COMPMID-420, COMPMID-414 - Port CLConvolutionLayer and CLFullyConnectedLayer to use 8 bit fixed point
Change-Id: I1cb1b4d7711ad7b569ee691e13a5df1b3430292b
Reviewed-on: http://mpd-gerrit.cambridge.arm.com/79565
Tested-by: Kaizen <jeremy.johnson+kaizengerrit@arm.com>
Reviewed-by: Georgios Pinitas <georgios.pinitas@arm.com>
diff --git a/arm_compute/core/CL/kernels/CLCol2ImKernel.h b/arm_compute/core/CL/kernels/CLCol2ImKernel.h
index d391cac..63b0b63 100644
--- a/arm_compute/core/CL/kernels/CLCol2ImKernel.h
+++ b/arm_compute/core/CL/kernels/CLCol2ImKernel.h
@@ -66,7 +66,7 @@
/** Set the input and output of the kernel.
*
- * @param[in] input The input tensor to convert. Data types supported: F16/F32
+ * @param[in] input The input tensor to convert. Data types supported: QS8/F16/F32
* @param[out] output The output tensor. 3 lower dimensions represent a single output [width, height, OFM],
* while the rest represent batch of outputs. Data types supported: Same as @p input
* @param[in] convolved_dims Output convolved dimensions.
diff --git a/arm_compute/core/CL/kernels/CLIm2ColKernel.h b/arm_compute/core/CL/kernels/CLIm2ColKernel.h
index b3b5cd8..e9f1a3f 100644
--- a/arm_compute/core/CL/kernels/CLIm2ColKernel.h
+++ b/arm_compute/core/CL/kernels/CLIm2ColKernel.h
@@ -69,7 +69,7 @@
/** Set the input and output of the kernel.
*
* @param[in] input The input tensor to convert. 3 lower dimensions represent a single input [width, height, IFM],
- * while every optional dimension from 4 and above represent a batch of inputs. Data types supported: F16/F32
+ * while every optional dimension from 4 and above represent a batch of inputs. Data types supported: QS8/F16/F32
* @param[out] output The output tensor. First 2 lower dimensions represent a transform of each 3D input,
* while every dimension above represents a batch. Data types supported: Same as @p input
* @param[in] kernel_dims The kernel dimensions (width and height).
diff --git a/arm_compute/runtime/CL/functions/CLConvolutionLayer.h b/arm_compute/runtime/CL/functions/CLConvolutionLayer.h
index 8030b40..50a7dc9 100644
--- a/arm_compute/runtime/CL/functions/CLConvolutionLayer.h
+++ b/arm_compute/runtime/CL/functions/CLConvolutionLayer.h
@@ -88,7 +88,7 @@
*
* @param[in] input Source tensor. 3 lower dimensions represent a single input [width, height, IFM],
* while every optional dimension from 4 and above represent a batch of inputs.
- * Data types supported: F16, F32.
+ * Data types supported: QS8/F16/F32.
* @param[in] weights Weights tensor. Weights are 4D tensor with dimensions [kernel_x, kernel_y, IFM, OFM]. Data type supported:Same as @p input.
* @param[in] biases Biases tensor. Shared biases supported. Biases are 1D tensor with dimensions [OFM]. Data type supported:Same as @p input.
* @param[out] output Destination tensor. 3 lower dimensions represent a single output [width, height, OFM], while the rest represent batch of outputs.
diff --git a/arm_compute/runtime/CL/functions/CLFullyConnectedLayer.h b/arm_compute/runtime/CL/functions/CLFullyConnectedLayer.h
index 826f445..807ff69 100644
--- a/arm_compute/runtime/CL/functions/CLFullyConnectedLayer.h
+++ b/arm_compute/runtime/CL/functions/CLFullyConnectedLayer.h
@@ -50,7 +50,7 @@
CLFullyConnectedLayerReshapeWeights();
/** Set the input and output tensors.
*
- * @param[in] input Weights tensor. The weights must be 2 dimensional. Data types supported: QS8/F32.
+ * @param[in] input Weights tensor. The weights must be 2 dimensional. Data types supported: QS8/F16/F32.
* @param[out] output Destination tensor. Data type supported: Same as @p input.
* @param[in] transpose_weights True if the weights must be transposed. Data types supported: Same as @p weights.
* @param[in] is_batched_fc_layer True if it is a batched fully connected layer
@@ -85,7 +85,7 @@
CLFullyConnectedLayer();
/** Set the input and output tensors.
*
- * @param[in] input Source tensor. Data type supported: F16/F32.
+ * @param[in] input Source tensor. Data type supported: QS8/F16/F32.
* @param[in] weights Weights tensor. The weights must be 2 dimensional. Data type supported: Same as @p input
* @param[in] biases Bias tensor. It can be nullptr. Data type supported:Same as @p input.
* @param[out] output Destination tensor. Data type supported: Same as @p input.
diff --git a/src/core/CL/cl_kernels/convolution_layer.cl b/src/core/CL/cl_kernels/convolution_layer.cl
index a5cbe3d..a875911 100644
--- a/src/core/CL/cl_kernels/convolution_layer.cl
+++ b/src/core/CL/cl_kernels/convolution_layer.cl
@@ -21,6 +21,7 @@
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
+#include "fixed_point.h"
#include "helpers.h"
/** This kernel reshapes the tensor's low three dimensions to single column
@@ -99,7 +100,7 @@
* @note The data type must be passed at compile time using -DDATA_TYPE: e.g. -DDATA_TYPE=float
* @note In case biases will be added to the convolution -DHAS_BIAS has to be passed to append the final matrix with 1 in each row.
*
- * @param[in] src_ptr Pointer to the source tensor. Supported data types: F16/F32
+ * @param[in] src_ptr Pointer to the source tensor. Supported data types: QS8/F16/F32
* @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
@@ -148,17 +149,21 @@
}
}
-#if defined(HAS_BIAS)
- *((__global DATA_TYPE *)output_ptr) = (DATA_TYPE)1;
-#endif /* HAS_BIAS */
+#ifdef HAS_BIAS
+#ifdef FIXED_POINT_POSITION
+ *((__global DATA_TYPE *)output_ptr) = (DATA_TYPE)(1 << FIXED_POINT_POSITION);
+#else // FIXED_POINT_POSITION
+ *((__global DATA_TYPE *)output_ptr) = 1.0f;
+#endif // FIXED_POINT_POSITION
+#endif // HAS_BIAS
}
-#endif //(CONVOLVED_WIDTH && STRIDE_X && STRIDE_Y && PAD_X && PAD_Y && KERNEL_WIDTH && KERNEL_HEIGHT && KERNEL_DEPTH && SRC_WIDTH && SRC_HEIGHT)
+#endif //defined(CONVOLVED_WIDTH) && defined(STRIDE_X) && defined(STRIDE_Y) && defined(PAD_X) && defined(PAD_Y) && defined(KERNEL_WIDTH) && defined(KERNEL_HEIGHT) && defined(KERNEL_DEPTH) && defined(SRC_WIDTH) && defined(SRC_HEIGHT)
/** This kernel performs a reshaping of the output of the convolution layer.
*
* @note The data type must be passed at compile time using -DDATA_TYPE: e.g. -DDATA_TYPE=float
*
- * @param[in] src_ptr Pointer to the source tensor. Supported data types: F16/F32
+ * @param[in] src_ptr Pointer to the source tensor. Supported data types: QS8/F16/F32
* @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
@@ -192,7 +197,7 @@
* @note Datatype should be given as a preprocessor argument using -DDATA_TYPE=type. e.g. -DDATA_TYPE=float
* @note In case biases will be added in late stage, -DHAS_BIAS has to be passed to append the final matrix with 1 in each row.
*
- * @param[in] src_ptr Pointer to the source tensor. Supported data types: F16/F32
+ * @param[in] src_ptr Pointer to the source tensor. Supported data types: QS8/F16/F32
* @param[in] src_stride_x Stride of the source tensor in X dimension (in bytes)
* @param[in] src_step_x src_stride_x * number of elements along X processed per workitem(in bytes)
* @param[in] src_stride_y Stride of the source tensor in Y dimension (in bytes)
@@ -225,7 +230,11 @@
if(get_global_id(0) == (get_global_size(0) - 1) && get_global_id(1) == (get_global_size(1) - 1) && get_global_id(2) == (get_global_size(2) - 1))
{
tmp_out_ptr += dst_stride_x;
+#ifdef FIXED_POINT_POSITION
+ *((__global DATA_TYPE *)tmp_out_ptr) = (DATA_TYPE)(1 << FIXED_POINT_POSITION);
+#else // FIXED_POINT_POSITION
*((__global DATA_TYPE *)tmp_out_ptr) = (DATA_TYPE)1;
+#endif // FIXED_POINT_POSITION
}
-#endif /* HAS_BIAS */
+#endif // HAS_BIAS
}
diff --git a/src/core/CL/cl_kernels/gemm.cl b/src/core/CL/cl_kernels/gemm.cl
index 46f1645..db15720 100644
--- a/src/core/CL/cl_kernels/gemm.cl
+++ b/src/core/CL/cl_kernels/gemm.cl
@@ -21,9 +21,12 @@
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
-#include "fixed_point.h"
#include "helpers.h"
+#ifdef FIXED_POINT_POSITION
+#include "fixed_point.h"
+#endif // FIXED_POINT_POSITION
+
/** This OpenCL kernel computes the "vector" 1x4 transposition of input matrix
*
* @param[in] src_ptr Pointer to the source matrix. Supported data types: U32/S32/F32
@@ -274,7 +277,11 @@
accum_value = vload16(0, (__global DATA_TYPE *)accum.ptr);
VEC_DATA_TYPE(DATA_TYPE, 16)
biases_value = vload16(0, (__global DATA_TYPE *)biases.ptr);
- accum_value = biases_value + accum_value;
+#ifdef FIXED_POINT_POSITION
+ accum_value = ADD_SAT_OP_EXPAND(biases_value, accum_value, DATA_TYPE, 16);
+#else // FIXED_POINT_POSITION
+ accum_value = biases_value + accum_value;
+#endif // FIXED_POINT_POSITION
// Store result in the accummulate buffer
vstore16(accum_value, 0, (__global DATA_TYPE *)accum.ptr);
diff --git a/src/core/CL/kernels/CLCol2ImKernel.cpp b/src/core/CL/kernels/CLCol2ImKernel.cpp
index 679943b..6b2a18b 100644
--- a/src/core/CL/kernels/CLCol2ImKernel.cpp
+++ b/src/core/CL/kernels/CLCol2ImKernel.cpp
@@ -43,9 +43,9 @@
void CLCol2ImKernel::configure(const ICLTensor *input, ICLTensor *output, std::pair<unsigned int, unsigned int> convolved_dims)
{
- ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F32);
- ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(output, 1, DataType::F32);
+ ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QS8, DataType::F16, DataType::F32);
ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
+ ARM_COMPUTE_ERROR_ON_MISMATCHING_FIXED_POINT(input, output);
_input = input;
_output = output;
diff --git a/src/core/CL/kernels/CLGEMMMatrixAccumulateBiasesKernel.cpp b/src/core/CL/kernels/CLGEMMMatrixAccumulateBiasesKernel.cpp
index 75c1a6e..a7ca6f2 100644
--- a/src/core/CL/kernels/CLGEMMMatrixAccumulateBiasesKernel.cpp
+++ b/src/core/CL/kernels/CLGEMMMatrixAccumulateBiasesKernel.cpp
@@ -53,6 +53,10 @@
std::set<std::string> build_opts;
build_opts.insert(("-DDATA_TYPE=" + get_cl_type_from_data_type(accum->info()->data_type())));
+ if(accum->info()->data_type() == DataType::QS8)
+ {
+ build_opts.emplace("-DFIXED_POINT_POSITION=" + support::cpp11::to_string(accum->info()->fixed_point_position()));
+ }
// Create kernel
_kernel = static_cast<cl::Kernel>(CLKernelLibrary::get().create_kernel("gemm_accumulate_biases", build_opts));
diff --git a/src/core/CL/kernels/CLGEMMTranspose1xWKernel.cpp b/src/core/CL/kernels/CLGEMMTranspose1xWKernel.cpp
index 73c8429..27b215f 100644
--- a/src/core/CL/kernels/CLGEMMTranspose1xWKernel.cpp
+++ b/src/core/CL/kernels/CLGEMMTranspose1xWKernel.cpp
@@ -56,7 +56,7 @@
ARM_COMPUTE_ERROR_ON_MISMATCHING_DIMENSIONS(output->info()->tensor_shape(), output_shape);
ARM_COMPUTE_ERROR_ON_MISMATCHING_FIXED_POINT(input, output);
- const unsigned int num_elems_processed_per_iteration = max_cl_vector_width / data_size_from_type(input->info()->data_type());
+ const unsigned int num_elems_processed_per_iteration = 16 / input->info()->element_size();
const float scale_x = num_elems_processed_per_iteration;
ARM_COMPUTE_ERROR_ON((0 == static_cast<int>(input->info()->dimension(0) * (1.f / scale_x))));
diff --git a/src/core/CL/kernels/CLIm2ColKernel.cpp b/src/core/CL/kernels/CLIm2ColKernel.cpp
index 092f495..51922e0 100644
--- a/src/core/CL/kernels/CLIm2ColKernel.cpp
+++ b/src/core/CL/kernels/CLIm2ColKernel.cpp
@@ -46,8 +46,9 @@
void CLIm2ColKernel::configure(const ICLTensor *input, ICLTensor *output, const Size2D &kernel_dims, const PadStrideInfo &conv_info, bool has_bias)
{
- ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F16, DataType::F32);
+ ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QS8, DataType::F16, DataType::F32);
ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, output);
+ ARM_COMPUTE_ERROR_ON_MISMATCHING_FIXED_POINT(input, output);
_input = input;
_output = output;
@@ -57,6 +58,11 @@
build_opts.emplace(("-DDATA_TYPE=" + get_cl_type_from_data_type(input->info()->data_type())));
build_opts.emplace((has_bias ? "-DHAS_BIAS" : ""));
+ if(input->info()->data_type() == DataType::QS8)
+ {
+ build_opts.emplace("-DFIXED_POINT_POSITION=" + support::cpp11::to_string(input->info()->fixed_point_position()));
+ }
+
int pad_x = 0;
int pad_y = 0;
int stride_x = 0;
diff --git a/src/runtime/CL/functions/CLConvolutionLayer.cpp b/src/runtime/CL/functions/CLConvolutionLayer.cpp
index b29bf8f..96d04dc 100644
--- a/src/runtime/CL/functions/CLConvolutionLayer.cpp
+++ b/src/runtime/CL/functions/CLConvolutionLayer.cpp
@@ -41,7 +41,7 @@
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::F16, DataType::F32);
+ ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(weights, 1, DataType::QS8, 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);
@@ -63,8 +63,9 @@
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);
TensorShape shape_wr(mat_weights_cols, mat_weights_rows);
- const DataType dt = weights->info()->data_type();
- TensorInfo info_wr(shape_wr, 1, dt);
+ const DataType dt = weights->info()->data_type();
+ const int fixed_point_position = weights->info()->fixed_point_position();
+ TensorInfo info_wr(shape_wr, 1, dt, fixed_point_position);
_weights_reshaped.allocator()->init(info_wr);
_weights_reshape_kernel.configure(weights, biases, &_weights_reshaped);
@@ -95,23 +96,27 @@
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::F16, DataType::F32);
+ ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QS8, DataType::F16, DataType::F32);
ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, weights, output);
+ ARM_COMPUTE_ERROR_ON_MISMATCHING_FIXED_POINT(input, weights, output);
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);
if(biases != nullptr)
{
- ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(biases, 1, DataType::F16, DataType::F32);
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();
+
_has_bias = (biases != nullptr);
_are_weights_reshaped = weights_info.are_reshaped();
- // Get parameters for conv_info
+ // Get parameters from conv_info
unsigned int stride_x = 0;
unsigned int stride_y = 0;
unsigned int pad_x = 0;
@@ -123,8 +128,8 @@
unsigned int conv_w = 0;
unsigned int conv_h = 0;
- const unsigned int kernel_width = _are_weights_reshaped ? weights_info.kernel_size().first : weights->info()->dimension(0);
- const unsigned int kernel_height = _are_weights_reshaped ? weights_info.kernel_size().second : weights->info()->dimension(1);
+ const unsigned int kernel_width = (_are_weights_reshaped) ? weights_info.kernel_size().first : weights->info()->dimension(0);
+ const unsigned int kernel_height = (_are_weights_reshaped) ? weights_info.kernel_size().second : weights->info()->dimension(1);
std::tie(conv_w, conv_h) = scaled_dimensions(input->info()->dimension(0), input->info()->dimension(1), kernel_width, kernel_height,
conv_info);
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");
@@ -132,9 +137,10 @@
// Check if its a "fully connected" convolution
_is_fully_connected_convolution = ((conv_w == 1) && (conv_h == 1));
- // Create tensor to store the reshaped weights
- size_t mat_weights_cols = weights->info()->dimension(3);
- size_t mat_weights_rows = weights->info()->dimension(0) * weights->info()->dimension(1) * weights->info()->dimension(2) + ((_has_bias) ? 1 : 0);
+ 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);
+
+ // Reshape weights if needed
if(_are_weights_reshaped)
{
mat_weights_cols = output->info()->dimension(2);
@@ -147,49 +153,48 @@
{
// Create tensor to store the reshaped weights
TensorShape shape_wr(mat_weights_cols, mat_weights_rows);
- TensorInfo info_wr(shape_wr, 1, weights->info()->data_type());
+ TensorInfo info_wr(shape_wr, 1, dt, fixed_point_position);
_weights_reshaped.allocator()->init(info_wr);
- _reshape_weights.configure(weights, biases, &_weights_reshaped, false);
- weights = &_weights_reshaped;
+ _reshape_weights.configure(weights, biases, &_weights_reshaped, false /* 1xW transpose */);
}
else
{
// Create tensor to store transposed weights
- TensorShape shape_wt(mat_weights_rows * 4, static_cast<size_t>(std::ceil(mat_weights_cols / 4.f)));
- TensorInfo info_wt(shape_wt, 1, weights->info()->data_type());
- _weights_transposed.allocator()->init(info_wt);
- _reshape_weights.configure(weights, biases, &_weights_transposed, true);
- weights = &_weights_transposed;
+ 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 = &_weights_reshaped;
}
+
// Create tensor to store im2col reshaped inputs
- const size_t mat_input_cols = mat_weights_rows;
- const size_t mat_input_rows = conv_w * conv_h;
- TensorShape shape_im2col = input->info()->tensor_shape();
+ const unsigned int mat_input_cols = mat_weights_rows;
+ const unsigned int mat_input_rows = conv_w * conv_h;
+ TensorShape shape_im2col = input->info()->tensor_shape();
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, input->info()->data_type()));
+ _input_im2col_reshaped.allocator()->init(TensorInfo(shape_im2col, 1, dt, fixed_point_position));
// Create tensor (interleave) to prepare input tensor for GEMM
if(!_is_fully_connected_convolution)
{
TensorShape shape_interleaved = shape_im2col;
shape_interleaved.set(0, shape_interleaved.x() * 4);
- shape_interleaved.set(1, std::ceil(static_cast<float>(shape_interleaved.y()) / 4.f));
- _input_interleaved_reshaped.allocator()->init(TensorInfo(shape_interleaved, 1, input->info()->data_type()));
+ shape_interleaved.set(1, std::ceil(shape_interleaved.y() / 4.f));
+ _input_interleaved_reshaped.allocator()->init(TensorInfo(shape_interleaved, 1, dt, fixed_point_position));
}
// Create GEMM output tensor
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, input->info()->data_type()));
+ _gemm_output.allocator()->init(TensorInfo(shape_gemm, 1, dt, fixed_point_position));
// Configure kernels
_input_im2col_kernel.configure(input, &_input_im2col_reshaped, Size2D(kernel_width, kernel_height), conv_info, _has_bias);
- _output_col2im_kernel.configure(&_gemm_output, output, std::make_pair(conv_w, conv_h));
-
if(_is_fully_connected_convolution)
{
_mm_kernel.configure(&_input_im2col_reshaped, weights, &_gemm_output, 1.0f);
@@ -199,19 +204,13 @@
_input_interleave_kernel.configure(&_input_im2col_reshaped, &_input_interleaved_reshaped);
_mm_kernel.configure(&_input_interleaved_reshaped, weights, &_gemm_output, 1.0f);
}
+ _output_col2im_kernel.configure(&_gemm_output, output, std::make_pair(conv_w, conv_h));
+ // Allocate intermediate tensor
if(!_are_weights_reshaped)
{
- if(!_is_fully_connected_convolution)
- {
- _weights_transposed.allocator()->allocate();
- }
- else
- {
- _weights_reshaped.allocator()->allocate();
- }
+ _weights_reshaped.allocator()->allocate();
}
-
_input_im2col_reshaped.allocator()->allocate();
if(!_is_fully_connected_convolution)
{
diff --git a/src/runtime/CL/functions/CLFullyConnectedLayer.cpp b/src/runtime/CL/functions/CLFullyConnectedLayer.cpp
index b51e709..11e670c 100644
--- a/src/runtime/CL/functions/CLFullyConnectedLayer.cpp
+++ b/src/runtime/CL/functions/CLFullyConnectedLayer.cpp
@@ -39,7 +39,7 @@
void CLFullyConnectedLayerReshapeWeights::configure(const ICLTensor *input, ICLTensor *output, bool transpose_weights, bool is_batched_fc_layer)
{
- ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QS8, DataType::F32);
+ ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QS8, DataType::F16, DataType::F32);
ARM_COMPUTE_ERROR_ON(output == nullptr);
ARM_COMPUTE_ERROR_ON(input->info()->num_dimensions() != 2);
ARM_COMPUTE_ERROR_ON((transpose_weights == false) && (is_batched_fc_layer == false));
@@ -196,8 +196,7 @@
void CLFullyConnectedLayer::configure(const ICLTensor *input, const ICLTensor *weights, const ICLTensor *biases, ICLTensor *output, bool transpose_weights, bool are_weights_reshaped)
{
- ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::F32);
- ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(weights, 1, DataType::F32);
+ ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input, 1, DataType::QS8, DataType::F16, DataType::F32);
ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(input, weights, output);
ARM_COMPUTE_ERROR_ON(weights->info()->num_dimensions() != 2);
diff --git a/tests/validation/CL/ConvolutionLayer.cpp b/tests/validation/CL/ConvolutionLayer.cpp
index fb06dd4..f613f77 100644
--- a/tests/validation/CL/ConvolutionLayer.cpp
+++ b/tests/validation/CL/ConvolutionLayer.cpp
@@ -47,6 +47,7 @@
namespace
{
const float tolerance_f32 = 1e-03f; /**< Tolerance value for comparing reference's output against implementation's output for DataType::F32 */
+const float tolerance_qs8 = 1.0f; /**< Tolerance value for comparing reference's output against implementation's output for DataType::QS8 */
CLTensor compute_convolution_layer(const TensorShape &input_shape, const TensorShape &weights_shape, const TensorShape &bias_shape, const TensorShape &output_shape, DataType dt,
const PadStrideInfo &conv_info, int fixed_point_position)
@@ -101,7 +102,7 @@
BOOST_TEST_DECORATOR(*boost::unit_test::label("precommit") * boost::unit_test::label("nightly"))
BOOST_DATA_TEST_CASE(Configuration,
- AlexNetConvolutionLayerDataset() * boost::unit_test::data::make({ DataType::F32 }),
+ AlexNetConvolutionLayerDataset() * boost::unit_test::data::make({ DataType::F32, DataType::QS8 }),
conv_set, dt)
{
// Set fixed point position data type allowed
@@ -185,6 +186,38 @@
}
BOOST_AUTO_TEST_SUITE_END()
+BOOST_AUTO_TEST_SUITE(Quantized)
+BOOST_TEST_DECORATOR(*boost::unit_test::label("precommit"))
+BOOST_DATA_TEST_CASE(SmallConvolutionLayer,
+ SmallConvolutionLayerDataset() * boost::unit_test::data::make(DataType::QS8) * boost::unit_test::data::xrange(4, 7),
+ conv_set, dt, fixed_point_position)
+{
+ // Compute function
+ CLTensor dst = compute_convolution_layer(conv_set.src_shape, conv_set.weights_shape, conv_set.bias_shape, conv_set.dst_shape, dt, conv_set.info, fixed_point_position);
+
+ // Compute reference
+ RawTensor ref_dst = Reference::compute_reference_convolution_layer(conv_set.src_shape, conv_set.weights_shape, conv_set.bias_shape, conv_set.dst_shape, dt, conv_set.info, fixed_point_position);
+
+ // Validate output
+ validate(CLAccessor(dst), ref_dst, tolerance_qs8);
+}
+
+BOOST_TEST_DECORATOR(*boost::unit_test::label("nightly"))
+BOOST_DATA_TEST_CASE(LargeConvolutionLayer,
+ AlexNetConvolutionLayerDataset() * boost::unit_test::data::make(DataType::QS8) * boost::unit_test::data::xrange(4, 7),
+ conv_set, dt, fixed_point_position)
+{
+ // Compute function
+ CLTensor dst = compute_convolution_layer(conv_set.src_shape, conv_set.weights_shape, conv_set.bias_shape, conv_set.dst_shape, dt, conv_set.info, fixed_point_position);
+
+ // Compute reference
+ RawTensor ref_dst = Reference::compute_reference_convolution_layer(conv_set.src_shape, conv_set.weights_shape, conv_set.bias_shape, conv_set.dst_shape, dt, conv_set.info, fixed_point_position);
+
+ // Validate output
+ validate(CLAccessor(dst), ref_dst, tolerance_qs8);
+}
+BOOST_AUTO_TEST_SUITE_END()
+
BOOST_AUTO_TEST_SUITE_END()
BOOST_AUTO_TEST_SUITE_END()
BOOST_AUTO_TEST_SUITE_END()
diff --git a/tests/validation/CL/FullyConnectedLayer.cpp b/tests/validation/CL/FullyConnectedLayer.cpp
new file mode 100644
index 0000000..4d00c30
--- /dev/null
+++ b/tests/validation/CL/FullyConnectedLayer.cpp
@@ -0,0 +1,222 @@
+/*
+ * Copyright (c) 2017 ARM Limited.
+ *
+ * SPDX-License-Identifier: MIT
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a copy
+ * of this software and associated documentation files (the "Software"), to
+ * deal in the Software without restriction, including without limitation the
+ * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or
+ * sell copies of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in all
+ * copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+ * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+ * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include "CL/CLAccessor.h"
+#include "TypePrinter.h"
+#include "dataset/FullyConnectedLayerDataset.h"
+#include "tests/Globals.h"
+#include "tests/Utils.h"
+#include "validation/Datasets.h"
+#include "validation/Reference.h"
+#include "validation/Validation.h"
+
+#include "arm_compute/core/Error.h"
+#include "arm_compute/core/Helpers.h"
+#include "arm_compute/runtime/CL/functions/CLFullyConnectedLayer.h"
+
+#include <random>
+
+using namespace arm_compute;
+using namespace arm_compute::test;
+using namespace arm_compute::test::cl;
+using namespace arm_compute::test::validation;
+
+namespace
+{
+const float tolerance_f32 = 1e-03f; /**< Tolerance value for comparing reference's output against implementation's output for DataType::F32 */
+const float tolerance_qs8 = 1.0f; /**< Tolerance value for comparing reference's output against implementation's output for DataType::QS8 */
+
+CLTensor compute_fully_connected_layer(const TensorShape &input_shape, const TensorShape &weights_shape, const TensorShape &bias_shape, const TensorShape &output_shape, DataType dt,
+ bool transpose_weights, int fixed_point_position)
+{
+ // Create tensors
+ CLTensor src = create_tensor<CLTensor>(input_shape, dt, 1, fixed_point_position);
+ CLTensor bias = create_tensor<CLTensor>(bias_shape, dt, 1, fixed_point_position);
+ CLTensor dst = create_tensor<CLTensor>(output_shape, dt, 1, fixed_point_position);
+
+ // Swap the first and second dimension of weights' shape if transpose_weights is true
+ TensorShape ws = weights_shape;
+ if(transpose_weights)
+ {
+ const size_t dimx = ws.x();
+ ws.set(0, ws.y());
+ ws.set(1, dimx);
+ }
+
+ CLTensor weights = create_tensor<CLTensor>(ws, dt, 1, fixed_point_position);
+
+ // Create and configure function.
+ // Note: We pass the weights already transposed
+ CLFullyConnectedLayer fc;
+ fc.configure(&src, &weights, &bias, &dst, false);
+
+ // Allocate tensors
+ src.allocator()->allocate();
+ weights.allocator()->allocate();
+ bias.allocator()->allocate();
+ dst.allocator()->allocate();
+
+ BOOST_TEST(!src.info()->is_resizable());
+ BOOST_TEST(!weights.info()->is_resizable());
+ BOOST_TEST(!bias.info()->is_resizable());
+ BOOST_TEST(!dst.info()->is_resizable());
+
+ // Fill tensors
+ if(dt == DataType::F32)
+ {
+ std::uniform_real_distribution<> distribution(-1.0f, 1.0f);
+ library->fill(CLAccessor(src), distribution, 0);
+ library->fill(CLAccessor(weights), distribution, 1);
+ library->fill(CLAccessor(bias), distribution, 2);
+ }
+ else
+ {
+ library->fill_tensor_uniform(CLAccessor(src), 0);
+ library->fill_tensor_uniform(CLAccessor(weights), 1);
+ library->fill_tensor_uniform(CLAccessor(bias), 2);
+ }
+
+ // Compute NEFullyConnectedLayer function
+ fc.run();
+
+ return dst;
+}
+} // namespace
+
+#ifndef DOXYGEN_SKIP_THIS
+BOOST_AUTO_TEST_SUITE(CL)
+BOOST_AUTO_TEST_SUITE(FullyConnectedLayer)
+
+BOOST_TEST_DECORATOR(*boost::unit_test::label("precommit") * boost::unit_test::label("nightly"))
+BOOST_DATA_TEST_CASE(Configuration,
+ SmallFullyConnectedLayerDataset() * boost::unit_test::data::make({ DataType::F32, DataType::QS8 }),
+ fc_set, dt)
+{
+ // Set fixed point position data type allowed
+ int fixed_point_position = (dt == DataType::F32) ? 0 : 3;
+
+ // Create tensors
+ CLTensor src = create_tensor<CLTensor>(fc_set.src_shape, dt, 1, fixed_point_position);
+ CLTensor bias = create_tensor<CLTensor>(fc_set.bias_shape, dt, 1, fixed_point_position);
+ CLTensor dst = create_tensor<CLTensor>(fc_set.dst_shape, dt, 1, fixed_point_position);
+
+ // Swap the first and second dimension of weights' shape if transpose_weights is true
+ TensorShape ws = fc_set.weights_shape;
+ if(fc_set.transpose_weights)
+ {
+ const size_t dimx = ws.x();
+ ws.set(0, ws.y());
+ ws.set(1, dimx);
+ }
+
+ CLTensor weights = create_tensor<CLTensor>(ws, dt, 1, fixed_point_position);
+
+ BOOST_TEST(src.info()->is_resizable());
+ BOOST_TEST(weights.info()->is_resizable());
+ BOOST_TEST(bias.info()->is_resizable());
+ BOOST_TEST(dst.info()->is_resizable());
+
+ // Create and configure function.
+ // Note: We pass the weights already transposed
+ CLFullyConnectedLayer fc;
+ fc.configure(&src, &weights, &bias, &dst, false);
+
+ // Validate valid region
+ const ValidRegion src_valid_region = shape_to_valid_region(fc_set.src_shape);
+ const ValidRegion weights_valid_region = shape_to_valid_region(ws);
+ const ValidRegion bias_valid_region = shape_to_valid_region(fc_set.bias_shape);
+ const ValidRegion dst_valid_region = shape_to_valid_region(fc_set.dst_shape);
+
+ validate(src.info()->valid_region(), src_valid_region);
+ validate(weights.info()->valid_region(), weights_valid_region);
+ validate(bias.info()->valid_region(), bias_valid_region);
+ validate(dst.info()->valid_region(), dst_valid_region);
+}
+
+BOOST_AUTO_TEST_SUITE(Float)
+BOOST_TEST_DECORATOR(*boost::unit_test::label("precommit"))
+BOOST_DATA_TEST_CASE(RunSmall,
+ SmallFullyConnectedLayerDataset() * boost::unit_test::data::make({ DataType::F32 }),
+ fc_set, dt)
+{
+ // Compute function
+ CLTensor dst = compute_fully_connected_layer(fc_set.src_shape, fc_set.weights_shape, fc_set.bias_shape, fc_set.dst_shape, dt, fc_set.transpose_weights, 0);
+
+ // Compute reference
+ RawTensor ref_dst = Reference::compute_reference_fully_connected_layer(fc_set.src_shape, fc_set.weights_shape, fc_set.bias_shape, fc_set.dst_shape, dt, fc_set.transpose_weights, 0);
+
+ // Validate output
+ validate(CLAccessor(dst), ref_dst, tolerance_f32);
+}
+
+BOOST_TEST_DECORATOR(*boost::unit_test::label("nightly"))
+BOOST_DATA_TEST_CASE(RunLarge,
+ LargeFullyConnectedLayerDataset() * boost::unit_test::data::make({ DataType::F32 }),
+ fc_set, dt)
+{
+ // Compute function
+ CLTensor dst = compute_fully_connected_layer(fc_set.src_shape, fc_set.weights_shape, fc_set.bias_shape, fc_set.dst_shape, dt, fc_set.transpose_weights, 0);
+
+ // Compute reference
+ RawTensor ref_dst = Reference::compute_reference_fully_connected_layer(fc_set.src_shape, fc_set.weights_shape, fc_set.bias_shape, fc_set.dst_shape, dt, fc_set.transpose_weights, 0);
+
+ // Validate output
+ validate(CLAccessor(dst), ref_dst, tolerance_f32);
+}
+BOOST_AUTO_TEST_SUITE_END()
+
+BOOST_AUTO_TEST_SUITE(Quantized)
+BOOST_TEST_DECORATOR(*boost::unit_test::label("precommit"))
+BOOST_DATA_TEST_CASE(RunSmall,
+ SmallFullyConnectedLayerDataset() * boost::unit_test::data::make({ DataType::QS8 }) * boost::unit_test::data::xrange(4, 7),
+ fc_set, dt, fixed_point_position)
+{
+ // Compute function
+ CLTensor dst = compute_fully_connected_layer(fc_set.src_shape, fc_set.weights_shape, fc_set.bias_shape, fc_set.dst_shape, dt, fc_set.transpose_weights, fixed_point_position);
+
+ // Compute reference
+ RawTensor ref_dst = Reference::compute_reference_fully_connected_layer(fc_set.src_shape, fc_set.weights_shape, fc_set.bias_shape, fc_set.dst_shape, dt, fc_set.transpose_weights, fixed_point_position);
+
+ // Validate output
+ validate(CLAccessor(dst), ref_dst, tolerance_qs8);
+}
+
+BOOST_TEST_DECORATOR(*boost::unit_test::label("nightly"))
+BOOST_DATA_TEST_CASE(RunLarge,
+ LargeFullyConnectedLayerDataset() * boost::unit_test::data::make({ DataType::QS8 }) * boost::unit_test::data::xrange(4, 7),
+ fc_set, dt, fixed_point_position)
+{
+ // Compute function
+ CLTensor dst = compute_fully_connected_layer(fc_set.src_shape, fc_set.weights_shape, fc_set.bias_shape, fc_set.dst_shape, dt, fc_set.transpose_weights, fixed_point_position);
+
+ // Compute reference
+ RawTensor ref_dst = Reference::compute_reference_fully_connected_layer(fc_set.src_shape, fc_set.weights_shape, fc_set.bias_shape, fc_set.dst_shape, dt, fc_set.transpose_weights, fixed_point_position);
+
+ // Validate output
+ validate(CLAccessor(dst), ref_dst, tolerance_qs8);
+}
+BOOST_AUTO_TEST_SUITE_END()
+
+BOOST_AUTO_TEST_SUITE_END()
+BOOST_AUTO_TEST_SUITE_END()
+#endif // DOXYGEN_SKIP_THIS