COMPMID-661: Convolution quantized (#32)

Change-Id: Id69df4ce98d1d89bdf9c9aa5c4d909659909b30f
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/110456
Tested-by: BSG Visual Compute Jenkins server to access repositories on http://mpd-gerrit.cambridge.arm.com <bsgcomp@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>

1 files changed, 115 insertions, 38 deletions

diff --git a/src/runtime/CL/functions/CLConvolutionLayer.cpp b/src/runtime/CL/functions/CLConvolutionLayer.cpp
index 8d45416b30..66548d19b2 100644
--- a/src/runtime/CL/functions/CLConvolutionLayer.cpp
+++ b/src/runtime/CL/functions/CLConvolutionLayer.cpp
@@ -27,6 +27,7 @@
 #include "arm_compute/core/Size2D.h"
 #include "arm_compute/core/Utils.h"
 #include "arm_compute/core/Validate.h"
+#include "arm_compute/core/utils/quantization/AsymmHelpers.h"
 #include "arm_compute/runtime/CL/CLScheduler.h"
 
 #include <cmath>
@@ -42,19 +43,22 @@ CLConvolutionLayerReshapeWeights::CLConvolutionLayerReshapeWeights(std::shared_p
 
 void CLConvolutionLayerReshapeWeights::configure(const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, bool transpose1xW)
 {
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(weights, 1, DataType::QS8, DataType::QS16, DataType::F16, DataType::F32);
+    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(weights, 1, DataType::QS8, DataType::QASYMM8, DataType::QS16, DataType::F16, DataType::F32);
     ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(weights, output);
     ARM_COMPUTE_ERROR_ON_MISMATCHING_FIXED_POINT(weights, output);
     ARM_COMPUTE_ERROR_ON(weights->info()->num_dimensions() > 4);
 
     if(biases != nullptr)
     {
+        ARM_COMPUTE_ERROR_ON(is_data_type_quantized_asymmetric(weights->info()->data_type()));
         ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(weights, biases);
         ARM_COMPUTE_ERROR_ON(biases->info()->dimension(0) != weights->info()->dimension(3));
         ARM_COMPUTE_ERROR_ON(biases->info()->num_dimensions() > 1);
     }
 
-    const bool _has_bias = (biases != nullptr);
+    const bool       append_biases = (biases != nullptr) && !is_data_type_quantized_asymmetric(weights->info()->data_type());
+    const unsigned   bias_element  = (append_biases) ? 1 : 0;
+    const ICLTensor *biases_to_use = (append_biases) ? biases : nullptr;
 
     _transpose1xW = transpose1xW;
 
@@ -62,7 +66,7 @@ void CLConvolutionLayerReshapeWeights::configure(const ICLTensor *weights, const
     {
         // Create tensor to store the reshaped weights
         const unsigned int mat_weights_cols = weights->info()->dimension(3);
-        const unsigned int mat_weights_rows = weights->info()->dimension(0) * weights->info()->dimension(1) * weights->info()->dimension(2) + (_has_bias ? 1 : 0);
+        const unsigned int mat_weights_rows = weights->info()->dimension(0) * weights->info()->dimension(1) * weights->info()->dimension(2) + bias_element;
         TensorShape        shape_wr(mat_weights_cols, mat_weights_rows);
         const DataType     dt                   = weights->info()->data_type();
         const int          fixed_point_position = weights->info()->fixed_point_position();
@@ -70,13 +74,13 @@ void CLConvolutionLayerReshapeWeights::configure(const ICLTensor *weights, const
 
         _weights_reshaped.allocator()->init(info_wr);
         _memory_group.manage(&_weights_reshaped);
-        _weights_reshape_kernel.configure(weights, biases, &_weights_reshaped);
+        _weights_reshape_kernel.configure(weights, biases_to_use, &_weights_reshaped);
         _weights_transposed_kernel.configure(&_weights_reshaped, output);
         _weights_reshaped.allocator()->allocate();
     }
     else
     {
-        _weights_reshape_kernel.configure(weights, biases, output);
+        _weights_reshape_kernel.configure(weights, biases_to_use, output);
     }
 }
 
@@ -95,36 +99,73 @@ void CLConvolutionLayerReshapeWeights::run()
 }
 
 CLConvolutionLayer::CLConvolutionLayer(std::shared_ptr<IMemoryManager> memory_manager)
-    : _memory_group(std::move(memory_manager)), _reshape_weights(), _input_im2col_kernel(), _input_interleave_kernel(), _mm_kernel(), _output_col2im_kernel(), _input_im2col_reshaped(),
-      _input_interleaved_reshaped(), _weights_reshaped(), _weights_transposed(), _gemm_output(), _has_bias(false), _is_fully_connected_convolution(false), _are_weights_reshaped(false)
+    : _memory_group(memory_manager), _reshape_weights(), _input_im2col_kernel(), _input_interleave_kernel(), _mm_kernel(), _mm_gemmlowp(memory_manager), _gemmlowp_output_stage(), _output_col2im_kernel(),
+      _input_im2col_reshaped(), _input_interleaved_reshaped(), _weights_reshaped(), _weights_transposed(), _gemm_output(), _tmp_output(), _append_bias(false), _is_fully_connected_convolution(false),
+      _are_weights_reshaped(false), _is_quantized(false)
 {
 }
 
+void CLConvolutionLayer::configure_mm(const ICLTensor *input, const ICLTensor *weights, ICLTensor *output, bool is_interleaved_transposed)
+{
+    if(_is_quantized)
+    {
+        // Since we need negative offsets for computing convolution, we need to change QuantizationInfo()
+        // Extract and negate input and weights offset
+        const QuantizationInfo input_quantization_info   = input->info()->quantization_info();
+        const QuantizationInfo weights_quantization_info = weights->info()->quantization_info();
+
+        input->info()->set_quantization_info(QuantizationInfo(input_quantization_info.scale, -input_quantization_info.offset));
+        weights->info()->set_quantization_info(QuantizationInfo(weights_quantization_info.scale, -weights_quantization_info.offset));
+
+        _mm_gemmlowp.configure(input, weights, output, GEMMInfo(false, false, true /* Reshape weights only for the first run*/));
+
+        // Revert back QuantizatioInfo as input and weights could be used in other convolution layers
+        input->info()->set_quantization_info(input_quantization_info);
+        weights->info()->set_quantization_info(weights_quantization_info);
+    }
+    else
+    {
+        _mm_kernel.configure(input, weights, output, 1.f, is_interleaved_transposed);
+    }
+}
+
 void CLConvolutionLayer::configure(const ICLTensor *input, const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, const PadStrideInfo &conv_info, const WeightsInfo &weights_info)
 {
-    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QS8, DataType::QS16, DataType::F16, DataType::F32);
+    ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QS8, DataType::QASYMM8, DataType::QS16, DataType::F16, DataType::F32);
     ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, weights);
     ARM_COMPUTE_ERROR_ON_MISMATCHING_FIXED_POINT(input, weights);
     ARM_COMPUTE_ERROR_ON(!weights_info.are_reshaped() && weights->info()->dimension(2) != input->info()->dimension(2));
     ARM_COMPUTE_ERROR_ON(weights->info()->num_dimensions() > 4);
+    ARM_COMPUTE_ERROR_ON(weights_info.are_reshaped() && is_data_type_quantized_asymmetric(input->info()->data_type()));
+
+    _is_quantized = is_data_type_quantized_asymmetric(input->info()->data_type());
 
     if(biases != nullptr)
     {
-        ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, biases);
+        if(_is_quantized)
+        {
+            ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(biases, 1, DataType::S32);
+        }
+        else
+        {
+            ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, biases);
+        }
         ARM_COMPUTE_ERROR_ON_MISMATCHING_FIXED_POINT(input, biases);
         ARM_COMPUTE_ERROR_ON(!weights_info.are_reshaped() && biases->info()->dimension(0) != weights->info()->dimension(3));
         ARM_COMPUTE_ERROR_ON(biases->info()->num_dimensions() > 1);
     }
 
-    const DataType dt                   = input->info()->data_type();
-    const int      fixed_point_position = input->info()->fixed_point_position();
+    const DataType dt = input->info()->data_type();
 
     // Set the GPU target for matrix multiply
     _mm_kernel.set_target(CLScheduler::get().target());
 
-    _has_bias             = (biases != nullptr);
+    _append_bias          = (biases != nullptr) && (!_is_quantized);
     _are_weights_reshaped = weights_info.are_reshaped();
 
+    const unsigned   bias_element  = (_append_bias) ? 1 : 0;
+    const ICLTensor *biases_to_use = (_append_bias) ? biases : nullptr;
+
     // Get parameters from conv_info
     unsigned int stride_x = 0;
     unsigned int stride_y = 0;
@@ -141,36 +182,36 @@ void CLConvolutionLayer::configure(const ICLTensor *input, const ICLTensor *weig
 
     // Check if its a "fully connected" convolution
     _is_fully_connected_convolution = ((conv_w == 1) && (conv_h == 1));
+    const bool run_interleaved      = (!_is_fully_connected_convolution && !_is_quantized);
 
     unsigned int mat_weights_cols = weights->info()->dimension(3);
-    unsigned int mat_weights_rows = weights->info()->dimension(0) * weights->info()->dimension(1) * weights->info()->dimension(2) + (_has_bias ? 1 : 0);
+    unsigned int mat_weights_rows = weights->info()->dimension(0) * weights->info()->dimension(1) * weights->info()->dimension(2) + bias_element;
 
     // Reshape weights if needed
     if(_are_weights_reshaped)
     {
         mat_weights_cols                         = weights_info.num_kernels();
         const unsigned int quarter_reshaped_cols = weights->info()->dimension(0) / 4;
-        mat_weights_rows                         = (_has_bias ? 1 + quarter_reshaped_cols : quarter_reshaped_cols);
+        mat_weights_rows                         = quarter_reshaped_cols + bias_element;
     }
     else
     {
-        if(_is_fully_connected_convolution)
+        if(_is_fully_connected_convolution || _is_quantized)
         {
             // Create tensor to store the reshaped weights
             TensorShape shape_wr(mat_weights_cols, mat_weights_rows);
-            TensorInfo  info_wr(shape_wr, 1, dt, fixed_point_position);
-            _weights_reshaped.allocator()->init(info_wr);
-            _reshape_weights.configure(weights, biases, &_weights_reshaped, false /* 1xW transpose */);
+            _weights_reshaped.allocator()->init(weights->info()->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(shape_wr));
+            _reshape_weights.configure(weights, biases_to_use, &_weights_reshaped, false /* 1xW transpose */);
         }
         else
         {
             // Create tensor to store transposed weights
             const float transpose_width = 16.0f / input->info()->element_size();
             TensorShape shape_wt(mat_weights_rows * static_cast<unsigned int>(transpose_width), static_cast<unsigned int>(std::ceil(mat_weights_cols / transpose_width)));
-            TensorInfo  info_wt(shape_wt, 1, dt, fixed_point_position);
-            _weights_reshaped.allocator()->init(info_wt);
-            _reshape_weights.configure(weights, biases, &_weights_reshaped, true /* 1xW transpose */);
+            _weights_reshaped.allocator()->init(weights->info()->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(shape_wt));
+            _reshape_weights.configure(weights, biases_to_use, &_weights_reshaped, true /* 1xW transpose */);
         }
+        _weights_reshaped.info()->set_quantization_info(weights->info()->quantization_info());
         weights = &_weights_reshaped;
     }
 
@@ -181,16 +222,16 @@ void CLConvolutionLayer::configure(const ICLTensor *input, const ICLTensor *weig
     shape_im2col.set(0, mat_input_cols);
     shape_im2col.set(1, mat_input_rows);
     shape_im2col.set(2, 1);
-    _input_im2col_reshaped.allocator()->init(TensorInfo(shape_im2col, 1, dt, fixed_point_position));
+    _input_im2col_reshaped.allocator()->init(input->info()->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(shape_im2col));
     _memory_group.manage(&_input_im2col_reshaped);
 
     // Create tensor (interleave) to prepare input tensor for GEMM
-    if(!_is_fully_connected_convolution)
+    if(run_interleaved)
     {
         TensorShape shape_interleaved = shape_im2col;
         shape_interleaved.set(0, shape_interleaved.x() * 4);
         shape_interleaved.set(1, std::ceil(shape_interleaved.y() / 4.f));
-        _input_interleaved_reshaped.allocator()->init(TensorInfo(shape_interleaved, 1, dt, fixed_point_position));
+        _input_interleaved_reshaped.allocator()->init(input->info()->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(shape_interleaved));
         _memory_group.manage(&_input_interleaved_reshaped);
     }
 
@@ -198,30 +239,51 @@ void CLConvolutionLayer::configure(const ICLTensor *input, const ICLTensor *weig
     TensorShape shape_gemm = _input_im2col_reshaped.info()->tensor_shape();
     shape_gemm.set(0, mat_weights_cols);
     shape_gemm.set(1, mat_input_rows);
-    _gemm_output.allocator()->init(TensorInfo(shape_gemm, 1, dt, fixed_point_position));
+    const DataType gemm_data_type = _is_quantized ? DataType::S32 : dt;
+    // GEMM output should be S32 for acquiring raw integer accumulator without quantized postprocessing for quantized asymmetric input.
+    TensorInfo info_gemm(input->info()->clone()->set_is_resizable(true).reset_padding().set_tensor_shape(shape_gemm).set_data_type(gemm_data_type).set_quantization_info(
+                             output->info()->quantization_info()));
+    _gemm_output.allocator()->init(info_gemm);
     _memory_group.manage(&_gemm_output);
 
     // Configure kernels
-
     _input_im2col_kernel.set_target(CLScheduler::get().target());
-    _input_im2col_kernel.configure(input, &_input_im2col_reshaped, Size2D(kernel_width, kernel_height), conv_info, _has_bias);
+    _input_im2col_kernel.configure(input, &_input_im2col_reshaped, Size2D(kernel_width, kernel_height), conv_info, _append_bias);
 
     // Configure matrix multiply
-    if(_is_fully_connected_convolution)
+    if(run_interleaved)
     {
-        // The matrix A and Matrix B have not been reshaped
-        _mm_kernel.configure(&_input_im2col_reshaped, weights, &_gemm_output, 1.0f, false);
+        _input_interleave_kernel.configure(&_input_im2col_reshaped, &_input_interleaved_reshaped);
+        configure_mm(&_input_interleaved_reshaped, weights, &_gemm_output);
+        _input_interleaved_reshaped.allocator()->allocate();
     }
     else
     {
-        _input_interleave_kernel.configure(&_input_im2col_reshaped, &_input_interleaved_reshaped);
-        _mm_kernel.configure(&_input_interleaved_reshaped, weights, &_gemm_output, 1.0f);
-        _input_interleaved_reshaped.allocator()->allocate();
+        configure_mm(&_input_im2col_reshaped, weights, &_gemm_output, false);
     }
     _input_im2col_reshaped.allocator()->allocate();
+
+    // Configure output stage for quantized case
+    if(_is_quantized)
+    {
+        float multiplier = input->info()->quantization_info().scale * weights->info()->quantization_info().scale / output->info()->quantization_info().scale;
+        int   output_multiplier, output_shift;
+        quantization::calculate_quantized_multiplier_less_than_one(multiplier, &output_multiplier, &output_shift);
+        _gemmlowp_output_stage.configure(&_gemm_output, biases, &_tmp_output, output_multiplier, output_shift, output->info()->quantization_info().offset);
+        _gemm_output.allocator()->allocate();
+    }
+
+    // Configure Col2Im
     _output_col2im_kernel.set_target(CLScheduler::get().target());
-    _output_col2im_kernel.configure(&_gemm_output, output, std::make_pair(conv_w, conv_h));
-    _gemm_output.allocator()->allocate();
+    _output_col2im_kernel.configure(_is_quantized ? &_tmp_output : &_gemm_output, output, std::make_pair(conv_w, conv_h));
+    if(_is_quantized)
+    {
+        _tmp_output.allocator()->allocate();
+    }
+    else
+    {
+        _gemm_output.allocator()->allocate();
+    }
 
     ARM_COMPUTE_ERROR_ON_MSG((output->info()->dimension(0) != conv_w) || (output->info()->dimension(1) != conv_h), "Output shape does not match the expected one");
 
@@ -243,15 +305,30 @@ void CLConvolutionLayer::run()
 
     _memory_group.acquire();
 
-    // Run input reshaping
+    // Run im2col
     CLScheduler::get().enqueue(_input_im2col_kernel);
-    if(!_is_fully_connected_convolution)
+
+    if(!_is_fully_connected_convolution && !_is_quantized)
     {
+        // Run interleave4x4
         CLScheduler::get().enqueue(_input_interleave_kernel);
     }
 
     // Runs matrix multiply on reshaped matrices
-    CLScheduler::get().enqueue(_mm_kernel);
+    if(_is_quantized)
+    {
+        _mm_gemmlowp.run();
+    }
+    else
+    {
+        CLScheduler::get().enqueue(_mm_kernel);
+    }
+
+    // Run output stage for quantized case
+    if(_is_quantized)
+    {
+        _gemmlowp_output_stage.run();
+    }
 
     // Reshape output matrix
     CLScheduler::get().enqueue(_output_col2im_kernel, false);