx_->allocator->Malloc(out_tensors_.at(0)->Size()));
if (input0_data_ == nullptr) {
@ -176,14 +177,14 @@ int MulInt8CPUKernel::Run() {
}
TileDimensionsInt8(static_cast<int8_t *>(in_tensors_.at(0)->MutableData()),
static_cast<int8_t *>(in_tensors_.at(1)->MutableData()), input0_data_, input1_data_, tile_para);
auto ret = static_cast<const lite::InnerContext *>(this->context_)
ret = static_cast<const lite::InnerContext *>(this->context_)
->thread_pool_->ParallelLaunch(MulInt8Run, this, thread_count_);
ctx_->allocator->Free(input0_data_);
ctx_->allocator->Free(input1_data_);
return ret;
}
auto ret = static_cast<const lite::InnerContext *>(this->context_)
ret = static_cast<const lite::InnerContext *>(this->context_)
->thread_pool_->ParallelLaunch(MulInt8Run, this, thread_count_);
return ret;
}

12
mindspore/lite/src/runtime/kernel/arm/int8/opt_op_handler.cc
View File

@ -15,23 +15,13 @@
*/
#include "src/runtime/kernel/arm/int8/opt_op_handler.h"
#include <stdlib.h>
#include <cstdlib>
#include "nnacl/op_base.h"
#ifdef __cplusplus
extern "C" {
#endif
extern void MatMulOptR4Int8Neon64(const int8_t *a, const int8_t *b, int *dst, int row4, int col4, int deep16,
const int *input_sum, const int *bias);
extern void MatmulInt8DpNeon64(const int8_t *a, const int8_t *b, int8_t *dst, int row8, int col8, int deep4,
const int *a_sums, const int *bias, int act_min, int act_max, int out_zp,
int *multiplier, int *left_shift, int *right_shift, int row, int col, int stride,
size_t peroc);
extern void MatmulInt8DpOpt(const int8_t *a, const int8_t *b, int8_t *dst, size_t row8, size_t col8, size_t deep4,
const int *a_sums, const int *bias, int act_min, int act_max, int out_zp, int *multiplier,
int *left_shift, int *right_shift, size_t stride, size_t peroc, int *filter_zp);
#ifdef ENABLE_ARM64
void MatMulR4Int8_optimize_handler(const int8_t *a, const int8_t *b, int *dst, int row4, int col4, int deep16,

8
mindspore/lite/src/runtime/kernel/arm/int8/opt_op_handler.h
View File

@ -22,6 +22,14 @@
#ifdef __cplusplus
extern "C" {
#endif
void MatMulOptR4Int8Neon64(const int8_t *a, const int8_t *b, int *dst, int row4, int col4, int deep16,
const int *input_sum, const int *bias);
void MatmulInt8DpNeon64(const int8_t *a, const int8_t *b, int8_t *dst, int row8, int col8, int deep4, const int *a_sums,
const int *bias, int act_min, int act_max, int out_zp, int *multiplier, int *left_shift,
int *right_shift, int row, int col, int stride, size_t peroc);
void MatmulInt8DpOpt(const int8_t *a, const int8_t *b, int8_t *dst, size_t row8, size_t col8, size_t deep4,
const int *a_sums, const int *bias, int act_min, int act_max, int out_zp, int *multiplier,
int *left_shift, int *right_shift, size_t stride, size_t peroc, int *filter_zp);
#ifdef ENABLE_ARM64
void IndirectGemmInt8_optimize_handler(int8_t *dst, const int8_t *src, const int8_t *weight, const int32_t *bias,
size_t ksize, size_t ic4, size_t output_channel, size_t offset,

1
mindspore/lite/src/runtime/kernel/arm/int8/pad_int8.cc
View File

@ -26,7 +26,6 @@ using mindspore::lite::RET_OK;
using mindspore::lite::KernelRegistrar;
using mindspore::schema::PrimitiveType_PadFusion;
namespace mindspore::kernel {
namespace {
constexpr size_t kMirrorPadInputSize = 2;
}

16
mindspore/lite/src/runtime/kernel/arm/int8/reduce_int8.cc
View File

@ -240,7 +240,7 @@ int ReduceInt8CPUKernel::CalculateQuantArgs() {
// (quant_out - zp_out)*scale_out = sum((quant_in -zp)*scale_in) * (1/num) for each axis in axes
// quant_out = sum(quant_in-zp) * (scale_in/scale_out) * (1/num)
if (mode_ == static_cast<int>(schema::ReduceMode_ReduceMean)) {
if (input->shape().size() == 4 && pattern_ == kernel::HW) {
if (input->shape().size() == DIMENSION_4D && pattern_ == kernel::HW) {
// special case, can use pattern
ReduceMean4DCalQuantParam();
pattern_impl_ = true;
@ -309,8 +309,8 @@ int ReduceInt8CPUKernel::CalculateQuantArgsReduceSumSquare() {
MS_LOG(ERROR) << "ReduceProd new QuantMultiplier failed.";
return RET_NULL_PTR;
}
double sumsquare_multiplier = quant_arg_.in_scale_ * quant_arg_.in_scale_ / quant_arg_.out_scale_;
QuantizeMultiplierSmallerThanOne(sumsquare_multiplier, &qm->multiplier_, &shift);
double last_sumsquare_multiplier = quant_arg_.in_scale_ * quant_arg_.in_scale_ / quant_arg_.out_scale_;
QuantizeMultiplierSmallerThanOne(last_sumsquare_multiplier, &qm->multiplier_, &shift);
qm->left_shift_ = shift < 0 ? -shift : 0;
qm->right_shift_ = shift > 0 ? shift : 0;
sum_square_multipliers_.push_back(qm);
@ -470,8 +470,9 @@ int ReduceInt8CPUKernel::Fast4DReduceMeanHWImpl() {
}
int ReduceInt8CPUKernel::Run() {
int ret;
if (!this->valid_shape_) {
auto ret = CalculateQuantArgs();
ret = CalculateQuantArgs();
if (ret != RET_OK) {
return ret;
}
@ -481,7 +482,7 @@ int ReduceInt8CPUKernel::Run() {
return Fast4DReduceMeanHWImpl();
}
auto ret = MallocTmpBuffer();
ret = MallocTmpBuffer();
if (ret != RET_OK) {
FreeTmpBuffer();
return ret;
@ -495,13 +496,14 @@ int ReduceInt8CPUKernel::Run() {
begin_src_data_[i] = static_cast<int32_t>(input_data[i]);
}
src_data_ = begin_src_data_;
int error_code = RET_ERROR;
for (size_t i = 0; i < data_buffers_.size(); ++i) {
GetQuantArgs(i);
dst_data_ = data_buffers_[i];
outer_size_ = outer_sizes_[i];
inner_size_ = inner_sizes_[i];
axis_size_ = axis_sizes_[i];
auto error_code = static_cast<const lite::InnerContext *>(this->context_)
error_code = static_cast<const lite::InnerContext *>(this->context_)
->thread_pool_->ParallelLaunch(ReduceInt8Impl, this, context_->thread_num_);
if (error_code != RET_OK) {
FreeTmpBuffer();
@ -517,7 +519,7 @@ int ReduceInt8CPUKernel::Run() {
axis_size_ = axis_sizes_.back();
last_dst_data_ = reinterpret_cast<int8_t *>(out_tensors_.at(0)->MutableData());
is_last_axis_ = true;
auto error_code = static_cast<const lite::InnerContext *>(this->context_)
error_code = static_cast<const lite::InnerContext *>(this->context_)
->thread_pool_->ParallelLaunch(ReduceInt8Impl, this, context_->thread_num_);
if (error_code != RET_OK) {
MS_LOG(ERROR) << "Reduce run error, error_code[" << error_code << "]";

1
mindspore/lite/src/runtime/kernel/arm/int8/reshape_int8.cc
View File

@ -29,7 +29,6 @@ using mindspore::lite::KernelRegistrar;
using mindspore::schema::PrimitiveType_Reshape;
namespace mindspore::kernel {
int ReshapeInt8CPUKernel::Init() {
auto *input_tensor = in_tensors_.at(kInputIndex);
auto in_quant_args = input_tensor->quant_params();

45
mindspore/lite/src/runtime/kernel/arm/int8/resize_int8.cc
View File

@ -33,6 +33,9 @@ using mindspore::lite::RET_OK;
using mindspore::lite::KernelRegistrar;
namespace mindspore::kernel {
namespace {
constexpr unsigned int OFFSET_BASE = 10;
} // namespace
void ResizeInt8CPUKernel::FreeResizeBiLinear() {
free(resize_quant_arg_.x_axis_index_);
free(resize_quant_arg_.x_axis_lower_);
@ -96,32 +99,32 @@ int ResizeInt8CPUKernel::Init() {
int ResizeInt8CPUKernel::InitResizeQuantArg() {
auto out_shape = out_tensors_.front()->shape();
resize_quant_arg_.x_axis_index_ = reinterpret_cast<int32_t *>(malloc(out_shape.at(2) * sizeof(int32_t)));
resize_quant_arg_.x_axis_index_ = reinterpret_cast<int32_t *>(malloc(out_shape.at(kNHWC_W) * sizeof(int32_t)));
if (resize_quant_arg_.x_axis_index_ == nullptr) {
MS_LOG(ERROR) << "malloc x axis index array failed.";
return RET_ERROR;
}
resize_quant_arg_.x_axis_lower_ = reinterpret_cast<int32_t *>(malloc(out_shape.at(2) * sizeof(int32_t)));
resize_quant_arg_.x_axis_lower_ = reinterpret_cast<int32_t *>(malloc(out_shape.at(kNHWC_W) * sizeof(int32_t)));
if (resize_quant_arg_.x_axis_lower_ == nullptr) {
MS_LOG(ERROR) << "malloc x_axis_lower_ array failed.";
return RET_ERROR;
}
resize_quant_arg_.x_axis_upper_ = reinterpret_cast<int32_t *>(malloc(out_shape.at(2) * sizeof(int32_t)));
resize_quant_arg_.x_axis_upper_ = reinterpret_cast<int32_t *>(malloc(out_shape.at(kNHWC_W) * sizeof(int32_t)));
if (resize_quant_arg_.x_axis_upper_ == nullptr) {
MS_LOG(ERROR) << "malloc x_axis_upper_ array failed.";
return RET_ERROR;
}
resize_quant_arg_.y_axis_index_ = reinterpret_cast<int32_t *>(malloc(out_shape.at(1) * sizeof(int32_t)));
resize_quant_arg_.y_axis_index_ = reinterpret_cast<int32_t *>(malloc(out_shape.at(kNHWC_H) * sizeof(int32_t)));
if (resize_quant_arg_.y_axis_index_ == nullptr) {
MS_LOG(ERROR) << "malloc y_axis_index_ array failed.";
return RET_ERROR;
}
resize_quant_arg_.y_axis_lower_ = reinterpret_cast<int32_t *>(malloc(out_shape.at(1) * sizeof(int32_t)));
resize_quant_arg_.y_axis_lower_ = reinterpret_cast<int32_t *>(malloc(out_shape.at(kNHWC_H) * sizeof(int32_t)));
if (resize_quant_arg_.y_axis_lower_ == nullptr) {
MS_LOG(ERROR) << "malloc y_axis_lower_ array failed.";
return RET_ERROR;
}
resize_quant_arg_.y_axis_upper_ = reinterpret_cast<int32_t *>(malloc(out_shape.at(1) * sizeof(int32_t)));
resize_quant_arg_.y_axis_upper_ = reinterpret_cast<int32_t *>(malloc(out_shape.at(kNHWC_H) * sizeof(int32_t)));
if (resize_quant_arg_.y_axis_upper_ == nullptr) {
MS_LOG(ERROR) << "malloc y_axis_upper_ array failed.";
return RET_ERROR;
@ -136,14 +139,14 @@ int ResizeInt8CPUKernel::CalRatio() {
auto out_tensor = out_tensors_.front();
auto out_width = out_tensor->Width();
auto out_height = out_tensor->Height();
resize_quant_arg_.ratio_x_ = ((1 << 10) * in_width + out_width / 2) / out_width;
resize_quant_arg_.ratio_y_ = ((1 << 10) * in_height + out_height / 2) / out_height;
resize_quant_arg_.ratio_x_ = ((1 << OFFSET_BASE) * in_width + out_width / 2) / out_width;
resize_quant_arg_.ratio_y_ = ((1 << OFFSET_BASE) * in_height + out_height / 2) / out_height;
bool align_corners = coordinate_transform_mode_ == schema::CoordinateTransformMode_ALIGN_CORNERS;
if (align_corners && out_width > 1) {
resize_quant_arg_.ratio_x_ = ((1 << 10) * (in_width - 1) + (out_width - 1) / 2) / (out_width - 1);
resize_quant_arg_.ratio_x_ = ((1 << OFFSET_BASE) * (in_width - 1) + (out_width - 1) / 2) / (out_width - 1);
}
if (align_corners && out_height > 1) {
resize_quant_arg_.ratio_y_ = ((1 << 10) * (in_height - 1) + (out_height - 1) / 2) / (out_height - 1);
resize_quant_arg_.ratio_y_ = ((1 << OFFSET_BASE) * (in_height - 1) + (out_height - 1) / 2) / (out_height - 1);
}
return RET_OK;
}
@ -152,46 +155,48 @@ int ResizeInt8CPUKernel::CalInterpolationRange() {
for (int i = 0; i < out_tensors_.front()->Height(); ++i) {
int32_t scaled_index = i * resize_quant_arg_.ratio_y_;
resize_quant_arg_.y_axis_index_[i] = scaled_index;
resize_quant_arg_.y_axis_lower_[i] = std::max(scaled_index / (1 << 10), 0);
resize_quant_arg_.y_axis_upper_[i] = std::min(scaled_index / (1 << 10) + 1, in_tensors_.front()->Height() - 1);
resize_quant_arg_.y_axis_lower_[i] = std::max(scaled_index / (1 << OFFSET_BASE), 0);
resize_quant_arg_.y_axis_upper_[i] =
std::min(scaled_index / (1 << OFFSET_BASE) + 1, in_tensors_.front()->Height() - 1);
}
for (int i = 0; i < out_tensors_.front()->Width(); ++i) {
int32_t scaled_index = i * resize_quant_arg_.ratio_x_;
resize_quant_arg_.x_axis_index_[i] = scaled_index;
resize_quant_arg_.x_axis_lower_[i] = std::max(scaled_index / (1 << 10), 0);
resize_quant_arg_.x_axis_upper_[i] = std::min(scaled_index / (1 << 10) + 1, in_tensors_.front()->Width() - 1);
resize_quant_arg_.x_axis_lower_[i] = std::max(scaled_index / (1 << OFFSET_BASE), 0);
resize_quant_arg_.x_axis_upper_[i] =
std::min(scaled_index / (1 << OFFSET_BASE) + 1, in_tensors_.front()->Width() - 1);
}
return RET_OK;
}
int ResizeInt8CPUKernel::InitResizeFloatQuantArg() {
auto out_shape = out_tensors_.front()->shape();
resize_float_quant_arg_.x_axis_index_ = reinterpret_cast<float *>(malloc(out_shape[2] * sizeof(float)));
resize_float_quant_arg_.x_axis_index_ = reinterpret_cast<float *>(malloc(out_shape[kNHWC_W] * sizeof(float)));
if (resize_float_quant_arg_.x_axis_index_ == nullptr) {
MS_LOG(ERROR) << "malloc x axis index array failed.";
return RET_ERROR;
}
resize_float_quant_arg_.x_axis_lower_ = reinterpret_cast<int32_t *>(malloc(out_shape[2] * sizeof(int32_t)));
resize_float_quant_arg_.x_axis_lower_ = reinterpret_cast<int32_t *>(malloc(out_shape[kNHWC_W] * sizeof(int32_t)));
if (resize_float_quant_arg_.x_axis_lower_ == nullptr) {
MS_LOG(ERROR) << "malloc x_axis_lower_ array failed.";
return RET_ERROR;
}
resize_float_quant_arg_.x_axis_upper_ = reinterpret_cast<int32_t *>(malloc(out_shape[2] * sizeof(int32_t)));
resize_float_quant_arg_.x_axis_upper_ = reinterpret_cast<int32_t *>(malloc(out_shape[kNHWC_W] * sizeof(int32_t)));
if (resize_float_quant_arg_.x_axis_upper_ == nullptr) {
MS_LOG(ERROR) << "malloc x_axis_upper_ array failed.";
return RET_ERROR;
}
resize_float_quant_arg_.y_axis_index_ = reinterpret_cast<float *>(malloc(out_shape[1] * sizeof(float)));
resize_float_quant_arg_.y_axis_index_ = reinterpret_cast<float *>(malloc(out_shape[kNHWC_H] * sizeof(float)));
if (resize_float_quant_arg_.y_axis_index_ == nullptr) {
MS_LOG(ERROR) << "malloc y_axis_index_ array failed.";
return RET_ERROR;
}
resize_float_quant_arg_.y_axis_lower_ = reinterpret_cast<int32_t *>(malloc(out_shape[1] * sizeof(int32_t)));
resize_float_quant_arg_.y_axis_lower_ = reinterpret_cast<int32_t *>(malloc(out_shape[kNHWC_H] * sizeof(int32_t)));
if (resize_float_quant_arg_.y_axis_lower_ == nullptr) {
MS_LOG(ERROR) << "malloc y_axis_lower_ array failed.";
return RET_ERROR;
}
resize_float_quant_arg_.y_axis_upper_ = reinterpret_cast<int32_t *>(malloc(out_shape[1] * sizeof(int32_t)));
resize_float_quant_arg_.y_axis_upper_ = reinterpret_cast<int32_t *>(malloc(out_shape[kNHWC_H] * sizeof(int32_t)));
if (resize_float_quant_arg_.y_axis_upper_ == nullptr) {
MS_LOG(ERROR) << "malloc y_axis_upper_ array failed.";
return RET_ERROR;

23
mindspore/lite/src/runtime/kernel/arm/int8/scale_int8.cc
View File

@ -1,5 +1,5 @@
/**
* Copyright 2020 Huawei Technologies Co., Ltd
* Copyright 2020-2021 Huawei Technologies Co., Ltd
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
@ -29,6 +29,7 @@ namespace mindspore::kernel {
namespace {
constexpr size_t kScaleInputsSize = 2;
constexpr size_t kScaleBiasInputsSize = 3;
constexpr int kOffsetIndex = 2;
} // namespace
ScaleInt8CPUKernel::~ScaleInt8CPUKernel() {
if (tile_para != nullptr) {
@ -68,16 +69,16 @@ int ScaleInt8CPUKernel::InitScaleOffset() {
}
scale_param_->const_offset_ = false;
if (in_tensors_.size() == 3) {
if (in_tensors_.size() == kScaleBiasInputsSize) {
has_bias_ = true;
auto offset_tensor = in_tensors_.at(2);
auto offset_tensor = in_tensors_.at(kOffsetIndex);
auto *offset_ptr = reinterpret_cast<int8_t *>(offset_tensor->data_c());
// offset may be const value ,can be processed in prepare stage
if (offset_ptr != nullptr) {
scale_param_->const_offset_ = true;
input2_data_ = offset_ptr;
// need broadcasting
if (in_tensors_.at(0)->ElementsNum() != in_tensors_.at(2)->ElementsNum()) {
if (in_tensors_.at(0)->ElementsNum() != in_tensors_.at(kOffsetIndex)->ElementsNum()) {
input2_data_ = reinterpret_cast<int8_t *>(malloc(out_tensors_.at(0)->Size()));
if (input2_data_ == nullptr) {
MS_LOG(ERROR) << "malloc input2_data_ failed.";
@ -88,7 +89,7 @@ int ScaleInt8CPUKernel::InitScaleOffset() {
return RET_ERROR;
}
malloced_offset_ = true;
TileOneDimensionInt8(reinterpret_cast<int8_t *>(in_tensors_.at(2)->data_c()),
TileOneDimensionInt8(reinterpret_cast<int8_t *>(in_tensors_.at(kOffsetIndex)->data_c()),
reinterpret_cast<int8_t *>(input2_data_), 0, tile_para->ndim_, tile_para->in_shape1_,
tile_para->in_strides1_, tile_para->out_strides_, tile_para->multiples1_);
}
@ -187,7 +188,7 @@ int ScaleInt8CPUKernel::InitQuantArgs() {
scale_param_->scale_mul_arg_.right_shift_ = shift < 0 ? -shift : 0;
if (in_tensors_.size() == kScaleBiasInputsSize) {
auto offset = in_tensors_.at(2);
auto offset = in_tensors_.at(kOffsetIndex);
auto offset_scale = offset->quant_params().front().scale;
scale_param_->offset_zp_ = offset->quant_params().front().zeroPoint;
@ -290,7 +291,7 @@ int ScaleInt8CPUKernel::Run() {
count_unit_ = thread_count_ > 1 ? UP_DIV(elements_num_, thread_count_) : elements_num_;
input0_data_ = reinterpret_cast<int8_t *>(in_tensors_.at(0)->data_c());
output_data_ = reinterpret_cast<int8_t *>(out_tensors_.at(0)->data_c());
int ret = RET_ERROR;
// need broadcasting
if (in_tensors_.at(0)->ElementsNum() != in_tensors_.at(1)->ElementsNum()) {
// scale is passed by previous node, need do broadcasting online
@ -314,12 +315,12 @@ int ScaleInt8CPUKernel::Run() {
input1_data_ = nullptr;
return RET_ERROR;
}
TileOneDimensionInt8(reinterpret_cast<int8_t *>(in_tensors_.at(2)->data_c()),
TileOneDimensionInt8(reinterpret_cast<int8_t *>(in_tensors_.at(kOffsetIndex)->data_c()),
reinterpret_cast<int8_t *>(input2_data_), 0, tile_para->ndim_, tile_para->in_shape1_,
tile_para->in_strides1_, tile_para->out_strides_, tile_para->multiples1_);
}
auto ret = static_cast<const lite::InnerContext *>(this->context_)
ret = static_cast<const lite::InnerContext *>(this->context_)
->thread_pool_->ParallelLaunch(ScaleRunInt8, this, op_parameter_->thread_num_);
// free memory malloced from memory pool
if (!scale_param_->const_scale_) {
@ -338,9 +339,9 @@ int ScaleInt8CPUKernel::Run() {
input1_data_ = reinterpret_cast<int8_t *>(in_tensors_.at(1)->data_c());
}
if (has_bias_ && !scale_param_->const_offset_) {
input2_data_ = reinterpret_cast<int8_t *>(in_tensors_.at(2)->data_c());
input2_data_ = reinterpret_cast<int8_t *>(in_tensors_.at(kOffsetIndex)->data_c());
}
auto ret = static_cast<const lite::InnerContext *>(this->context_)
ret = static_cast<const lite::InnerContext *>(this->context_)
->thread_pool_->ParallelLaunch(ScaleRunInt8, this, op_parameter_->thread_num_);
if (ret != RET_OK) {
MS_LOG(ERROR) << "Scale error error_code[" << ret << "]";

1
mindspore/lite/src/runtime/kernel/arm/int8/slice_int8.cc
View File

@ -27,7 +27,6 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_SliceFusion;
namespace mindspore::kernel {
int SliceInt8CPUKernel::Init() {
auto input = in_tensors_.at(0);
auto output = out_tensors_.at(0);

1
mindspore/lite/src/runtime/kernel/arm/int8/split_int8.cc
View File

@ -30,7 +30,6 @@ using mindspore::lite::KernelRegistrar;
using mindspore::schema::PrimitiveType_Split;
namespace mindspore::kernel {
int SplitInt8CPUKernel::Init() {
auto ret = SplitBaseCPUKernel::Init();
if (ret != RET_OK) {

8
mindspore/lite/src/runtime/kernel/arm/int8/transpose_int8.cc
View File

@ -80,15 +80,15 @@ int TransposeInt8CPUKernel::DoTranspose(int task_id) {
void TransposeInt8CPUKernel::GetNHNCTransposeFunc(lite::Tensor *in_tensor, lite::Tensor *out_tensor,
TransposeParameter *param) {
auto out_shape = out_tensor->shape();
if (in_tensor->shape().size() == 4 && param->perm_[0] == 0 && param->perm_[1] == 2 && param->perm_[2] == 3 &&
param->perm_[3] == 1) {
if (in_tensor->shape().size() == DIMENSION_4D && param->perm_[0] == 0 && param->perm_[1] == 2 &&
param->perm_[2] == 3 && param->perm_[3] == 1) {
nhnc_param_[0] = out_shape[0];
nhnc_param_[1] = out_shape[1] * out_shape[2];
nhnc_param_[2] = out_shape[3];
NHNCTransposeFunc_ = PackNCHWToNHWCInt8;
}
if (in_tensor->shape().size() == 4 && param->perm_[0] == 0 && param->perm_[1] == 3 && param->perm_[2] == 1 &&
param->perm_[3] == 2) {
if (in_tensor->shape().size() == DIMENSION_4D && param->perm_[0] == 0 && param->perm_[1] == 3 &&
param->perm_[2] == 1 && param->perm_[3] == 2) {
nhnc_param_[0] = out_shape[0];
nhnc_param_[1] = out_shape[2] * out_shape[3];
nhnc_param_[2] = out_shape[1];

3
mindspore/lite/src/runtime/kernel/arm/string/normalize.cc
View File

@ -37,7 +37,6 @@ const std::map<std::string, std::string> *kRegexTransforms = new (std::nothrow)
{"i'm", "i am"},
});
const int32_t kMaxStringLength = 300;
} // namespace
int NormalizeCPUKernel::Init() {
@ -49,7 +48,7 @@ int NormalizeCPUKernel::Init() {
int NormalizeCPUKernel::ReSize() { return RET_OK; }
std::string NormalizeCPUKernel::Trim(const std::string &str, const std::string &pattern /*= " \t\n\v\f\r"*/) {
std::string NormalizeCPUKernel::Trim(const std::string &str, const std::string &pattern /* = " \t\n\v\f\r" */) {
auto begin = str.find_first_not_of(pattern);
if (begin == std::string::npos) {
MS_LOG(WARNING) << "Meaningless input string!";

14
mindspore/lite/src/runtime/kernel/arm/string/predict.cc
View File

@ -24,6 +24,12 @@ using mindspore::lite::RET_OK;
using mindspore::schema::PrimitiveType_CustomPredict;
namespace mindspore::kernel {
namespace {
constexpr int INPUT_INDEX = 0;
constexpr int KEY_INDEX = 1;
constexpr int LABEL_INDEX = 2;
constexpr int WEIGHT_INDEX = 3;
} // namespace
int PredictCPUKernel::Init() {
if (!InferShapeDone()) {
return RET_OK;
@ -35,10 +41,10 @@ int PredictCPUKernel::ReSize() { return RET_OK; }
std::vector<LabelInfo> PredictCPUKernel::GetLabelInfo() {
std::vector<LabelInfo> label_info_vec;
auto input_tensor = in_tensors_.at(0);
auto keys_tensor = in_tensors_.at(1);
auto labels_tensor = in_tensors_.at(2);
auto weights_tensor = in_tensors_.at(3);
auto input_tensor = in_tensors_.at(INPUT_INDEX);
auto keys_tensor = in_tensors_.at(KEY_INDEX);
auto labels_tensor = in_tensors_.at(LABEL_INDEX);
auto weights_tensor = in_tensors_.at(WEIGHT_INDEX);
int32_t *input = reinterpret_cast<int32_t *>(input_tensor->MutableData());
int32_t *key_begin = reinterpret_cast<int32_t *>(keys_tensor->MutableData());

1
mindspore/lite/src/runtime/kernel/cuda/cuda_subgraph.cc
View File

@ -22,5 +22,4 @@
namespace mindspore::kernel {
using mindspore::lite::RET_ERROR;
using mindspore::lite::RET_OK;
} // namespace mindspore::kernel

20
mindspore/lite/src/runtime/kernel/npu/activation_npu.cc
View File

@ -43,6 +43,26 @@ int ActivationNPUKernel::SetNPUInputs(const std::vector<lite::Tensor *> &inputs,
return RET_ERROR;
}
act_->set_input_x(*npu_inputs[0]);
/*
* mode : Activation mode, with options as follows:
* 0 : Sigmoid
* 1 : ReLU
* 2 : Tanh
* 3 : Clipped ReLU
* 4 : ELU
* 5 : PReLU
* 6 : Abs
* 7 : Relu1
* 8 : Softsign
* 9 : Softplus
* 10 : Hardsigmoid
* 11 : Threshold ReLU
* 12 : Selu
* 13 : Linear
* 14 : Relu6
* 15 : GeLU.
* Defaults to 1 (ReLU). 1.
*/
switch (act_param_->type_) {
case schema::ActivationType_SIGMOID:
act_->set_attr_mode(0);

8
mindspore/lite/src/runtime/kernel/npu/arithmetic_npu.cc
View File

@ -44,6 +44,10 @@ using mindspore::schema::PrimitiveType_NotEqual;
using mindspore::schema::PrimitiveType_SubFusion;
namespace mindspore::kernel {
namespace {
constexpr int RELU_MODE = 1;
constexpr int RELU6_MODE = 14;
} // namespace
int ArithmeticNPUKernel::IsSupport(const std::vector<lite::Tensor *> &inputs,
const std::vector<lite::Tensor *> &outputs, OpParameter *opParameter) {
if (inputs[0]->shape() != inputs[1]->shape()) {
@ -84,9 +88,9 @@ int ArithmeticNPUKernel::SetActivation() {
}
act_->set_input_x(*op_);
if (activation_type_ == ActivationType_RELU) {
act_->set_attr_mode(1);
act_->set_attr_mode(RELU_MODE);
} else if (activation_type_ == ActivationType_RELU6) {
act_->set_attr_mode(14);
act_->set_attr_mode(RELU6_MODE);
} else {
MS_LOG(ERROR) << "Unsupported activation type for op " << name_;
return RET_ERROR;

30
mindspore/lite/src/runtime/kernel/npu/convolution_base_npu.cc
View File

@ -19,6 +19,16 @@
#include "nnacl/pack.h"
namespace mindspore::kernel {
namespace {
constexpr int BATCH_INDEX = 0;
constexpr int HEIGHT_INDEX = 1;
constexpr int WIDTH_INDEX = 2;
constexpr int CHANNEL_INDEX = 3;
constexpr size_t WITH_BIAS_SIZE = 3;
constexpr int BIAS_INDEX = 2;
constexpr int RELU_MODE = 1;
constexpr int RELU6_MODE = 14;
} // namespace
ConvolutionBaseNPUKernel::~ConvolutionBaseNPUKernel() {
if (act_ != nullptr) {
delete act_;
@ -47,14 +57,16 @@ int ConvolutionBaseNPUKernel::InitWeightConst(const std::vector<lite::Tensor *>
MS_LOG(ERROR) << "Malloc buffer failed.";
return RET_ERROR;
}
PackNHWCToNCHWFp32(nhwc_data, nchw_data, w_shape[0], w_shape[1] * w_shape[2], w_shape[3], 0, 0);
PackNHWCToNCHWFp32(nhwc_data, nchw_data, w_shape[BATCH_INDEX], w_shape[HEIGHT_INDEX] * w_shape[WIDTH_INDEX],
w_shape[CHANNEL_INDEX], 0, 0);
std::shared_ptr<ge::Tensor> weight_tensor = std::shared_ptr<ge::Tensor>(new (std::nothrow) ge::Tensor());
std::shared_ptr<ge::Tensor> weight_tensor = std::make_shared<ge::Tensor>();
if (weight_tensor == nullptr) {
MS_LOG(ERROR) << "new weight_tensor failed.";
return RET_ERROR;
}
ge::TensorDesc tensor_desc(lite::ConverterToNPUShape({w_shape[0], w_shape[3], w_shape[1], w_shape[2]}),
ge::TensorDesc tensor_desc(lite::ConverterToNPUShape({w_shape[BATCH_INDEX], w_shape[CHANNEL_INDEX],
w_shape[HEIGHT_INDEX], w_shape[WIDTH_INDEX]}),
ge::FORMAT_NCHW, lite::ConverterToNPUDataType(inputs[1]->data_type()));
weight_tensor->SetTensorDesc(tensor_desc);
weight_tensor->SetData(reinterpret_cast<const uint8_t *>(nchw_data), inputs[1]->Size());
@ -65,16 +77,16 @@ int ConvolutionBaseNPUKernel::InitWeightConst(const std::vector<lite::Tensor *>
}
int ConvolutionBaseNPUKernel::InitBiasConst(const std::vector<lite::Tensor *> &inputs) {
if (inputs.size() >= 3) {
if (inputs.size() >= WITH_BIAS_SIZE) {
bias_ = new (std::nothrow) hiai::op::Const(name_ + "_b");
if (bias_ == nullptr) {
MS_LOG(ERROR) << "New bias const failed.";
return RET_ERROR;
}
inputs[2]->set_format(schema::Format_NCHW);
auto bias_tensor = mindspore::lite::ConverterToNPUTensor(inputs[2]);
inputs[BIAS_INDEX]->set_format(schema::Format_NCHW);
auto bias_tensor = mindspore::lite::ConverterToNPUTensor(inputs[BIAS_INDEX]);
bias_->set_attr_value(bias_tensor);
inputs[2]->set_format(schema::Format_NHWC);
inputs[BIAS_INDEX]->set_format(schema::Format_NHWC);
}
return RET_OK;
}
@ -87,9 +99,9 @@ int ConvolutionBaseNPUKernel::SetActivation(const ge::Operator *input, ActType a
}
act_->set_input_x(*input);
if (act_type == ActType_Relu) {
act_->set_attr_mode(1);
act_->set_attr_mode(RELU_MODE);
} else if (act_type == ActType_Relu6) {
act_->set_attr_mode(14);
act_->set_attr_mode(RELU6_MODE);
} else {
MS_LOG(ERROR) << "Unsupported activation type for convolution.";
return RET_ERROR;

8
mindspore/lite/src/runtime/kernel/npu/gather_npu.cc
View File

@ -22,14 +22,18 @@ using mindspore::lite::KernelRegistrar;
using mindspore::schema::PrimitiveType_Gather;
namespace mindspore::kernel {
namespace {
constexpr size_t WITH_AXES_SIZE = 3;
constexpr int AXIS_INDEX = 2;
} // namespace
int GatherNPUKernel::IsSupport(const std::vector<lite::Tensor *> &inputs, const std::vector<lite::Tensor *> &outputs,
OpParameter *opParameter) {
if (inputs[1]->data_type() != kNumberTypeInt32) {
MS_LOG(WARNING) << "Gather indices only support Int32";
return RET_ERROR;
}
if (inputs.size() >= 3 && inputs[2]->ElementsNum() == 1) {
axis_ = static_cast<int *>(inputs[2]->data_c())[0];
if (inputs.size() >= WITH_AXES_SIZE && inputs[AXIS_INDEX]->ElementsNum() == 1) {
axis_ = static_cast<int *>(inputs[AXIS_INDEX]->data_c())[0];
} else {
MS_LOG(WARNING) << "NPU axis is attribute.";
return RET_ERROR;

20
mindspore/lite/src/runtime/kernel/npu/instance_norm_npu.cc
View File

@ -24,6 +24,10 @@ using mindspore::lite::KernelRegistrar;
using mindspore::schema::PrimitiveType_InstanceNorm;
namespace mindspore::kernel {
namespace {
constexpr int GAMMA_INDEX = 1;
constexpr int BETA_INDEX = 2;
} // namespace
int InstanceNormNPUKernel::IsSupport(const std::vector<lite::Tensor *> &inputs,
const std::vector<lite::Tensor *> &outputs, OpParameter *opParameter) {
return RET_OK;
@ -39,16 +43,16 @@ int InstanceNormNPUKernel::SetNPUInputs(const std::vector<lite::Tensor *> &input
}
op_->set_input_x(*npu_inputs[0]);
auto gamma_shape = inputs[1]->shape();
std::shared_ptr<ge::Tensor> gamma_tensor = std::shared_ptr<ge::Tensor>(new (std::nothrow) ge::Tensor());
auto gamma_shape = inputs[GAMMA_INDEX]->shape();
std::shared_ptr<ge::Tensor> gamma_tensor = std::make_shared<ge::Tensor>();
if (gamma_tensor == nullptr) {
MS_LOG(ERROR) << "new gamma_tensor failed.";
return RET_ERROR;
}
ge::TensorDesc gamma_tensor_desc(lite::ConverterToNPUShape({1, gamma_shape[0], 1, 1}), ge::FORMAT_NCHW,
lite::ConverterToNPUDataType(inputs[1]->data_type()));
lite::ConverterToNPUDataType(inputs[GAMMA_INDEX]->data_type()));
gamma_tensor->SetTensorDesc(gamma_tensor_desc);
gamma_tensor->SetData(reinterpret_cast<const uint8_t *>(inputs[1]->data_c()), inputs[1]->Size());
gamma_tensor->SetData(reinterpret_cast<const uint8_t *>(inputs[GAMMA_INDEX]->data_c()), inputs[GAMMA_INDEX]->Size());
gamma_ = new (std::nothrow) hiai::op::Const(name_ + "_gamma");
if (gamma_ == nullptr) {
MS_LOG(ERROR) << "New gamma_ const failed.";
@ -57,16 +61,16 @@ int InstanceNormNPUKernel::SetNPUInputs(const std::vector<lite::Tensor *> &input
gamma_->set_attr_value(gamma_tensor);
op_->set_input_gamma(*gamma_);
auto beta_shape = inputs[2]->shape();
std::shared_ptr<ge::Tensor> beta_tensor = std::shared_ptr<ge::Tensor>(new (std::nothrow) ge::Tensor());
auto beta_shape = inputs[BETA_INDEX]->shape();
std::shared_ptr<ge::Tensor> beta_tensor = std::make_shared<ge::Tensor>();
if (beta_tensor == nullptr) {
MS_LOG(ERROR) << "new beta_tensor failed.";
return RET_ERROR;
}
ge::TensorDesc beta_tensor_desc(lite::ConverterToNPUShape({1, beta_shape[0], 1, 1}), ge::FORMAT_NCHW,
lite::ConverterToNPUDataType(inputs[2]->data_type()));
lite::ConverterToNPUDataType(inputs[BETA_INDEX]->data_type()));
beta_tensor->SetTensorDesc(beta_tensor_desc);
beta_tensor->SetData(reinterpret_cast<const uint8_t *>(inputs[2]->data_c()), inputs[2]->Size());
beta_tensor->SetData(reinterpret_cast<const uint8_t *>(inputs[BETA_INDEX]->data_c()), inputs[BETA_INDEX]->Size());
beta_ = new (std::nothrow) hiai::op::Const(name_ + "_beta");
if (beta_ == nullptr) {
MS_LOG(ERROR) << "New beta_ const failed.";

3
mindspore/lite/src/runtime/kernel/npu/pad_npu.cc
View File

@ -47,10 +47,11 @@ int PadNPUKernel::SetNPUInputs(const std::vector<lite::Tensor *> &inputs, const
MS_LOG(ERROR) << name_ << " op is nullptr";
return RET_ERROR;
}
// padding shape is spatial_dim x 2.
int size = static_cast<int>(param_->padding_length / 2);
ge::TensorDesc padding_tensor_desc(ge::Shape({size, 2}), ge::FORMAT_NCHW, ge::DT_INT32);
ge::TensorPtr padding_tensor = std::make_shared<hiai::Tensor>(padding_tensor_desc);
padding_tensor->SetData(reinterpret_cast<uint8_t *>(param_->paddings_), 2 * size * sizeof(int));
padding_tensor->SetData(reinterpret_cast<uint8_t *>(param_->paddings_), param_->padding_length * sizeof(int));
hiai_paddings_ = new hiai::op::Const(name_ + "paddings");
hiai_paddings_->set_attr_value(padding_tensor);

17
mindspore/lite/src/runtime/kernel/npu/pooling_npu.cc
View File

@ -23,6 +23,13 @@ using mindspore::schema::PrimitiveType_AvgPoolFusion;
using mindspore::schema::PrimitiveType_MaxPoolFusion;
namespace mindspore::kernel {
namespace {
constexpr int MAX_MODE = 0;
constexpr int AVG_MODE = 1;
constexpr int L2_MODE = 2;
constexpr int PAD_MODE_SAME = 6;
constexpr int PAD_MODE_VALID = 5;
} // namespace
int PoolingNPUKernel::IsSupport(const std::vector<lite::Tensor *> &inputs, const std::vector<lite::Tensor *> &outputs,
OpParameter *opParameter) {
if (pooling_param_->pad_l_ > pooling_param_->stride_w_ || pooling_param_->pad_u_ > pooling_param_->stride_h_) {
@ -34,20 +41,20 @@ int PoolingNPUKernel::IsSupport(const std::vector<lite::Tensor *> &inputs, const
int PoolingNPUKernel::SetPoolingParam() {
if (pooling_param_->pool_mode_ == PoolMode_MaxPool) {
pooling_->set_attr_mode(0);
pooling_->set_attr_mode(MAX_MODE);
} else if (pooling_param_->pool_mode_ == PoolMode_AvgPool) {
pooling_->set_attr_mode(1);
pooling_->set_attr_mode(AVG_MODE);
} else {
pooling_->set_attr_mode(2);
pooling_->set_attr_mode(L2_MODE);
}
pooling_->set_attr_global_pooling(pooling_param_->global_);
pooling_->set_attr_window({pooling_param_->window_h_, pooling_param_->window_w_});
pooling_->set_attr_stride({pooling_param_->stride_h_, pooling_param_->stride_w_});
if (pooling_param_->pad_mode_ == Pad_same) {
pooling_->set_attr_pad_mode(6);
pooling_->set_attr_pad_mode(PAD_MODE_SAME);
pooling_->set_attr_pad({0, 0, 0, 0});
} else if (pooling_param_->pad_mode_ == Pad_valid) {
pooling_->set_attr_pad_mode(5);
pooling_->set_attr_pad_mode(PAD_MODE_VALID);
pooling_->set_attr_pad({0, 0, 0, 0});
} else {
pooling_->set_attr_pad_mode(0);

35
mindspore/lite/src/runtime/kernel/npu/resize_npu.cc
View File

@ -25,6 +25,9 @@ using mindspore::lite::KernelRegistrar;
using mindspore::schema::PrimitiveType_Resize;
namespace mindspore::kernel {
namespace {
constexpr size_t SIZE_TENSOR_DIMS = 2;
} // namespace
int ResizeNPUKernel::IsSupport(const std::vector<lite::Tensor *> &inputs, const std::vector<lite::Tensor *> &outputs,
OpParameter *opParameter) {
if (resize_parameter_->method_ != schema::ResizeMethod_LINEAR &&
@ -45,32 +48,32 @@ int ResizeNPUKernel::SetNPUInputs(const std::vector<lite::Tensor *> &inputs, con
ge::TensorPtr sizeTensor = std::make_shared<hiai::Tensor>(sizeTensorDesc);
vector<int32_t> dataValue = {static_cast<int32_t>(resize_parameter_->new_height_),
static_cast<int32_t>(resize_parameter_->new_width_)};
sizeTensor->SetData(reinterpret_cast<uint8_t *>(dataValue.data()), 2 * sizeof(int32_t));
sizeTensor->SetData(reinterpret_cast<uint8_t *>(dataValue.data()), SIZE_TENSOR_DIMS * sizeof(int32_t));
out_size_ = new (std::nothrow) hiai::op::Const(name_ + "_size");
out_size_->set_attr_value(sizeTensor);
if (resize_parameter_->method_ == schema::ResizeMethod_LINEAR) {
auto op = new (std::nothrow) hiai::op::ResizeBilinearV2(name_);
if (op == nullptr) {
MS_LOG(ERROR) << " op is nullptr.";
auto linear_op = new (std::nothrow) hiai::op::ResizeBilinearV2(name_);
if (linear_op == nullptr) {
MS_LOG(ERROR) << " linear_op is nullptr.";
return RET_ERROR;
}
op->set_attr_align_corners(resize_parameter_->coordinate_transform_mode_ ==
linear_op->set_attr_align_corners(resize_parameter_->coordinate_transform_mode_ ==
schema::CoordinateTransformMode_ALIGN_CORNERS);
op->set_input_x(*npu_inputs[0]);
op->set_input_size(*out_size_);
op->set_attr_half_pixel_centers(resize_parameter_->preserve_aspect_ratio_);
op_ = op;
linear_op->set_input_x(*npu_inputs[0]);
linear_op->set_input_size(*out_size_);
linear_op->set_attr_half_pixel_centers(resize_parameter_->preserve_aspect_ratio_);
op_ = linear_op;
} else if (resize_parameter_->method_ == schema::ResizeMethod_NEAREST) {
auto op = new (std::nothrow) hiai::op::ResizeNearestNeighborV2(name_);
if (op == nullptr) {
MS_LOG(ERROR) << " op is nullptr.";
auto nearest_op = new (std::nothrow) hiai::op::ResizeNearestNeighborV2(name_);
if (nearest_op == nullptr) {
MS_LOG(ERROR) << " nearest_op is nullptr.";
return RET_ERROR;
}
op->set_attr_align_corners(resize_parameter_->coordinate_transform_mode_ ==
nearest_op->set_attr_align_corners(resize_parameter_->coordinate_transform_mode_ ==
schema::CoordinateTransformMode_ALIGN_CORNERS);
op->set_input_x(*npu_inputs[0]);
op->set_input_size(*out_size_);
op_ = op;
nearest_op->set_input_x(*npu_inputs[0]);
nearest_op->set_input_size(*out_size_);
op_ = nearest_op;
} else {
MS_LOG(WARNING) << "Unsupported resize method type:" << resize_parameter_->method_;
return RET_ERROR;

28
mindspore/lite/src/runtime/kernel/npu/scale_npu.cc
View File

@ -25,12 +25,20 @@ using mindspore::schema::Format_NHWC;
using mindspore::schema::PrimitiveType_ScaleFusion;
namespace mindspore::kernel {
namespace {
constexpr int DIMS_4D = 4;
constexpr int BIAS_INDEX = 2;
constexpr size_t NONE_BIAS_SIZE = 2;
constexpr int RELU_MODE = 1;
constexpr int RELU6_MODE = 14;
} // namespace
int ScaleNPUKernel::IsSupport(const std::vector<lite::Tensor *> &inputs, const std::vector<lite::Tensor *> &outputs,
OpParameter *opParameter) {
if (scale_parameter_->axis_ < 0) {
scale_parameter_->axis_ = scale_parameter_->axis_ + inputs[0]->shape().size();
}
if (inputs.size() > 1 && inputs[0]->shape().size() == 4 && inputs[0]->format() == schema::Format_NHWC) {
if (inputs.size() > 1 && inputs[0]->shape().size() == DIMS_4D && inputs[0]->format() == schema::Format_NHWC) {
// scale now only supports on axis 3
if (scale_parameter_->axis_ != 3) {
MS_LOG(ERROR) << "Npu scale axis attr only support on channel, now is " << scale_parameter_->axis_;
return RET_ERROR;
@ -56,7 +64,7 @@ int ScaleNPUKernel::SetNPUInputs(const std::vector<lite::Tensor *> &inputs, cons
MS_ASSERT(inputs.size() > 1);
auto scale_shape = inputs.at(1)->shape();
std::shared_ptr<ge::Tensor> scale_tensor = std::shared_ptr<ge::Tensor>(new (std::nothrow) ge::Tensor());
std::shared_ptr<ge::Tensor> scale_tensor = std::make_shared<ge::Tensor>();
if (scale_tensor == nullptr) {
MS_LOG(ERROR) << "new scale_tensor failed.";
return RET_ERROR;
@ -73,17 +81,19 @@ int ScaleNPUKernel::SetNPUInputs(const std::vector<lite::Tensor *> &inputs, cons
scale_->set_attr_value(scale_tensor);
op_->set_input_scale(*scale_);
if (inputs.size() > 2 && inputs[2] != nullptr) {
auto bias_shape = inputs[2]->shape();
std::shared_ptr<ge::Tensor> bias_tensor = std::shared_ptr<ge::Tensor>(new (std::nothrow) ge::Tensor());
// inputs size can be larger than 2 when optional bias is provided.
// bias index 2
if (inputs.size() > NONE_BIAS_SIZE && inputs[BIAS_INDEX] != nullptr) {
auto bias_shape = inputs[BIAS_INDEX]->shape();
std::shared_ptr<ge::Tensor> bias_tensor = std::make_shared<ge::Tensor>();
if (bias_tensor == nullptr) {
MS_LOG(ERROR) << "new bias_tensor failed.";
return RET_ERROR;
}
ge::TensorDesc bias_tensor_desc(lite::ConverterToNPUShape({1, bias_shape[0], 1, 1}), ge::FORMAT_NCHW,
lite::ConverterToNPUDataType(inputs[2]->data_type()));
lite::ConverterToNPUDataType(inputs[BIAS_INDEX]->data_type()));
bias_tensor->SetTensorDesc(bias_tensor_desc);
bias_tensor->SetData(reinterpret_cast<const uint8_t *>(inputs[2]->data_c()), inputs[2]->Size());
bias_tensor->SetData(reinterpret_cast<const uint8_t *>(inputs[BIAS_INDEX]->data_c()), inputs[BIAS_INDEX]->Size());
bias_ = new (std::nothrow) hiai::op::Const(name_ + "_beta");
if (bias_ == nullptr) {
MS_LOG(ERROR) << "New beta_ const failed.";
@ -120,9 +130,9 @@ int ScaleNPUKernel::SetActivation(const ge::Operator *input, int act_type) {
}
act_->set_input_x(*input);
if (act_type == schema::ActivationType_RELU) {
act_->set_attr_mode(1);
act_->set_attr_mode(RELU_MODE);
} else if (act_type == schema::ActivationType_RELU6) {
act_->set_attr_mode(14);
act_->set_attr_mode(RELU6_MODE);
} else {
MS_LOG(ERROR) << "Unsupported activation type for scale.";
return RET_ERROR;

11
mindspore/lite/src/runtime/kernel/npu/slice_npu.cc
View File

@ -22,6 +22,11 @@ using mindspore::lite::KernelRegistrar;
using mindspore::schema::PrimitiveType_SliceFusion;
namespace mindspore::kernel {
namespace {
constexpr int INPUT_INDEX = 0;
constexpr int OFFSET_INDEX = 1;
constexpr int SIZE_INDEX = 2;
} // namespace
int SliceNPUKernel::IsSupport(const std::vector<lite::Tensor *> &inputs, const std::vector<lite::Tensor *> &outputs,
OpParameter *opParameter) {
return RET_OK;
@ -34,9 +39,9 @@ int SliceNPUKernel::SetNPUInputs(const std::vector<lite::Tensor *> &inputs, cons
MS_LOG(ERROR) << name_ << " op is nullptr";
return RET_ERROR;
}
op_->set_input_x(*npu_inputs[0]);
op_->set_input_offsets(*npu_inputs[1]);
op_->set_input_size(*npu_inputs[2]);
op_->set_input_x(*npu_inputs[INPUT_INDEX]);
op_->set_input_offsets(*npu_inputs[OFFSET_INDEX]);
op_->set_input_size(*npu_inputs[SIZE_INDEX]);
return RET_OK;
}

20
mindspore/lite/src/runtime/kernel/npu/strided_slice_npu.cc
View File

@ -22,14 +22,20 @@ using mindspore::lite::KernelRegistrar;
using mindspore::schema::PrimitiveType_StridedSlice;
namespace mindspore::kernel {
namespace {
constexpr int INPUT_INDEX = 0;
constexpr int BEGIN_INDEX = 1;
constexpr int END_INDEX = 2;
constexpr int AXES_INDEX = 3;
} // namespace
int StridedSliceNPUKernel::IsSupport(const std::vector<lite::Tensor *> &inputs,
const std::vector<lite::Tensor *> &outputs, OpParameter *opParameter) {
// Only onnx StridedSlice has 5 inputs, of which the 4th input is axes and the 5th input is strides.
if (inputs.size() == 5) {
vector<int> axes;
size_t size = inputs[3]->shape()[0];
size_t size = inputs[AXES_INDEX]->shape()[0];
axes.resize(size);
memcpy(axes.data(), inputs[3]->data_c(), sizeof(int) * size);
memcpy(axes.data(), inputs[AXES_INDEX]->data_c(), sizeof(int) * size);
for (int i = 0; i < axes.size(); ++i) {
if (i != axes[i]) {
MS_LOG(ERROR) << "Does not support setting axis, so the axis must be continuous.";
@ -49,14 +55,16 @@ int StridedSliceNPUKernel::SetNPUInputs(const std::vector<lite::Tensor *> &input
MS_LOG(ERROR) << name_ << " op is nullptr";
return RET_ERROR;
}
op_->set_input_x(*npu_inputs[0]);
op_->set_input_begin(*npu_inputs[1]);
op_->set_input_end(*npu_inputs[2]);
op_->set_input_x(*npu_inputs[INPUT_INDEX]);
op_->set_input_begin(*npu_inputs[BEGIN_INDEX]);
op_->set_input_end(*npu_inputs[END_INDEX]);
// The strides position of onnx is the 5th, and the others are the 4th.
// For onnx models, input size 5.
if (npu_inputs.size() == 5) {
// The strides position of onnx models is the 5th, index 4.
op_->set_input_strides(*npu_inputs[4]);
} else {
// The strides position of other models are the 4th, index 3.
op_->set_input_strides(*npu_inputs[3]);
}
op_->set_attr_begin_mask(param_->begins_mask_);

6
mindspore/lite/src/runtime/kernel/npu/transpose_npu.cc
View File

@ -22,13 +22,17 @@ using mindspore::lite::KernelRegistrar;
using mindspore::schema::PrimitiveType_Transpose;
namespace mindspore::kernel {
namespace {
constexpr size_t TRANSPOSE_INPUT_SIZE = 2;
} // namespace
int TransposeNPUKernel::IsSupport(const std::vector<lite::Tensor *> &inputs, const std::vector<lite::Tensor *> &outputs,
OpParameter *opParameter) {
if (conjugate_) {
MS_LOG(ERROR) << "Unsupported conjugate transpose.";
return RET_ERROR;
}
if (inputs.size() >= 2 && inputs[1]->data_c() != nullptr) {
if (inputs.size() == TRANSPOSE_INPUT_SIZE && inputs[1]->data_c() != nullptr) {
for (int i = 0; i < inputs[1]->ElementsNum(); i++) {
perm_.push_back(static_cast<int *>(inputs[1]->data_c())[i]);
}