From bc4d7c2d0c3484152256d5c9dbb61e6a149bdc20 Mon Sep 17 00:00:00 2001 From: Michalis Spyrou Date: Tue, 3 Dec 2019 15:11:09 +0000 Subject: COMPMID-2797 Add support for QASYMM8_SIGNED in NEArithmeticAddition Change-Id: Ie1abf7839d40ffcdc9748c5a3baee7c00cabad21 Signed-off-by: Michalis Spyrou Reviewed-on: https://review.mlplatform.org/c/2413 Reviewed-by: Michele Di Giorgio Comments-Addressed: Arm Jenkins Tested-by: Arm Jenkins --- src/core/CL/CLHelpers.cpp | 1 - .../NEON/kernels/NEArithmeticAdditionKernel.cpp | 198 ++++++++++++++++++++- 2 files changed, 196 insertions(+), 3 deletions(-) (limited to 'src') diff --git a/src/core/CL/CLHelpers.cpp b/src/core/CL/CLHelpers.cpp index 47472a3dae..9754bebd18 100644 --- a/src/core/CL/CLHelpers.cpp +++ b/src/core/CL/CLHelpers.cpp @@ -44,7 +44,6 @@ std::string get_cl_type_from_data_type(const DataType &dt) case DataType::S8: case DataType::QASYMM8_SIGNED: case DataType::QSYMM8: - case DataType::QASYMM8_SIGNED: case DataType::QSYMM8_PER_CHANNEL: return "char"; case DataType::U16: diff --git a/src/core/NEON/kernels/NEArithmeticAdditionKernel.cpp b/src/core/NEON/kernels/NEArithmeticAdditionKernel.cpp index 733d1372a1..947be18b80 100644 --- a/src/core/NEON/kernels/NEArithmeticAdditionKernel.cpp +++ b/src/core/NEON/kernels/NEArithmeticAdditionKernel.cpp @@ -335,6 +335,193 @@ void add_QASYMM8_QASYMM8_QASYMM8(const ITensor *in1, const ITensor *in2, ITensor } } +void add_QASYMM8_SIGNED_QASYMM8_SIGNED_QASYMM8_SIGNED(const ITensor *in1, const ITensor *in2, ITensor *out, ConvertPolicy policy, const Window &window) +{ + ARM_COMPUTE_UNUSED(policy); + + // Create input windows + Window input1_win = window.broadcast_if_dimension_le_one(in1->info()->tensor_shape()); + Window input2_win = window.broadcast_if_dimension_le_one(in2->info()->tensor_shape()); + + // Clear X Dimension on execution window as we handle manually + Window win = window; + win.set(Window::DimX, Window::Dimension(0, 1, 1)); + + const int window_step_x = 16; + const auto window_start_x = static_cast(window.x().start()); + const auto window_end_x = static_cast(window.x().end()); + const bool is_broadcast_across_x = (input1_win.x().step() == 0) || (input2_win.x().step() == 0); + + const UniformQuantizationInfo iq1_info = in1->info()->quantization_info().uniform(); + const UniformQuantizationInfo iq2_info = in2->info()->quantization_info().uniform(); + const UniformQuantizationInfo oq_info = out->info()->quantization_info().uniform(); + + const float32x4_t vscale1 = vdupq_n_f32(iq1_info.scale); + const float32x4_t vscale2 = vdupq_n_f32(iq2_info.scale); + const float32x4_t invvscaleo = vdupq_n_f32(1.f / oq_info.scale); + const int32x4_t voffset1 = vdupq_n_s32(iq1_info.offset); + const int32x4_t voffset2 = vdupq_n_s32(iq2_info.offset); + const float32x4_t voffseto = vdupq_n_f32(oq_info.offset); + + if(is_broadcast_across_x) + { + const bool is_broadcast_input_2 = input2_win.x().step() == 0; + Window broadcast_win = is_broadcast_input_2 ? input2_win : input1_win; + Window non_broadcast_win = !is_broadcast_input_2 ? input2_win : input1_win; + const ITensor *broadcast_tensor = is_broadcast_input_2 ? in2 : in1; + const ITensor *non_broadcast_tensor = !is_broadcast_input_2 ? in2 : in1; + const UniformQuantizationInfo broadcast_qinfo = broadcast_tensor->info()->quantization_info().uniform(); + const UniformQuantizationInfo non_broadcast_qinfo = non_broadcast_tensor->info()->quantization_info().uniform(); + + // Clear X Dimension on execution window as we handle manually + non_broadcast_win.set(Window::DimX, Window::Dimension(0, 1, 1)); + + Iterator broadcast_input(broadcast_tensor, broadcast_win); + Iterator non_broadcast_input(non_broadcast_tensor, non_broadcast_win); + Iterator output(out, win); + + execute_window_loop(win, [&](const Coordinates &) + { + const auto non_broadcast_input_ptr = reinterpret_cast(non_broadcast_input.ptr()); + const auto output_ptr = reinterpret_cast(output.ptr()); + + const int8_t broadcast_value = *reinterpret_cast(broadcast_input.ptr()); + const int8x16_t broadcast_value_vec = vdupq_n_s8(broadcast_value); + + const float32x4x4_t bf = + { + { + vmulq_f32(vcvtq_f32_s32(vsubq_s32(vmovl_s16(vget_low_s16(vmovl_s8(vget_low_s8(broadcast_value_vec)))), voffset2)), vscale2), + vmulq_f32(vcvtq_f32_s32(vsubq_s32(vmovl_s16(vget_high_s16(vmovl_s8(vget_low_s8(broadcast_value_vec)))), voffset2)), vscale2), + vmulq_f32(vcvtq_f32_s32(vsubq_s32(vmovl_s16(vget_low_s16(vmovl_s8(vget_high_s8(broadcast_value_vec)))), voffset2)), vscale2), + vmulq_f32(vcvtq_f32_s32(vsubq_s32(vmovl_s16(vget_high_s16(vmovl_s8(vget_high_s8(broadcast_value_vec)))), voffset2)), vscale2), + } + }; + const float bfs = static_cast(broadcast_value - broadcast_qinfo.offset) * broadcast_qinfo.scale; + + // Compute S elements per iteration + int x = window_start_x; + for(; x <= (window_end_x - window_step_x); x += window_step_x) + { + const int8x16_t a = vld1q_s8(non_broadcast_input_ptr + x); + const float32x4x4_t af = + { + { + vmulq_f32(vcvtq_f32_s32(vsubq_s32(vmovl_s16(vget_low_s16(vmovl_s8(vget_low_s8(a)))), voffset1)), vscale1), + vmulq_f32(vcvtq_f32_s32(vsubq_s32(vmovl_s16(vget_high_s16(vmovl_s8(vget_low_s8(a)))), voffset1)), vscale1), + vmulq_f32(vcvtq_f32_s32(vsubq_s32(vmovl_s16(vget_low_s16(vmovl_s8(vget_high_s8(a)))), voffset1)), vscale1), + vmulq_f32(vcvtq_f32_s32(vsubq_s32(vmovl_s16(vget_high_s16(vmovl_s8(vget_high_s8(a)))), voffset1)), vscale1), + } + }; + + const int32x4x4_t rf = + { + { +#ifdef __aarch64__ + vcvtnq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[0], bf.val[0]), invvscaleo)), + vcvtnq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[1], bf.val[1]), invvscaleo)), + vcvtnq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[2], bf.val[2]), invvscaleo)), + vcvtnq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[3], bf.val[3]), invvscaleo)), +#else //__aarch64__ + vcvtq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[0], bf.val[0]), invvscaleo)), + vcvtq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[1], bf.val[1]), invvscaleo)), + vcvtq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[2], bf.val[2]), invvscaleo)), + vcvtq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[3], bf.val[3]), invvscaleo)), +#endif //__aarch64__ + } + }; + + const int8x8_t pa = vqmovn_s16(vcombine_s16(vqmovn_s32(rf.val[0]), vqmovn_s32(rf.val[1]))); + const int8x8_t pb = vqmovn_s16(vcombine_s16(vqmovn_s32(rf.val[2]), vqmovn_s32(rf.val[3]))); + vst1q_s8(output_ptr + x, vcombine_s8(pa, pb)); + } + + // Compute left-over elements + for(; x < window_end_x; ++x) + { + const float afs = static_cast(*(non_broadcast_input_ptr + x) - non_broadcast_qinfo.offset) * non_broadcast_qinfo.scale; + *(output_ptr + x) = quantize_qasymm8_signed((afs + bfs), oq_info); + } + }, + broadcast_input, non_broadcast_input, output); + } + else + { + // Clear X Dimension on execution window as we handle manually + input1_win.set(Window::DimX, Window::Dimension(0, 1, 1)); + input2_win.set(Window::DimX, Window::Dimension(0, 1, 1)); + + Iterator input1(in1, input1_win); + Iterator input2(in2, input2_win); + Iterator output(out, win); + + execute_window_loop(win, [&](const Coordinates &) + { + const auto input1_ptr = reinterpret_cast(input1.ptr()); + const auto input2_ptr = reinterpret_cast(input2.ptr()); + const auto output_ptr = reinterpret_cast(output.ptr()); + + // Compute S elements per iteration + int x = window_start_x; + for(; x <= (window_end_x - window_step_x); x += window_step_x) + { + const int8x16_t a = vld1q_s8(input1_ptr + x); + const int8x16_t b = vld1q_s8(input2_ptr + x); + + const float32x4x4_t af = + { + { + vmulq_f32(vcvtq_f32_s32(vsubq_s32(vmovl_s16(vget_low_s16(vmovl_s8(vget_low_s8(a)))), voffset1)), vscale1), + vmulq_f32(vcvtq_f32_s32(vsubq_s32(vmovl_s16(vget_high_s16(vmovl_s8(vget_low_s8(a)))), voffset1)), vscale1), + vmulq_f32(vcvtq_f32_s32(vsubq_s32(vmovl_s16(vget_low_s16(vmovl_s8(vget_high_s8(a)))), voffset1)), vscale1), + vmulq_f32(vcvtq_f32_s32(vsubq_s32(vmovl_s16(vget_high_s16(vmovl_s8(vget_high_s8(a)))), voffset1)), vscale1), + } + }; + + const float32x4x4_t bf = + { + { + vmulq_f32(vcvtq_f32_s32(vsubq_s32(vmovl_s16(vget_low_s16(vmovl_s8(vget_low_s8(b)))), voffset2)), vscale2), + vmulq_f32(vcvtq_f32_s32(vsubq_s32(vmovl_s16(vget_high_s16(vmovl_s8(vget_low_s8(b)))), voffset2)), vscale2), + vmulq_f32(vcvtq_f32_s32(vsubq_s32(vmovl_s16(vget_low_s16(vmovl_s8(vget_high_s8(b)))), voffset2)), vscale2), + vmulq_f32(vcvtq_f32_s32(vsubq_s32(vmovl_s16(vget_high_s16(vmovl_s8(vget_high_s8(b)))), voffset2)), vscale2), + } + }; + + const int32x4x4_t rf = + { + { +#ifdef __aarch64__ + vcvtnq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[0], bf.val[0]), invvscaleo)), + vcvtnq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[1], bf.val[1]), invvscaleo)), + vcvtnq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[2], bf.val[2]), invvscaleo)), + vcvtnq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[3], bf.val[3]), invvscaleo)), +#else //__aarch64__ + vcvtq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[0], bf.val[0]), invvscaleo)), + vcvtq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[1], bf.val[1]), invvscaleo)), + vcvtq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[2], bf.val[2]), invvscaleo)), + vcvtq_s32_f32(vmlaq_f32(voffseto, vaddq_f32(af.val[3], bf.val[3]), invvscaleo)), +#endif //__aarch64__ + } + }; + + const int8x8_t pa = vqmovn_s16(vcombine_s16(vqmovn_s32(rf.val[0]), vqmovn_s32(rf.val[1]))); + const int8x8_t pb = vqmovn_s16(vcombine_s16(vqmovn_s32(rf.val[2]), vqmovn_s32(rf.val[3]))); + vst1q_s8(output_ptr + x, vcombine_s8(pa, pb)); + } + + // Compute left-over elements + for(; x < window_end_x; ++x) + { + const float afs = static_cast((*(input1_ptr + x)) - iq1_info.offset) * iq1_info.scale; + const float bfs = static_cast((*(input2_ptr + x)) - iq2_info.offset) * iq2_info.scale; + *(output_ptr + x) = quantize_qasymm8_signed((afs + bfs), out->info()->quantization_info()); + } + }, + input1, input2, output); + } +} + void add_QSYMM16_QSYMM16_QSYMM16(const ITensor *in1, const ITensor *in2, ITensor *out, ConvertPolicy policy, const Window &window) { ARM_COMPUTE_UNUSED(policy); @@ -641,8 +828,8 @@ Status validate_arguments(const ITensorInfo &input1, const ITensorInfo &input2, ARM_COMPUTE_UNUSED(policy); ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(&input1); - ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(&input1, 1, DataType::U8, DataType::QASYMM8, DataType::S16, DataType::QSYMM16, DataType::F16, DataType::F32); - ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(&input2, 1, DataType::U8, DataType::QASYMM8, DataType::S16, DataType::QSYMM16, DataType::F16, DataType::F32); + ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(&input1, 1, DataType::U8, DataType::QASYMM8, DataType::QASYMM8_SIGNED, DataType::S16, DataType::QSYMM16, DataType::F16, DataType::F32); + ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(&input2, 1, DataType::U8, DataType::QASYMM8, DataType::QASYMM8_SIGNED, DataType::S16, DataType::QSYMM16, DataType::F16, DataType::F32); const TensorShape out_shape = TensorShape::broadcast_shape(input1.tensor_shape(), input2.tensor_shape()); @@ -663,6 +850,7 @@ Status validate_arguments(const ITensorInfo &input1, const ITensorInfo &input2, && !(input1.data_type() == DataType::F32 && input2.data_type() == DataType::F32 && output.data_type() == DataType::F32) && !(input1.data_type() == DataType::F16 && input2.data_type() == DataType::F16 && output.data_type() == DataType::F16) && !(input1.data_type() == DataType::QASYMM8 && input2.data_type() == DataType::QASYMM8 && output.data_type() == DataType::QASYMM8) + && !(input1.data_type() == DataType::QASYMM8_SIGNED && input2.data_type() == DataType::QASYMM8_SIGNED && output.data_type() == DataType::QASYMM8_SIGNED) && !(input1.data_type() == DataType::QSYMM16 && input2.data_type() == DataType::QSYMM16 && output.data_type() == DataType::QSYMM16), "You called addition with the wrong image formats"); @@ -699,6 +887,10 @@ std::pair validate_and_configure_window(ITensorInfo &input1, ITe { set_data_type_if_unknown(output, DataType::QASYMM8); } + else if(input1.data_type() == DataType::QASYMM8_SIGNED) + { + set_data_type_if_unknown(output, DataType::QASYMM8_SIGNED); + } else if(input1.data_type() == DataType::QSYMM16) { set_data_type_if_unknown(output, DataType::QSYMM16); @@ -733,6 +925,8 @@ void NEArithmeticAdditionKernel::configure(const ITensor *input1, const ITensor { { "add_wrap_QASYMM8_QASYMM8_QASYMM8", &add_QASYMM8_QASYMM8_QASYMM8 }, { "add_saturate_QASYMM8_QASYMM8_QASYMM8", &add_QASYMM8_QASYMM8_QASYMM8 }, + { "add_wrap_QASYMM8_SIGNED_QASYMM8_SIGNED_QASYMM8_SIGNED", &add_QASYMM8_SIGNED_QASYMM8_SIGNED_QASYMM8_SIGNED }, + { "add_saturate_QASYMM8_SIGNED_QASYMM8_SIGNED_QASYMM8_SIGNED", &add_QASYMM8_SIGNED_QASYMM8_SIGNED_QASYMM8_SIGNED }, { "add_wrap_QSYMM16_QSYMM16_QSYMM16", &add_QSYMM16_QSYMM16_QSYMM16 }, { "add_saturate_QSYMM16_QSYMM16_QSYMM16", &add_QSYMM16_QSYMM16_QSYMM16 }, { "add_wrap_U8_U8_U8", &add_same }, -- cgit v1.2.1