diff options
author | Georgios Pinitas <georgios.pinitas@arm.com> | 2019-05-21 13:32:43 +0100 |
---|---|---|
committer | Georgios Pinitas <georgios.pinitas@arm.com> | 2019-06-03 14:51:29 +0000 |
commit | 4c5469b192665c94118a8a558787cb9cec2d0765 (patch) | |
tree | 168aa969de8243bdbb1f25247dd9f54d037ae32c /src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NCHWKernel.cpp | |
parent | 43a129e94df41f9ac8bc78b702da5a387ada0494 (diff) | |
download | ComputeLibrary-4c5469b192665c94118a8a558787cb9cec2d0765.tar.gz |
COMPMID-2225: Add interface support for new quantized data types.
Add support for:
-QSYMM8, 8-bit quantized symmetric
-QSYMM8_PER_CHANNEL, 8-bit quantized symmetric with per channel quantization
Change-Id: I00c4ff98e44af37419470af61419ee95d0de2463
Signed-off-by: Georgios Pinitas <georgios.pinitas@arm.com>
Reviewed-on: https://review.mlplatform.org/c/1236
Tested-by: Arm Jenkins <bsgcomp@arm.com>
Reviewed-by: Gian Marco Iodice <gianmarco.iodice@arm.com>
Comments-Addressed: Arm Jenkins <bsgcomp@arm.com>
Diffstat (limited to 'src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NCHWKernel.cpp')
-rw-r--r-- | src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NCHWKernel.cpp | 22 |
1 files changed, 13 insertions, 9 deletions
diff --git a/src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NCHWKernel.cpp b/src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NCHWKernel.cpp index cd25bb1e7f..615327a7cc 100644 --- a/src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NCHWKernel.cpp +++ b/src/core/CL/kernels/CLDepthwiseConvolutionLayer3x3NCHWKernel.cpp @@ -251,30 +251,34 @@ void CLDepthwiseConvolutionLayer3x3NCHWKernel::configure(const ICLTensor *input, if(is_qasymm) { - float multiplier = _input->info()->quantization_info().scale * _weights->info()->quantization_info().scale / _output->info()->quantization_info().scale; + const UniformQuantizationInfo iq_info = _input->info()->quantization_info().uniform(); + const UniformQuantizationInfo wq_info = _weights->info()->quantization_info().uniform(); + const UniformQuantizationInfo oq_info = _output->info()->quantization_info().uniform(); + + float multiplier = iq_info.scale * wq_info.scale / oq_info.scale; int output_multiplier = 0; int output_shift = 0; quantization::calculate_quantized_multiplier_less_than_one(multiplier, &output_multiplier, &output_shift); build_opts.add_option("-DCONV_STRIDE_Y=" + support::cpp11::to_string(_conv_stride_y)); - build_opts.add_option("-DINPUT_OFFSET=" + support::cpp11::to_string(-_input->info()->quantization_info().offset)); - build_opts.add_option("-DWEIGHTS_OFFSET=" + support::cpp11::to_string(-_weights->info()->quantization_info().offset)); - build_opts.add_option("-DOUTPUT_OFFSET=" + support::cpp11::to_string(_output->info()->quantization_info().offset)); - build_opts.add_option("-DK_OFFSET=" + support::cpp11::to_string(9 * input->info()->quantization_info().offset * weights->info()->quantization_info().offset)); + build_opts.add_option("-DINPUT_OFFSET=" + support::cpp11::to_string(-iq_info.offset)); + build_opts.add_option("-DWEIGHTS_OFFSET=" + support::cpp11::to_string(-wq_info.offset)); + build_opts.add_option("-DOUTPUT_OFFSET=" + support::cpp11::to_string(oq_info.offset)); + build_opts.add_option("-DK_OFFSET=" + support::cpp11::to_string(9 * iq_info.offset * wq_info.offset)); build_opts.add_option("-DOUTPUT_MULTIPLIER=" + support::cpp11::to_string(output_multiplier)); build_opts.add_option("-DOUTPUT_SHIFT=" + support::cpp11::to_string(output_shift)); if(act_info.enabled()) { - const int a_val = output->info()->quantization_info().quantize(act_info.a(), RoundingPolicy::TO_NEAREST_UP); - const int b_val = output->info()->quantization_info().quantize(act_info.b(), RoundingPolicy::TO_NEAREST_UP); - const int o1 = output->info()->quantization_info().offset; + const int a_val = quantize_qasymm8(act_info.a(), oq_info); + const int b_val = quantize_qasymm8(act_info.b(), oq_info); + const int o1 = oq_info.offset; build_opts.add_option("-DA_VAL=" + support::cpp11::to_string(a_val)); build_opts.add_option("-DB_VAL=" + support::cpp11::to_string(b_val)); build_opts.add_option("-DCONST_0=" + support::cpp11::to_string(o1)); - const float s1 = input->info()->quantization_info().scale; + const float s1 = iq_info.scale; build_opts.add_option("-DS1_VAL=" + float_to_string_with_full_precision(s1)); build_opts.add_option("-DO1_VAL=" + support::cpp11::to_string(o1)); } |