Skip to content

P
Paddle-Lite

PaddlePaddle / Paddle-Lite

通知 331
Star 4
Fork 1
P
Paddle-Lite
未验证 提交 d14e57f7 编写于 作者: L Leonardo-Ding 提交者： GitHub
浏览文件
操作

[ARM] optimize depthwise int8 f3s1 arm neon kernel,test=develop (#4125)

上级 42cefe1b
展开全部 显示空白变更内容
内联 并排
Showing with 6412 addition and 357 deletion
lite/backends/arm/math/conv_depthwise.h
浏览文件 @ d14e57f7
......@@ -106,6 +106,42 @@ void conv_depthwise_3x3s1_int8(Dtype* dout,
int padh,
ARMContext* ctx);
void conv_depthwise_3x3s1_int8_int8_impl(int8_t* dout,
const int8_t* din,
const int8_t* weights,
const float* scale,
const float* bias,
bool flag_bias,
int flag_act,
float* alpha,
int num,
int chin,
int hin,
int win,
int hout,
int wout,
int padw,
int padh,
ARMContext* ctx);
void conv_depthwise_3x3s1_int8_float_impl(float* dout,
const int8_t* din,
const int8_t* weights,
const float* scale,
const float* bias,
bool flag_bias,
int flag_act,
float* alpha,
int num,
int chin,
int hin,
int win,
int hout,
int wout,
int padw,
int padh,
ARMContext* ctx);
template <typename Dtype>
void conv_depthwise_3x3s2_int8(Dtype* dout,
const int8_t* din,
......
lite/backends/arm/math/conv_impl.cc
浏览文件 @ d14e57f7
......@@ -814,7 +814,15 @@ void conv_depthwise_3x3_int8_fp32(const void* din,
alpha[3] = local_alpha;
}
}
bool support_act_type = flag_act <= 1;
bool support_pad_type =
(paddings[0] == paddings[1]) && (paddings[2] == paddings[3]) &&
(paddings[0] == paddings[2]) && (paddings[0] == 0 || paddings[0] == 1);
bool support_stride_type = (param.strides[0] == 1 && param.strides[1] == 1);
bool support_width_type = w_in > 9 ? true : false;
if (stride == 1) {
if (!support_act_type || !support_pad_type || !support_stride_type ||
!support_width_type) {
conv_depthwise_3x3s1_int8(reinterpret_cast<float*>(dout),
reinterpret_cast<const int8_t*>(din),
reinterpret_cast<const int8_t*>(weights),
......@@ -832,6 +840,26 @@ void conv_depthwise_3x3_int8_fp32(const void* din,
pad_w,
pad_h,
ctx);
} else {
conv_depthwise_3x3s1_int8_float_impl(
reinterpret_cast<float*>(dout),
reinterpret_cast<const int8_t*>(din),
reinterpret_cast<const int8_t*>(weights),
scale,
bias,
flag_bias,
flag_act,
alpha,
num,
ch_in,
h_in,
w_in,
h_out,
w_out,
pad_w,
pad_h,
ctx);
}
} else if (stride == 2) {
conv_depthwise_3x3s2_int8(reinterpret_cast<float*>(dout),
reinterpret_cast<const int8_t*>(din),
......@@ -897,7 +925,15 @@ void conv_depthwise_3x3_int8_int8(const void* din,
alpha[3] = local_alpha;
}
}
bool support_act_type = flag_act <= 1;
bool support_pad_type =
(paddings[0] == paddings[1]) && (paddings[2] == paddings[3]) &&
(paddings[0] == paddings[2]) && (paddings[0] == 0 || paddings[0] == 1);
bool support_stride_type = (param.strides[0] == 1 && param.strides[1] == 1);
bool support_width_type = w_in > 9 ? true : false;
if (stride == 1) {
if (!support_act_type || !support_pad_type || !support_stride_type ||
!support_width_type) {
conv_depthwise_3x3s1_int8(reinterpret_cast<int8_t*>(dout),
reinterpret_cast<const int8_t*>(din),
reinterpret_cast<const int8_t*>(weights),
......@@ -915,6 +951,26 @@ void conv_depthwise_3x3_int8_int8(const void* din,
pad_w,
pad_h,
ctx);
} else {
conv_depthwise_3x3s1_int8_int8_impl(
reinterpret_cast<int8_t*>(dout),
reinterpret_cast<const int8_t*>(din),
reinterpret_cast<const int8_t*>(weights),
scale,
bias,
flag_bias,
flag_act,
alpha,
num,
ch_in,
h_in,
w_in,
h_out,
w_out,
pad_w,
pad_h,
ctx);
}
} else if (stride == 2) {
conv_depthwise_3x3s2_int8(reinterpret_cast<int8_t*>(dout),
reinterpret_cast<const int8_t*>(din),
......
lite/kernels/arm/conv_depthwise.cc
浏览文件 @ d14e57f7
......@@ -31,7 +31,6 @@ void DepthwiseConv<PRECISION(kFloat), PRECISION(kFloat)>::PrepareForRun() {
auto paddings = *param.paddings;
// select dw conv kernel
if (kw == 3) {
// VLOG(5) << "invoke 3x3 dw conv fp32";
bool pads_less = ((paddings[1] < 2) && (paddings[3] < 2));
if (pads_less && paddings[0] == paddings[2] &&
(paddings[0] == 0 || paddings[0] == 1)) {
......@@ -54,7 +53,6 @@ void DepthwiseConv<PRECISION(kFloat), PRECISION(kFloat)>::PrepareForRun() {
kernel_func_name_ = "conv_depthwise_3x3_fp32";
#endif
} else if (kw == 5) {
// VLOG(5) << "invoke 5x5 dw conv fp32";
auto strides = param.strides;
if ((strides[0] == 1 && strides[1] == 1) ||
(strides[0] == 2 && strides[1] == 2)) {
......@@ -104,23 +102,44 @@ void DepthwiseConv<PRECISION(kInt8), PRECISION(kFloat)>::PrepareForRun() {
w_scale_[i] = scale[i] * in_scale;
}
}
auto paddings = *param.paddings;
auto strides = param.strides;
auto x_dims = param.x->dims();
int iw = x_dims[3];
int ih = x_dims[2];
auto act_param = param.activation_param;
bool has_act = act_param.has_active;
lite_api::ActivationType act_type = act_param.active_type;
// no activation and relu activation is supported now
bool support_act_type =
(has_act == false) ||
(has_act == true && act_type == lite_api::ActivationType::kRelu);
bool support_pad_type =
(paddings[0] == paddings[1]) && (paddings[2] == paddings[3]) &&
(paddings[0] == paddings[2]) && (paddings[0] == 0 || paddings[0] == 1);
bool support_stride_type = (strides[0] == 1 && strides[1] == 1);
bool support_width_type = iw > 9 ? true : false;
/// select dw conv kernel
if (kw == 3) {
// trans weights
// VLOG(5) << "invoke 3x3 dw conv int8 kernel fp32 out";
impl_ = lite::arm::math::conv_depthwise_3x3_int8_fp32;
#ifdef LITE_WITH_PROFILE
kernel_func_name_ = "conv_depthwise_3x3_int8_fp32";
#endif
if (!support_act_type || !support_pad_type || !support_stride_type ||
!support_width_type) {
int cround = ROUNDUP(w_dims[0], 8);
weights_.Resize({cround / 8, 1, kh * kw, 8});
auto wptr = param.filter->data<int8_t>();
auto wptr_new = weights_.mutable_data<int8_t>();
lite::arm::math::conv_trans_weights_numc(wptr, wptr_new, oc, 1, 8, 9);
flag_trans_weights_ = true;
} else {
flag_trans_weights_ = false;
}
} else if (kw == 5) {
// trans weights
// VLOG(5) << "invoke 5x5 dw conv int8 kernel fp32 out";
impl_ = lite::arm::math::conv_depthwise_5x5_int8_fp32;
#ifdef LITE_WITH_PROFILE
kernel_func_name_ = "conv_depthwise_5x5_int8_fp32";
......@@ -175,23 +194,45 @@ void DepthwiseConv<PRECISION(kInt8), PRECISION(kInt8)>::PrepareForRun() {
param.activation_param.Relu_clipped_coef =
param.activation_param.Relu_clipped_coef / param.output_scale;
}
auto paddings = *param.paddings;
auto strides = param.strides;
auto x_dims = param.x->dims();
int iw = x_dims[3];
int ih = x_dims[2];
auto act_param = param.activation_param;
bool has_act = act_param.has_active;
lite_api::ActivationType act_type = act_param.active_type;
// no activation and relu activation is supported now
bool support_act_type =
(has_act == false) ||
(has_act == true && act_type == lite_api::ActivationType::kRelu);
bool support_pad_type =
(paddings[0] == paddings[1]) && (paddings[2] == paddings[3]) &&
(paddings[0] == paddings[2]) && (paddings[0] == 0 || paddings[0] == 1);
bool support_stride_type = (strides[0] == 1 && strides[1] == 1);
bool support_width_type = iw > 9 ? true : false;
/// select dw conv kernel
if (kw == 3) {
// trans weights
// VLOG(5) << "invoke 3x3 dw conv int8 kernel int8 out";
impl_ = lite::arm::math::conv_depthwise_3x3_int8_int8;
#ifdef LITE_WITH_PROFILE
kernel_func_name_ = "conv_depthwise_3x3_int8_int8";
#endif
if (!support_act_type || !support_pad_type || !support_stride_type ||
!support_width_type) {
int cround = ROUNDUP(w_dims[0], 8);
weights_.Resize({cround / 8, 1, kh * kw, 8});
auto wptr = param.filter->data<int8_t>();
auto wptr_new = weights_.mutable_data<int8_t>();
lite::arm::math::conv_trans_weights_numc(wptr, wptr_new, oc, 1, 8, 9);
flag_trans_weights_ = true;
} else {
flag_trans_weights_ = false;
}
} else if (kw == 5) {
// trans weights
// VLOG(5) << "invoke 5x5 dw conv int8 kernel int8 out";
impl_ = lite::arm::math::conv_depthwise_5x5_int8_int8;
#ifdef LITE_WITH_PROFILE
kernel_func_name_ = "conv_depthwise_5x5_int8_int8";
......@@ -283,7 +324,7 @@ void DepthwiseConv<PRECISION(kInt8), PRECISION(kFloat)>::Run() {
auto w_dims = param.filter->dims();
auto o_dims = param.output->dims();
int iw = x_dims[3]; // nchw
int iw = x_dims[3];
int ih = x_dims[2];
int ic = x_dims[1];
int bs = x_dims[0];
......@@ -333,7 +374,7 @@ void DepthwiseConv<PRECISION(kInt8), PRECISION(kInt8)>::Run() {
auto w_dims = param.filter->dims();
auto o_dims = param.output->dims();
int iw = x_dims[3]; // nchw
int iw = x_dims[3];
int ih = x_dims[2];
int ic = x_dims[1];
int bs = x_dims[0];
......
lite/tests/math/conv_int8_compute_test.cc
浏览文件 @ d14e57f7
......@@ -125,7 +125,7 @@ void release_param(ConvParam* param) {
#ifdef LITE_WITH_ARM
#include "lite/backends/arm/math/funcs.h"
void test_conv_int8(const std::vector<DDim>& input_dims,
void test_conv_int8(const DDim& dim_in,
const DDim& weight_dim,
int group,
const std::vector<int>& strides,
......@@ -237,24 +237,21 @@ void test_conv_int8(const std::vector<DDim>& input_dims,
conv_int8_fp32.SetContext(std::move(ctx2));
/// set param and context
for (auto& dim_in : input_dims) {
param_int8_out.x->Resize(dim_in);
DDim out_tmp_dims = compute_out_dim(dim_in, param_int8_out);
if (out_tmp_dims[2] < 1 || out_tmp_dims[3] < 1) {
continue;
return;
}
param_fp32_out.x->Resize(dim_in);
param_int8_out.output->Resize(out_tmp_dims);
param_fp32_out.output->Resize(out_tmp_dims);
break;
}
conv_int8_int8.SetParam(param_int8_out);
conv_int8_fp32.SetParam(param_fp32_out);
/// prepare for run
conv_int8_int8.PrepareForRun();
conv_int8_fp32.PrepareForRun();
for (auto& dim_in : input_dims) {
CHECK_EQ(weight_dim[1] * group, dim_in[1])
<< "input channel must equal to weights channel";
DDim dim_out = compute_out_dim(dim_in, param_int8_out);
......@@ -333,7 +330,7 @@ void test_conv_int8(const std::vector<DDim>& input_dims,
weight_dim[3] / group;
/// warm up
for (int i = 0; i < FLAGS_warmup; ++i) {
conv_int8_int8.Launch();
conv_int8_fp32.Launch();
}
/// compute fp32 output
Timer t0;
......@@ -343,13 +340,13 @@ void test_conv_int8(const std::vector<DDim>& input_dims,
t0.Stop();
}
LOG(INFO) << "int8 conv, fp32 output: output shape" << dim_out
<< ",running time, avg: " << t0.LapTimes().Avg()
<< ", min time: " << t0.LapTimes().Min()
<< ",running time, avg: " << t0.LapTimes().Avg() << " ms"
<< ", min time: " << t0.LapTimes().Min() << " ms"
<< ", total GOPS: " << 1e-9 * gops
<< " GOPS, avg GOPs: " << 1e-6 * gops / t0.LapTimes().Avg()
<< " GOPs, max GOPs: " << 1e-6 * gops / t0.LapTimes().Min();
/// compute int8 output
// compute int8 output
t0.Reset();
for (int i = 0; i < FLAGS_repeats; ++i) {
t0.Start();
......@@ -386,9 +383,8 @@ void test_conv_int8(const std::vector<DDim>& input_dims,
release_param(&param_fp32_out);
LOG(FATAL) << "test int8 conv, fp32 out: input: " << dim_in
<< ", output: " << dim_out
<< ", weight dim: " << weight_dim
<< ", pad: " << pads[0] << ", " << pads[1] << ", "
<< pads[2] << ", " << pads[3]
<< ", weight dim: " << weight_dim << ", pad: " << pads[0]
<< ", " << pads[1] << ", " << pads[2] << ", " << pads[3]
<< ", stride: " << strides[0] << ", " << strides[1]
<< ", dila_: " << dilas[0] << ", " << dilas[1]
<< ", group: " << group
......@@ -398,7 +394,7 @@ void test_conv_int8(const std::vector<DDim>& input_dims,
}
}
}
/// compare result int8 output
// compare result int8 output
if (FLAGS_check_result) {
double max_ratio = 0;
double max_diff = 0;
......@@ -431,8 +427,7 @@ void test_conv_int8(const std::vector<DDim>& input_dims,
count += 1;
}
}
check =
check &&
check = check &&
count < std::max(10, static_cast<int>(0.01 * tdiff.numel()));
if (!check) {
LOG(WARNING) << "int8 basic result";
......@@ -445,9 +440,8 @@ void test_conv_int8(const std::vector<DDim>& input_dims,
release_param(&param_fp32_out);
LOG(FATAL) << "test int8 conv, int8 out: input: " << dim_in
<< ", output: " << dim_out
<< ", weight dim: " << weight_dim
<< ", pad: " << pads[0] << ", " << pads[1] << ", "
<< pads[2] << ", " << pads[3]
<< ", weight dim: " << weight_dim << ", pad: " << pads[0]
<< ", " << pads[1] << ", " << pads[2] << ", " << pads[3]
<< ", stride: " << strides[0] << ", " << strides[1]
<< ", dila_: " << dilas[0] << ", " << dilas[1]
<< ", bias: " << (flag_bias ? "true" : "false")
......@@ -466,12 +460,11 @@ void test_conv_int8(const std::vector<DDim>& input_dims,
<< ", power_mode: " << cls << " successed!!\n";
}
}
}
release_param(&param_int8_out);
release_param(&param_fp32_out);
}
#else
void test_conv_int8(const std::vector<DDim>& input_dims,
void test_conv_int8(const DDim& dims_in,
const DDim& weight_dim,
int group,
const std::vector<int>& strides,
......@@ -493,13 +486,10 @@ TEST(TestConv3x3DWInt8, test_conv3x3_depthwise) {
for (auto& flag_bias : {false, true}) {
for (auto& flag_act : {0, 1, 2, 4}) {
for (auto& c : {1, 3, 5, 8, 16, 32}) {
std::vector<DDim> dims;
DDim weights_dim({c, 1, 3, 3});
for (auto& batch : {1, 2}) {
for (auto& h : {1, 3, 15, 33}) {
dims.push_back(DDim({batch, c, h, h}));
}
}
DDim dims({batch, c, h, h});
test_conv_int8(dims,
weights_dim,
c,
......@@ -508,7 +498,7 @@ TEST(TestConv3x3DWInt8, test_conv3x3_depthwise) {
{1, 1},
flag_bias,
flag_act,
{4},
{FLAGS_threads},
{FLAGS_power_mode},
FLAGS_clipped_coef,
FLAGS_leakey_relu_alpha);
......@@ -518,6 +508,8 @@ TEST(TestConv3x3DWInt8, test_conv3x3_depthwise) {
}
}
}
}
}
}
#endif /// 3x3dw
......@@ -529,13 +521,10 @@ TEST(TestConv5x5DWInt8, test_conv5x5_depthwise) {
for (auto& flag_bias : {false, true}) {
for (auto& flag_act : {0, 1, 2, 4}) {
for (auto& c : {1, 5, 15, 33}) {
std::vector<DDim> dims;
DDim weights_dim({c, 1, 5, 5});
for (auto& batch : {1, 2}) {
for (auto& h : {1, 3, 15, 33, 112, 224}) {
dims.push_back(DDim({batch, c, h, h}));
}
}
DDim dims({batch, c, h, h});
test_conv_int8(dims,
weights_dim,
c,
......@@ -554,6 +543,8 @@ TEST(TestConv5x5DWInt8, test_conv5x5_depthwise) {
}
}
}
}
}
}
#endif /// 5x5dw
......@@ -565,16 +556,13 @@ TEST(TestConv1x1s1Int8, test_conv1x1s1) {
for (auto& g : {1, 2}) {
for (auto& flag_bias : {false, true}) {
for (auto& flag_act : {0, 1, 2, 4}) {
std::vector<DDim> dims;
if (cin % g != 0 || cout % g != 0) {
continue;
}
DDim weights_dim({cout, cin / g, 1, 1});
for (auto& batch : {1, 2}) {
for (auto& h : {1, 9, 16, 33}) {
dims.push_back(DDim({batch, cin, h, h}));
}
}
DDim dims({batch, cin, h, h});
test_conv_int8(dims,
weights_dim,
g,
......@@ -593,6 +581,8 @@ TEST(TestConv1x1s1Int8, test_conv1x1s1) {
}
}
}
}
}
}
#endif /// conv1x1s1
......@@ -606,18 +596,16 @@ TEST(TestConv3x3s1Int8, test_conv_3x3s1) {
for (auto& pad_left : {1, 2}) {
for (auto& pad_right : {1, 2}) {
for (auto& flag_bias : {false, true}) {
for (auto& flag_act : {0, 1, 2, 4}) {
std::vector<DDim> dims;
for (auto& flag_act : {0, 1}) {
DDim weights_dim({cout, cin, 3, 3});
for (auto& batch : {1, 2}) {
for (auto& h : {1, 7, 17, 33}) {
dims.push_back(DDim({batch, cin, h, h}));
}
}
DDim dims({batch, cin, h, h});
if (cin == 1 && cout == 1) {
continue;
}
test_conv_int8(dims,
test_conv_int8(
dims,
weights_dim,
1,
{1, 1},
......@@ -638,6 +626,8 @@ TEST(TestConv3x3s1Int8, test_conv_3x3s1) {
}
}
}
}
}
}
#endif /// conv3x3s1
......@@ -652,14 +642,12 @@ TEST(TestConv3x3s2Int8, test_conv_3x3s2) {
for (auto& pad_right : {1, 2}) {
for (auto& flag_bias : {false, true}) {
for (auto& flag_act : {0, 1, 2, 4}) {
std::vector<DDim> dims;
DDim weights_dim({cout, cin, 3, 3});
for (auto& batch : {1, 2}) {
for (auto& h : {1, 7, 19, 33}) {
dims.push_back(DDim({batch, cin, h, h}));
}
}
test_conv_int8(dims,
DDim dims({batch, cin, h, h});
test_conv_int8(
dims,
weights_dim,
1,
{2, 2},
......@@ -680,6 +668,8 @@ TEST(TestConv3x3s2Int8, test_conv_3x3s2) {
}
}
}
}
}
}
#endif /// conv3x3s2
......@@ -702,19 +692,18 @@ TEST(TestConvRandInt8, test_conv_rand) {
if (cin % g != 0 || cout % g != 0) {
break;
}
std::vector<DDim> dims;
DDim weights_dim({cout, cin / g, kh, kw});
for (auto& batch : {1, 2}) {
for (auto& h : {1, 3, 5, 19}) {
dims.push_back(DDim({batch, cin, h, h}));
}
}
test_conv_int8(
dims,
DDim dims({batch, cin, h, h});
test_conv_int8(dims,
weights_dim,
g,
{stride, stride},
{pad_top, pad_bottom, pad_left, pad_right},
{pad_top,
pad_bottom,
pad_left,
pad_right},
{dila, dila},
flag_bias,
flag_act,
......@@ -736,6 +725,8 @@ TEST(TestConvRandInt8, test_conv_rand) {
}
}
}
}
}
}
#endif /// random param conv
......
Markdown is supported
0% 或 .
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 或