m53297601
/
mindspore
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

fix bug of int8 conv 3x3

Browse Source
pull/5225/head
fuzhiye 5 years ago
parent 13e9f5b86e
commit 05be89372d
8 changed files with 52 additions and 47 deletions
Show Stats Download Patch File Download Diff File

5
mindspore/lite/nnacl/fp16/conv_fp16.c
Unescape View File

@ -369,9 +369,10 @@ void ConvFp16(float16_t *input_data, float16_t *packed_input, float16_t *packed_
out_channel * sizeof(float16_t), 0, 0, relu, relu6);
} else {
// res part
IndirectGemmFp16_16x8(tmp_out_block, gemm_input, packed_weight, bias_data, conv_depth, ic4, out_channel,
float16_t *tmp_out_ptr = tmp_out_block + task_id * tile_n * out_channel;
IndirectGemmFp16_16x8(tmp_out_ptr, gemm_input, packed_weight, bias_data, conv_depth, ic4, out_channel,
out_channel * sizeof(float16_t), 0, 0, relu, relu6);
memcpy(output_data + out_offset, tmp_out_block, real_cal_num * out_channel * sizeof(float16_t));
memcpy(output_data + out_offset, tmp_out_ptr, real_cal_num * out_channel * sizeof(float16_t));
}
}
}

7
mindspore/lite/nnacl/fp32/conv.c
Unescape View File

@ -248,9 +248,10 @@ void ConvFp32(float *input_data, float *packed_input, float *packed_weight, cons
relu, relu6);
} else {
// res part
gemm_func(tmp_out_block, gemm_input, packed_weight, bias_data, conv_depth, ic4, out_channel, output_offset, 0,
0, relu, relu6);
memcpy(output_data + out_offset, tmp_out_block, real_cal_num * out_channel * sizeof(float));
float *tmp_out_ptr = tmp_out_block + task_id * TILE_NUM * out_channel;
gemm_func(tmp_out_ptr, gemm_input, packed_weight, bias_data, conv_depth, ic4, out_channel, output_offset, 0, 0,
relu, relu6);
memcpy(output_data + out_offset, tmp_out_ptr, real_cal_num * out_channel * sizeof(float));
}
}
}

28
mindspore/lite/nnacl/int8/conv_int8.c
Unescape View File

@ -199,7 +199,7 @@ void IndirectGemmInt8Opt(int8_t *dst, int32_t *tmp_dst, const int8_t *src, const
}
}
void Conv3x3Uint8Gemm(int32_t *dst, const int16_t *src, const int16_t *weight, int oc, int ic8, size_t real_cal_num) {
void Conv3x3Int8Gemm(int32_t *dst, const int16_t *src, const int16_t *weight, int oc, int ic8, size_t real_cal_num) {
int oc4 = UP_DIV(oc, C4NUM);
#ifdef ENABLE_ARM64
IndirectGemmInt16to32_8x4(dst, src, weight, 16, ic8, oc4, oc4 * 4 * 16 * sizeof(int32_t));
@ -298,9 +298,10 @@ void ConvInt8(int8_t *input_data, int8_t *packed_input, int8_t *packed_weight, c
out_channel, tmp_input_sum, conv_param);
} else {
// res part
IndirectGemmInt8(tmp_out, tmp_dst + tmp_dst_offset, gemm_input, packed_weight, bias_data, ic4, kernel_plane,
int8_t *tmp_out_ptr = tmp_out + task_id * tile_n * out_channel;
IndirectGemmInt8(tmp_out_ptr, tmp_dst + tmp_dst_offset, gemm_input, packed_weight, bias_data, ic4, kernel_plane,
out_channel, tmp_input_sum, conv_param);
memcpy(output_data + out_offset, tmp_out, real_cal_num * out_channel);
memcpy(output_data + out_offset, tmp_out_ptr, real_cal_num * out_channel);
}
}
}
@ -360,9 +361,10 @@ void ConvInt8Opt(int8_t *input_data, int8_t *packed_input, int8_t *packed_weight
kernel_plane, out_channel, tmp_input_sum, conv_param, gemm_func);
} else {
// res part
IndirectGemmInt8Opt(tmp_out, tmp_dst + tmp_dst_offset, gemm_input, packed_weight, bias_data, ic4, kernel_plane,
out_channel, tmp_input_sum, conv_param, gemm_func);
memcpy(output_data + out_offset, tmp_out, real_cal_num * out_channel);
int8_t *tmp_out_ptr = tmp_out + task_id * tile_n * out_channel;
IndirectGemmInt8Opt(tmp_out_ptr, tmp_dst + tmp_dst_offset, gemm_input, packed_weight, bias_data, ic4,
kernel_plane, out_channel, tmp_input_sum, conv_param, gemm_func);
memcpy(output_data + out_offset, tmp_out_ptr, real_cal_num * out_channel);
}
}
}
@ -412,15 +414,15 @@ void Conv3x3Int8(int16_t *input_data, int16_t *transed_weight, const int32_t *bi
int start_index = thread_id * TILE_NUM;
int real_cal_num = (output_count - start_index) < TILE_NUM ? (output_count - start_index) : TILE_NUM;
Conv3x3Uint8InputTransform(input_data + in_batch_offset, tile_buffer + task_id * tile_buffer_offset,
block_unit_buffer + task_id * block_unit_buffer_offset, start_index, real_cal_num,
out_w_block, conv_param);
Conv3x3Int8InputTransform(input_data + in_batch_offset, tile_buffer + task_id * tile_buffer_offset,
block_unit_buffer + task_id * block_unit_buffer_offset, start_index, real_cal_num,
out_w_block, conv_param);
Conv3x3Uint8Gemm(tmp_dst_buffer + task_id * tmp_dst_buffer_offset, tile_buffer + task_id * tile_buffer_offset,
transed_weight, output_channel, ic8, real_cal_num);
Conv3x3Int8Gemm(tmp_dst_buffer + task_id * tmp_dst_buffer_offset, tile_buffer + task_id * tile_buffer_offset,
transed_weight, output_channel, ic8, real_cal_num);
Conv3x3Uint8OutputTransform(tmp_dst_buffer + task_id * tmp_dst_buffer_offset, tmp_out + tmp_out_batch_offset,
bias_data, start_index, real_cal_num, out_w_block, conv_param);
Conv3x3Int8OutputTransform(tmp_dst_buffer + task_id * tmp_dst_buffer_offset, tmp_out + tmp_out_batch_offset,
bias_data, start_index, real_cal_num, out_w_block, conv_param);
}
}
}

44
mindspore/lite/nnacl/winograd_transform.c
Unescape View File

@ -695,7 +695,7 @@ void Conv3x3Fp32OutputTransform(const float *gemm_out, float *out_data, const fl
}
// int8 conv3x3
void Conv3x3Uint8InputUnit(int16_t *tmp_data, int16_t *trans_input_data, size_t step, int input_zp) {
void Conv3x3Int8InputUnit(int16_t *tmp_data, int16_t *trans_input_data, size_t step, int input_zp) {
#ifdef ENABLE_ARM
int16x8_t zp = vdupq_n_s16(input_zp);
@ -864,7 +864,7 @@ void Conv3x3Uint8InputUnit(int16_t *tmp_data, int16_t *trans_input_data, size_t
#endif
}
void Conv3x3Uint8InputTransform(const int16_t *input_data, int16_t *trans_input, int16_t *tmp_data, int start_index,
void Conv3x3Int8InputTransform(const int16_t *input_data, int16_t *trans_input, int16_t *tmp_data, int start_index,
int real_cal_num, int out_w_block, ConvParameter *conv_param) {
// input data format : nhwc
int input_channel = conv_param->input_channel_;
@ -904,7 +904,7 @@ void Conv3x3Uint8InputTransform(const int16_t *input_data, int16_t *trans_input,
int dst_ic8_offset = dst_plane_offset + ic * TILE_NUM * C8NUM;
size_t dst_step = ic8 * C8NUM * TILE_NUM;
int16_t *trans_input_ptr = trans_input + dst_ic8_offset;
Conv3x3Uint8InputUnit(tmp_data, trans_input_ptr, dst_step, input_zp);
Conv3x3Int8InputUnit(tmp_data, trans_input_ptr, dst_step, input_zp);
}
}
}
@ -1175,7 +1175,7 @@ void Conv3x3Int8FilterTransform(const int16_t *weight_data, int16_t *trans_weigh
}
}
void Conv3x3Uint8OutputUnit(const int32_t *gemm_out, const int32_t *bias_data, int8_t *output_data, bool h_not_bound,
void Conv3x3Int8OutputUnit(const int32_t *gemm_out, const int32_t *bias_data, int8_t *output_data, bool h_not_bound,
bool w_not_bound, int output_w, int real_num, int oc_start, ConvParameter *conv_param) {
int32_t *left_shift = conv_param->conv_quant_arg_.left_shift_;
int32_t *right_shift = conv_param->conv_quant_arg_.right_shift_;
@ -1267,27 +1267,27 @@ void Conv3x3Uint8OutputUnit(const int32_t *gemm_out, const int32_t *bias_data, i
d11 = vmaxq_s32(d11, output_min);
d11 = vminq_s32(d11, output_max);
(output_data)[0] = (uint8_t)d00[0];
(output_data + 1)[0] = (uint8_t)d00[1];
(output_data + 2)[0] = (uint8_t)d00[2];
(output_data + 3)[0] = (uint8_t)d00[3];
(output_data)[0] = (int8_t)d00[0];
(output_data + 1)[0] = (int8_t)d00[1];
(output_data + 2)[0] = (int8_t)d00[2];
(output_data + 3)[0] = (int8_t)d00[3];
if (w_not_bound) {
*(output_data + 4) = (uint8_t)d01[0];
*(output_data + 5) = (uint8_t)d01[1];
*(output_data + 6) = (uint8_t)d01[2];
*(output_data + 7) = (uint8_t)d01[3];
*(output_data + 4) = (int8_t)d01[0];
*(output_data + 5) = (int8_t)d01[1];
*(output_data + 6) = (int8_t)d01[2];
*(output_data + 7) = (int8_t)d01[3];
}
if (h_not_bound) {
*(output_data + output_w * 4) = (uint8_t)d10[0];
*(output_data + output_w * 4 + 1) = (uint8_t)d10[1];
*(output_data + output_w * 4 + 2) = (uint8_t)d10[2];
*(output_data + output_w * 4 + 3) = (uint8_t)d10[3];
*(output_data + output_w * 4) = (int8_t)d10[0];
*(output_data + output_w * 4 + 1) = (int8_t)d10[1];
*(output_data + output_w * 4 + 2) = (int8_t)d10[2];
*(output_data + output_w * 4 + 3) = (int8_t)d10[3];
if (w_not_bound) {
*(output_data + output_w * 4 + 4) = (uint8_t)d11[0];
*(output_data + output_w * 4 + 5) = (uint8_t)d11[1];
*(output_data + output_w * 4 + 6) = (uint8_t)d11[2];
*(output_data + output_w * 4 + 7) = (uint8_t)d11[3];
*(output_data + output_w * 4 + 4) = (int8_t)d11[0];
*(output_data + output_w * 4 + 5) = (int8_t)d11[1];
*(output_data + output_w * 4 + 6) = (int8_t)d11[2];
*(output_data + output_w * 4 + 7) = (int8_t)d11[3];
}
}
#else
@ -1456,7 +1456,7 @@ void Conv3x3Uint8OutputUnit(const int32_t *gemm_out, const int32_t *bias_data, i
#endif
}
void Conv3x3Uint8OutputTransform(const int32_t *gemm_out, int8_t *out_data, const int32_t *bias_data, int start_index,
void Conv3x3Int8OutputTransform(const int32_t *gemm_out, int8_t *out_data, const int32_t *bias_data, int start_index,
int real_cal_num, int out_w_block, ConvParameter *conv_param) {
int output_channel = conv_param->output_channel_;
int output_w = conv_param->output_w_;
@ -1483,7 +1483,7 @@ void Conv3x3Uint8OutputTransform(const int32_t *gemm_out, int8_t *out_data, cons
int real_num = (output_channel - j * C4NUM) < C4NUM ? (output_channel - j * C4NUM) : C4NUM;
bool w_not_bound = out_w_index * OUPUT_UNIT + 1 < output_w;
bool h_not_bound = out_h_index * OUPUT_UNIT + 1 < output_h;
Conv3x3Uint8OutputUnit(src_ptr, bias_ptr, dst_ptr, h_not_bound, w_not_bound, output_w, real_num, j * C4NUM,
Conv3x3Int8OutputUnit(src_ptr, bias_ptr, dst_ptr, h_not_bound, w_not_bound, output_w, real_num, j * C4NUM,
conv_param);
}
}

8
mindspore/lite/nnacl/winograd_transform.h
Unescape View File

@ -56,18 +56,18 @@ void Conv3x3Fp32OutputTransform(const float *gemm_out, float *out_data, const fl
int real_cal_num, int out_w_block, ConvParameter *conv_param);
// for int8 convolution 3x3 filter/input/output transform
void Conv3x3Uint8InputUnit(int16_t *tmp_data, int16_t *trans_input_data, size_t step, int input_zp);
void Conv3x3Int8InputUnit(int16_t *tmp_data, int16_t *trans_input_data, size_t step, int input_zp);
void Conv3x3Uint8InputTransform(const int16_t *input_data, int16_t *trans_input, int16_t *tmp_data, int start_index,
void Conv3x3Int8InputTransform(const int16_t *input_data, int16_t *trans_input, int16_t *tmp_data, int start_index,
int real_cal_num, int out_w_block, ConvParameter *conv_param);
void Conv3x3Int8FilterTransform(const int16_t *weight_data, int16_t *trans_weight, int iC8, int output_channel,
int kernel_plane);
void Conv3x3Uint8OutputUnit(const int32_t *gemm_out, const int32_t *bias_data, int8_t *output_data, bool h_not_bound,
void Conv3x3Int8OutputUnit(const int32_t *gemm_out, const int32_t *bias_data, int8_t *output_data, bool h_not_bound,
bool w_not_bound, int output_w, int real_num, int oc_start, ConvParameter *conv_param);
void Conv3x3Uint8OutputTransform(const int32_t *gemm_out, int8_t *out_data, const int32_t *bias_data, int start_index,
void Conv3x3Int8OutputTransform(const int32_t *gemm_out, int8_t *out_data, const int32_t *bias_data, int start_index,
int real_cal_num, int out_w_block, ConvParameter *conv_param);
#ifdef __cplusplus
}

2
mindspore/lite/src/runtime/kernel/arm/fp16/convolution_fp16.cc
Unescape View File

@ -112,7 +112,7 @@ int ConvolutionFP16CPUKernel::InitTmpBuffer() {
}
memset(nhwc4_input_, 0, nhwc4_input_size);
tmp_output_block_ = reinterpret_cast<float16_t *>(malloc(cal_num * out_channel * sizeof(float16_t)));
tmp_output_block_ = reinterpret_cast<float16_t *>(malloc(thread_count_ * cal_num * out_channel * sizeof(float16_t)));
if (tmp_output_block_ == nullptr) {
MS_LOG(ERROR) << "malloc tmp_output_block_ failed.";
return RET_ERROR;

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

@ -83,7 +83,8 @@ int ConvolutionCPUKernel::InitTmpBuffer() {
int out_channel = conv_param_->output_channel_;
MS_ASSERT(ctx_->allocator != nullptr);
tmp_output_block_ = reinterpret_cast<float *>(ctx_->allocator->Malloc(TILE_NUM * out_channel * sizeof(float)));
tmp_output_block_ =
reinterpret_cast<float *>(ctx_->allocator->Malloc(thread_count_ * TILE_NUM * out_channel * sizeof(float)));
if (tmp_output_block_ == nullptr) {
MS_LOG(ERROR) << "malloc tmp output block failed.";
return RET_ERROR;

2
mindspore/lite/src/runtime/kernel/arm/int8/convolution_int8.cc
Unescape View File

@ -338,7 +338,7 @@ int ConvolutionInt8CPUKernel::RunImpl(int task_id) {
return RET_OK;
}
int ConvolutionInt8Impl(int task_id, LiteParallelGroupEnv *mpenv, void *cdata) {
int ConvolutionInt8Impl(int task_id, LiteParallelGroupEnv *penv, void *cdata) {
auto conv = reinterpret_cast<ConvolutionInt8CPUKernel *>(cdata);
auto error_code = conv->RunImpl(task_id);
if (error_code != RET_OK) {

Write Preview
Loading…
Cancel
Save