Author: reinwald
Date: Thu Dec 21 19:49:38 2017
New Revision: 1818972
URL: http://svn.apache.org/viewvc?rev=1818972&view=rev
Log:
Add 1.0.0 release notes, and updated roadmap
Modified:
systemml/site/roadmap.html
Modified: systemml/site/roadmap.html
URL:
http://svn.apache.org/viewvc/systemml/site/roadmap.html?rev=1818972&r1=1818971&r2=1818972&view=diff
==============================================================================
--- systemml/site/roadmap.html (original)
+++ systemml/site/roadmap.html Thu Dec 21 19:49:38 2017
@@ -175,19 +175,75 @@
<section class="full-stripe full-stripe--alternate">
<div class="ml-container ml-container--narrow content-group">
<div class="col col-12">
- <h2>Planned for Future SystemML 1.0</h2>
+ <h2>Planned for Future SystemML 1.1</h2>
<ul>
- <li>Compression (additional operations)</li>
- <li>Experimental Features:
+ <li>Algorithms & builtin functions
<ul>
- <li>Deep Learning</li>
- <li>Single GPU support</li>
- <li>Native BLAS support</li>
- <li>Code Generation</li>
+ <li>NN layers-based factorization machines with regression &
classification capabilities</li>
+ <li>NN optimization test suite with well known optimization test
functions</li>
+ <li>Model selection & hyper parameter tuning</li>
+ <li>Additional distribution functions, e.g. weibull, gamma</li>
+ <li>Generalization of operations, such as xor, and other
operations</li>
+ </ul>
+ </li>
+ <li>Enhanced Deep Learning support
+ <ul>
+ <li>Coherent sparse operations on CPU/GPU</li>
+ <li>Coherent single-precision support on CPU/GPU</li>
+ <li>Distributed DL operations</li>
+ </ul>
+ </li>
+ <li>GPU Support
+ <ul>
+ <li>Full compiler integration (cost-based, automatic
placement)</li>
+ <li>Multi GPUs</li>
+ <li>Distributed GPUs</li>
+ </ul>
+ </li>
+ <li>Code generation
+ <ul>
+ <li>Deep learning operationss</li>
+ <li>Heterogeneous HW, incl GPUs</li>
+ </ul>
+ </li>
+ <li>Compressed Linear Algebra
+ <ul>
+ <li>Matrix-matrix multiplications</li>
+ <li>Deep learning operations</li>
+ <li>Ultra-sparse datasets</li>
+ </ul>
+ </li>
+ <li>Misc Runtime
+ <ul>
+ <li>Large dense matrix blocks > 16GB</li>
+ <li>NUMA-awareness (thread pools, matrix partitioning)</li>
+ <li>Unified memory management (ops, bufferpool,
RDDs/broadcasts)</li>
+ <li>Support additional external formats such as feather format for
matrices and frames</li>
+ <li>Parfor support for broadcasts</li>
+ <li>Extended support for multi-threaded operations</li>
+ <li>Boolean matrices</li>
+ </ul>
+ </li>
+ <li>Misc Compiler
+ <ul>
+ <li>Support single-output UDFs in expressions</li>
+ <li>Consolidate replicated compilation chain (e.g., diff APIs)</li>
+ <li>Holistic sum-product optimization and operator fusion</li>
+ <li>Extended sparsity estimators</li>
+ <li>Rewrites and compiler improvements for mini-batching including
prefetching</li>
+ <li>Parfor optimizer support for shared reads</li>
+ <li>SPOOF compiler improvement</li>
+ </ul>
+ </li>
+ <li>APIs
+ <ul>
+ <li>Python Binding for JMLC API</li>
+ <li>Consistency Python/Java APIs</li>
</ul>
+
</li>
- <li>Rigorous Performance and Scalability Testing (Bug Fixes)</li>
- <li>Remove Deprecated APIs/Functions</li>
+
+
</ul>
</div>
@@ -211,6 +267,40 @@
<h2>Current Release</h2>
<ul>
<li>
+ <strong>SystemML 1.0.0 (<a
href="release-notes/systemml-release-notes-1.0.0.html">released</a> in
December, 2017)
+ <a
href="https://issues.apache.org/jira/secure/ReleaseNote.jspa?projectId=12319522&version=12338328";>details</a></strong>
+ <ul>
+ <li>Enhanced Deep Learning support with enhanced NN layers and
functions, Caffe2DML, and operator implementation</li>
+ <li>Native BLAS support</li>
+ <li>Additional algorithms: autoencoder, enhanced PCA</li>
+ <li>Enhanced rewrites, IPA, vectorization, and instruction
generation</li>
+ <li>Enhanced JMLC API, e.g. prepared scripts with thread affinity
for outputs and configs, script cloning, configuration management</li>
+ <li>SystemML Lite artifact</li>
+ <li>Compression on by default</li>
+ </ul>
+ </li>
+ <li>
+ Experimental Features
+ <ul>
+ <li>Keras2DML.</li>
+ <li>Enhanced code generation, code gen optimizer, and
multi-threaded codegen operators/li>
+ <li>Enhanced GPU support</li>
+ </ul>
+ </li>
+ <li>
+ Removals
+ <ul>
+ <li>Dropped JDK 7 support</li>
+ </ul>
+ </li>
+ </ul>
+ </div>
+
+ <div class="col col-12">
+ <h2>Prior Releases</h2>
+
+ <ul>
+ <li>
<strong>SystemML 0.15.0 (<a
href="release-notes/systemml-release-notes-0.15.0.html">released</a> in
September, 2017)
<a
href="https://issues.apache.org/jira/secure/ReleaseNote.jspa?projectId=12319522&version=12341587";>details</a></strong>
<ul>
@@ -233,10 +323,6 @@
</ul>
</li>
</ul>
- </div>
-
- <div class="col col-12">
- <h2>Prior Releases</h2>
<ul>
<li>
@@ -250,8 +336,9 @@
<li>
Experimental Features
<ul>
- <li>New Code Generation capabilities for automatic operator fusion
(basic code generator, compiler integration, runtime integration, in-memory
source code compilation, extended explain tool, support for right indexing and
replace in cellwise and row aggregate templates, support for row, column, or no
aggregation in rowwise template).
- Note code generation provides significant performance gains with
fewer read/write intermediates, reduced scans of inputs and intermediates, and
enhanced sparsity exploitation. To enable this feature, set codegen.enabled
property to true in SystemML-config.xml file.</li>
+ <li>New Code Generation capabilities for automatic operator fusion
(basic code generator, compiler integration, runtime integration, in-memory
source code compilation, extended explain tool, support for right indexing and
replace in cellwise and
+ row aggregate templates, support for row, column, or no
aggregation in rowwise template). Note code generation provides significant
performance gains with fewer read/write intermediates, reduced scans of inputs
and intermediates, and enhanced
+ sparsity exploitation. To enable this feature, set
codegen.enabled property to true in SystemML-config.xml file.</li>
<li>New instructions and operators for GPU support
(relu_maxpooling, conv2d_bias_add, bias_multiply)</li>
</ul>
</li>
@@ -270,30 +357,30 @@
<a
href="https://issues.apache.org/jira/secure/ReleaseNote.jspa?projectId=12319522&version=12339247";>details</a></strong>
<ul>
<li>Updated build for Spark 2.1.0</li>
- <li>New simplification rewrites for stratstats</li>
- <li>New fused operator tack+* in CP and Spark</li>
+ <li>New simplification rewrites for stratstats</li>
+ <li>New fused operator tack+* in CP and Spark</li>
<li>New dmlFromResource capability in Python (equivalent to
Scala)</li>
- <li>Add input float support to MLContext</li>
+ <li>Add input float support to MLContext</li>
</ul>
</li>
<li>
Documentation Enhancements
<ul>
<li>Deploy versioned documentation to main project website</li>
- <li>Add python mlcontext example to engine dev
guide</li>
- <li>Add MLContext info functionality to docs</li>
- <li>Update DML Language Reference for write description
parameter</li>
+ <li>Add python mlcontext example to engine dev guide</li>
+ <li>Add MLContext info functionality to docs</li>
+ <li>Update DML Language Reference for write description
parameter</li>
</ul>
</li>
<li>
Deprecations, Removals
<ul>
<li>Deprecate old MLContext API</li>
- <li>Deprecate parfor perftesttool</li>
- <li>Deprecate SQLContext methods</li>
- <li>Replace deprecated Accumulator with
AccumulatorV2</li>
- <li>Replace append with cbind for matrices</li>
- <li>Migrate Vector and LabeledPoint classes from mllib
to ml</li>
+ <li>Deprecate parfor perftesttool</li>
+ <li>Deprecate SQLContext methods</li>
+ <li>Replace deprecated Accumulator with AccumulatorV2</li>
+ <li>Replace append with cbind for matrices</li>
+ <li>Migrate Vector and LabeledPoint classes from mllib to ml</li>
</ul>
</li>
<li>
@@ -312,8 +399,8 @@
<li>Support pip install of new python package</li>
<li>Allow NumPy arrays, Pandas DataFrame and SciPy matrices as
input to MLContext</li>
<li>Improve SystemML Python DSL for NumPy</li>
- <li>Updated build for Spark 1.6.0</li>
- <li>DML utility script to shuffle input dataset</li>
+ <li>Updated build for Spark 1.6.0</li>
+ <li>DML utility script to shuffle input dataset</li>
</ul>
</li>
<li>
