[DOC] Document update (#329)

README.md

@@ -4,23 +4,25 @@
 [Installation](docs/how_to/install.md) |
 [Documentation](http://docs.tvmlang.org) |
 [Tutorials](http://tutorials.tvmlang.org) |
+[Operator Inventory](topi) |
 [FAQ](docs/faq.md) |
 [Contributors](CONTRIBUTORS.md) |
 [Release Notes](NEWS.md)
 
-TVM
-===
+TVM: Tensor IR Stack for Deep Learning Systems
+==============================================
+TVM is a Tensor intermediate representation(IR) stack for deep learning systems. It is designed to close the gap between the
+productivity-focused deep learning frameworks, and the performance- and efficiency-focused hardware backends.
+TVM works with deep learning frameworks to provide end to end compilation to different backends.
 
-TVM is a low level domain specific language(DSL) for compiling tensor computation pipelines.
-It is designed to compile multi-dimensional tensor algebra pipelines which
-are crucial to deep learning frameworks.
+License
+-------
+© Contributors, 2017. Licensed under an [Apache-2.0](https://github.com/dmlc/tvm/blob/master/LICENSE) license.
 
 Contribute to TVM
 -----------------
-Your help is very valuable to make the package better for everyone.
+TVM adopts apache committer model, we aim to create an open source project that is maintained and owned by the community.
+
 - [Contributor Guide](docs/how_to/contribute.md)
 - Please add your name to [CONTRIBUTORS.md](CONTRIBUTORS.md)
 - Please also update [NEWS.md](NEWS.md) on changes and improvements in API and codes.
-
-## Documentation
-The current documentation can be build locally via sphinx. See [docs](docs) folder for details.

docs/api/python/topi.rst

@@ -13,8 +13,16 @@ Index
    topi.tanh
    topi.log
    topi.sqrt
+   topi.sigmoid
+   topi.broadcast_to
+   topi.max
+   topi.sum
+   topi.min
    topi.nn.relu
+   topi.nn.dilate
    topi.nn.scale_shift
+   topi.nn.conv2d_nchw
+   topi.nn.conv2d_hwcn
    topi.nn.depthwise_conv2d
 
 
@@ -22,7 +30,11 @@ Index
 
 .. autosummary::
 
-   topi.cuda.schedule_depthwise_conv2d_map
+   topi.cuda.schedule_conv2d_nchw
+   topi.cuda.schedule_conv2d_hwcn
+   topi.cuda.schedule_depthwise_conv2d
+   topi.cuda.schedule_reduce
+   topi.cuda.schedule_broadcast_to
 
 
 topi
@@ -31,15 +43,27 @@ topi
 .. autofunction:: topi.tanh
 .. autofunction:: topi.log
 .. autofunction:: topi.sqrt
+.. autofunction:: topi.sigmoid
+.. autofunction:: topi.broadcast_to
+.. autofunction:: topi.max
+.. autofunction:: topi.sum
+.. autofunction:: topi.min
 
 topi.nn
 ~~~~~~~
 .. autofunction:: topi.nn.relu
+.. autofunction:: topi.nn.dilate
 .. autofunction:: topi.nn.scale_shift
+.. autofunction:: topi.nn.conv2d_nchw
+.. autofunction:: topi.nn.conv2d_hwcn
 .. autofunction:: topi.nn.depthwise_conv2d
 
 topi.cuda
 ~~~~~~~~~
 .. automodule:: topi.cuda
 
-.. autofunction:: topi.cuda.schedule_depthwise_conv2d_map
+.. autofunction:: topi.cuda.schedule_conv2d_nchw
+.. autofunction:: topi.cuda.schedule_conv2d_hwcn
+.. autofunction:: topi.cuda.schedule_depthwise_conv2d
+.. autofunction:: topi.cuda.schedule_reduce
+.. autofunction:: topi.cuda.schedule_broadcast_to

docs/faq.md

@@ -6,17 +6,27 @@ How to Install
 --------------
 See [Installation](https://github.com/dmlc/tvm/blob/master/docs/how_to/install.md)
 
-TVM's relation to XLA
----------------------
-They has different abstraction level.
-XLA is a higher level tensor algebra DSL, the system defines codegen and loop transformation
-rules for each kernels. TVM is an low level array index based DSL that give the loop transformation
-primitives to the user. In terms of design philosophy, TVM aims to be directly used by developers
-and provide general support for different framework via DLPack.
-See also [This Issue](https://github.com/dmlc/tvm/issues/151)
+TVM's relation to Other IR/DSL Projects
+---------------------------------------
+There are usually two levels of abstractions of IR in the deep learning systems.
+NNVM, TensorFlow's XLA and Intel's ngraph uses computation graph representation.
+This representation is high level, and can be helpful to perform generic optimizations
+such as memory reuse, layout transformation and automatic differentiation.
+
+TVM adopts a low level representation, that explicitly express the choice of memory
+layout, parallelization pattern, locality and hardware primtives etc.
+This level of IR is closer to directly target hardwares.
+The low level IR adopt ideas from existing image processing languages like Halide, darkroom
+and loop transformation tools like loopy and polyhedra based analysis.
+We specifically focus of expressing deep learning workloads(e.g. recurrence),
+optimization for different hardware backends and embedding with frameworks to provide
+end-to-end compilation stack.
+
 
 TVM's relation to libDNN cuDNN
 ------------------------------
 TVM can incorporate these library as external calls. One goal of TVM is to be able to
 generate high performing kernels. We will evolve TVM an incremental manner as
 we learn from the technics of manual kernel crafting and add these as primitives in DSL.
+See also [TVM Operator Inventory](https://github.com/dmlc/tvm/tree/master/topi) for
+recipes of operators in TVM.

docs/how_to/install.md

@@ -8,7 +8,7 @@ scratch on various systems. It consists of two steps:
 
 To get started, clone tvm repo from github. It is important to clone the submodules along, with ```--recursive``` option.
 ```bash
-git clone --recursive ssh://git@github.com/dmlc/tvm
+git clone --recursive https://github.com/dmlc/tvm
 ```
 For windows users who use github tools, you can open the git shell, and type the following command.
 ```bash

topi/README.md

-# TVM Operator Inventory
+# TOPI: TVM Operator Inventory
 
-topi is the operator collection library for TVM intended at sharing the effort of crafting and
-optimizing tvm generated kernels. The goal:
+TOPI is the operator collection library for TVM intended at sharing the effort of crafting
+and optimizing tvm generated kernels. The goal:
 
 - Provide sugars for operator declaration
 - Give common primitives for fused op creation.
@@ -21,6 +21,7 @@ optimizing tvm generated kernels. The goal:
   - Some kernels have requirements on shape and data layout, assert them
 - Data layout aware, if not specified in argument or in function, assume NCHW by default.
 
+
 ## Testcase
 - Add testcases to testout the schedule and dataflow in the TOPI workflow
 - Only do correctness testing without attaching compiler flags and only run it once.

topi/python/topi/__init__.py

 # pylint: disable=redefined-builtin, wildcard-import
 """TVM Operator Inventory.
 
-TOPI is the operator collection library for TVM intended at sharing the effort of crafting and
-optimizing tvm generated kernels.
+TOPI is the operator collection library for TVM, to provide sugars
+for constructing compute declaration as well as optimized schedules.
+
+Some of the schedule function may have been specially optimized for a
+specific workload.
 """
 from __future__ import absolute_import as _abs