blob: ac5a2e4c6ac8c81b61ab1775fd74d422ec26d2ca [file] [log] [blame]
/*
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you 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.
*/
package org.apache.sysml.runtime.io;
import java.io.File;
import java.io.IOException;
import org.apache.hadoop.fs.FileSystem;
import org.apache.hadoop.fs.Path;
import org.apache.hadoop.io.SequenceFile;
import org.apache.hadoop.mapred.JobConf;
import org.apache.sysml.conf.ConfigurationManager;
import org.apache.sysml.runtime.DMLRuntimeException;
import org.apache.sysml.runtime.controlprogram.ParForProgramBlock.PDataPartitionFormat;
import org.apache.sysml.runtime.matrix.data.MatrixBlock;
import org.apache.sysml.runtime.matrix.data.MatrixIndexes;
import org.apache.sysml.runtime.matrix.mapred.DistributedCacheInput;
import org.apache.sysml.runtime.matrix.mapred.MRConfigurationNames;
import org.apache.sysml.runtime.matrix.mapred.MRJobConfiguration;
import org.apache.sysml.runtime.util.MapReduceTool;
public class WriterBinaryBlock extends MatrixWriter
{
protected int _replication = -1;
public WriterBinaryBlock( int replication ) {
_replication = replication;
}
@Override
public final void writeMatrixToHDFS(MatrixBlock src, String fname, long rlen, long clen, int brlen, int bclen, long nnz)
throws IOException, DMLRuntimeException
{
//prepare file access
JobConf job = new JobConf(ConfigurationManager.getCachedJobConf());
FileSystem fs = FileSystem.get(job);
Path path = new Path( fname );
//if the file already exists on HDFS, remove it.
MapReduceTool.deleteFileIfExistOnHDFS( fname );
//set up preferred custom serialization framework for binary block format
if( MRJobConfiguration.USE_BINARYBLOCK_SERIALIZATION )
MRJobConfiguration.addBinaryBlockSerializationFramework( job );
//core write sequential/parallel
if( src.isDiag() )
writeDiagBinaryBlockMatrixToHDFS(path, job, fs, src, rlen, clen, brlen, bclen);
else
writeBinaryBlockMatrixToHDFS(path, job, fs, src, rlen, clen, brlen, bclen);
IOUtilFunctions.deleteCrcFilesFromLocalFileSystem(fs, path);
}
@Override
@SuppressWarnings("deprecation")
public final void writeEmptyMatrixToHDFS(String fname, long rlen, long clen, int brlen, int bclen)
throws IOException, DMLRuntimeException
{
JobConf job = new JobConf(ConfigurationManager.getCachedJobConf());
Path path = new Path( fname );
FileSystem fs = FileSystem.get(job);
SequenceFile.Writer writer = new SequenceFile.Writer(fs, job, path,
MatrixIndexes.class, MatrixBlock.class);
MatrixIndexes index = new MatrixIndexes(1, 1);
MatrixBlock block = new MatrixBlock((int)Math.min(rlen, brlen),
(int)Math.min(clen, bclen), true);
writer.append(index, block);
writer.close();
IOUtilFunctions.deleteCrcFilesFromLocalFileSystem(fs, path);
}
/**
*
* @param path
* @param job
* @param src
* @param rlen
* @param clen
* @param brlen
* @param bclen
* @throws IOException
* @throws DMLRuntimeException
*/
protected void writeBinaryBlockMatrixToHDFS( Path path, JobConf job, FileSystem fs, MatrixBlock src, long rlen, long clen, int brlen, int bclen )
throws IOException, DMLRuntimeException
{
//sequential write
writeBinaryBlockMatrixToSequenceFile(path, job, fs, src, brlen, bclen, 0, (int)rlen);
}
/**
*
* @param path
* @param job
* @param fs
* @param src
* @param brlen
* @param bclen
* @param rl
* @param ru
* @throws DMLRuntimeException
* @throws IOException
*/
@SuppressWarnings("deprecation")
protected final void writeBinaryBlockMatrixToSequenceFile( Path path, JobConf job, FileSystem fs, MatrixBlock src, int brlen, int bclen, int rl, int ru )
throws DMLRuntimeException, IOException
{
boolean sparse = src.isInSparseFormat();
int rlen = src.getNumRows();
int clen = src.getNumColumns();
// 1) create sequence file writer, with right replication factor
// (config via MRConfigurationNames.DFS_REPLICATION not possible since sequence file internally calls fs.getDefaultReplication())
SequenceFile.Writer writer = null;
if( _replication > 0 ) //if replication specified (otherwise default)
{
//copy of SequenceFile.Writer(fs, job, path, MatrixIndexes.class, MatrixBlock.class), except for replication
writer = new SequenceFile.Writer(fs, job, path, MatrixIndexes.class, MatrixBlock.class, job.getInt(MRConfigurationNames.IO_FILE_BUFFER_SIZE, 4096),
(short)_replication, fs.getDefaultBlockSize(), null, new SequenceFile.Metadata());
}
else
{
writer = new SequenceFile.Writer(fs, job, path, MatrixIndexes.class, MatrixBlock.class);
}
try
{
// 2) bound check for src block
if( src.getNumRows() > rlen || src.getNumColumns() > clen )
{
throw new IOException("Matrix block [1:"+src.getNumRows()+",1:"+src.getNumColumns()+"] " +
"out of overall matrix range [1:"+rlen+",1:"+clen+"].");
}
//3) reblock and write
MatrixIndexes indexes = new MatrixIndexes();
if( rlen <= brlen && clen <= bclen && rl == 0 ) //opt for single block
{
//directly write single block
indexes.setIndexes(1, 1);
writer.append(indexes, src);
}
else //general case
{
//initialize blocks for reuse (at most 4 different blocks required)
MatrixBlock[] blocks = createMatrixBlocksForReuse(rlen, clen, brlen, bclen, sparse, src.getNonZeros());
//create and write subblocks of matrix
for(int blockRow = rl/brlen; blockRow < (int)Math.ceil(ru/(double)brlen); blockRow++)
for(int blockCol = 0; blockCol < (int)Math.ceil(src.getNumColumns()/(double)bclen); blockCol++)
{
int maxRow = (blockRow*brlen + brlen < src.getNumRows()) ? brlen : src.getNumRows() - blockRow*brlen;
int maxCol = (blockCol*bclen + bclen < src.getNumColumns()) ? bclen : src.getNumColumns() - blockCol*bclen;
int row_offset = blockRow*brlen;
int col_offset = blockCol*bclen;
//get reuse matrix block
MatrixBlock block = getMatrixBlockForReuse(blocks, maxRow, maxCol, brlen, bclen);
//copy submatrix to block
src.sliceOperations( row_offset, row_offset+maxRow-1,
col_offset, col_offset+maxCol-1, block );
//append block to sequence file
indexes.setIndexes(blockRow+1, blockCol+1);
writer.append(indexes, block);
//reset block for later reuse
block.reset();
}
}
}
finally {
IOUtilFunctions.closeSilently(writer);
}
}
/**
*
* @param path
* @param job
* @param src
* @param rlen
* @param clen
* @param brlen
* @param bclen
* @param replication
* @throws IOException
* @throws DMLRuntimeException
*/
@SuppressWarnings("deprecation")
protected final void writeDiagBinaryBlockMatrixToHDFS( Path path, JobConf job, FileSystem fs, MatrixBlock src, long rlen, long clen, int brlen, int bclen )
throws IOException, DMLRuntimeException
{
boolean sparse = src.isInSparseFormat();
// 1) create sequence file writer, with right replication factor
// (config via MRConfigurationNames.DFS_REPLICATION not possible since sequence file internally calls fs.getDefaultReplication())
SequenceFile.Writer writer = null;
if( _replication > 0 ) //if replication specified (otherwise default)
{
//copy of SequenceFile.Writer(fs, job, path, MatrixIndexes.class, MatrixBlock.class), except for replication
writer = new SequenceFile.Writer(fs, job, path, MatrixIndexes.class, MatrixBlock.class, job.getInt(MRConfigurationNames.IO_FILE_BUFFER_SIZE, 4096),
(short)_replication, fs.getDefaultBlockSize(), null, new SequenceFile.Metadata());
}
else
{
writer = new SequenceFile.Writer(fs, job, path, MatrixIndexes.class, MatrixBlock.class);
}
try
{
// 2) bound check for src block
if( src.getNumRows() > rlen || src.getNumColumns() > clen )
{
throw new IOException("Matrix block [1:"+src.getNumRows()+",1:"+src.getNumColumns()+"] " +
"out of overall matrix range [1:"+rlen+",1:"+clen+"].");
}
//3) reblock and write
MatrixIndexes indexes = new MatrixIndexes();
if( rlen <= brlen && clen <= bclen ) //opt for single block
{
//directly write single block
indexes.setIndexes(1, 1);
writer.append(indexes, src);
}
else //general case
{
//initialize blocks for reuse (at most 4 different blocks required)
MatrixBlock[] blocks = createMatrixBlocksForReuse(rlen, clen, brlen, bclen, sparse, src.getNonZeros());
MatrixBlock emptyBlock = new MatrixBlock();
//create and write subblocks of matrix
for(int blockRow = 0; blockRow < (int)Math.ceil(src.getNumRows()/(double)brlen); blockRow++)
for(int blockCol = 0; blockCol < (int)Math.ceil(src.getNumColumns()/(double)bclen); blockCol++)
{
int maxRow = (blockRow*brlen + brlen < src.getNumRows()) ? brlen : src.getNumRows() - blockRow*brlen;
int maxCol = (blockCol*bclen + bclen < src.getNumColumns()) ? bclen : src.getNumColumns() - blockCol*bclen;
MatrixBlock block = null;
if( blockRow==blockCol ) //block on diagonal
{
int row_offset = blockRow*brlen;
int col_offset = blockCol*bclen;
//get reuse matrix block
block = getMatrixBlockForReuse(blocks, maxRow, maxCol, brlen, bclen);
//copy submatrix to block
src.sliceOperations( row_offset, row_offset+maxRow-1,
col_offset, col_offset+maxCol-1, block );
}
else //empty block (not on diagonal)
{
block = emptyBlock;
block.reset(maxRow, maxCol);
}
//append block to sequence file
indexes.setIndexes(blockRow+1, blockCol+1);
writer.append(indexes, block);
//reset block for later reuse
if( blockRow!=blockCol )
block.reset();
}
}
}
finally {
IOUtilFunctions.closeSilently(writer);
}
}
/**
*
* @param path
* @param job
* @param src
* @param rlen
* @param clen
* @param brlen
* @param bclen
* @param pformat
* @throws IOException
* @throws DMLRuntimeException
*/
@SuppressWarnings("deprecation")
public final void writePartitionedBinaryBlockMatrixToHDFS( Path path, JobConf job, MatrixBlock src, long rlen, long clen, int brlen, int bclen, PDataPartitionFormat pformat )
throws IOException, DMLRuntimeException
{
boolean sparse = src.isInSparseFormat();
FileSystem fs = FileSystem.get(job);
//set up preferred custom serialization framework for binary block format
if( MRJobConfiguration.USE_BINARYBLOCK_SERIALIZATION )
MRJobConfiguration.addBinaryBlockSerializationFramework( job );
//initialize blocks for reuse (at most 4 different blocks required)
MatrixBlock[] blocks = createMatrixBlocksForReuse(rlen, clen, brlen, bclen, sparse, src.getNonZeros());
switch( pformat )
{
case ROW_BLOCK_WISE_N:
{
long numBlocks = ((rlen-1)/brlen)+1;
long numPartBlocks = (long)Math.ceil(((double)DistributedCacheInput.PARTITION_SIZE)/clen/brlen);
int count = 0;
for( int k = 0; k<numBlocks; k+=numPartBlocks )
{
// 1) create sequence file writer, with right replication factor
// (config via MRConfigurationNames.DFS_REPLICATION not possible since sequence file internally calls fs.getDefaultReplication())
Path path2 = new Path(path.toString()+File.separator+(++count));
SequenceFile.Writer writer = new SequenceFile.Writer(fs, job, path2, MatrixIndexes.class, MatrixBlock.class);
//3) reblock and write
try
{
MatrixIndexes indexes = new MatrixIndexes();
//create and write subblocks of matrix
for(int blockRow = k; blockRow < Math.min((int)Math.ceil(src.getNumRows()/(double)brlen),k+numPartBlocks); blockRow++)
for(int blockCol = 0; blockCol < (int)Math.ceil(src.getNumColumns()/(double)bclen); blockCol++)
{
int maxRow = (blockRow*brlen + brlen < src.getNumRows()) ? brlen : src.getNumRows() - blockRow*brlen;
int maxCol = (blockCol*bclen + bclen < src.getNumColumns()) ? bclen : src.getNumColumns() - blockCol*bclen;
int row_offset = blockRow*brlen;
int col_offset = blockCol*bclen;
//get reuse matrix block
MatrixBlock block = getMatrixBlockForReuse(blocks, maxRow, maxCol, brlen, bclen);
//copy submatrix to block
src.sliceOperations( row_offset, row_offset+maxRow-1,
col_offset, col_offset+maxCol-1, block );
//append block to sequence file
indexes.setIndexes(blockRow+1, blockCol+1);
writer.append(indexes, block);
//reset block for later reuse
block.reset();
}
}
finally
{
IOUtilFunctions.closeSilently(writer);
}
}
break;
}
case COLUMN_BLOCK_WISE_N:
{
long numBlocks = ((clen-1)/bclen)+1;
long numPartBlocks = (long)Math.ceil(((double)DistributedCacheInput.PARTITION_SIZE)/rlen/bclen);
int count = 0;
for( int k = 0; k<numBlocks; k+=numPartBlocks )
{
// 1) create sequence file writer, with right replication factor
// (config via MRConfigurationNames.DFS_REPLICATION not possible since sequence file internally calls fs.getDefaultReplication())
Path path2 = new Path(path.toString()+File.separator+(++count));
SequenceFile.Writer writer = new SequenceFile.Writer(fs, job, path2, MatrixIndexes.class, MatrixBlock.class);
//3) reblock and write
try
{
MatrixIndexes indexes = new MatrixIndexes();
//create and write subblocks of matrix
for(int blockRow = 0; blockRow < (int)Math.ceil(src.getNumRows()/(double)brlen); blockRow++)
for(int blockCol = k; blockCol < Math.min((int)Math.ceil(src.getNumColumns()/(double)bclen),k+numPartBlocks); blockCol++)
{
int maxRow = (blockRow*brlen + brlen < src.getNumRows()) ? brlen : src.getNumRows() - blockRow*brlen;
int maxCol = (blockCol*bclen + bclen < src.getNumColumns()) ? bclen : src.getNumColumns() - blockCol*bclen;
int row_offset = blockRow*brlen;
int col_offset = blockCol*bclen;
//get reuse matrix block
MatrixBlock block = getMatrixBlockForReuse(blocks, maxRow, maxCol, brlen, bclen);
//copy submatrix to block
src.sliceOperations( row_offset, row_offset+maxRow-1,
col_offset, col_offset+maxCol-1, block );
//append block to sequence file
indexes.setIndexes(blockRow+1, blockCol+1);
writer.append(indexes, block);
//reset block for later reuse
block.reset();
}
}
finally
{
IOUtilFunctions.closeSilently(writer);
}
}
break;
}
default:
throw new DMLRuntimeException("Unsupported partition format for distributed cache input: "+pformat);
}
}
}