TReduceMax.cpp 6.67 KB
Newer Older
xiaotong committed
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
/* NiuTrans.Tensor - an open-source tensor library
* Copyright (C) 2017, Natural Language Processing Lab, Northestern University.
* 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.
*/

/*
liyinqiao committed
19
* $Created by: Xu Chen (email: hello_master1954@163.com) 2018-06-30
xiaotong committed
20 21
*/

liyinqiao committed
22
#include "../core/utilities/CheckData.h"
liyinqiao committed
23
#include "TReduceMax.h"
xiaotong committed
24

liyinqiao committed
25
namespace nts { // namespace nts(NiuTrans.Tensor)
liyinqiao committed
26 27 28 29

/* 
case 1: get the max value of the items along a dimension of the tensor. 
In this case,
liyinqiao committed
30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62
(2, 4) -> (4), dim = 0
(2, 4) -> (2), dim = 1
*/
bool TestReduceMax1()
{
    /* a input tensor of size (2, 4) */
    int sOrder = 2;
    int * sDimSize = new int[sOrder];
    sDimSize[0] = 2;
    sDimSize[1] = 4;

    int sUnitNum = 1;
    for (int i = 0; i < sOrder; i++)
        sUnitNum *= sDimSize[i];

    /* a output tensor of size (4) */
    int tOrder1 = 1;
    int * tDimSize1 = new int[tOrder1];
    tDimSize1[0] = 4;

    int tUnitNum1 = 1;
    for (int i = 0; i < tOrder1; i++)
        tUnitNum1 *= tDimSize1[i];

    /* a output tensor of size (2) */
    int tOrder2 = 1;
    int * tDimSize2 = new int[tOrder2];
    tDimSize2[0] = 2;

    int tUnitNum2 = 1;
    for (int i = 0; i < tOrder2; i++)
        tUnitNum2 *= tDimSize2[i];

liyinqiao committed
63 64 65 66
    DTYPE sData[2][4] = { {0.0F, 5.0F, 2.0F, 3.0F},
                          {4.0F, 1.0F, 6.0F, 7.0F} };
    DTYPE answer1[4] = {4.0F, 5.0F, 6.0F, 7.0F};
    DTYPE answer2[2] = {5.0F, 7.0F};
liyinqiao committed
67 68 69 70 71

    /* CPU test */
    bool cpuTest = true;

    /* create tensors */
72 73 74
    XTensor * s = NewTensorV2(sOrder, sDimSize);
    XTensor * t1 = NewTensorV2(tOrder1, tDimSize1);
    XTensor * t2 = NewTensorV2(tOrder2, tDimSize2);
75 76
    XTensor tUser1;
    XTensor tUser2;
liyinqiao committed
77 78 79 80 81 82 83

    /* initialize variables */
    s->SetData(sData, sUnitNum);
    t1->SetZeroAll();
    t2->SetZeroAll();

    /* call ReduceMax function */
84 85
    _ReduceMax(s, t1, 0);
    _ReduceMax(s, t2, 1);
86 87
    tUser1 = ReduceMax(*s, 0);
    tUser2 = ReduceMax(*s, 1);
liyinqiao committed
88 89

    /* check results */
liyinqiao committed
90 91
    cpuTest = _CheckData(t1, answer1, tUnitNum1) && _CheckData(&tUser1, answer1, tUnitNum1)
        && _CheckData(t2, answer2, tUnitNum2) && _CheckData(&tUser2, answer2, tUnitNum2);
xiaotong committed
92 93

#ifdef USE_CUDA
liyinqiao committed
94 95 96 97
    /* GPU test */
    bool gpuTest = true;

    /* create tensors */
98 99 100
    XTensor * sGPU = NewTensorV2(sOrder, sDimSize, X_FLOAT, 1.0F, 0);
    XTensor * tGPU1 = NewTensorV2(tOrder1, tDimSize1, X_FLOAT, 1.0F, 0);
    XTensor * tGPU2 = NewTensorV2(tOrder2, tDimSize2, X_FLOAT, 1.0F, 0);
101 102
    XTensor tUserGPU1;
    XTensor tUserGPU2;
liyinqiao committed
103 104 105 106 107 108 109

    /* initialize variables */
    sGPU->SetData(sData, sUnitNum);
    tGPU1->SetZeroAll();
    tGPU2->SetZeroAll();

    /* call ReduceMax function */
110 111
    _ReduceMax(sGPU, tGPU1, 0);
    _ReduceMax(sGPU, tGPU2, 1);
112 113
    tUserGPU1 = ReduceMax(*sGPU, 0);
    tUserGPU2 = ReduceMax(*sGPU, 1);
liyinqiao committed
114 115

    /* check results */
liyinqiao committed
116 117
    gpuTest = _CheckData(tGPU1, answer1, tUnitNum1) && _CheckData(&tUserGPU1, answer1, tUnitNum1)
        && _CheckData(tGPU2, answer2, tUnitNum2) && _CheckData(&tUserGPU2, answer2, tUnitNum2);
liyinqiao committed
118 119 120 121 122 123 124 125 126 127 128 129 130

    /* destroy variables */
    delete s;
    delete t1;
    delete t2;
    delete sGPU;
    delete tGPU1;
    delete tGPU2;
    delete[] sDimSize;
    delete[] tDimSize1;
    delete[] tDimSize2;

    return cpuTest && gpuTest;
xiaotong committed
131
#else
liyinqiao committed
132 133 134 135 136 137 138 139 140
    /* destroy variables */
    delete s;
    delete t1;
    delete t2;
    delete[] sDimSize;
    delete[] tDimSize1;
    delete[] tDimSize2;

    return cpuTest;
xiaotong committed
141
#endif // USE_CUDA
liyinqiao committed
142
}
xiaotong committed
143

144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227
/*
case 2: get the max value of the items along a dimension of the scalar tensor.
In this case,
(4) -> scalar, dim = 0
*/
bool TestReduceMax2()
{
    /* a input tensor of size (4) */
    int sOrder = 1;
    int * sDimSize = new int[sOrder];
    sDimSize[0] = 4;

    int sUnitNum = 1;
    for (int i = 0; i < sOrder; i++)
        sUnitNum *= sDimSize[i];

    /* a output scalar tensor */
    int tOrder = 0;
    int * tDimSize = new int[MAX_TENSOR_DIM_NUM];
    int tUnitNum = 1;

    DTYPE sData[4] = {0.0F, 5.0F, 2.0F, 3.0F};
    DTYPE answer[1] = {5.0F};

    /* CPU test */
    bool cpuTest = true;

    /* create tensors */
    XTensor * s = NewTensorV2(sOrder, sDimSize);
    XTensor * t = NewTensorV2(tOrder, tDimSize);
    XTensor tUser;

    /* initialize variables */
    s->SetData(sData, sUnitNum);
    t->SetZeroAll();

    /* call ReduceMax function */
    _ReduceMax(s, t, 0);
    tUser = ReduceMax(*s, 0);

    /* check results */
    cpuTest = _CheckData(t, answer, tUnitNum) && _CheckData(&tUser, answer, tUnitNum);

#ifdef USE_CUDA
    /* GPU test */
    bool gpuTest = true;

    /* create tensors */
    XTensor * sGPU = NewTensorV2(sOrder, sDimSize, X_FLOAT, 1.0F, 0);
    XTensor * tGPU = NewTensorV2(tOrder, tDimSize, X_FLOAT, 1.0F, 0);
    XTensor tUserGPU;

    /* initialize variables */
    sGPU->SetData(sData, sUnitNum);
    tGPU->SetZeroAll();
    tGPU->SetZeroAll();

    /* call ReduceMax function */
    _ReduceMax(sGPU, tGPU, 0);
    tUserGPU = ReduceMax(*sGPU, 0);

    /* check results */
    gpuTest = _CheckData(tGPU, answer, tUnitNum) && _CheckData(&tUserGPU, answer, tUnitNum);

    /* destroy variables */
    delete s;
    delete t;
    delete sGPU;
    delete tGPU;
    delete[] sDimSize;
    delete[] tDimSize;

    return cpuTest && gpuTest;
#else
    /* destroy variables */
    delete s;
    delete t;
    delete[] sDimSize;
    delete[] tDimSize;

    return cpuTest;
#endif // USE_CUDA
}

liyinqiao committed
228 229 230 231
/* other cases */
/*
TODO!!
*/
xiaotong committed
232

liyinqiao committed
233 234 235 236 237 238 239 240 241 242 243
/* test for ReduceMax Function */
bool TestReduceMax()
{
    XPRINT(0, stdout, "[TEST ReduceMax] get the max value of the items along a dimension of the tensor\n");
    bool returnFlag = true, caseFlag = true;

    /* case 1 test */
    caseFlag = TestReduceMax1();
    if (!caseFlag) {
        returnFlag = false;
        XPRINT(0, stdout, ">> case 1 failed!\n");
xiaotong committed
244
    }
liyinqiao committed
245 246
    else
        XPRINT(0, stdout, ">> case 1 passed!\n");
xiaotong committed
247

248 249 250 251 252 253 254 255 256
    /* case 2 test */
    caseFlag = TestReduceMax2();
    if (!caseFlag) {
        returnFlag = false;
        XPRINT(0, stdout, ">> case 2 failed!\n");
    }
    else
    XPRINT(0, stdout, ">> case 2 passed!\n");

liyinqiao committed
257
    /* other cases test */
xiaotong committed
258 259 260 261
    /*
    TODO!!
    */

liyinqiao committed
262 263 264 265 266
    if (returnFlag) {
        XPRINT(0, stdout, ">> All Passed!\n");
    }
    else
        XPRINT(0, stdout, ">> Failed!\n");
xiaotong committed
267

liyinqiao committed
268
    XPRINT(0, stdout, "\n");
xiaotong committed
269

liyinqiao committed
270
    return returnFlag;
xiaotong committed
271 272
    }

liyinqiao committed
273
} // namespace nts(NiuTrans.Tensor)