Examples with KahanObject org.apache.sysml.runtime.instructions.cp.KahanObject used on opensource projects

Example 21 with KahanObject

use of org.apache.sysml.runtime.instructions.cp.KahanObject in project incubator-systemml by apache.

the class LibMatrixAgg method aggregateBinaryMatrixSparseGeneric.

private static void aggregateBinaryMatrixSparseGeneric(MatrixBlock in, MatrixBlock aggVal, MatrixBlock aggCorr) throws DMLRuntimeException {
    if (in.isEmptyBlock(false))
        return;
    SparseBlock a = in.getSparseBlock();
    KahanObject buffer1 = new KahanObject(0, 0);
    KahanPlus akplus = KahanPlus.getKahanPlusFnObject();
    final int m = in.rlen;
    final int rlen = Math.min(a.numRows(), m);
    for (int i = 0; i < rlen; i++) {
        if (!a.isEmpty(i)) {
            int apos = a.pos(i);
            int alen = a.size(i);
            int[] aix = a.indexes(i);
            double[] avals = a.values(i);
            for (int j = apos; j < apos + alen; j++) {
                int jix = aix[j];
                buffer1._sum = aggVal.quickGetValue(i, jix);
                buffer1._correction = aggCorr.quickGetValue(i, jix);
                akplus.execute2(buffer1, avals[j]);
                aggVal.quickSetValue(i, jix, buffer1._sum);
                aggCorr.quickSetValue(i, jix, buffer1._correction);
            }
        }
    }
    //note: nnz of aggVal/aggCorr maintained internally 
    aggVal.examSparsity();
    aggCorr.examSparsity();
}

Example 22 with KahanObject

use of org.apache.sysml.runtime.instructions.cp.KahanObject in project incubator-systemml by apache.

the class LibMatrixAgg method cumaggregateUnaryMatrixDense.

private static void cumaggregateUnaryMatrixDense(MatrixBlock in, MatrixBlock out, AggType optype, ValueFunction vFn, double[] agg, int rl, int ru) throws DMLRuntimeException {
    final int m = in.rlen;
    final int n = in.clen;
    double[] a = in.getDenseBlock();
    double[] c = out.getDenseBlock();
    switch(optype) {
        case //CUMSUM
        CUM_KAHAN_SUM:
            {
                KahanObject kbuff = new KahanObject(0, 0);
                KahanPlus kplus = KahanPlus.getKahanPlusFnObject();
                d_ucumkp(a, agg, c, m, n, kbuff, kplus, rl, ru);
                break;
            }
        case //CUMPROD
        CUM_PROD:
            {
                d_ucumm(a, agg, c, m, n, rl, ru);
                break;
            }
        case CUM_MIN:
        case CUM_MAX:
            {
                double init = Double.MAX_VALUE * ((optype == AggType.CUM_MAX) ? -1 : 1);
                d_ucummxx(a, agg, c, m, n, init, (Builtin) vFn, rl, ru);
                break;
            }
        default:
            throw new DMLRuntimeException("Unsupported cumulative aggregation type: " + optype);
    }
}

Example 23 with KahanObject

use of org.apache.sysml.runtime.instructions.cp.KahanObject in project incubator-systemml by apache.

the class CompressedMatrixBlock method aggregateUnaryOperations.

@Override
public MatrixValue aggregateUnaryOperations(AggregateUnaryOperator op, MatrixValue result, int blockingFactorRow, int blockingFactorCol, MatrixIndexes indexesIn, boolean inCP) throws DMLRuntimeException {
    //call uncompressed matrix mult if necessary
    if (!isCompressed()) {
        return super.aggregateUnaryOperations(op, result, blockingFactorRow, blockingFactorCol, indexesIn, inCP);
    }
    //check for supported operations
    if (!(op.aggOp.increOp.fn instanceof KahanPlus || op.aggOp.increOp.fn instanceof KahanPlusSq || (op.aggOp.increOp.fn instanceof Builtin && (((Builtin) op.aggOp.increOp.fn).getBuiltinCode() == BuiltinCode.MIN || ((Builtin) op.aggOp.increOp.fn).getBuiltinCode() == BuiltinCode.MAX)))) {
        throw new DMLRuntimeException("Unary aggregates other than sum/sumsq/min/max not supported yet.");
    }
    Timing time = LOG.isDebugEnabled() ? new Timing(true) : null;
    //prepare output dimensions
    CellIndex tempCellIndex = new CellIndex(-1, -1);
    op.indexFn.computeDimension(rlen, clen, tempCellIndex);
    if (op.aggOp.correctionExists) {
        switch(op.aggOp.correctionLocation) {
            case LASTROW:
                tempCellIndex.row++;
                break;
            case LASTCOLUMN:
                tempCellIndex.column++;
                break;
            case LASTTWOROWS:
                tempCellIndex.row += 2;
                break;
            case LASTTWOCOLUMNS:
                tempCellIndex.column += 2;
                break;
            default:
                throw new DMLRuntimeException("unrecognized correctionLocation: " + op.aggOp.correctionLocation);
        }
    }
    // initialize and allocate the result
    if (result == null)
        result = new MatrixBlock(tempCellIndex.row, tempCellIndex.column, false);
    else
        result.reset(tempCellIndex.row, tempCellIndex.column, false);
    MatrixBlock ret = (MatrixBlock) result;
    ret.allocateDenseBlock();
    //special handling init value for rowmins/rowmax
    if (op.indexFn instanceof ReduceCol && op.aggOp.increOp.fn instanceof Builtin) {
        double val = Double.MAX_VALUE * ((((Builtin) op.aggOp.increOp.fn).getBuiltinCode() == BuiltinCode.MAX) ? -1 : 1);
        Arrays.fill(ret.getDenseBlock(), val);
    }
    //core unary aggregate
    if (op.getNumThreads() > 1 && getExactSizeOnDisk() > MIN_PAR_AGG_THRESHOLD) {
        //multi-threaded execution of all groups 
        ArrayList<ColGroup>[] grpParts = createStaticTaskPartitioning((op.indexFn instanceof ReduceCol) ? 1 : op.getNumThreads(), false);
        ColGroupUncompressed uc = getUncompressedColGroup();
        try {
            //compute uncompressed column group in parallel (otherwise bottleneck)
            if (uc != null)
                ret = (MatrixBlock) uc.getData().aggregateUnaryOperations(op, ret, blockingFactorRow, blockingFactorCol, indexesIn, false);
            //compute all compressed column groups
            ExecutorService pool = Executors.newFixedThreadPool(op.getNumThreads());
            ArrayList<UnaryAggregateTask> tasks = new ArrayList<UnaryAggregateTask>();
            if (op.indexFn instanceof ReduceCol && grpParts.length > 0) {
                int blklen = BitmapEncoder.getAlignedBlocksize((int) (Math.ceil((double) rlen / op.getNumThreads())));
                for (int i = 0; i < op.getNumThreads() & i * blklen < rlen; i++) tasks.add(new UnaryAggregateTask(grpParts[0], ret, i * blklen, Math.min((i + 1) * blklen, rlen), op));
            } else
                for (ArrayList<ColGroup> grp : grpParts) tasks.add(new UnaryAggregateTask(grp, ret, 0, rlen, op));
            List<Future<MatrixBlock>> rtasks = pool.invokeAll(tasks);
            pool.shutdown();
            //aggregate partial results
            if (op.indexFn instanceof ReduceAll) {
                if (op.aggOp.increOp.fn instanceof KahanFunction) {
                    KahanObject kbuff = new KahanObject(ret.quickGetValue(0, 0), 0);
                    for (Future<MatrixBlock> rtask : rtasks) {
                        double tmp = rtask.get().quickGetValue(0, 0);
                        ((KahanFunction) op.aggOp.increOp.fn).execute2(kbuff, tmp);
                    }
                    ret.quickSetValue(0, 0, kbuff._sum);
                } else {
                    double val = ret.quickGetValue(0, 0);
                    for (Future<MatrixBlock> rtask : rtasks) {
                        double tmp = rtask.get().quickGetValue(0, 0);
                        val = op.aggOp.increOp.fn.execute(val, tmp);
                    }
                    ret.quickSetValue(0, 0, val);
                }
            }
        } catch (Exception ex) {
            throw new DMLRuntimeException(ex);
        }
    } else {
        //process UC column group
        for (ColGroup grp : _colGroups) if (grp instanceof ColGroupUncompressed)
            grp.unaryAggregateOperations(op, ret);
        //process OLE/RLE column groups
        aggregateUnaryOperations(op, _colGroups, ret, 0, rlen);
    }
    //special handling zeros for rowmins/rowmax
    if (op.indexFn instanceof ReduceCol && op.aggOp.increOp.fn instanceof Builtin) {
        int[] rnnz = new int[rlen];
        for (ColGroup grp : _colGroups) grp.countNonZerosPerRow(rnnz, 0, rlen);
        Builtin builtin = (Builtin) op.aggOp.increOp.fn;
        for (int i = 0; i < rlen; i++) if (rnnz[i] < clen)
            ret.quickSetValue(i, 0, builtin.execute2(ret.quickGetValue(i, 0), 0));
    }
    //drop correction if necessary
    if (op.aggOp.correctionExists && inCP)
        ret.dropLastRowsOrColums(op.aggOp.correctionLocation);
    //post-processing
    ret.recomputeNonZeros();
    if (LOG.isDebugEnabled())
        LOG.debug("Compressed uagg k=" + op.getNumThreads() + " in " + time.stop());
    return ret;
}

Example 24 with KahanObject

use of org.apache.sysml.runtime.instructions.cp.KahanObject in project incubator-systemml by apache.

the class ColGroupRLE method computeColSums.

@Override
protected final void computeColSums(MatrixBlock result, KahanFunction kplus) {
    KahanObject kbuff = new KahanObject(0, 0);
    final int numCols = getNumCols();
    final int numVals = getNumValues();
    for (int k = 0; k < numVals; k++) {
        int boff = _ptr[k];
        int blen = len(k);
        int valOff = k * numCols;
        int curRunEnd = 0;
        int count = 0;
        for (int bix = 0; bix < blen; bix += 2) {
            int curRunStartOff = curRunEnd + _data[boff + bix];
            curRunEnd = curRunStartOff + _data[boff + bix + 1];
            count += curRunEnd - curRunStartOff;
        }
        //scale counts by all values
        for (int j = 0; j < numCols; j++) {
            kbuff.set(result.quickGetValue(0, _colIndexes[j]), result.quickGetValue(1, _colIndexes[j]));
            kplus.execute3(kbuff, _values[valOff + j], count);
            result.quickSetValue(0, _colIndexes[j], kbuff._sum);
            result.quickSetValue(1, _colIndexes[j], kbuff._correction);
        }
    }
}

Example 25 with KahanObject

use of org.apache.sysml.runtime.instructions.cp.KahanObject in project incubator-systemml by apache.

the class GroupedAggMRCombiner method reduce.

@Override
public void reduce(TaggedMatrixIndexes key, Iterator<WeightedCell> values, OutputCollector<TaggedMatrixIndexes, WeightedCell> out, Reporter reporter) throws IOException {
    long start = System.currentTimeMillis();
    //get aggregate operator
    GroupedAggregateInstruction ins = grpaggInstructions.get(key.getTag());
    Operator op = ins.getOperator();
    boolean isPartialAgg = true;
    //combine iterator to single value
    try {
        if (//everything except sum
        op instanceof CMOperator) {
            if (((CMOperator) op).isPartialAggregateOperator()) {
                cmObj.reset();
                CM lcmFn = cmFn.get(key.getTag());
                //partial aggregate cm operator 
                while (values.hasNext()) {
                    WeightedCell value = values.next();
                    lcmFn.execute(cmObj, value.getValue(), value.getWeight());
                }
                outCell.setValue(cmObj.getRequiredPartialResult(op));
                outCell.setWeight(cmObj.getWeight());
            } else //forward tuples to reducer
            {
                isPartialAgg = false;
                while (values.hasNext()) out.collect(key, values.next());
            }
        } else if (//sum
        op instanceof AggregateOperator) {
            AggregateOperator aggop = (AggregateOperator) op;
            if (aggop.correctionExists) {
                KahanObject buffer = new KahanObject(aggop.initialValue, 0);
                KahanPlus.getKahanPlusFnObject();
                //partial aggregate with correction
                while (values.hasNext()) {
                    WeightedCell value = values.next();
                    aggop.increOp.fn.execute(buffer, value.getValue() * value.getWeight());
                }
                outCell.setValue(buffer._sum);
                outCell.setWeight(1);
            } else //no correction
            {
                double v = aggop.initialValue;
                //partial aggregate without correction
                while (values.hasNext()) {
                    WeightedCell value = values.next();
                    v = aggop.increOp.fn.execute(v, value.getValue() * value.getWeight());
                }
                outCell.setValue(v);
                outCell.setWeight(1);
            }
        } else
            throw new IOException("Unsupported operator in instruction: " + ins);
    } catch (Exception ex) {
        throw new IOException(ex);
    }
    //collect the output (to reducer)
    if (isPartialAgg)
        out.collect(key, outCell);
    reporter.incrCounter(Counters.COMBINE_OR_REDUCE_TIME, System.currentTimeMillis() - start);
}