| 1 |
// Copyright 2002 - 2008, 2010, 2011 National Technology Engineering |
| 2 |
// Solutions of Sandia, LLC (NTESS). Under the terms of Contract |
| 3 |
// DE-NA0003525 with NTESS, the U.S. Government retains certain rights |
| 4 |
// in this software. |
| 5 |
// |
| 6 |
// Redistribution and use in source and binary forms, with or without |
| 7 |
// modification, are permitted provided that the following conditions are |
| 8 |
// met: |
| 9 |
// |
| 10 |
// * Redistributions of source code must retain the above copyright |
| 11 |
// notice, this list of conditions and the following disclaimer. |
| 12 |
// |
| 13 |
// * Redistributions in binary form must reproduce the above |
| 14 |
// copyright notice, this list of conditions and the following |
| 15 |
// disclaimer in the documentation and/or other materials provided |
| 16 |
// with the distribution. |
| 17 |
// |
| 18 |
// * Neither the name of NTESS nor the names of its contributors |
| 19 |
// may be used to endorse or promote products derived from this |
| 20 |
// software without specific prior written permission. |
| 21 |
// |
| 22 |
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS |
| 23 |
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT |
| 24 |
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR |
| 25 |
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT |
| 26 |
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, |
| 27 |
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT |
| 28 |
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, |
| 29 |
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY |
| 30 |
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT |
| 31 |
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE |
| 32 |
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. |
| 33 |
// |
| 34 |
|
| 35 |
#include "gtest/gtest.h" |
| 36 |
#include "stk_util/parallel/GenerateParallelConsistentIDs.hpp" // for generate_parallel_consist... |
| 37 |
#include "stk_util/parallel/Parallel.hpp" // for MPI_COMM_WORLD, MPI_Allre... |
| 38 |
#include "stk_util/parallel/ParallelVectorConcat.hpp" // for parallel_vector_concat |
| 39 |
#include "stk_util/util/ReportHandler.hpp" // for ThrowRequireMsg |
| 40 |
#include <algorithm> // for sort, binary_search |
| 41 |
#include <cstdint> // for uint64_t |
| 42 |
#include <memory> // for allocator_traits<>::value... |
| 43 |
#include <utility> // for pair |
| 44 |
#include <vector> // for vector |
| 45 |
|
| 46 |
|
| 47 |
//------------------------------------------------------------------------------------------------------------------- |
| 48 |
// Helper functions for test error checking, consolidate pairs of local order and resultant ids into a single |
| 49 |
// global sorted list. |
| 50 |
|
| 51 |
std::vector<std::pair<uint64_t, uint64_t> > GenerateGlobalPairedList(std::vector<uint64_t>& localOrderArray, std::vector<uint64_t>& newIdsLocal, stk::ParallelMachine comm) { |
| 52 |
|
| 53 |
STK_ThrowRequireMsg(localOrderArray.size() == newIdsLocal.size(),"Inconsistent Sizes in GenerateGlobalPairedList"); |
| 54 |
|
| 55 |
std::vector<std::pair<uint64_t, uint64_t> > idOrderPairsLocal; |
| 56 |
std::vector<std::pair<uint64_t, uint64_t> > idOrderPairsGlobal; |
| 57 |
for(uint64_t i = 0; i < localOrderArray.size(); ++i) { |
| 58 |
idOrderPairsLocal.emplace_back(localOrderArray[i], newIdsLocal[i]); |
| 59 |
} |
| 60 |
|
| 61 |
stk::parallel_vector_concat(comm, idOrderPairsLocal, idOrderPairsGlobal); |
| 62 |
|
| 63 |
std::sort(idOrderPairsGlobal.begin(), idOrderPairsGlobal.end()); |
| 64 |
|
| 65 |
return idOrderPairsGlobal; |
| 66 |
} |
| 67 |
|
| 68 |
//------------------------------------------------------------------------------------------------------------------- |
| 69 |
// Verify that a result generated from the consistent ids function meets all properties nessecary of consistent ids |
| 70 |
void VerifyConsistentIds(uint64_t maxAllowableId, |
| 71 |
std::vector<uint64_t>& existingIdsLocal, |
| 72 |
std::vector<uint64_t>& localOrderArray, |
| 73 |
std::vector<uint64_t>& newIdsLocal, |
| 74 |
stk::ParallelMachine comm) { |
| 75 |
|
| 76 |
EXPECT_EQ(localOrderArray.size(), newIdsLocal.size()); |
| 77 |
// |
| 78 |
// Generate a global list of all ids created matched with and sorted by their order array entry |
| 79 |
// |
| 80 |
std::vector<std::pair<uint64_t, uint64_t> > idOrderPairsGlobal = GenerateGlobalPairedList(localOrderArray, newIdsLocal, comm); |
| 81 |
|
| 82 |
std::vector<uint64_t> existingIdsGlobal; |
| 83 |
stk::parallel_vector_concat(comm, existingIdsLocal, existingIdsGlobal); |
| 84 |
std::sort(existingIdsGlobal.begin(), existingIdsGlobal.end()); |
| 85 |
// |
| 86 |
// Verify properties of the returned ids |
| 87 |
// |
| 88 |
for(uint64_t i=0; i<idOrderPairsGlobal.size(); ++i) { |
| 89 |
uint64_t curNewID = idOrderPairsGlobal[i].second; |
| 90 |
// 1) Verify all ids must be less than or equal to maxAllowableId |
| 91 |
EXPECT_LE(curNewID, maxAllowableId); |
| 92 |
// 2) Verify no overlap between existing and new id set |
| 93 |
EXPECT_FALSE(std::binary_search(existingIdsGlobal.begin(), existingIdsGlobal.end(), curNewID)); |
| 94 |
// 3) Verify ids come back in order dictated by the ordering array |
| 95 |
if(i > 0) { |
| 96 |
uint64_t oldNewID = idOrderPairsGlobal[i-1].second; |
| 97 |
EXPECT_LT(oldNewID, curNewID); |
| 98 |
EXPECT_LT(idOrderPairsGlobal[i-1].first, idOrderPairsGlobal[i].first); |
| 99 |
} |
| 100 |
} |
| 101 |
|
| 102 |
} |
| 103 |
|
| 104 |
//------------------------------------------------------------------------------ |
| 105 |
// |
| 106 |
// Test the generate_parallel_consistent_ids function in stk_util/parallel |
| 107 |
// |
| 108 |
TEST(UnitTestParallel, testGenerateParallelConsistentIDs) { |
| 109 |
int mpi_rank = stk::parallel_machine_rank(MPI_COMM_WORLD); |
| 110 |
int mpi_size = stk::parallel_machine_size(MPI_COMM_WORLD); |
| 111 |
// |
| 112 |
// Test 1, nominal element death case |
| 113 |
// Each processor starts with 1000 unique densly packed ids ordered 1->N. Add 10 new ids per processor |
| 114 |
// produced by kill every 100th element creating 10 new elements per processor. |
| 115 |
// |
| 116 |
uint64_t maxAllowableId = ~0U; |
| 117 |
|
| 118 |
std::vector<uint64_t> newIds1; |
| 119 |
std::vector<uint64_t> localOrderArray1; |
| 120 |
|
| 121 |
{ |
| 122 |
std::vector<uint64_t> existingIds(1000); |
| 123 |
uint64_t firstId = mpi_rank*1000; |
| 124 |
|
| 125 |
|
| 126 |
for(uint64_t i=0; i<1000; ++i) { |
| 127 |
uint64_t currentElemId = firstId+i; |
| 128 |
existingIds[i] = currentElemId; |
| 129 |
if(i%100 == 0) { |
| 130 |
// local order array is the 'source element' which provides a unique repeatable ordering |
| 131 |
localOrderArray1.push_back(currentElemId); |
| 132 |
} |
| 133 |
} |
| 134 |
newIds1 = stk::generate_parallel_consistent_ids(maxAllowableId, existingIds, localOrderArray1, MPI_COMM_WORLD); |
| 135 |
|
| 136 |
uint64_t localSize = newIds1.size(); |
| 137 |
uint64_t globalSize; |
| 138 |
MPI_Allreduce(&localSize, &globalSize, 1, MPI_UNSIGNED_LONG, MPI_SUM, MPI_COMM_WORLD); |
| 139 |
EXPECT_EQ(globalSize, (uint64_t)mpi_size*10); |
| 140 |
VerifyConsistentIds(maxAllowableId, existingIds, localOrderArray1, newIds1, MPI_COMM_WORLD); |
| 141 |
} |
| 142 |
// |
| 143 |
// Test 2, same total source data as case 1, but source data scattered among processors |
| 144 |
// |
| 145 |
std::vector<uint64_t> newIds2; |
| 146 |
std::vector<uint64_t> localOrderArray2; |
| 147 |
{ |
| 148 |
std::vector<uint64_t> existingIds; |
| 149 |
uint64_t destProc = mpi_size/2; |
| 150 |
for(int i=0; i<1000*mpi_size; ++i) { |
| 151 |
destProc += i; |
| 152 |
if(destProc >= (uint64_t)mpi_size) { |
| 153 |
destProc = destProc%mpi_size; |
| 154 |
} |
| 155 |
if(destProc == (uint64_t)mpi_rank) { |
| 156 |
existingIds.push_back(i); |
| 157 |
if(i%100 == 0) { |
| 158 |
localOrderArray2.push_back(i); |
| 159 |
} |
| 160 |
} |
| 161 |
} |
| 162 |
newIds2 = stk::generate_parallel_consistent_ids(maxAllowableId, existingIds, localOrderArray2, MPI_COMM_WORLD); |
| 163 |
|
| 164 |
uint64_t localSize = newIds2.size(); |
| 165 |
uint64_t globalSize; |
| 166 |
MPI_Allreduce(&localSize, &globalSize, 1, MPI_UNSIGNED_LONG, MPI_SUM, MPI_COMM_WORLD); |
| 167 |
EXPECT_EQ(globalSize, (uint64_t)mpi_size*10); |
| 168 |
VerifyConsistentIds(maxAllowableId, existingIds, localOrderArray2, newIds2, MPI_COMM_WORLD); |
| 169 |
} |
| 170 |
// |
| 171 |
// Test 3, verify test 2 and test 3 generate the same ids in identical order, confirming the id generation scheme is decomposition independent |
| 172 |
// |
| 173 |
std::vector<std::pair<uint64_t, uint64_t> > idOrderPairsGlobal1 = GenerateGlobalPairedList(localOrderArray1, newIds1, MPI_COMM_WORLD); |
| 174 |
std::vector<std::pair<uint64_t, uint64_t> > idOrderPairsGlobal2 = GenerateGlobalPairedList(localOrderArray2, newIds2, MPI_COMM_WORLD); |
| 175 |
EXPECT_EQ(idOrderPairsGlobal1.size(), idOrderPairsGlobal2.size()); |
| 176 |
for(uint64_t i=0; i < idOrderPairsGlobal1.size(); ++i) { |
| 177 |
EXPECT_EQ(idOrderPairsGlobal1[i].first, idOrderPairsGlobal2[i].first); |
| 178 |
EXPECT_EQ(idOrderPairsGlobal1[i].second, idOrderPairsGlobal2[i].second); |
| 179 |
} |
| 180 |
// |
| 181 |
// Test 4, sparse fill |
| 182 |
// |
| 183 |
{ |
| 184 |
std::vector<uint64_t> newIds; |
| 185 |
std::vector<uint64_t> existingIds(1000); |
| 186 |
std::vector<uint64_t> localOrderArray; |
| 187 |
|
| 188 |
uint64_t firstId = mpi_rank*1000; |
| 189 |
|
| 190 |
for(uint64_t i=0; i<1000; ++i) { |
| 191 |
uint64_t currentElemId = firstId+i; |
| 192 |
existingIds[i] = currentElemId; |
| 193 |
} |
| 194 |
|
| 195 |
if(mpi_rank == 0) { |
| 196 |
existingIds.push_back(~0U); |
| 197 |
} |
| 198 |
|
| 199 |
|
| 200 |
for(int i=0; i<10; ++i) { |
| 201 |
localOrderArray.push_back(mpi_rank*10 + i); |
| 202 |
} |
| 203 |
|
| 204 |
|
| 205 |
newIds = stk::generate_parallel_consistent_ids(maxAllowableId, existingIds, localOrderArray, MPI_COMM_WORLD); |
| 206 |
|
| 207 |
|
| 208 |
uint64_t localSize = newIds.size(); |
| 209 |
uint64_t globalSize; |
| 210 |
MPI_Allreduce(&localSize, &globalSize, 1, MPI_UNSIGNED_LONG, MPI_SUM, MPI_COMM_WORLD); |
| 211 |
EXPECT_EQ(globalSize, (uint64_t)mpi_size*10); |
| 212 |
VerifyConsistentIds(maxAllowableId, existingIds, localOrderArray, newIds, MPI_COMM_WORLD); |
| 213 |
} |
| 214 |
|
| 215 |
} |