blob: 4f9440c780be52d75df6783da8318a4f6d57b6ef [file] [log] [blame]
/*
* Copyright 2024 Google LLC
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
// Depthwise convolution based on Kelvin ops
// Data types: input: s8, filter: s8, bias s32
#include "tensorflow/lite/kernels/internal/reference/integer_ops/depthwise_conv.h"
#include "tflm/opt/conv_util.h"
namespace kelvin::opt {
namespace {
// Reorders a vector to match the pattern after double-widening.
// N must be a multiple of 4.
void VectorSwizzle(const int32_t* input, int32_t* output, int N) {
assert(N >= 4 && N % 4 == 0);
const int32_t(&in)[N] = *(int32_t(*)[N])input;
int32_t(&out)[N] = *(int32_t(*)[N]) output;
const int32_t* p_in = in;
for (int i = 0; i < N / 4; ++i) {
int32_t* out0 = out + i + 0;
int32_t* out1 = out + i + 16;
int32_t* out2 = out + i + 8;
int32_t* out3 = out + i + 24;
*out0 = *p_in++;
*out1 = *p_in++;
*out2 = *p_in++;
*out3 = *p_in++;
}
}
// special case of input depth = 32n, filter shape of 3x3, strides of 1
void DepthwiseConvS83x3D32_Stride1(
const tflite::DepthwiseParams& params, const int32_t* output_multiplier,
const int32_t* output_shift, const tflite::RuntimeShape& input_shape,
const int8_t* input_data, const tflite::RuntimeShape& filter_shape,
const int8_t* filter_data, const tflite::RuntimeShape& bias_shape,
const int32_t* bias_data, const tflite::RuntimeShape& output_shape,
int8_t* output_data
) {
const int stride_width = params.stride_width;
const int stride_height = params.stride_height;
const int pad_width = params.padding_values.width;
const int pad_height = params.padding_values.height;
const int32_t input_offset = params.input_offset;
const int32_t output_offset = params.output_offset;
const int32_t output_activation_min = params.quantized_activation_min;
const int32_t output_activation_max = params.quantized_activation_max;
const int batches = MatchingDim(input_shape, 0, output_shape, 0);
const int input_height = input_shape.Dims(1);
const int input_width = input_shape.Dims(2);
const int input_depth = input_shape.Dims(3);
const int output_height = output_shape.Dims(1);
const int output_width = output_shape.Dims(2);
const int output_depth = output_shape.Dims(3);
int32_t swizzled_bias_data[32];
int32_t swizzled_shift_multi[32];
int32_t swizzled_output_multi[32];
for (int in_channel = 0; in_channel + 32 <= input_depth; in_channel += 32) {
const int output_channel = in_channel;
int8_t* p_output = output_data + output_channel;
VectorSwizzle(bias_data + output_channel, swizzled_bias_data, 32);
VectorSwizzle(output_multiplier + output_channel, swizzled_output_multi, 32);
VectorSwizzle(output_shift + output_channel, swizzled_shift_multi, 32);
vld_w_x_m(v52, swizzled_bias_data);
vld_w_x_m(v56, swizzled_output_multi);
vld_w_x_m(v60, swizzled_shift_multi);
vrsub_w_vx_m(v60, v60, 0);
union {
vdwconv_u8_t dwconv;
uint32_t raw;
} cmds;
cmds.raw = 0;
cmds.dwconv.sdata1 = true;
cmds.dwconv.sbias1 = input_offset;
cmds.dwconv.sdata2 = true;
cmds.dwconv.sbias2 = 0;
cmds.dwconv.mode = 0;
cmds.dwconv.sparsity = 0;
cmds.dwconv.regbase = 0;
#define FLT_0_0 v0
#define FLT_0_1 v3
#define FLT_0_2 v6
#define FLT_1_0 v1
#define FLT_1_1 v4
#define FLT_1_2 v7
#define FLT_2_0 v2
#define FLT_2_1 v5
#define FLT_2_2 v8
#define INPUT_0_0 v9
#define INPUT_0_1 v12
#define INPUT_0_2 v15
#define INPUT_0_3 v18
#define INPUT_0_4 v21
#define INPUT_0_5 v24
#define INPUT_1_0 v10
#define INPUT_1_1 v13
#define INPUT_1_2 v16
#define INPUT_1_3 v19
#define INPUT_1_4 v22
#define INPUT_1_5 v25
#define INPUT_2_0 v11
#define INPUT_2_1 v14
#define INPUT_2_2 v17
#define INPUT_2_3 v20
#define INPUT_2_4 v23
#define INPUT_2_5 v26
#define INPUT_PTRS(_strides) \
const int in_y_origin = (out_y * stride_height) - pad_height; \
const int in_x_origin = (out_x * stride_width) - pad_width; \
const int8_t* p_in_0 = input_data + \
(batch * input_height * input_width * input_depth) + \
(in_y_origin * input_width * input_depth) + \
((in_x_origin + _strides) * input_depth) + \
in_channel; \
const int8_t* p_in_1 = p_in_0 + (input_width * input_depth); \
const int8_t* p_in_2 = p_in_1 + (input_width * input_depth); \
(void)p_in_2;
#define COMPUTE() \
adwinit_v(v48, v48); \
adwconv_vxv(v48, INPUT_0_0, cmds, FLT_0_0); \
adwconv_vxv(v48, INPUT_0_1, cmds, FLT_0_1); \
vdwconv_vxv(v48, INPUT_0_2, cmds, FLT_0_2);
// Don't reorder me, otherwise data will not be
// loaded in the correct order
// (we can reuse the p_flt* due to the `p` vld variant).
const int8_t* p_flt0 = filter_data + in_channel;
const int8_t* p_flt1 = p_flt0 + input_depth;
const int32_t stride = 2 * input_depth;
vld_b_sp_xx(FLT_0_0, p_flt0, stride);
vld_b_sp_xx(FLT_0_1, p_flt1, stride);
vld_b_sp_xx(FLT_0_2, p_flt0, stride);
vld_b_sp_xx(FLT_1_0, p_flt1, stride);
vld_b_sp_xx(FLT_1_1, p_flt0, stride);
vld_b_sp_xx(FLT_1_2, p_flt1, stride);
vld_b_sp_xx(FLT_2_0, p_flt0, stride);
vld_b_sp_xx(FLT_2_1, p_flt1, stride);
vld_b_sp_xx(FLT_2_2, p_flt0, stride);
for (int batch = 0; batch < batches; ++batch) {
int out_y = 0;
for (; out_y < pad_height; ++out_y) {
int out_x = 0;
vdup_b_x(INPUT_0_0, -input_offset);
vdup_b_x(INPUT_0_1, -input_offset);
vdup_b_x(INPUT_0_2, -input_offset);
for (; out_x < pad_width; ++out_x) {
INPUT_PTRS(1);
vmv_v_m(v48, v52);
vdup_b_x(INPUT_1_0, -input_offset);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vdup_b_x(INPUT_2_0, -input_offset);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
COMPUTE();
INT32_TO_INT8_OUTPUT_PIPELINE_INPLACE(
v48, v56, v60,
output_activation_min,
output_activation_max,
output_offset);
vsraqs_b_vx(v48, v48, 0);
vst_b_x(v48, p_output);
p_output += output_depth;
}
for (; out_x < output_width - pad_width; ++out_x) {
INPUT_PTRS(0);
vmv_v_m(v48, v52);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
COMPUTE();
INT32_TO_INT8_OUTPUT_PIPELINE_INPLACE(
v48, v56, v60,
output_activation_min,
output_activation_max,
output_offset);
vsraqs_b_vx(v48, v48, 0);
vst_b_x(v48, p_output);
p_output += output_depth;
}
for (; out_x < output_width; ++out_x) {
INPUT_PTRS(0);
vmv_v_m(v48, v52);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vdup_b_x(INPUT_1_2, -input_offset);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vdup_b_x(INPUT_2_2, -input_offset);
COMPUTE();
INT32_TO_INT8_OUTPUT_PIPELINE_INPLACE(
v48, v56, v60,
output_activation_min,
output_activation_max,
output_offset);
vsraqs_b_vx(v48, v48, 0);
vst_b_x(v48, p_output);
p_output += output_depth;
}
}
for (; out_y < output_height - pad_height; ++out_y) {
int out_x = 0;
for (; out_x < pad_width; ++out_x) {
INPUT_PTRS(1);
vmv_v_m(v48, v52);
vdup_b_x(INPUT_0_0, -input_offset);
vdup_b_x(INPUT_1_0, -input_offset);
vdup_b_x(INPUT_2_0, -input_offset);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
COMPUTE();
INT32_TO_INT8_OUTPUT_PIPELINE_INPLACE(
v48, v56, v60,
output_activation_min,
output_activation_max,
output_offset);
vsraqs_b_vx(v48, v48, 0);
vst_b_x(v48, p_output);
p_output += output_depth;
}
for (; out_x + 4 <= output_width - pad_width; out_x += 4) {
INPUT_PTRS(0);
// Initialize accumulators w/ bias data.
vmv_v_m(v36, v52);
vmv_v_m(v40, v52);
vmv_v_m(v44, v52);
vmv_v_m(v48, v52);
vld_b_sp_xx(INPUT_0_0, p_in_0, stride_width * input_depth);
vld_b_sp_xx(INPUT_1_0, p_in_1, stride_width * input_depth);
vld_b_sp_xx(INPUT_2_0, p_in_2, stride_width * input_depth);
vld_b_sp_xx(INPUT_0_1, p_in_0, stride_width * input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, stride_width * input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, stride_width * input_depth);
vld_b_sp_xx(INPUT_0_2, p_in_0, stride_width * input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, stride_width * input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, stride_width * input_depth);
adwinit_v(v48, v48);
adwconv_vxv(v48, INPUT_0_0, cmds, FLT_0_0);
adwconv_vxv(v48, INPUT_0_1, cmds, FLT_0_1);
vdwconv_vxv(v48, INPUT_0_2, cmds, FLT_0_2);
vld_b_sp_xx(INPUT_0_3, p_in_0, stride_width * input_depth);
vld_b_sp_xx(INPUT_1_3, p_in_1, stride_width * input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, stride_width * input_depth);
adwinit_v(v44, v44);
adwconv_vxv(v44, INPUT_0_1, cmds, FLT_0_0);
adwconv_vxv(v44, INPUT_0_2, cmds, FLT_0_1);
vdwconv_vxv(v44, INPUT_0_3, cmds, FLT_0_2);
vld_b_sp_xx(INPUT_0_4, p_in_0, stride_width * input_depth);
vld_b_sp_xx(INPUT_1_4, p_in_1, stride_width * input_depth);
vld_b_sp_xx(INPUT_2_4, p_in_2, stride_width * input_depth);
adwinit_v(v40, v40);
adwconv_vxv(v40, INPUT_0_2, cmds, FLT_0_0);
adwconv_vxv(v40, INPUT_0_3, cmds, FLT_0_1);
vdwconv_vxv(v40, INPUT_0_4, cmds, FLT_0_2);
vld_b_sp_xx(INPUT_0_5, p_in_0, stride_width * input_depth);
vld_b_sp_xx(INPUT_1_5, p_in_1, stride_width * input_depth);
vld_b_sp_xx(INPUT_2_5, p_in_2, stride_width * input_depth);
adwinit_v(v36, v36);
adwconv_vxv(v36, INPUT_0_3, cmds, FLT_0_0);
adwconv_vxv(v36, INPUT_0_4, cmds, FLT_0_1);
vdwconv_vxv(v36, INPUT_0_5, cmds, FLT_0_2);
INT32_TO_INT8_OUTPUT_PIPELINE_INPLACE4(
v48, v44, v40, v36, v56, v60,
output_activation_min,
output_activation_max,
output_offset);
vsraqs_b_vx(v48, v48, 0);
vsraqs_b_vx(v44, v44, 0);
vsraqs_b_vx(v40, v40, 0);
vsraqs_b_vx(v36, v36, 0);
vst_b_x(v48, p_output);
p_output += output_depth;
vst_b_x(v44, p_output);
p_output += output_depth;
vst_b_x(v40, p_output);
p_output += output_depth;
vst_b_x(v36, p_output);
p_output += output_depth;
}
for (; out_x < output_width - pad_width; ++out_x) {
INPUT_PTRS(0);
vmv_v_m(v48, v52);
vld_b_sp_xx(INPUT_0_0, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
COMPUTE();
INT32_TO_INT8_OUTPUT_PIPELINE_INPLACE(
v48, v56, v60,
output_activation_min,
output_activation_max,
output_offset);
vsraqs_b_vx(v48, v48, 0);
vst_b_x(v48, p_output);
p_output += output_depth;
}
for (; out_x < output_width; ++out_x) {
INPUT_PTRS(0);
vmv_v_m(v48, v52);
vdup_b_x(INPUT_0_2, -input_offset);
vdup_b_x(INPUT_1_2, -input_offset);
vdup_b_x(INPUT_2_2, -input_offset);
vld_b_sp_xx(INPUT_0_0, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
COMPUTE();
INT32_TO_INT8_OUTPUT_PIPELINE_INPLACE(
v48, v56, v60,
output_activation_min,
output_activation_max,
output_offset);
vsraqs_b_vx(v48, v48, 0);
vst_b_x(v48, p_output);
p_output += output_depth;
}
}
for (; out_y < output_height; ++out_y) {
vdup_b_x(INPUT_2_0, -input_offset);
vdup_b_x(INPUT_2_1, -input_offset);
vdup_b_x(INPUT_2_2, -input_offset);
int out_x = 0;
for (; out_x < pad_width; ++out_x) {
INPUT_PTRS(1);
vmv_v_m(v48, v52);
vdup_b_x(INPUT_0_0, -input_offset);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vdup_b_x(INPUT_1_0, -input_offset);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
COMPUTE();
INT32_TO_INT8_OUTPUT_PIPELINE_INPLACE(
v48, v56, v60,
output_activation_min,
output_activation_max,
output_offset);
vsraqs_b_vx(v48, v48, 0);
vst_b_x(v48, p_output);
p_output += output_depth;
}
for (; out_x < output_width - pad_width; ++out_x) {
INPUT_PTRS(0);
vmv_v_m(v48, v52);
vld_b_sp_xx(INPUT_0_0, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
COMPUTE();
INT32_TO_INT8_OUTPUT_PIPELINE_INPLACE(
v48, v56, v60,
output_activation_min,
output_activation_max,
output_offset);
vsraqs_b_vx(v48, v48, 0);
vst_b_x(v48, p_output);
p_output += output_depth;
}
for (; out_x < output_width; ++out_x) {
INPUT_PTRS(0);
vmv_v_m(v48, v52);
vld_b_sp_xx(INPUT_0_0, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vdup_b_x(INPUT_0_2, -input_offset);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vdup_b_x(INPUT_1_2, -input_offset);
COMPUTE();
INT32_TO_INT8_OUTPUT_PIPELINE_INPLACE(
v48, v56, v60,
output_activation_min,
output_activation_max,
output_offset);
vsraqs_b_vx(v48, v48, 0);
vst_b_x(v48, p_output);
p_output += output_depth;
}
}
}
}
#undef FLT_0_0
#undef FLT_0_1
#undef FLT_0_2
#undef FLT_1_0
#undef FLT_1_1
#undef FLT_1_2
#undef FLT_2_0
#undef FLT_2_1
#undef FLT_2_2
#undef INPUT_0_0
#undef INPUT_0_1
#undef INPUT_0_2
#undef INPUT_0_3
#undef INPUT_0_4
#undef INPUT_0_5
#undef INPUT_1_0
#undef INPUT_1_1
#undef INPUT_1_2
#undef INPUT_1_3
#undef INPUT_1_4
#undef INPUT_1_5
#undef INPUT_2_0
#undef INPUT_2_1
#undef INPUT_2_2
#undef INPUT_2_3
#undef INPUT_2_4
#undef INPUT_2_5
#undef COMPUTE
#undef INPUT_PTRS
}
// special case of input depth = 32n, filter shape of 3x3
void DepthwiseConvS83x3D32(
const tflite::DepthwiseParams& params, const int32_t* output_multiplier,
const int32_t* output_shift, const tflite::RuntimeShape& input_shape,
const int8_t* input_data, const tflite::RuntimeShape& filter_shape,
const int8_t* filter_data, const tflite::RuntimeShape& bias_shape,
const int32_t* bias_data, const tflite::RuntimeShape& output_shape,
int8_t* output_data
) {
const int stride_width = params.stride_width;
const int stride_height = params.stride_height;
const int pad_width = params.padding_values.width;
const int pad_height = params.padding_values.height;
const int32_t input_offset = params.input_offset;
const int32_t output_offset = params.output_offset;
const int32_t output_activation_min = params.quantized_activation_min;
const int32_t output_activation_max = params.quantized_activation_max;
const int batches = MatchingDim(input_shape, 0, output_shape, 0);
const int input_height = input_shape.Dims(1);
const int input_width = input_shape.Dims(2);
const int input_depth = input_shape.Dims(3);
const int output_height = output_shape.Dims(1);
const int output_width = output_shape.Dims(2);
const int output_depth = output_shape.Dims(3);
int32_t swizzled_bias_data[32];
int32_t swizzled_shift_multi[32];
int32_t swizzled_output_multi[32];
for (int in_channel = 0; in_channel + 32 <= input_depth; in_channel += 32) {
const int output_channel = in_channel;
VectorSwizzle(bias_data + output_channel, swizzled_bias_data, 32);
VectorSwizzle(output_multiplier + output_channel, swizzled_output_multi, 32);
VectorSwizzle(output_shift + output_channel, swizzled_shift_multi, 32);
vld_w_x_m(v52, swizzled_bias_data);
vld_w_x_m(v56, swizzled_output_multi);
vld_w_x_m(v60, swizzled_shift_multi);
vrsub_w_vx_m(v60, v60, 0);
union {
vdwconv_u8_t dwconv;
uint32_t raw;
} cmds;
cmds.raw = 0;
cmds.dwconv.sdata1 = true;
cmds.dwconv.sbias1 = input_offset;
cmds.dwconv.sdata2 = true;
cmds.dwconv.sbias2 = 0;
cmds.dwconv.mode = 0;
cmds.dwconv.sparsity = 0;
cmds.dwconv.regbase = 0;
// Don't reorder me, otherwise data will not be
// loaded in the correct order
// (we can reuse the p_flt* due to the `p` vld variant).
const int8_t* p_flt0 = filter_data + in_channel;
const int8_t* p_flt1 = p_flt0 + input_depth;
const int32_t stride = 2 * input_depth;
vld_b_sp_xx(v6, p_flt0, stride);
vld_b_sp_xx(v7, p_flt1, stride);
vld_b_sp_xx(v8, p_flt0, stride);
vld_b_sp_xx(v9, p_flt1, stride);
vld_b_sp_xx(v10, p_flt0, stride);
vld_b_sp_xx(v11, p_flt1, stride);
vld_b_sp_xx(v12, p_flt0, stride);
vld_b_sp_xx(v13, p_flt1, stride);
vld_b_sp_xx(v14, p_flt0, stride);
for (int batch = 0; batch < batches; ++batch) {
const int8_t* p_output = output_data + (batch * output_width * output_height * output_depth) + output_channel;
for (int out_y = 0; out_y < output_height; ++out_y) {
const int in_y_origin = (out_y * stride_height) - pad_height;
const int y_offset = (output_depth * output_width * out_y);
for (int out_x = 0; out_x < output_width; ++out_x) {
const int in_x_origin = (out_x * stride_width) - pad_width;
// Initialize accumulators w/ bias data.
vmv_v_m(v48, v52);
bool top_pad = in_y_origin < 0;
bool left_pad = in_x_origin < 0;
bool bottom_pad = (in_y_origin + 2) >= input_height;
bool right_pad = (in_x_origin + 2) >= input_width;
bool padding_required = top_pad || left_pad || bottom_pad || right_pad;
const int8_t* p_in_0 = input_data +
(batch * input_height * input_width * input_depth) +
(in_y_origin * input_width * input_depth) +
(in_x_origin * input_depth) +
in_channel;
const int8_t* p_in_1 = p_in_0 + (input_width * input_depth);
const int8_t* p_in_2 = p_in_1 + (input_width * input_depth);
if (!padding_required) {
vld_b_sp_xx(v15, p_in_0, input_depth);
vld_b_sp_xx(v16, p_in_0, input_depth);
vld_b_sp_xx(v17, p_in_0, input_depth);
vld_b_sp_xx(v18, p_in_1, input_depth);
vld_b_sp_xx(v19, p_in_1, input_depth);
vld_b_sp_xx(v20, p_in_1, input_depth);
vld_b_sp_xx(v21, p_in_2, input_depth);
vld_b_sp_xx(v22, p_in_2, input_depth);
vld_b_sp_xx(v23, p_in_2, input_depth);
} else {
// Top row
if (top_pad || left_pad) {
vdup_b_x(v15, -input_offset);
} else {
vld_b_x(v15, p_in_0);
}
if (top_pad) {
vdup_b_x(v16, -input_offset);
} else {
vld_b_x(v16, p_in_0 + input_depth);
}
if (top_pad || right_pad) {
vdup_b_x(v17, -input_offset);
} else {
vld_b_x(v17, p_in_0 + (2 * input_depth));
}
// Middle row
if (left_pad) {
vdup_b_x(v18, -input_offset);
} else {
vld_b_x(v18, p_in_1);
}
vld_b_x(v19, p_in_1 + input_depth);
if (right_pad) {
vdup_b_x(v20, -input_offset);
} else {
vld_b_x(v20, p_in_1 + (2 * input_depth));
}
// Bottom row
if (bottom_pad || left_pad) {
vdup_b_x(v21, -input_offset);
} else {
vld_b_x(v21, p_in_2);
}
if (bottom_pad) {
vdup_b_x(v22, -input_offset);
} else {
vld_b_x(v22, p_in_2 + input_depth);
}
if (bottom_pad || right_pad) {
vdup_b_x(v23, -input_offset);
} else {
vld_b_x(v23, p_in_2 + (2 * input_depth));
}
}
adwinit_v(v48, v48);
adwconv_vxv(v48, v15, cmds, v6);
adwconv_vxv(v48, v18, cmds, v9);
vdwconv_vxv(v48, v21, cmds, v12);
vdmulh_w_rn_vv_m(v48, v48, v56);
vsha_w_r_vv_m(v48, v48, v60);
vadd_w_vx_m(v48, v48, output_offset);
vmax_w_vx_m(v48, v48, output_activation_min);
vmin_w_vx_m(v48, v48, output_activation_max);
vsraqs_b_vx(v48, v48, 0);
vst_b_x(v48, p_output + (out_x * output_depth) + y_offset);
}
}
}
}
}
// special case of input depth = 32n, filter shape of 5x5, stride == 1
void DepthwiseConvS85x5D32_Stride1(
const tflite::DepthwiseParams& params, const int32_t* output_multiplier,
const int32_t* output_shift, const tflite::RuntimeShape& input_shape,
const int8_t* input_data, const tflite::RuntimeShape& filter_shape,
const int8_t* filter_data, const tflite::RuntimeShape& bias_shape,
const int32_t* bias_data, const tflite::RuntimeShape& output_shape,
int8_t* output_data
) {
const int stride_width = params.stride_width;
const int stride_height = params.stride_height;
const int pad_width = params.padding_values.width;
const int pad_height = params.padding_values.height;
assert(pad_width == 2);
assert(pad_height == 2);
const int32_t input_offset = params.input_offset;
const int32_t output_offset = params.output_offset;
const int32_t output_activation_min = params.quantized_activation_min;
const int32_t output_activation_max = params.quantized_activation_max;
const int batches = MatchingDim(input_shape, 0, output_shape, 0);
const int input_height = input_shape.Dims(1);
const int input_width = input_shape.Dims(2);
const int input_depth = input_shape.Dims(3);
const int output_height = output_shape.Dims(1);
const int output_width = output_shape.Dims(2);
const int output_depth = output_shape.Dims(3);
int32_t swizzled_bias_data[32];
int32_t swizzled_shift_multi[32];
int32_t swizzled_output_multi[32];
// INPUT_Y_X
#define INPUT_0_0 v26
#define INPUT_0_1 v29
#define INPUT_0_2 v32
#define INPUT_0_3 v35
#define INPUT_0_4 v38
#define INPUT_1_0 v27
#define INPUT_1_1 v30
#define INPUT_1_2 v33
#define INPUT_1_3 v36
#define INPUT_1_4 v39
#define INPUT_2_0 v28
#define INPUT_2_1 v31
#define INPUT_2_2 v34
#define INPUT_2_3 v37
#define INPUT_2_4 v40
#define INPUT_3_0 v41
#define INPUT_3_1 v42
#define INPUT_3_2 v43
#define INPUT_3_3 v44
#define INPUT_3_4 v45
#define INPUT_4_0 v47
#define INPUT_4_1 v48
#define INPUT_4_2 v49
#define INPUT_4_3 v50
#define INPUT_4_4 v51
#define INPUT_0_5 v53
#define INPUT_1_5 v54
#define INPUT_2_5 v55
#define INPUT_3_5 v46
#define INPUT_4_5 v52
#define FLT_0_0 v0
#define FLT_0_1 v3
#define FLT_0_2 v6
#define FLT_0_3 v9
#define FLT_0_4 v12
#define FLT_1_0 v1
#define FLT_1_1 v4
#define FLT_1_2 v7
#define FLT_1_3 v10
#define FLT_1_4 v13
#define FLT_2_0 v2
#define FLT_2_1 v5
#define FLT_2_2 v8
#define FLT_2_3 v11
#define FLT_2_4 v14
#define FLT_3_0 v15
#define FLT_3_1 v16
#define FLT_3_2 v17
#define FLT_3_3 v18
#define FLT_3_4 v19
#define FLT_HOLE v20
#define FLT_4_0 v21
#define FLT_4_1 v22
#define FLT_4_2 v23
#define FLT_4_3 v24
#define FLT_4_4 v25
#define COMPUTE() \
vld_w_x_m(v60, swizzled_bias_data); \
adwinit_v(v60, v60); \
/* 0,0 1,0 2,0 */ \
adwconv_vxv(v60, INPUT_0_0, cmds, FLT_0_0); \
/* 0,1 1,1 2,1 */ \
adwconv_vxv(v60, INPUT_0_1, cmds, FLT_0_1); \
/* 0,2 1,2 2,2*/ \
adwconv_vxv(v60, INPUT_0_2, cmds, FLT_0_2); \
/* 0,3 1,3 2,3 */ \
adwconv_vxv(v60, INPUT_0_3, cmds, FLT_0_3); \
/* 0,4 1,4 2,4 */ \
adwconv_vxv(v60, INPUT_0_4, cmds, FLT_0_4); \
/* 3,0 3,1 3,2 */ \
adwconv_vxv(v60, INPUT_3_0, cmds, FLT_3_0); \
/* 3,3 3,4 hole */ \
adwconv_vxv(v60, INPUT_3_3, cmds, FLT_3_3); \
/* hole 4,0 4,1*/ \
adwconv_vxv(v60, INPUT_3_5, cmds, FLT_HOLE); \
/* 4,2 4,3 4,4*/ \
vdwconv_vxv(v60, INPUT_4_2, cmds, FLT_4_2); \
INT32_TO_INT8_OUTPUT_PIPELINE_INPLACE(v60, v56, v52, \
output_activation_min, \
output_activation_max, \
output_offset); \
vsraqs_b_vx(v60, v60, 0); \
vst_b_x(v60, p_output);
#define INPUT_PTRS(_strides) \
const int in_x_origin = (out_x * stride_width) - pad_width; \
const int in_y_origin = (out_y * stride_height) - pad_height; \
const int8_t* p_in_0 = input_data + \
(batch * input_height * input_width * input_depth) + \
(in_y_origin * input_width * input_depth) + \
((in_x_origin + _strides) * input_depth) + \
in_channel; \
const int8_t* p_in_1 = p_in_0 + (input_width * input_depth); \
const int8_t* p_in_2 = p_in_1 + (input_width * input_depth); \
const int8_t* p_in_3 = p_in_2 + (input_width * input_depth); \
const int8_t* p_in_4 = p_in_3 + (input_width * input_depth); \
(void)p_in_4;
for (int in_channel = 0; in_channel + 32 <= input_depth; in_channel += 32) {
const int output_channel = in_channel;
VectorSwizzle(bias_data + output_channel, swizzled_bias_data, 32);
VectorSwizzle(output_multiplier + output_channel, swizzled_output_multi, 32);
VectorSwizzle(output_shift + output_channel, swizzled_shift_multi, 32);
union {
vdwconv_u8_t dwconv;
uint32_t raw;
} cmds;
cmds.raw = 0;
cmds.dwconv.sdata1 = true;
cmds.dwconv.sbias1 = input_offset;
cmds.dwconv.sdata2 = true;
cmds.dwconv.sbias2 = 0;
cmds.dwconv.mode = 0;
cmds.dwconv.sparsity = 0;
cmds.dwconv.regbase = 0;
// Don't reorder me!
const int8_t* p_flt0 = filter_data + in_channel;
const int32_t stride = input_depth;
vld_b_sp_xx(FLT_0_0, p_flt0, stride);
vld_b_sp_xx(FLT_0_1, p_flt0, stride);
vld_b_sp_xx(FLT_0_2, p_flt0, stride);
vld_b_sp_xx(FLT_0_3, p_flt0, stride);
vld_b_sp_xx(FLT_0_4, p_flt0, stride);
vld_b_sp_xx(FLT_1_0, p_flt0, stride);
vld_b_sp_xx(FLT_1_1, p_flt0, stride);
vld_b_sp_xx(FLT_1_2, p_flt0, stride);
vld_b_sp_xx(FLT_1_3, p_flt0, stride);
vld_b_sp_xx(FLT_1_4, p_flt0, stride);
vld_b_sp_xx(FLT_2_0, p_flt0, stride);
vld_b_sp_xx(FLT_2_1, p_flt0, stride);
vld_b_sp_xx(FLT_2_2, p_flt0, stride);
vld_b_sp_xx(FLT_2_3, p_flt0, stride);
vld_b_sp_xx(FLT_2_4, p_flt0, stride);
vld_b_sp_xx(FLT_3_0, p_flt0, stride);
vld_b_sp_xx(FLT_3_1, p_flt0, stride);
vld_b_sp_xx(FLT_3_2, p_flt0, stride);
vld_b_sp_xx(FLT_3_3, p_flt0, stride);
vld_b_sp_xx(FLT_3_4, p_flt0, stride);
vld_b_sp_xx(FLT_4_0, p_flt0, stride);
vld_b_sp_xx(FLT_4_1, p_flt0, stride);
vld_b_sp_xx(FLT_4_2, p_flt0, stride);
vld_b_sp_xx(FLT_4_3, p_flt0, stride);
vld_b_sp_xx(FLT_4_4, p_flt0, stride);
vdup_b_x(FLT_HOLE, 0);
vld_w_x_m(v56, swizzled_output_multi);
vld_w_x_m(v52, swizzled_shift_multi);
vrsub_w_vx_m(v52, v52, 0);
for (int batch = 0; batch < batches; ++batch) {
const int8_t* p_output = output_data + (batch * output_height * output_width * output_depth) + output_channel;
int out_y = 0;
// Done
{ // out_y = 0;
int out_x = 0;
vdup_b_x(INPUT_0_0, -input_offset);
vdup_b_x(INPUT_0_1, -input_offset);
vdup_b_x(INPUT_0_2, -input_offset);
vdup_b_x(INPUT_0_3, -input_offset);
vdup_b_x(INPUT_0_4, -input_offset);
vdup_b_x(INPUT_1_0, -input_offset);
vdup_b_x(INPUT_1_1, -input_offset);
vdup_b_x(INPUT_1_2, -input_offset);
vdup_b_x(INPUT_1_3, -input_offset);
vdup_b_x(INPUT_1_4, -input_offset);
{ // out_x == 0
INPUT_PTRS(2);
vdup_b_x(INPUT_2_0, -input_offset);
vdup_b_x(INPUT_2_1, -input_offset);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_4, p_in_2, input_depth);
vdup_b_x(INPUT_3_0, -input_offset);
vdup_b_x(INPUT_3_1, -input_offset);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_3, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_4, p_in_3, input_depth);
vdup_b_x(INPUT_4_0, -input_offset);
vdup_b_x(INPUT_4_1, -input_offset);
vld_b_sp_xx(INPUT_4_2, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_3, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_4, p_in_4, input_depth);
COMPUTE();
p_output += output_depth;
++out_x;
}
{ // out_x == 1
INPUT_PTRS(1);
vdup_b_x(INPUT_2_0, -input_offset);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_4, p_in_2, input_depth);
vdup_b_x(INPUT_3_0, -input_offset);
vld_b_sp_xx(INPUT_3_1, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_3, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_4, p_in_3, input_depth);
vdup_b_x(INPUT_4_0, -input_offset);
vld_b_sp_xx(INPUT_4_1, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_2, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_3, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_4, p_in_4, input_depth);
COMPUTE();
p_output += output_depth;
++out_x;
}
for (; out_x < output_width - pad_width; ++out_x) {
INPUT_PTRS(0);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_4, p_in_2, input_depth);
vld_b_sp_xx(INPUT_3_0, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_1, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_3, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_4, p_in_3, input_depth);
vld_b_sp_xx(INPUT_4_0, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_1, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_2, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_3, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_4, p_in_4, input_depth);
COMPUTE();
p_output += output_depth;
}
{ // out_x == output_width - 2
INPUT_PTRS(0);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vdup_b_x(INPUT_2_4, -input_offset);
vld_b_sp_xx(INPUT_3_0, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_1, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_3, p_in_3, input_depth);
vdup_b_x(INPUT_3_4, -input_offset);
vld_b_sp_xx(INPUT_4_0, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_1, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_2, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_3, p_in_4, input_depth);
vdup_b_x(INPUT_4_4, -input_offset);
COMPUTE();
p_output += output_depth;
++out_x;
}
{ // out_x == output_width - 1
INPUT_PTRS(0);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vdup_b_x(INPUT_2_3, -input_offset);
vdup_b_x(INPUT_2_4, -input_offset);
vld_b_sp_xx(INPUT_3_0, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_1, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vdup_b_x(INPUT_3_3, -input_offset);
vdup_b_x(INPUT_3_4, -input_offset);
vld_b_sp_xx(INPUT_4_0, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_1, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_2, p_in_4, input_depth);
vdup_b_x(INPUT_4_3, -input_offset);
vdup_b_x(INPUT_4_4, -input_offset);
COMPUTE();
p_output += output_depth;
++out_x;
}
++out_y;
}
// Done
{ // out_y = 1;
int out_x = 0;
vdup_b_x(INPUT_0_0, -input_offset);
vdup_b_x(INPUT_0_1, -input_offset);
vdup_b_x(INPUT_0_2, -input_offset);
vdup_b_x(INPUT_0_3, -input_offset);
vdup_b_x(INPUT_0_4, -input_offset);
{ // out_x = 0;
INPUT_PTRS(2);
vdup_b_x(INPUT_1_0, -input_offset);
vdup_b_x(INPUT_1_1, -input_offset);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_3, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_4, p_in_1, input_depth);
vdup_b_x(INPUT_2_0, -input_offset);
vdup_b_x(INPUT_2_1, -input_offset);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_4, p_in_2, input_depth);
vdup_b_x(INPUT_3_0, -input_offset);
vdup_b_x(INPUT_3_1, -input_offset);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_3, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_4, p_in_3, input_depth);
vdup_b_x(INPUT_4_0, -input_offset);
vdup_b_x(INPUT_4_1, -input_offset);
vld_b_sp_xx(INPUT_4_2, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_3, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_4, p_in_4, input_depth);
COMPUTE();
p_output += output_depth;
++out_x;
}
{ // out_x = 1;
INPUT_PTRS(1);
vdup_b_x(INPUT_1_0, -input_offset);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_3, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_4, p_in_1, input_depth);
vdup_b_x(INPUT_2_0, -input_offset);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_4, p_in_2, input_depth);
vdup_b_x(INPUT_3_0, -input_offset);
vld_b_sp_xx(INPUT_3_1, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_3, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_4, p_in_3, input_depth);
vdup_b_x(INPUT_4_0, -input_offset);
vld_b_sp_xx(INPUT_4_1, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_2, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_3, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_4, p_in_4, input_depth);
COMPUTE();
p_output += output_depth;
++out_x;
}
for (; out_x < output_width - pad_width; ++out_x) {
INPUT_PTRS(0);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_3, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_4, p_in_1, input_depth);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_4, p_in_2, input_depth);
vld_b_sp_xx(INPUT_3_0, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_1, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_3, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_4, p_in_3, input_depth);
vld_b_sp_xx(INPUT_4_0, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_1, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_2, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_3, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_4, p_in_4, input_depth);
COMPUTE();
p_output += output_depth;
}
{ // out_x = output_width - 2
INPUT_PTRS(0);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_3, p_in_1, input_depth);
vdup_b_x(INPUT_1_4, -input_offset);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vdup_b_x(INPUT_2_4, -input_offset);
vld_b_sp_xx(INPUT_3_0, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_1, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_3, p_in_3, input_depth);
vdup_b_x(INPUT_3_4, -input_offset);
vld_b_sp_xx(INPUT_4_0, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_1, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_2, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_3, p_in_4, input_depth);
vdup_b_x(INPUT_4_4, -input_offset);
COMPUTE();
p_output += output_depth;
++out_x;
}
{ // out_x = output_width - 1
INPUT_PTRS(0);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vdup_b_x(INPUT_1_3, -input_offset);
vdup_b_x(INPUT_1_4, -input_offset);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vdup_b_x(INPUT_2_3, -input_offset);
vdup_b_x(INPUT_2_4, -input_offset);
vld_b_sp_xx(INPUT_3_0, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_1, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vdup_b_x(INPUT_3_3, -input_offset);
vdup_b_x(INPUT_3_4, -input_offset);
vld_b_sp_xx(INPUT_4_0, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_1, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_2, p_in_4, input_depth);
vdup_b_x(INPUT_4_3, -input_offset);
vdup_b_x(INPUT_4_4, -input_offset);
COMPUTE();
p_output += output_depth;
}
++out_y;
}
// Done
for (; out_y < output_height - pad_height; ++out_y) {
int out_x = 0;
{ // out_x == 0
INPUT_PTRS(2);
vdup_b_x(INPUT_0_0, -input_offset);
vdup_b_x(INPUT_0_1, -input_offset);
vdup_b_x(INPUT_1_0, -input_offset);
vdup_b_x(INPUT_1_1, -input_offset);
vdup_b_x(INPUT_2_0, -input_offset);
vdup_b_x(INPUT_2_1, -input_offset);
vdup_b_x(INPUT_3_0, -input_offset);
vdup_b_x(INPUT_3_1, -input_offset);
vdup_b_x(INPUT_4_0, -input_offset);
vdup_b_x(INPUT_4_1, -input_offset);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_3, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_4, p_in_0, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_3, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_4, p_in_1, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_4, p_in_2, input_depth);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_3, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_4, p_in_3, input_depth);
vld_b_sp_xx(INPUT_4_2, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_3, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_4, p_in_4, input_depth);
COMPUTE();
p_output += output_depth;
++out_x;
}
{ // out_x == 1
INPUT_PTRS(1);
vdup_b_x(INPUT_0_0, -input_offset);
vdup_b_x(INPUT_1_0, -input_offset);
vdup_b_x(INPUT_2_0, -input_offset);
vdup_b_x(INPUT_3_0, -input_offset);
vdup_b_x(INPUT_4_0, -input_offset);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_3, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_4, p_in_0, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_3, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_4, p_in_1, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_4, p_in_2, input_depth);
vld_b_sp_xx(INPUT_3_1, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_3, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_4, p_in_3, input_depth);
vld_b_sp_xx(INPUT_4_1, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_2, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_3, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_4, p_in_4, input_depth);
COMPUTE();
p_output += input_depth;
++out_x;
}
for (; out_x + 4 <= output_width - pad_width; out_x += 4) {
INPUT_PTRS(0);
vld_w_x_m(v60, swizzled_bias_data);
adwinit_v(v60, v60);
// Load top 3x8, in column-major order
vld_b_sp_xx(v26, p_in_0, input_depth);
vld_b_sp_xx(v27, p_in_1, input_depth);
vld_b_sp_xx(v28, p_in_2, input_depth);
vld_b_sp_xx(v29, p_in_0, input_depth);
vld_b_sp_xx(v30, p_in_1, input_depth);
vld_b_sp_xx(v31, p_in_2, input_depth);
vld_b_sp_xx(v32, p_in_0, input_depth);
vld_b_sp_xx(v33, p_in_1, input_depth);
vld_b_sp_xx(v34, p_in_2, input_depth);
vld_b_sp_xx(v35, p_in_0, input_depth);
vld_b_sp_xx(v36, p_in_1, input_depth);
vld_b_sp_xx(v37, p_in_2, input_depth);
vld_b_sp_xx(v38, p_in_0, input_depth);
vld_b_sp_xx(v39, p_in_1, input_depth);
vld_b_sp_xx(v40, p_in_2, input_depth);
vld_b_sp_xx(v41, p_in_0, input_depth);
vld_b_sp_xx(v42, p_in_1, input_depth);
vld_b_sp_xx(v43, p_in_2, input_depth);
vld_b_sp_xx(v44, p_in_0, input_depth);
vld_b_sp_xx(v45, p_in_1, input_depth);
vld_b_sp_xx(v46, p_in_2, input_depth);
vld_b_sp_xx(v47, p_in_0, input_depth);
vld_b_sp_xx(v48, p_in_1, input_depth);
vld_b_sp_xx(v49, p_in_2, input_depth);
// Compute 3x5, starting from 0,3
adwconv_vxv(v60, v35, cmds, FLT_0_0);
adwconv_vxv(v60, v38, cmds, FLT_0_1);
adwconv_vxv(v60, v41, cmds, FLT_0_2);
adwconv_vxv(v60, v44, cmds, FLT_0_3);
vdwconv_vxv(v60, v47, cmds, FLT_0_4);
// Compute 3x5, starting from 0,2
vld_w_x_m(v56, swizzled_bias_data);
adwinit_v(v56, v56);
adwconv_vxv(v56, v32, cmds, FLT_0_0);
adwconv_vxv(v56, v35, cmds, FLT_0_1);
adwconv_vxv(v56, v38, cmds, FLT_0_2);
adwconv_vxv(v56, v41, cmds, FLT_0_3);
vdwconv_vxv(v56, v44, cmds, FLT_0_4);
// Compute 3x5, starting from 0,1
vld_w_x_m(v52, swizzled_bias_data);
adwinit_v(v52, v52);
adwconv_vxv(v52, v29, cmds, FLT_0_0);
adwconv_vxv(v52, v32, cmds, FLT_0_1);
adwconv_vxv(v52, v35, cmds, FLT_0_2);
adwconv_vxv(v52, v38, cmds, FLT_0_3);
vdwconv_vxv(v52, v41, cmds, FLT_0_4);
// Compute 3x5, starting from 0,3
vld_w_x_m(v48, swizzled_bias_data);
adwinit_v(v48, v48);
adwconv_vxv(v48, v26, cmds, FLT_0_0);
adwconv_vxv(v48, v29, cmds, FLT_0_1);
adwconv_vxv(v48, v32, cmds, FLT_0_2);
adwconv_vxv(v48, v35, cmds, FLT_0_3);
vdwconv_vxv(v48, v38, cmds, FLT_0_4);
// Load bottom 2x8, row major
vld_b_sp_xx(v26, p_in_3, input_depth);
vld_b_sp_xx(v27, p_in_3, input_depth);
vld_b_sp_xx(v28, p_in_3, input_depth);
vld_b_sp_xx(v29, p_in_3, input_depth);
vld_b_sp_xx(v30, p_in_3, input_depth);
vld_b_sp_xx(v31, p_in_3, input_depth);
vld_b_sp_xx(v32, p_in_3, input_depth);
vld_b_sp_xx(v33, p_in_3, input_depth);
vld_b_sp_xx(v34, p_in_4, input_depth);
vld_b_sp_xx(v35, p_in_4, input_depth);
vld_b_sp_xx(v36, p_in_4, input_depth);
vld_b_sp_xx(v37, p_in_4, input_depth);
vld_b_sp_xx(v38, p_in_4, input_depth);
vld_b_sp_xx(v39, p_in_4, input_depth);
vld_b_sp_xx(v40, p_in_4, input_depth);
vld_b_sp_xx(v41, p_in_4, input_depth);
// Compute bottom 2x5, starting at 3,3
adwinit_v(v60, v60);
adwconv_vxv(v60, v29, cmds, FLT_3_0);
adwconv_vxv(v60, v32, cmds, FLT_3_3);
adwconv_vxv(v60, v36, cmds, FLT_HOLE);
vdwconv_vxv(v60, v39, cmds, FLT_4_2);
// Compute bottom 2x5, starting at 3,2
adwinit_v(v56, v56);
adwconv_vxv(v56, v28, cmds, FLT_3_0);
adwconv_vxv(v56, v31, cmds, FLT_3_3);
adwconv_vxv(v56, v35, cmds, FLT_HOLE);
vdwconv_vxv(v56, v38, cmds, FLT_4_2);
// Compute bottom 2x5, starting at 3,1
adwinit_v(v52, v52);
adwconv_vxv(v52, v27, cmds, FLT_3_0);
adwconv_vxv(v52, v30, cmds, FLT_3_3);
adwconv_vxv(v52, v34, cmds, FLT_HOLE);
vdwconv_vxv(v52, v37, cmds, FLT_4_2);
// Compute bottom 2x5, starting at 3,0
adwinit_v(v48, v48);
adwconv_vxv(v48, v26, cmds, FLT_3_0);
adwconv_vxv(v48, v29, cmds, FLT_3_3);
adwconv_vxv(v48, v33, cmds, FLT_HOLE);
vdwconv_vxv(v48, v36, cmds, FLT_4_2);
// Load output parameters
vld_w_x_m(v40, swizzled_output_multi);
vld_w_x_m(v44, swizzled_shift_multi);
vrsub_w_vx_m(v44, v44, 0);
// Compute final outputs, for both 5x5 patches, and store.
INT32_TO_INT8_OUTPUT_PIPELINE_INPLACE4(v60, v56, v52, v48, v40, v44, output_activation_min, output_activation_max, output_offset);
vsraqs_b_vx(v48, v48, 0);
vst_b_x(v48, p_output);
p_output += output_depth;
vsraqs_b_vx(v52, v52, 0);
vst_b_x(v52, p_output);
p_output += output_depth;
vsraqs_b_vx(v56, v56, 0);
vst_b_x(v56, p_output);
p_output += output_depth;
vsraqs_b_vx(v60, v60, 0);
vst_b_x(v60, p_output);
p_output += output_depth;
}
// These were clobbered due to the different compute pattern
// in the previous loop, so re-load them.
vld_w_x_m(v56, swizzled_output_multi);
vld_w_x_m(v52, swizzled_shift_multi);
vrsub_w_vx_m(v52, v52, 0);
for (; out_x < output_width - pad_width; ++out_x) {
INPUT_PTRS(0);
vld_b_sp_xx(INPUT_0_0, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_3, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_4, p_in_0, input_depth);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_3, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_4, p_in_1, input_depth);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_4, p_in_2, input_depth);
vld_b_sp_xx(INPUT_3_0, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_1, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_3, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_4, p_in_3, input_depth);
vld_b_sp_xx(INPUT_4_0, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_1, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_2, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_3, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_4, p_in_4, input_depth);
COMPUTE();
p_output += output_depth;
}
{ // out_x == output_width - 2
INPUT_PTRS(0);
vld_b_sp_xx(INPUT_0_0, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_3, p_in_0, input_depth);
vdup_b_x(INPUT_0_4, -input_offset);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_3, p_in_1, input_depth);
vdup_b_x(INPUT_1_4, -input_offset);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vdup_b_x(INPUT_2_4, -input_offset);
vld_b_sp_xx(INPUT_3_0, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_1, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_3, p_in_3, input_depth);
vdup_b_x(INPUT_3_4, -input_offset);
vld_b_sp_xx(INPUT_4_0, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_1, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_2, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_3, p_in_4, input_depth);
vdup_b_x(INPUT_4_4, -input_offset);
COMPUTE();
p_output += output_depth;
++out_x;
}
{ // out_x == output_width - 1
INPUT_PTRS(0);
vld_b_sp_xx(INPUT_0_0, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vdup_b_x(INPUT_0_3, -input_offset);
vdup_b_x(INPUT_0_4, -input_offset);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vdup_b_x(INPUT_1_3, -input_offset);
vdup_b_x(INPUT_1_4, -input_offset);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vdup_b_x(INPUT_2_3, -input_offset);
vdup_b_x(INPUT_2_4, -input_offset);
vld_b_sp_xx(INPUT_3_0, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_1, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vdup_b_x(INPUT_3_3, -input_offset);
vdup_b_x(INPUT_3_4, -input_offset);
vld_b_sp_xx(INPUT_4_0, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_1, p_in_4, input_depth);
vld_b_sp_xx(INPUT_4_2, p_in_4, input_depth);
vdup_b_x(INPUT_4_3, -input_offset);
vdup_b_x(INPUT_4_4, -input_offset);
COMPUTE();
p_output += output_depth;
}
}
// Done
{ // out_y == output_height - 2
int out_x = 0;
vdup_b_x(INPUT_4_0, -input_offset);
vdup_b_x(INPUT_4_1, -input_offset);
vdup_b_x(INPUT_4_2, -input_offset);
vdup_b_x(INPUT_4_3, -input_offset);
vdup_b_x(INPUT_4_4, -input_offset);
{ // out_x == 0
INPUT_PTRS(2);
vdup_b_x(INPUT_0_0, -input_offset);
vdup_b_x(INPUT_0_1, -input_offset);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_3, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_4, p_in_0, input_depth);
vdup_b_x(INPUT_1_0, -input_offset);
vdup_b_x(INPUT_1_1, -input_offset);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_3, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_4, p_in_1, input_depth);
vdup_b_x(INPUT_2_0, -input_offset);
vdup_b_x(INPUT_2_1, -input_offset);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_4, p_in_2, input_depth);
vdup_b_x(INPUT_3_0, -input_offset);
vdup_b_x(INPUT_3_1, -input_offset);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_3, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_4, p_in_3, input_depth);
COMPUTE();
p_output += output_depth;
++out_x;
}
{ // out_x == 1
INPUT_PTRS(1);
vdup_b_x(INPUT_0_0, -input_offset);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_3, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_4, p_in_0, input_depth);
vdup_b_x(INPUT_1_0, -input_offset);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_3, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_4, p_in_1, input_depth);
vdup_b_x(INPUT_2_0, -input_offset);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_4, p_in_2, input_depth);
vdup_b_x(INPUT_3_0, -input_offset);
vld_b_sp_xx(INPUT_3_1, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_3, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_4, p_in_3, input_depth);
COMPUTE();
p_output += output_depth;
++out_x;
}
for (; out_x < output_width - pad_width; ++out_x) {
INPUT_PTRS(0);
vld_b_sp_xx(INPUT_0_0, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_3, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_4, p_in_0, input_depth);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_3, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_4, p_in_1, input_depth);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_4, p_in_2, input_depth);
vld_b_sp_xx(INPUT_3_0, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_1, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_3, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_4, p_in_3, input_depth);
COMPUTE();
p_output += output_depth;
}
{ // out_x == output_width - 2
INPUT_PTRS(0);
vld_b_sp_xx(INPUT_0_0, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_3, p_in_0, input_depth);
vdup_b_x(INPUT_0_4, -input_offset);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_3, p_in_1, input_depth);
vdup_b_x(INPUT_1_4, -input_offset);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vdup_b_x(INPUT_2_4, -input_offset);
vld_b_sp_xx(INPUT_3_0, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_1, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_3, p_in_3, input_depth);
vdup_b_x(INPUT_3_4, -input_offset);
COMPUTE();
p_output += output_depth;
++out_x;
}
{ // out_x == output_width - 1
INPUT_PTRS(0);
vld_b_sp_xx(INPUT_0_0, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vdup_b_x(INPUT_0_3, -input_offset);
vdup_b_x(INPUT_0_4, -input_offset);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vdup_b_x(INPUT_1_3, -input_offset);
vdup_b_x(INPUT_1_4, -input_offset);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vdup_b_x(INPUT_2_3, -input_offset);
vdup_b_x(INPUT_2_4, -input_offset);
vld_b_sp_xx(INPUT_3_0, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_1, p_in_3, input_depth);
vld_b_sp_xx(INPUT_3_2, p_in_3, input_depth);
vdup_b_x(INPUT_3_3, -input_offset);
vdup_b_x(INPUT_3_4, -input_offset);
COMPUTE();
p_output += output_depth;
++out_x;
}
++out_y;
}
// Done
{ // out_y == output_height - 1
int out_x = 0;
vdup_b_x(INPUT_3_0, -input_offset);
vdup_b_x(INPUT_3_1, -input_offset);
vdup_b_x(INPUT_3_2, -input_offset);
vdup_b_x(INPUT_3_3, -input_offset);
vdup_b_x(INPUT_3_4, -input_offset);
vdup_b_x(INPUT_4_0, -input_offset);
vdup_b_x(INPUT_4_1, -input_offset);
vdup_b_x(INPUT_4_2, -input_offset);
vdup_b_x(INPUT_4_3, -input_offset);
vdup_b_x(INPUT_4_4, -input_offset);
{ // out_x == 0
INPUT_PTRS(2);
vdup_b_x(INPUT_0_0, -input_offset);
vdup_b_x(INPUT_0_1, -input_offset);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_3, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_4, p_in_0, input_depth);
vdup_b_x(INPUT_1_0, -input_offset);
vdup_b_x(INPUT_1_1, -input_offset);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_3, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_4, p_in_1, input_depth);
vdup_b_x(INPUT_2_0, -input_offset);
vdup_b_x(INPUT_2_1, -input_offset);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_4, p_in_2, input_depth);
COMPUTE();
p_output += output_depth;
++out_x;
}
{ // out_x == 1
INPUT_PTRS(1);
vdup_b_x(INPUT_0_0, -input_offset);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_3, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_4, p_in_0, input_depth);
vdup_b_x(INPUT_1_0, -input_offset);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_3, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_4, p_in_1, input_depth);
vdup_b_x(INPUT_2_0, -input_offset);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_4, p_in_2, input_depth);
COMPUTE();
p_output += output_depth;
++out_x;
}
for (; out_x < output_width - pad_width; ++out_x) {
INPUT_PTRS(0);
vld_b_sp_xx(INPUT_0_0, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_3, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_4, p_in_0, input_depth);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_3, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_4, p_in_1, input_depth);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_4, p_in_2, input_depth);
COMPUTE();
p_output += output_depth;
}
{ // out_x == output_width - 2
INPUT_PTRS(0);
vld_b_sp_xx(INPUT_0_0, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_3, p_in_0, input_depth);
vdup_b_x(INPUT_0_4, -input_offset);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_3, p_in_1, input_depth);
vdup_b_x(INPUT_1_4, -input_offset);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_3, p_in_2, input_depth);
vdup_b_x(INPUT_2_4, -input_offset);
COMPUTE();
p_output += output_depth;
++out_x;
}
{ // out_x == output_width - 1
INPUT_PTRS(0);
vld_b_sp_xx(INPUT_0_0, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_1, p_in_0, input_depth);
vld_b_sp_xx(INPUT_0_2, p_in_0, input_depth);
vdup_b_x(INPUT_0_3, -input_offset);
vdup_b_x(INPUT_0_4, -input_offset);
vld_b_sp_xx(INPUT_1_0, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_1, p_in_1, input_depth);
vld_b_sp_xx(INPUT_1_2, p_in_1, input_depth);
vdup_b_x(INPUT_1_3, -input_offset);
vdup_b_x(INPUT_1_4, -input_offset);
vld_b_sp_xx(INPUT_2_0, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_1, p_in_2, input_depth);
vld_b_sp_xx(INPUT_2_2, p_in_2, input_depth);
vdup_b_x(INPUT_2_3, -input_offset);
vdup_b_x(INPUT_2_4, -input_offset);
COMPUTE();
p_output += output_depth;
}
}
}
}
#undef INPUT_PTRS
#undef COMPUTE
#undef INPUT_0_0
#undef INPUT_0_1
#undef INPUT_0_2
#undef INPUT_0_3
#undef INPUT_0_4
#undef INPUT_1_0
#undef INPUT_1_1
#undef INPUT_1_2
#undef INPUT_1_3
#undef INPUT_1_4
#undef INPUT_2_0
#undef INPUT_2_1
#undef INPUT_2_2
#undef INPUT_2_3
#undef INPUT_2_4
#undef INPUT_3_0
#undef INPUT_3_1
#undef INPUT_3_2
#undef INPUT_3_3
#undef INPUT_3_4
#undef INPUT_4_0
#undef INPUT_4_1
#undef INPUT_4_2
#undef INPUT_4_3
#undef INPUT_4_4
#undef INPUT_0_5
#undef INPUT_1_5
#undef INPUT_2_5
#undef INPUT_3_5
#undef INPUT_4_5
#undef FLT_0_0
#undef FLT_0_1
#undef FLT_0_2
#undef FLT_0_3
#undef FLT_0_4
#undef FLT_1_0
#undef FLT_1_1
#undef FLT_1_2
#undef FLT_1_3
#undef FLT_1_4
#undef FLT_2_0
#undef FLT_2_1
#undef FLT_2_2
#undef FLT_2_3
#undef FLT_2_4
#undef FLT_3_0
#undef FLT_3_1
#undef FLT_3_2
#undef FLT_3_3
#undef FLT_3_4
#undef FLT_HOLE
#undef FLT_4_0
#undef FLT_4_1
#undef FLT_4_2
#undef FLT_4_3
#undef FLT_4_4
}
// special case of input depth = 32n, filter shape of 5x5
void DepthwiseConvS85x5D32(
const tflite::DepthwiseParams& params, const int32_t* output_multiplier,
const int32_t* output_shift, const tflite::RuntimeShape& input_shape,
const int8_t* input_data, const tflite::RuntimeShape& filter_shape,
const int8_t* filter_data, const tflite::RuntimeShape& bias_shape,
const int32_t* bias_data, const tflite::RuntimeShape& output_shape,
int8_t* output_data
) {
const int stride_width = params.stride_width;
const int stride_height = params.stride_height;
const int pad_width = params.padding_values.width;
const int pad_height = params.padding_values.height;
const int32_t input_offset = params.input_offset;
const int32_t output_offset = params.output_offset;
const int32_t output_activation_min = params.quantized_activation_min;
const int32_t output_activation_max = params.quantized_activation_max;
const int batches = MatchingDim(input_shape, 0, output_shape, 0);
const int input_height = input_shape.Dims(1);
const int input_width = input_shape.Dims(2);
const int input_depth = input_shape.Dims(3);
const int filter_height = filter_shape.Dims(1);
const int filter_width = filter_shape.Dims(2);
const int output_height = output_shape.Dims(1);
const int output_width = output_shape.Dims(2);
const int output_depth = output_shape.Dims(3);
int32_t swizzled_bias_data[32];
int32_t swizzled_shift_multi[32];
int32_t swizzled_output_multi[32];
for (int in_channel = 0; in_channel + 32 <= input_depth; in_channel += 32) {
const int output_channel = in_channel;
VectorSwizzle(bias_data + output_channel, swizzled_bias_data, 32);
VectorSwizzle(output_multiplier + output_channel, swizzled_output_multi, 32);
VectorSwizzle(output_shift + output_channel, swizzled_shift_multi, 32);
vld_w_x_m(v52, swizzled_bias_data);
vld_w_x_m(v56, swizzled_output_multi);
vld_w_x_m(v60, swizzled_shift_multi);
vrsub_w_vx_m(v60, v60, 0);
// Don't reorder me!
const int8_t* p_flt = filter_data + in_channel;
vld_b_sp_xx(v6, p_flt, input_depth);
vld_b_sp_xx(v7, p_flt, input_depth);
vld_b_sp_xx_m(v8, p_flt, input_depth);
vld_b_sp_xx_m(v12, p_flt, input_depth);
vld_b_sp_xx_m(v16, p_flt, input_depth);
vld_b_sp_xx_m(v20, p_flt, input_depth);
vld_b_sp_xx_m(v24, p_flt, input_depth);
vld_b_sp_xx(v28, p_flt, input_depth);
vld_b_sp_xx(v29, p_flt, input_depth);
vld_b_sp_xx(v30, p_flt, input_depth);
for (int batch = 0; batch < batches; ++batch) {
const int8_t* p_input = input_data + (batch * input_width * input_height * input_depth) + in_channel;
const int8_t* p_output = output_data + (batch * output_width * output_height * output_depth) + output_channel;
for (int out_y = 0; out_y < output_height; ++out_y) {
const int out_y_offset = (out_y * output_width * output_depth);
for (int out_x = 0; out_x < output_width; ++out_x) {
const int in_x_origin = (out_x * stride_width) - pad_width;
const int in_y_origin = (out_y * stride_height) - pad_height;
// Initialize accumulators w/ bias_data
vmv_v_m(v48, v52);
for (int filter_y = 0; filter_y < filter_height; ++filter_y) {
const int in_y = in_y_origin + filter_y;
if ((in_y < 0) || (in_y >= input_height)) {
continue;
}
switch (filter_y) {
case 0:
vaddw_h_vx(v31, v6, 0);
vaddw_h_vx(v33, v7, 0);
vaddw_h_vx(v35, v8, 0);
vaddw_h_vx(v37, v9, 0);
vaddw_h_vx(v39, v10, 0);
break;
case 1:
vaddw_h_vx(v31, v11, 0);
vaddw_h_vx(v33, v12, 0);
vaddw_h_vx(v35, v13, 0);
vaddw_h_vx(v37, v14, 0);
vaddw_h_vx(v39, v15, 0);
break;
case 2:
vaddw_h_vx(v31, v16, 0);
vaddw_h_vx(v33, v17, 0);
vaddw_h_vx(v35, v18, 0);
vaddw_h_vx(v37, v19, 0);
vaddw_h_vx(v39, v20, 0);
break;
case 3:
vaddw_h_vx(v31, v21, 0);
vaddw_h_vx(v33, v22, 0);
vaddw_h_vx(v35, v23, 0);
vaddw_h_vx(v37, v24, 0);
vaddw_h_vx(v39, v25, 0);
break;
case 4:
vaddw_h_vx(v31, v26, 0);
vaddw_h_vx(v33, v27, 0);
vaddw_h_vx(v35, v28, 0);
vaddw_h_vx(v37, v29, 0);
vaddw_h_vx(v39, v30, 0);
break;
}
const int in_y_offset = in_y * input_width * input_depth;
for (int filter_x = 0; filter_x < filter_width; ++filter_x) {
const int in_x = in_x_origin + filter_x;
if ((in_x < 0) || (in_x >= input_width)) {
continue;
}
vld_b_x(v0, p_input + (in_x * input_depth) + in_y_offset);
vaddw_h_vx(v0, v0, 0);
vadd_h_vx(v0, v0, static_cast<int16_t>(input_offset));
vadd_h_vx(v1, v1,
static_cast<int16_t>(input_offset)); // v0 v1 input
switch (filter_x) {
case 0:
vmulw_w_vv(v2, v1, v32);
vmulw_w_vv(v0, v0, v31);
break;
case 1:
vmulw_w_vv(v2, v1, v34);
vmulw_w_vv(v0, v0, v33);
break;
case 2:
vmulw_w_vv(v2, v1, v36);
vmulw_w_vv(v0, v0, v35);
break;
case 3:
vmulw_w_vv(v2, v1, v38);
vmulw_w_vv(v0, v0, v37);
break;
case 4:
vmulw_w_vv(v2, v1, v40);
vmulw_w_vv(v0, v0, v39);
break;
}
vadd_w_vv_m(v48, v48, v0);
}
}
vdmulh_w_rn_vv_m(v48, v48, v56);
vsha_w_r_vv_m(v48, v48, v60);
vadd_w_vx_m(v48, v48, output_offset);
vmax_w_vx_m(v48, v48, output_activation_min);
vmin_w_vx_m(v48, v48, output_activation_max);
vsraqs_b_vx(v48, v48, 0);
vst_b_x(v48, p_output + out_y_offset + (out_x * output_depth));
}
}
}
}
}
// special case of input depth = 32n
void DepthwiseConvS8D32(
const tflite::DepthwiseParams& params, const int32_t* output_multiplier,
const int32_t* output_shift, const tflite::RuntimeShape& input_shape,
const int8_t* input_data, const tflite::RuntimeShape& filter_shape,
const int8_t* filter_data, const tflite::RuntimeShape& bias_shape,
const int32_t* bias_data, const tflite::RuntimeShape& output_shape,
int8_t* output_data
) {
const int stride_width = params.stride_width;
const int stride_height = params.stride_height;
const int pad_width = params.padding_values.width;
const int pad_height = params.padding_values.height;
const int32_t input_offset = params.input_offset;
const int32_t output_offset = params.output_offset;
const int32_t output_activation_min = params.quantized_activation_min;
const int32_t output_activation_max = params.quantized_activation_max;
const int batches = MatchingDim(input_shape, 0, output_shape, 0);
const int input_height = input_shape.Dims(1);
const int input_width = input_shape.Dims(2);
const int input_depth = input_shape.Dims(3);
const int filter_height = filter_shape.Dims(1);
const int filter_width = filter_shape.Dims(2);
const int output_height = output_shape.Dims(1);
const int output_width = output_shape.Dims(2);
int32_t swizzled_bias_data[32];
int32_t swizzled_shift_multi[32];
int32_t swizzled_output_multi[32];
for (int in_channel = 0; in_channel + 32 <= input_depth; in_channel += 32) {
const int output_channel = in_channel;
VectorSwizzle(bias_data + output_channel, swizzled_bias_data, 32);
VectorSwizzle(output_multiplier + output_channel, swizzled_output_multi, 32);
VectorSwizzle(output_shift + output_channel, swizzled_shift_multi, 32);
vld_w_x_m(v20, swizzled_bias_data);
vld_w_x_m(v24, swizzled_output_multi);
vld_w_x_m(v28, swizzled_shift_multi);
vrsub_w_vx_m(v28, v28, 0);
for (int batch = 0; batch < batches; ++batch) {
for (int out_y = 0; out_y < output_height; ++out_y) {
for (int out_x = 0; out_x < output_width; ++out_x) {
const int in_x_origin = (out_x * stride_width) - pad_width;
const int in_y_origin = (out_y * stride_height) - pad_height;
vdup_w_x_m(v48, 0);
for (int filter_y = 0; filter_y < filter_height; ++filter_y) {
const int in_y = in_y_origin + filter_y;
if ((in_y < 0) || (in_y >= input_height)) {
continue;
}
for (int filter_x = 0; filter_x < filter_width; ++filter_x) {
const int in_x = in_x_origin + filter_x;
if ((in_x < 0) || (in_x >= input_width)) {
continue;
}
vld_b_x(v0, &input_data[tflite::Offset(input_shape, batch, in_y,
in_x, in_channel)]); // xp
vld_b_x(v4, &filter_data[tflite::Offset(filter_shape, 0, filter_y,
filter_x, in_channel)]);
vaddw_h_vx(v0, v0, 0);
vadd_h_vx(v0, v0, static_cast<int16_t>(input_offset));
vadd_h_vx(v1, v1,
static_cast<int16_t>(input_offset)); // v0 v1 input
vaddw_h_vx(v4, v4, static_cast<int16_t>(0));
vmulw_w_vv(v8, v0, v4);
vmulw_w_vv(v10, v1, v5);
vadd_w_vv_m(v48, v48, v8);
}
}
vadd_w_vv_m(v48, v48, v20); // add bias
vdmulh_w_rn_vv_m(v48, v48, v24);
vsha_w_r_vv_m(v48, v48, v28);
vadd_w_vx_m(v48, v48, output_offset);
vmax_w_vx_m(v48, v48, output_activation_min);
vmin_w_vx_m(v48, v48, output_activation_max);
vsraqs_b_vx(v48, v48, 0);
vst_b_x(v48, &output_data[tflite::Offset(output_shape, batch, out_y,
out_x, output_channel)]);
}
}
}
}
}
// generic implementation based on Kelvin ops
void DepthwiseConvS8Generic(
const tflite::DepthwiseParams& params, const int32_t* output_multiplier,
const int32_t* output_shift, const tflite::RuntimeShape& input_shape,
const int8_t* input_data, const tflite::RuntimeShape& filter_shape,
const int8_t* filter_data, const tflite::RuntimeShape& bias_shape,
const int32_t* bias_data, const tflite::RuntimeShape& output_shape,
int8_t* output_data) {
// TBD: Use Kelvin implementation to replace the below
tflite::reference_integer_ops::DepthwiseConvPerChannel(
params, output_multiplier, output_shift, input_shape, input_data,
filter_shape, filter_data, bias_shape, bias_data, output_shape,
output_data);
return;
}
} // namespace
void DepthwiseConvS8(
const tflite::DepthwiseParams& params, const int32_t* output_multiplier,
const int32_t* output_shift, const tflite::RuntimeShape& input_shape,
const int8_t* input_data, const tflite::RuntimeShape& filter_shape,
const int8_t* filter_data, const tflite::RuntimeShape& bias_shape,
const int32_t* bias_data, const tflite::RuntimeShape& output_shape,
int8_t* output_data) {
// Get parameters.
// TODO(b/141565753): Re-introduce ScopedProfilingLabel on Micro.
const int stride_width = params.stride_width;
const int stride_height = params.stride_height;
const int pad_width = params.padding_values.width;
const int pad_height = params.padding_values.height;
const int filter_height = filter_shape.Dims(1);
const int filter_width = filter_shape.Dims(2);
const int dilation_width_factor = params.dilation_width_factor;
const int dilation_height_factor = params.dilation_height_factor;
const int depth_multiplier = params.depth_multiplier;
const int32_t output_activation_min = params.quantized_activation_min;
const int32_t output_activation_max = params.quantized_activation_max;
// Check dimensions of the tensors.
TFLITE_DCHECK_EQ(input_shape.DimensionsCount(), 4);
TFLITE_DCHECK_EQ(filter_shape.DimensionsCount(), 4);
TFLITE_DCHECK_EQ(output_shape.DimensionsCount(), 4);
TFLITE_DCHECK_LE(output_activation_min, output_activation_max);
const int output_depth = MatchingDim(filter_shape, 3, output_shape, 3);
const int input_depth = input_shape.Dims(3);
TFLITE_DCHECK_EQ(output_depth, input_depth * depth_multiplier);
TFLITE_DCHECK_EQ(bias_shape.FlatSize(), output_depth);
#define RUN_KERNEL(kernel) { \
kernel(params, output_multiplier, output_shift, input_shape, input_data, \
filter_shape, filter_data, bias_shape, bias_data, output_shape, \
output_data \
); \
return; \
}
if (depth_multiplier == 1 &&
dilation_height_factor == 1 && dilation_width_factor == 1 &&
stride_height <= 2 && stride_width <= 2) {
// special case of output depth = 32n
if (output_depth % 32 == 0) {
if (filter_width == 5 && filter_height == 5) {
if (stride_width <= 1 && stride_height <= 1 && params.padding_type == tflite::PaddingType::kSame) {
RUN_KERNEL(DepthwiseConvS85x5D32_Stride1);
}
RUN_KERNEL(DepthwiseConvS85x5D32);
} if (filter_width == 3 && filter_height == 3 && pad_width <= 1 && pad_height <= 1 && stride_width == 1 && stride_height == 1) {
RUN_KERNEL(DepthwiseConvS83x3D32_Stride1);
} if (filter_width == 3 && filter_height == 3 && pad_width <= 1 && pad_height <= 1) {
RUN_KERNEL(DepthwiseConvS83x3D32);
}
RUN_KERNEL(DepthwiseConvS8D32);
}
RUN_KERNEL(DepthwiseConvS8Generic);
}
RUN_KERNEL(tflite::reference_integer_ops::DepthwiseConvPerChannel);
#undef RUN_KERNEL
}
} // namespace kelvin::opt