COMPMID-816 - Enabled CLConvolutionLayer to use CLGEMM function instead
of CLGEMMMatrixMultiplyKernel kernel.
Change-Id: If035fa3d1fb3ff4012442bcd908c370d21aa6657
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/115990
Tested-by: Jenkins <bsgcomp@arm.com>
Reviewed-by: Pablo Tello <pablo.tello@arm.com>
Reviewed-by: Anthony Barbier <anthony.barbier@arm.com>
diff --git a/arm_compute/runtime/CL/functions/CLConvolutionLayer.h b/arm_compute/runtime/CL/functions/CLConvolutionLayer.h
index 3fe6604..f6672ce 100644
--- a/arm_compute/runtime/CL/functions/CLConvolutionLayer.h
+++ b/arm_compute/runtime/CL/functions/CLConvolutionLayer.h
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
@@ -36,6 +36,7 @@
#include "arm_compute/core/Types.h"
#include "arm_compute/runtime/CL/CLMemoryGroup.h"
#include "arm_compute/runtime/CL/CLTensor.h"
+#include "arm_compute/runtime/CL/functions/CLGEMM.h"
#include "arm_compute/runtime/CL/functions/CLGEMMLowpMatrixMultiplyCore.h"
#include "arm_compute/runtime/CL/functions/CLGEMMLowpOutputStage.h"
#include "arm_compute/runtime/IMemoryManager.h"
@@ -76,15 +77,20 @@
bool _transpose1xW;
};
-/** Basic function to compute the convolution layer. This function calls the following OpenCL kernels:
+/** Basic function to compute the convolution layer. This function calls the following OpenCL kernels/functions:
*
- * -# @ref CLWeightsReshapeKernel (executed only once for each configuration)
- * -# @ref CLGEMMTranspose1xWKernel (executed only once for each configuration)
+ * Note: weights already reshaped for quantized asymmetric is not supported
+ *
* -# @ref CLIm2ColKernel
- * -# @ref CLGEMMInterleave4x4Kernel
- * -# @ref CLGEMMMatrixMultiplyKernel or @ref CLGEMMLowpMatrixMultiplyCore (if quantized asymmetric)
+ * -# @ref CLGEMMLowpMatrixMultiplyCore (if quantized asymmetric)
* -# @ref CLGEMMLowpQuantizeDownInt32ToUint8Scale (if quantized asymmetric)
* -# @ref CLCol2ImKernel
+ *
+ * if the weights are already reshaped:
+ * -# @ref CLGEMMInterleave4x4Kernel
+ * -# @ref CLGEMMMatrixMultiplyKernel
+ * else
+ * -# @ref CLGEMM
*/
class CLConvolutionLayer : public IFunction
{
@@ -119,20 +125,21 @@
* except for input of QASYMM8 type where output should be of S32 type.
* @param is_interleaved_transposed Flag that signals if matrix is interleaved transposed
*/
- void configure_mm(const ICLTensor *input, const ICLTensor *weights, ICLTensor *output, bool is_interleaved_transposed = true);
+ void configure_mm(const ICLTensor *input, const ICLTensor *weights, ICLTensor *output, bool is_interleaved_transposed, bool are_weights_reshaped);
private:
CLMemoryGroup _memory_group;
CLConvolutionLayerReshapeWeights _reshape_weights;
- CLIm2ColKernel _input_im2col_kernel;
- CLGEMMInterleave4x4Kernel _input_interleave_kernel;
+ CLIm2ColKernel _im2col_kernel;
+ CLGEMMInterleave4x4Kernel _interleave_kernel;
CLGEMMMatrixMultiplyKernel _mm_kernel;
+ CLGEMM _mm_gemm;
CLGEMMLowpMatrixMultiplyCore _mm_gemmlowp;
CLGEMMLowpQuantizeDownInt32ToUint8ScaleByFixedPoint _gemmlowp_output_stage;
- CLCol2ImKernel _output_col2im_kernel;
+ CLCol2ImKernel _col2im_kernel;
- CLTensor _input_im2col_reshaped;
- CLTensor _input_interleaved_reshaped;
+ CLTensor _im2col_output;
+ CLTensor _interleave_output;
CLTensor _weights_reshaped;
CLTensor _weights_transposed;
CLTensor _gemm_output;
diff --git a/src/runtime/CL/functions/CLConvolutionLayer.cpp b/src/runtime/CL/functions/CLConvolutionLayer.cpp
index 2c1ddc3..d115397 100644
--- a/src/runtime/CL/functions/CLConvolutionLayer.cpp
+++ b/src/runtime/CL/functions/CLConvolutionLayer.cpp
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
@@ -87,7 +87,6 @@
{
_memory_group.acquire();
- cl::CommandQueue q = CLScheduler::get().queue();
CLScheduler::get().enqueue(_weights_reshape_kernel);
if(_transpose1xW)
{
@@ -98,33 +97,49 @@
}
CLConvolutionLayer::CLConvolutionLayer(std::shared_ptr<IMemoryManager> memory_manager)
- : _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(), _are_weights_reshaped(false), _is_quantized(false),
+ : _memory_group(memory_manager), _reshape_weights(), _im2col_kernel(), _interleave_kernel(), _mm_kernel(), _mm_gemm(memory_manager), _mm_gemmlowp(memory_manager), _gemmlowp_output_stage(),
+ _col2im_kernel(), _im2col_output(), _interleave_output(), _weights_reshaped(), _weights_transposed(), _gemm_output(), _tmp_output(), _are_weights_reshaped(false), _is_quantized(false),
_is_interleaved_transposed(false)
{
}
-void CLConvolutionLayer::configure_mm(const ICLTensor *input, const ICLTensor *weights, ICLTensor *output, bool is_interleaved_transposed)
+void CLConvolutionLayer::configure_mm(const ICLTensor *input, const ICLTensor *weights, ICLTensor *output, bool is_interleaved_transposed, bool are_weights_reshaped)
{
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();
+ if(are_weights_reshaped)
+ {
+ ARM_COMPUTE_ERROR("Weights already reshaped are not suppported with gemmlowp");
+ }
+ else
+ {
+ // 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));
+ 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*/));
+ _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);
+ // 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);
+ if(are_weights_reshaped)
+ {
+ // Configure matrix multiply kernel
+ _mm_kernel.configure(input, weights, output, 1.f, is_interleaved_transposed);
+ }
+ else
+ {
+ // Configure matrix multiply function
+ _mm_gemm.configure(input, weights, nullptr, output, 1.0f, 0.0f, GEMMInfo(false, false, true /* Reshape weights only for the first run*/));
+ }
}
}
@@ -133,6 +148,7 @@
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() && CLScheduler::get().target() == GPUTarget::BIFROST);
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()));
@@ -158,8 +174,8 @@
// Set the GPU target for matrix multiply and im2col and col2im
_mm_kernel.set_target(CLScheduler::get().target());
- _input_im2col_kernel.set_target(CLScheduler::get().target());
- _output_col2im_kernel.set_target(CLScheduler::get().target());
+ _im2col_kernel.set_target(CLScheduler::get().target());
+ _col2im_kernel.set_target(CLScheduler::get().target());
const bool append_bias = (biases != nullptr) && (!_is_quantized);
_are_weights_reshaped = weights_info.are_reshaped();
@@ -183,7 +199,7 @@
// Check if its a "fully connected" convolution
const bool is_fully_connected_convolution = ((conv_w == 1) && (conv_h == 1));
- _is_interleaved_transposed = (!is_fully_connected_convolution && !_is_quantized);
+ _is_interleaved_transposed = (!is_fully_connected_convolution) && (!_is_quantized) && (_are_weights_reshaped);
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) + bias_element;
@@ -205,8 +221,9 @@
}
else
{
- // _weights_reshaped will be auto configured in the kernel
- _reshape_weights.configure(weights, biases_to_use, &_weights_reshaped, _is_interleaved_transposed /* 1xW transpose */);
+ // _weights_reshaped will be auto configured in the kernel.
+ // Just append biases and do not transpose 1xW as it will be reshaped in CLGEMM
+ _reshape_weights.configure(weights, biases_to_use, &_weights_reshaped, false);
weights = &_weights_reshaped;
}
@@ -221,11 +238,11 @@
// FIXME: input->clone() doesn't work with subtensors for grouped convolutions.
TensorInfo im2col_reshaped_info(shape_im2col, 1, dt, input->info()->fixed_point_position());
im2col_reshaped_info.set_quantization_info(input->info()->quantization_info());
- _input_im2col_reshaped.allocator()->init(im2col_reshaped_info);
- _memory_group.manage(&_input_im2col_reshaped);
+ _im2col_output.allocator()->init(im2col_reshaped_info);
+ _memory_group.manage(&_im2col_output);
// Create GEMM output tensor
- TensorShape shape_gemm = _input_im2col_reshaped.info()->tensor_shape();
+ TensorShape shape_gemm = _im2col_output.info()->tensor_shape();
shape_gemm.set(0, mat_weights_cols);
shape_gemm.set(1, mat_input_rows);
const DataType gemm_data_type = _is_quantized ? DataType::S32 : dt;
@@ -237,24 +254,24 @@
_memory_group.manage(&_gemm_output);
// Configure im2col
- _input_im2col_kernel.configure(input, &_input_im2col_reshaped, Size2D(kernel_width, kernel_height), conv_info, append_bias);
+ _im2col_kernel.configure(input, &_im2col_output, Size2D(kernel_width, kernel_height), conv_info, append_bias);
// Configure matrix multiply
if(_is_interleaved_transposed)
{
// Configure GEMMInterleave4x4. _input_interleaved_reshaped will be auto configured in the kernel
- _input_interleave_kernel.configure(&_input_im2col_reshaped, &_input_interleaved_reshaped);
- _memory_group.manage(&_input_interleaved_reshaped);
+ _interleave_kernel.configure(&_im2col_output, &_interleave_output);
+ _memory_group.manage(&_interleave_output);
// Configure GEMM
- configure_mm(&_input_interleaved_reshaped, weights, &_gemm_output);
- _input_interleaved_reshaped.allocator()->allocate();
+ configure_mm(&_interleave_output, weights, &_gemm_output, true, _are_weights_reshaped);
+ _interleave_output.allocator()->allocate();
}
else
{
- configure_mm(&_input_im2col_reshaped, weights, &_gemm_output, false);
+ configure_mm(&_im2col_output, weights, &_gemm_output, false, _are_weights_reshaped);
}
- _input_im2col_reshaped.allocator()->allocate();
+ _im2col_output.allocator()->allocate();
// Configure output stage for quantized case
if(_is_quantized)
@@ -267,7 +284,7 @@
}
// Configure Col2Im
- _output_col2im_kernel.configure(_is_quantized ? &_tmp_output : &_gemm_output, output, std::make_pair(conv_w, conv_h));
+ _col2im_kernel.configure(_is_quantized ? &_tmp_output : &_gemm_output, output, std::make_pair(conv_w, conv_h));
if(_is_quantized)
{
_tmp_output.allocator()->allocate();
@@ -298,32 +315,39 @@
_memory_group.acquire();
// Run im2col
- CLScheduler::get().enqueue(_input_im2col_kernel);
+ CLScheduler::get().enqueue(_im2col_kernel);
+ // Note: _is_interleaved_transposed is true only if the weights passed to the function have been passed already reshaped
+ // and if we do not have QASYMM8 data type. If this flag is true, we need to run the
+ // gemm kernel instead of gemm function
if(_is_interleaved_transposed)
{
- // Run interleave4x4
- CLScheduler::get().enqueue(_input_interleave_kernel);
- }
+ // Run interleave4x4 kernel
+ CLScheduler::get().enqueue(_interleave_kernel);
- // Runs matrix multiply on reshaped matrices
- if(_is_quantized)
- {
- _mm_gemmlowp.run();
+ // Run matrix multiply kernel
+ CLScheduler::get().enqueue(_mm_kernel);
}
else
{
- CLScheduler::get().enqueue(_mm_kernel);
- }
+ // Runs CLGEMM or CLGEMMLowpMatrixMultiplyCore functions
+ if(_is_quantized)
+ {
+ // Run gemmlowp
+ _mm_gemmlowp.run();
- // Run output stage for quantized case
- if(_is_quantized)
- {
- _gemmlowp_output_stage.run();
+ // Run output stage
+ _gemmlowp_output_stage.run();
+ }
+ else
+ {
+ // Run gemm
+ _mm_gemm.run();
+ }
}
// Reshape output matrix
- CLScheduler::get().enqueue(_output_col2im_kernel, false);
+ CLScheduler::get().enqueue(_col2im_kernel, false);
_memory_group.release();
}
diff --git a/src/runtime/CL/functions/CLGEMM.cpp b/src/runtime/CL/functions/CLGEMM.cpp
index 1657bdc..c676a10 100644
--- a/src/runtime/CL/functions/CLGEMM.cpp
+++ b/src/runtime/CL/functions/CLGEMM.cpp
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2017, 2018 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
@@ -80,7 +80,7 @@
matrix_a = &_tmp_a;
matrix_b = &_tmp_b;
- // _tmp_a and _tmp_n will be auto configured in _interleave_kernel and in _transpose_kernel
+ // _tmp_a and _tmp_b will be auto configured in _interleave_kernel and in _transpose_kernel
// Configure interleave kernel
_interleave_kernel.configure(a, &_tmp_a);
diff --git a/tests/validation/CL/ConvolutionLayer.cpp b/tests/validation/CL/ConvolutionLayer.cpp
index 56e10f0..8b9db91 100644
--- a/tests/validation/CL/ConvolutionLayer.cpp
+++ b/tests/validation/CL/ConvolutionLayer.cpp
@@ -1,5 +1,5 @@
/*
- * Copyright (c) 2017 ARM Limited.
+ * Copyright (c) 2017-2018 ARM Limited.
*
* SPDX-License-Identifier: MIT
*
@@ -114,7 +114,7 @@
TEST_SUITE(Float)
TEST_SUITE(FP16)
FIXTURE_DATA_TEST_CASE(RunSmall, CLConvolutionLayerFixture<half>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallConvolutionLayerDataset(),
- framework::dataset::make("ReshapeWeights", { true, false })),
+ framework::dataset::make("ReshapeWeights", { true })),
framework::dataset::make("DataType",
DataType::F16)))
{
@@ -122,7 +122,7 @@
validate(CLAccessor(_target), _reference, tolerance_f16, tolerance_num);
}
FIXTURE_DATA_TEST_CASE(RunLarge, CLConvolutionLayerFixture<half>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeConvolutionLayerDataset(),
- framework::dataset::make("ReshapeWeights", { true, false })),
+ framework::dataset::make("ReshapeWeights", { true })),
framework::dataset::make("DataType",
DataType::F16)))
{
@@ -133,7 +133,7 @@
TEST_SUITE(FP32)
FIXTURE_DATA_TEST_CASE(RunSmall, CLConvolutionLayerFixture<float>, framework::DatasetMode::PRECOMMIT, combine(combine(datasets::SmallConvolutionLayerDataset(),
- framework::dataset::make("ReshapeWeights", { true, false })),
+ framework::dataset::make("ReshapeWeights", { true })),
framework::dataset::make("DataType",
DataType::F32)))
{
@@ -141,7 +141,7 @@
validate(CLAccessor(_target), _reference, tolerance_f32);
}
FIXTURE_DATA_TEST_CASE(RunLarge, CLConvolutionLayerFixture<float>, framework::DatasetMode::NIGHTLY, combine(combine(datasets::LargeConvolutionLayerDataset(),
- framework::dataset::make("ReshapeWeights", { true, false })),
+ framework::dataset::make("ReshapeWeights", { true })),
framework::dataset::make("DataType",
DataType::F32)))
{
@@ -158,7 +158,7 @@
TEST_SUITE(QS8)
// We test for fixed point precision [4,6]
FIXTURE_DATA_TEST_CASE(RunSmall, CLConvolutionLayerFixedPointFixture<int8_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallConvolutionLayerDataset(),
- framework::dataset::make("ReshapeWeights", { true, false })),
+ framework::dataset::make("ReshapeWeights", { true })),
framework::dataset::make("DataType",
DataType::QS8)),
framework::dataset::make("FractionalBits", 4, 7)))
@@ -167,7 +167,7 @@
validate(CLAccessor(_target), _reference, tolerance_fixed);
}
FIXTURE_DATA_TEST_CASE(RunLarge, CLConvolutionLayerFixedPointFixture<int8_t>, framework::DatasetMode::NIGHTLY, combine(combine(combine(datasets::LargeConvolutionLayerDataset(),
- framework::dataset::make("ReshapeWeights", { true, false })),
+ framework::dataset::make("ReshapeWeights", { true })),
framework::dataset::make("DataType",
DataType::QS8)),
framework::dataset::make("FractionalBits", 4, 7)))
@@ -180,7 +180,7 @@
TEST_SUITE(QS16)
// Testing for fixed point position [1,14)
FIXTURE_DATA_TEST_CASE(RunSmall, CLConvolutionLayerFixedPointFixture<int16_t>, framework::DatasetMode::PRECOMMIT, combine(combine(combine(datasets::SmallConvolutionLayerDataset(),
- framework::dataset::make("ReshapeWeights", { true, false })),
+ framework::dataset::make("ReshapeWeights", { true })),
framework::dataset::make("DataType",
DataType::QS16)),
framework::dataset::make("FractionalBits", 1, 14)))
@@ -189,7 +189,7 @@
validate(CLAccessor(_target), _reference, tolerance_fixed);
}
FIXTURE_DATA_TEST_CASE(RunLarge, CLConvolutionLayerFixedPointFixture<int16_t>, framework::DatasetMode::NIGHTLY, combine(combine(combine(datasets::LargeConvolutionLayerDataset(),
- framework::dataset::make("ReshapeWeights", { true, false })),
+ framework::dataset::make("ReshapeWeights", { true })),
framework::dataset::make("DataType",
DataType::QS16)),
framework::dataset::make("FractionalBits", 1, 14)))