Add CLPool3d Int8 Support
- Adds Qasymm8 and Qasymm8_signed support to the 3d pool operator
Resolves: COMPMID-4669
Signed-off-by: Mohammed Suhail Munshi <MohammedSuhail.Munshi@arm.com>
Change-Id: I36038c2b7c4f36baf67f7aae801356890e104538
Reviewed-on: https://eu-gerrit-1.euhpc.arm.com/c/VisualCompute/ComputeLibrary/+/410496
Tested-by: bsgcomp <bsgcomp@arm.com>
Reviewed-by: Sheri Zhang <sheri.zhang@arm.com>
Comments-Addressed: bsgcomp <bsgcomp@arm.com>
Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/7391
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gunes Bayir <gunes.bayir@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
diff --git a/src/gpu/cl/kernels/ClPool3dKernel.cpp b/src/gpu/cl/kernels/ClPool3dKernel.cpp
index 929ccf7..a090ac5 100644
--- a/src/gpu/cl/kernels/ClPool3dKernel.cpp
+++ b/src/gpu/cl/kernels/ClPool3dKernel.cpp
@@ -49,19 +49,22 @@
ARM_COMPUTE_RETURN_ERROR_ON_F16_UNSUPPORTED(src);
ARM_COMPUTE_RETURN_ERROR_ON_MSG((pool_info.stride.x() == 0 || pool_info.stride.y() == 0 || pool_info.stride.z() == 0), "Strides cannot be zero.");
- ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(src, 1, DataType::F16, DataType::F32);
+ ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(src, 1, DataType::F16, DataType::F32, DataType::QASYMM8_SIGNED, DataType::QASYMM8);
+ ARM_COMPUTE_RETURN_ERROR_ON_MSG((!is_data_type_float(src->data_type())) && (!pool_info.exclude_padding
+ && (pool_info.pool_type == PoolingType::AVG)),
+ "Exclude padding is unsupported for non-float types for Avg op");
- const auto data_layout = src->data_layout();
- const int idx_width = get_data_layout_dimension_index(data_layout, DataLayoutDimension::WIDTH);
- const int idx_height = get_data_layout_dimension_index(data_layout, DataLayoutDimension::HEIGHT);
- const int idx_depth = get_data_layout_dimension_index(data_layout, DataLayoutDimension::DEPTH);
- const bool is_global_pooling = pool_info.is_global_pooling;
+ const auto data_layout = src->data_layout();
+ const int idx_width = get_data_layout_dimension_index(data_layout, DataLayoutDimension::WIDTH);
+ const int idx_height = get_data_layout_dimension_index(data_layout, DataLayoutDimension::HEIGHT);
+ const int idx_depth = get_data_layout_dimension_index(data_layout, DataLayoutDimension::DEPTH);
+ const bool is_global_pooling = pool_info.is_global_pooling;
const unsigned int pool_size_x = is_global_pooling ? src->dimension(idx_width) : pool_info.pool_size.width;
const unsigned int pool_size_y = is_global_pooling ? src->dimension(idx_height) : pool_info.pool_size.height;
const unsigned int pool_size_z = is_global_pooling ? src->dimension(idx_depth) : pool_info.pool_size.depth;
- int output_width = 0;
- int output_height = 0;
- int output_depth = 0;
+ int output_width = 0;
+ int output_height = 0;
+ int output_depth = 0;
bool round_type_ceil_with_asymm_padding = (pool_info.round_type == DimensionRoundingType::CEIL) && (!is_symmetric(pool_info.padding));
ARM_COMPUTE_RETURN_ERROR_ON_MSG(round_type_ceil_with_asymm_padding, "Cannot use dimension round type CEIL when padding is asymmetric.");
@@ -143,10 +146,31 @@
build_opts.add_option("-DSRC_HEIGHT=" + support::cpp11::to_string(src->dimension(idx_height)));
build_opts.add_option("-DSRC_DEPTH=" + support::cpp11::to_string(src->dimension(idx_depth)));
+ // If datatype is quantized add relevant parameters
+ if(is_data_type_quantized_asymmetric(data_type) && src->quantization_info() != dst->quantization_info())
+ {
+ const UniformQuantizationInfo iq_info = src->quantization_info().uniform();
+ const UniformQuantizationInfo oq_info = dst->quantization_info().uniform();
+
+ build_opts.add_option("-DOFFSET_IN1=" + float_to_string_with_full_precision(iq_info.offset));
+ build_opts.add_option("-DOFFSET_OUT=" + float_to_string_with_full_precision(oq_info.offset));
+ build_opts.add_option("-DSCALE_IN1=" + float_to_string_with_full_precision(iq_info.scale));
+ build_opts.add_option("-DSCALE_OUT=" + float_to_string_with_full_precision(oq_info.scale));
+ }
+
// Set the initial value for the pooling operation accordingly with the data type
if(pool_type == PoolingType::MAX)
{
- build_opts.add_option("-DINITIAL_VALUE=" + float_to_string_with_full_precision(std::numeric_limits<float>::lowest()));
+ if(is_data_type_quantized(data_type))
+ {
+ PixelValue type_min{};
+ std::tie(type_min, std::ignore) = get_min_max(data_type);
+ build_opts.add_option("-DINITIAL_VALUE=" + support::cpp11::to_string(type_min.get<int32_t>()));
+ }
+ else
+ {
+ build_opts.add_option("-DINITIAL_VALUE=" + float_to_string_with_full_precision(std::numeric_limits<float>::lowest()));
+ }
}
else
{
@@ -164,6 +188,11 @@
{
acc_data_type = DataType::F32;
}
+ else if(is_data_type_quantized(data_type) && pool_type != PoolingType::MAX) // Use S32 for avg pooling to allow for integer division
+ {
+ acc_data_type = DataType::S32;
+ }
+
build_opts.add_option("-DACC_DATA_TYPE=" + get_cl_type_from_data_type(acc_data_type));
build_opts.add_option_if(use_fp_mixed_precision, "-DFP_MIXED_PRECISION");
build_opts.add_option_if(exclude_padding, "-DEXCLUDE_PADDING");
@@ -172,8 +201,10 @@
build_opts.add_option("-DDST_CHANNELS=" + support::cpp11::to_string(dst->dimension(idx_channel)));
build_opts.add_option("-DDST_BATCH_SIZE=" + support::cpp11::to_string(dst->dimension(idx_batch_size)));
build_opts.add_option("-DVEC_SIZE_LEFTOVER=" + support::cpp11::to_string(src->dimension(0) % _num_elems_processed_per_iteration));
- std::string kernel_name = "pooling_3d_layer_MxN_ndhwc";
- _kernel = create_kernel(compile_context, kernel_name, build_opts.options());
+
+ // if datatype is quantized use quantized kernel function
+ std::string kernel_name = (is_data_type_quantized_asymmetric(data_type) ? "pooling_3d_layer_MxN_ndhwc_quantized" : "pooling_3d_layer_MxN_ndhwc");
+ _kernel = create_kernel(compile_context, kernel_name, build_opts.options());
// Configure kernel window
Window win = calculate_max_window(*dst, Steps(_num_elems_processed_per_iteration));