quantize_mkldnn_op.cc 6.1 KB
Newer Older
X
xiaoli.liu@intel.com 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License. */

#include "mkldnn.hpp"
#include "paddle/fluid/framework/tensor.h"
#include "paddle/fluid/operators/quantize_op.h"
#include "paddle/fluid/platform/mkldnn_helper.h"
X
xiaoli.liu@intel.com 已提交
19
#include "paddle/fluid/platform/mkldnn_reuse.h"
X
xiaoli.liu@intel.com 已提交
20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38

namespace paddle {
namespace operators {

using mkldnn::memory;
using mkldnn::primitive;
using mkldnn::reorder;
using platform::to_void_cast;
using Tensor = framework::Tensor;
using framework::DataLayout;
using mkldnn::stream;
using platform::GetMKLDNNFormat;

template <typename T>
class QuantOpKernel : public framework::OpKernel<T> {
 public:
  void Compute(const framework::ExecutionContext& ctx) const override {
    auto* input = ctx.Input<Tensor>("Input");
    auto scale_data = ctx.Attr<float>("Scale");
39 40
    auto scale_shift = ctx.Attr<float>("Shift");
    bool with_shift = scale_shift != 0.0f;
X
xiaoli.liu@intel.com 已提交
41
    auto* output = ctx.Output<Tensor>("Output");
42 43 44 45 46 47 48 49 50 51 52 53

    PADDLE_ENFORCE_NE(
        scale_data, 0.0f,
        platform::errors::InvalidArgument("Quantization scale cannot be 0.0"));
    PADDLE_ENFORCE_GE(scale_shift, 0,
                      platform::errors::Unimplemented(
                          "Quantization shift must be nonnegative."));
    PADDLE_ENFORCE_LE(
        scale_shift, 255,
        platform::errors::Unimplemented(
            "Quantization shift must be less than or equal to 255."));

X
xiaoli.liu@intel.com 已提交
54 55 56 57 58
    auto& dev_ctx =
        ctx.template device_context<platform::MKLDNNDeviceContext>();
    const auto& engine = dev_ctx.GetEngine();

    std::vector<primitive> pipeline;
A
Adam 已提交
59 60
    auto src_tz = paddle::framework::vectorize<int64_t>(input->dims());
    auto dst_tz = paddle::framework::vectorize<int64_t>(output->dims());
X
xiaoli.liu@intel.com 已提交
61 62 63

    const T* input_data = input->data<T>();

64
    bool is_negative_input = ctx.Attr<bool>("is_negative_input");
65
    bool bfloat16 = ctx.Attr<bool>("bfloat16");
66

67 68 69 70 71
    std::string key =
        platform::CreateKey(dev_ctx, src_tz, scale_data, scale_shift,
                            is_negative_input, ctx.OutputName("Output"));
    key = platform::ExtendKeyWithThreadInfoIfNeeded(dev_ctx, key);

72 73 74
    const std::string key_prim = key + "@r";
    const std::string key_src_mem = key + "@s";
    const std::string key_dst_mem = key + "@d";
75 76

    std::shared_ptr<mkldnn::memory> src_memory;
X
xiaoli.liu@intel.com 已提交
77
    std::shared_ptr<mkldnn::memory> dst_memory;
78 79 80 81
    std::shared_ptr<reorder> reorder_p;
    reorder_p = std::static_pointer_cast<reorder>(dev_ctx.GetBlob(key_prim));

    if (reorder_p == nullptr) {
82 83 84
      std::string out_layout = ctx.Attr<std::string>("output_format");
      MKLDNNMemoryFormat out_format =
          platform::data_format_to_memory_format(out_layout);
85 86 87 88
      mkldnn::primitive_attr attri;
      int mask = 0;
      attri.set_output_scales(mask, {scale_data});

89 90 91 92 93 94 95 96 97
      if (with_shift) {
        mkldnn::post_ops post_operations;
        post_operations.append_sum();
        attri.set_post_ops(post_operations);
        uint8_t* output_data = output->mutable_data<uint8_t>(ctx.GetPlace());
        // memset casts scale_shift to unsigned char (uint8_t) internally
        std::memset(output_data, scale_shift, output->numel());
      }

98 99
      auto src_md = platform::MKLDNNMemDesc({src_tz}, memory::data_type::f32,
                                            input->format());
A
Adam 已提交
100 101
      src_memory = std::make_shared<mkldnn::memory>(
          src_md, engine, to_void_cast<T>(input_data));
102

A
Adam 已提交
103
      std::shared_ptr<mkldnn::memory::desc> dst_md;
104 105 106
      if (bfloat16) {
        platform::SetDstMemoryQuantized<paddle::platform::bfloat16>(
            ctx, output, dst_tz, engine, dst_md, dst_memory, out_format);
107
      } else if (is_negative_input && !with_shift) {
108
        platform::SetDstMemoryQuantized<int8_t>(ctx, output, dst_tz, engine,
109
                                                dst_md, dst_memory, out_format);
110
      } else {
111 112
        platform::SetDstMemoryQuantized<uint8_t>(
            ctx, output, dst_tz, engine, dst_md, dst_memory, out_format);
113 114
      }
      auto reorder_pd = std::shared_ptr<reorder::primitive_desc>(
A
Adam 已提交
115 116
          new reorder::primitive_desc(*src_memory, *dst_memory, attri));
      reorder_p = std::shared_ptr<reorder>(new reorder(*reorder_pd));
117 118 119 120

      dev_ctx.SetBlob(key_prim, reorder_p);
      dev_ctx.SetBlob(key_src_mem, src_memory);
      dev_ctx.SetBlob(key_dst_mem, dst_memory);
X
xiaoli.liu@intel.com 已提交
121
    } else {
122 123 124 125 126 127 128
      src_memory = std::static_pointer_cast<mkldnn::memory>(
          dev_ctx.GetBlob(key_src_mem));
      src_memory->set_data_handle(to_void_cast<T>(input_data));

      dst_memory = std::static_pointer_cast<mkldnn::memory>(
          dev_ctx.GetBlob(key_dst_mem));
      auto place = ctx.GetPlace();
129 130 131 132

      if (bfloat16) {
        dst_memory->set_data_handle(
            output->mutable_data<paddle::platform::bfloat16>(place));
133 134 135 136
      } else if (with_shift || !is_negative_input) {
        uint8_t* output_data = output->mutable_data<uint8_t>(ctx.GetPlace());
        if (with_shift) std::memset(output_data, scale_shift, output->numel());
        dst_memory->set_data_handle(output_data);
137
      } else {
138 139
        dst_memory->set_data_handle(
            output->mutable_data<int8_t>(ctx.GetPlace()));
140
      }
X
xiaoli.liu@intel.com 已提交
141
    }
142

143
    auto& astream = platform::MKLDNNDeviceContext::tls().get_stream();
144 145 146 147 148 149
    {
      platform::RecordEvent record_reorder("int_reorder",
                                           platform::EventRole::kUniqueOp);
      reorder_p->execute(astream, *src_memory, *dst_memory);
      astream.wait();
    }
A
Adam 已提交
150

X
xiaoli.liu@intel.com 已提交
151 152 153 154 155 156 157 158 159 160
    output->set_layout(DataLayout::kMKLDNN);
    output->set_format(GetMKLDNNFormat(*dst_memory));
  }
};
}  // namespace operators
}  // namespace paddle
namespace ops = paddle::operators;

REGISTER_OP_KERNEL(quantize, MKLDNN, ::paddle::platform::CPUPlace,
                   ops::QuantOpKernel<float>);