compute_primitives.h 8.8 KB
Newer Older
F
Feng Xing 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
// Copyright (c) 2021 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.

#pragma once

N
niuliling123 已提交
17 18 19 20 21 22 23
#ifdef PADDLE_WITH_CUDA
#include <cuda_fp16.h>
#endif
#ifdef PADDLE_WITH_HIP
#include <hip/hip_fp16.h>
#endif

24 25
// #include <algorithm>
#include "paddle/fluid/platform/cuda_device_function.h"
N
niuliling123 已提交
26 27
#include "paddle/fluid/platform/float16.h"

F
Feng Xing 已提交
28 29
namespace paddle {
namespace operators {
N
niuliling123 已提交
30 31 32
namespace kernel_primitives {
namespace details {

33
#ifdef __HIPCC__
34
constexpr int kReduceMaxThread = 256;
35 36
constexpr int kWarpSize = 64;
#else
37
constexpr int kReduceMaxThread = 128;
38 39 40
constexpr int kWarpSize = 32;
#endif

41 42
// kGlobalMode: block reduce, each block gets an output;
// kLocalMode: thread reduce, each thread gets an output;
43 44
enum ReduceMode { kGlobalMode, kLocalMode };

N
niuliling123 已提交
45 46 47 48 49 50 51 52 53 54 55 56
template <typename T>
class MPTypeTrait {
 public:
  using Type = T;
};

template <>
class MPTypeTrait<platform::float16> {
 public:
  using Type = float;
};

57 58 59 60 61 62 63
/**
 * @brief will be used in BlockYReduce, get the index of reduce_num in shared
 * memory
 */
__device__ __forceinline__ int SharedMemoryIndex(int index) {
  return (threadIdx.y + index) * blockDim.x + threadIdx.x;
}
N
niuliling123 已提交
64

65 66 67 68 69 70 71
template <typename T, typename ReduceOp>
__device__ __forceinline__ T WarpReduce(T val, ReduceOp reducer) {
  unsigned mask = 0u;
  CREATE_SHFL_MASK(mask, true);
  for (int stride = details::kWarpSize / 2; stride > 0; stride >>= 1) {
    T temp = paddle::platform::CudaShuffleDownSync(mask, val, stride);
    val = reducer(val, temp);
N
niuliling123 已提交
72
  }
73 74
  return val;
}
N
niuliling123 已提交
75

76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106
/* e.g.
 * |---------block---------|
 * |warp0|warp1|warp2|warp3|
 * |0~31|32~63|64~95|96~127|  ---->blockDim.x = 128
 *  \|/  \|/   \|/    \|/     ---->1. First WarpReduce in each warp
 * res0  res1  res2  res3     ---->2. Store result of each warp to shared memory
 *   \    \    /     /        ---->3. Load the result above from shared memory
 *        res                         to warp0 and process the second WarpReduce
 */

/**
 * @brief BlockXReduce reduce along blockDim.x
 */
template <typename T, typename ReduceOp>
__device__ __forceinline__ T BlockXReduce(T val, ReduceOp reducer) {
  __syncthreads();
  using details::kWarpSize;
  __shared__ T shared[2 * kWarpSize];
  int block_dim_x = blockDim.x;
  if (blockDim.x > kWarpSize) {
    block_dim_x = blockDim.x / kWarpSize;
    int lane = threadIdx.x % kWarpSize;
    int tid = threadIdx.y * blockDim.x + threadIdx.x;
    int wid = tid / kWarpSize;
    int bid = threadIdx.y;
    val = WarpReduce(val, reducer);
    if (lane == 0) {
      shared[wid] = val;
    }
    __syncthreads();
    val = shared[bid * block_dim_x + lane];
N
niuliling123 已提交
107
  }
108 109 110 111 112 113 114 115 116 117 118 119 120 121 122

  unsigned mask = 0u;
  CREATE_SHFL_MASK(mask, true);
  for (int stride = 1; stride < block_dim_x; stride <<= 1) {
    T temp = paddle::platform::CudaShuffleDownSync(mask, val, stride);
    val = reducer(val, temp);
  }
  return val;
}

/**
 * @brief BlockYReduce reduce along blockDim.y
 */
template <typename T, typename ReduceOp>
__device__ __forceinline__ T BlockYReduce(T val, ReduceOp reducer) {
123
  __shared__ T shared_memory[details::kReduceMaxThread];
124 125 126 127 128 129 130 131 132 133 134 135 136
  shared_memory[SharedMemoryIndex(0)] = val;
  for (int stride = blockDim.y / 2; stride > 0; stride >>= 1) {
    __syncthreads();
    if (threadIdx.y < stride && threadIdx.y + stride < blockDim.y) {
      T temp = shared_memory[SharedMemoryIndex(stride)];
      val = reducer(val, temp);
    }
    shared_memory[SharedMemoryIndex(0)] = val;
  }
  return val;
}

}  // namespace details
N
niuliling123 已提交
137 138 139 140

/*************************** Compute Function****************************/

/**
141
 * @brief binary function, in1 and in2 have same shape
N
niuliling123 已提交
142
 * @param:
143 144 145 146 147 148
 * T: data type of in1, in2
 * OutT: data type of out
 * NX: the cols of in1, in2
 * NY: the rows of in1, in2
 * BlockSize: the config of this device
 * OpFunc: compute functor eg: in1 + in2, in1 - in2
N
niuliling123 已提交
149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164
 */
template <typename T, typename OutT, int NX, int NY, int BlockSize,
          class OpFunc>
__device__ __forceinline__ void ElementwiseBinary(OutT* out, const T* in1,
                                                  const T* in2,
                                                  OpFunc compute) {
  T args[2];
#pragma unroll
  for (int idx = 0; idx < NX * NY; ++idx) {
    args[0] = in1[idx];
    args[1] = in2[idx];
    out[idx] = static_cast<OutT>(compute(args));
  }
}

/**
165
 * @brief ternary function, in1, in2 and in3 have same shape
N
niuliling123 已提交
166
 * @param:
167 168 169 170 171 172
 * T: data type of in1, in2, in3
 * OutT: data type of out
 * NX: the cols of in1, in2
 * NY: the rows of in1, in2
 * BlockSize: the config of this device
 * OpFunc: compute functor eg: out = in1 * in2 + in3
N
niuliling123 已提交
173 174 175
 */
template <typename T, typename OutT, int NX, int NY, int BlockSize,
          class OpFunc>
176 177 178 179
__device__ __forceinline__ void ElementwiseTernary(OutT* out, const T* in1,
                                                   const T* in2, const T* in3,
                                                   OpFunc compute) {
  T args[3];
N
niuliling123 已提交
180 181
#pragma unroll
  for (int idx = 0; idx < NX * NY; ++idx) {
182 183 184 185
    args[0] = in1[idx];
    args[1] = in2[idx];
    args[2] = in3[idx];
    out[idx] = static_cast<OutT>(compute(args));
N
niuliling123 已提交
186
  }
F
Feng Xing 已提交
187
}
N
niuliling123 已提交
188 189

/**
190 191
 * @brief cycle binary function, in1's shape size is [1, NX], in2's shape size
 * is [NY, NX], out's shape size is [NY, NX]
N
niuliling123 已提交
192
 * @param:
193 194 195 196 197 198
 * T: data type of in1, in2
 * OutT: data type of out
 * NX: the cols of in1, in2
 * NY: the rows of in1, in2
 * BlockSize: the config of this device
 * OpFunc: compute functor eg: in1 + in2, in1 - in2
N
niuliling123 已提交
199 200 201 202 203 204 205 206 207 208 209 210 211
 */
template <typename T, typename OutT, int NX, int NY, int BlockSize,
          class OpFunc>
__device__ __forceinline__ void CycleBinary(OutT* out, const T* in1,
                                            const T* in2, OpFunc compute) {
#pragma unroll
  for (int idx = 0; idx < NX; idx++) {
#pragma unroll
    for (int idy = 0; idy < NY; idy++) {
      out[idx + idy * NX] =
          static_cast<OutT>(compute(in1[idx], in2[idx + idy * NX]));
    }
  }
F
Feng Xing 已提交
212
}
N
niuliling123 已提交
213 214

/**
215
 * @brief unary function
N
niuliling123 已提交
216
 * @param:
217 218 219 220 221 222
 * T: data type of in
 * OutT: data type of out
 * NX: the cols of in
 * NY: the rows of in
 * BlockSize: the config of this device
 * OpFunc: compute functor eg: relu, exp
N
niuliling123 已提交
223 224 225 226 227 228 229 230 231 232 233
 */
template <typename T, typename OutT, int NX, int NY, int BlockSize,
          class OpFunc>
__device__ __forceinline__ void ElementwiseUnary(OutT* out, const T* in,
                                                 OpFunc compute) {
#pragma unroll
  for (int idx = 0; idx < NX * NY; idx++) {
    out[idx] = static_cast<OutT>(compute(in + idx));
  }
}

234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286
/**
 * @brief reduce function, in's shape size is [NX, NY].
 * If ReduceMode == kLocalMode then reduce NX, the shape of out is [NY, 1],
 * if ReduceMode == kGlobalMode then reduce between different threads, the
 * shape of out is [NY, NX]. If reduce_last_dim is false and reduce_num was
 * split, BlockYReduce will be called. If reduce_last_dim is true and
 * reduce_num was split, BlockXReduce will be called
 * @typename:
 * T: data type of in
 * NX: the cols of in
 * NY: the rows of in
 * BlockSize: the config of this device
 * OpFunc: reduce functor, eg: CustomSum, CustomMean in reduce_functor_op.h
 * @param:
 * reducer: reduce functor, eg: CustomSum<T>()
 * reduce_last_dim: if in's last dim need to be reduce then reduce_last_dim =
 * true
 */
template <typename T, int NX, int NY, int BlockSize, class OpFunc,
          details::ReduceMode Mode>
__device__ __forceinline__ void Reduce(T* out, const T* in, OpFunc reducer,
                                       bool reduce_last_dim) {
  int block_index = blockDim.y;

  if (Mode == details::ReduceMode::kGlobalMode) {
    bool block_reduce_y = (!reduce_last_dim) && (block_index > 1);
    // when reduce is not required for the last dim, and reduce num has been
    // split into multiple threads
    if (block_reduce_y) {
#pragma unroll
      for (int i = 0; i < NY * NX; i++) {  // reduce along blockdim.y
        out[i] = details::BlockYReduce<T, OpFunc>(out[i], reducer);
      }
    }

    // when last dimension need to be reduced
    if (reduce_last_dim) {
#pragma unroll
      for (int i = 0; i < NY * NX; i++) {  // reduce along blockDim.x
        out[i] = details::BlockXReduce<T, OpFunc>(out[i], reducer);
      }
    }
  } else {  // else  kLocalMode
#pragma unroll
    for (int i = 0; i < NY; ++i) {
#pragma unroll
      for (int j = 0; j < NX; ++j) {
        out[i] = reducer(out[i], in[i * NX + j]);
      }
    }
  }
}

N
niuliling123 已提交
287 288 289
}  // namespace kernel_primitives
}  // namespace operators
}  // namespace paddle