spark-reviews mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From mengxr <...@git.apache.org>
Subject [GitHub] spark pull request: [SPARK-3974][MLlib] Distributed Block Matrix A...
Date Tue, 27 Jan 2015 02:02:04 GMT
Github user mengxr commented on a diff in the pull request:

    https://github.com/apache/spark/pull/3200#discussion_r23582380
  
    --- Diff: mllib/src/main/scala/org/apache/spark/mllib/linalg/distributed/BlockMatrix.scala
---
    @@ -0,0 +1,242 @@
    +/*
    + * 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.spark.mllib.linalg.distributed
    +
    +import breeze.linalg.{DenseMatrix => BDM}
    +
    +import org.apache.spark.{Logging, Partitioner}
    +import org.apache.spark.mllib.linalg._
    +import org.apache.spark.mllib.rdd.RDDFunctions._
    +import org.apache.spark.rdd.RDD
    +import org.apache.spark.storage.StorageLevel
    +
    +/**
    + * A grid partitioner, which stores every block in a separate partition.
    + *
    + * @param numRowBlocks Number of blocks that form the rows of the matrix.
    + * @param numColBlocks Number of blocks that form the columns of the matrix.
    + */
    +private[mllib] class GridPartitioner(
    +    val numRowBlocks: Int,
    +    val numColBlocks: Int,
    +    val numParts: Int) extends Partitioner {
    +  // Having the number of partitions greater than the number of sub matrices does not
help
    +  override val numPartitions = math.min(numParts, numRowBlocks * numColBlocks)
    +
    +  /**
    +   * Returns the index of the partition the SubMatrix belongs to. Tries to achieve block
wise
    +   * partitioning.
    +   *
    +   * @param key The key for the SubMatrix. Can be its position in the grid (its column
major index)
    +   *            or a tuple of three integers that are the final row index after the multiplication,
    +   *            the index of the block to multiply with, and the final column index after
the
    +   *            multiplication.
    +   * @return The index of the partition, which the SubMatrix belongs to.
    +   */
    +  override def getPartition(key: Any): Int = {
    +    key match {
    +      case (blockRowIndex: Int, blockColIndex: Int) =>
    +        getBlockId(blockRowIndex, blockColIndex)
    +      case (blockRowIndex: Int, innerIndex: Int, blockColIndex: Int) =>
    +        getBlockId(blockRowIndex, blockColIndex)
    +      case _ =>
    +        throw new IllegalArgumentException(s"Unrecognized key. key: $key")
    +    }
    +  }
    +
    +  /** Partitions sub-matrices as blocks with neighboring sub-matrices. */
    +  private def getBlockId(blockRowIndex: Int, blockColIndex: Int): Int = {
    +    val totalBlocks = numRowBlocks * numColBlocks
    +    // Gives the number of blocks that need to be in each partition
    +    val partitionRatio = math.ceil(totalBlocks * 1.0 / numPartitions).toInt
    +    // Number of neighboring blocks to take in each row
    +    val subBlocksPerRow = math.ceil(numRowBlocks * 1.0 / partitionRatio).toInt
    +    // Number of neighboring blocks to take in each column
    +    val subBlocksPerCol = math.ceil(numColBlocks * 1.0 / partitionRatio).toInt
    +    // Coordinates of the block
    +    val i = blockRowIndex / subBlocksPerRow
    +    val j = blockColIndex / subBlocksPerCol
    +    val blocksPerRow = math.ceil(numRowBlocks * 1.0 / subBlocksPerRow).toInt
    +    j * blocksPerRow + i
    +  }
    +
    +  /** Checks whether the partitioners have the same characteristics */
    +  override def equals(obj: Any): Boolean = {
    +    obj match {
    +      case r: GridPartitioner =>
    +        (this.numRowBlocks == r.numRowBlocks) && (this.numColBlocks == r.numColBlocks)
&&
    +          (this.numPartitions == r.numPartitions)
    +      case _ =>
    +        false
    +    }
    +  }
    +}
    +
    +/**
    + * Represents a distributed matrix in blocks of local matrices.
    + *
    + * @param rdd The RDD of SubMatrices (local matrices) that form this matrix
    + * @param nRows Number of rows of this matrix
    + * @param nCols Number of columns of this matrix
    + * @param numRowBlocks Number of blocks that form the rows of this matrix
    --- End diff --
    
    Could it be derived from `nRows` and `rowsPerBlock`? Having `nRows`, `numRowBlocks`, and
`rowsPerBlock` would leave space for inconsistent inputs.


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastructure@apache.org or file a JIRA ticket
with INFRA.
---

---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org


Mime
View raw message