[DOCS] Jenkins deployment of docs, add FAQ (#157)

Jenkinsfile

@@ -153,7 +153,19 @@ stage('Integration Test') {
         timeout(time: max_time, unit: 'MINUTES') {
           sh "${docker_run} gpu ./tests/scripts/task_python_docs.sh"
         }
+        pack_lib('mydocs', 'docs.tgz')
       }
     }
   }
 }
+
+stage('Deploy') {
+    node('docker' && 'doc') {
+      ws('workspace/tvm/deploy-docs') {
+        if (env.BRANCH_NAME == "master") {
+           unpack_lib('mydocs', 'docs.tgz')
+           sh "tar xf docs.tgz -C /var/docs"
+        }
+      }
+    }
+}

README.md

-[![Build Status](https://travis-ci.com/dmlc/tvm.svg?token=mkPkB7rD7Myq9dXZa9GV&branch=master)](https://travis-ci.com/dmlc/tvm)
 [![GitHub license](http://dmlc.github.io/img/apache2.svg)](./LICENSE)
+[![Build Status](http://mode-gpu.cs.washington.edu:8080/buildStatus/icon?job=dmlc/tvm/master)](http://mode-gpu.cs.washington.edu:8080/job/dmlc/job/tvm/job/master/)
 
 [Installation](docs/how_to/install.md) |
-[Documentation](docs) |
-[Tutorials](tutorials) |
+[Documentation](http://docs.tvmlang.org) |
+[Tutorials](http://tutorials.tvmlang.org) |
+[FAQ](docs/faq.md) |
 [Contributors](CONTRIBUTORS.md) |
 [Release Notes](NEWS.md)
 

docs/README.txt

 The documentation of tvm is generated with recommonmark and sphinx.
 
+- A hosted version of doc is at http://docs.tvmlang.org
 - pip install sphinx>=1.5.5 sphinx-gallery sphinx_rtd_theme matplotlib Image recommonmark
 - Build tvm first in the root folder.
 - To build locally, you need to enable USE_CUDA, USE_OPENCL, LLVM_CONFIG in config.mk and then type "make html" in this folder.

docs/faq.md

+Frequently Asked Questions
+==========================
+This document contains frequently asked questions.
+
+How to Install
+--------------
+See [Installation](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 libDNN cuDNN
+------------------------------
+TVM can incoporate 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.

docs/index.rst

@@ -12,6 +12,7 @@ Contents
    self
    how_to/install
    tutorials/index
+   faq
    api/python/index
    how_to/contribute
    genindex

tests/ci_build/Dockerfile.cpu

@@ -5,6 +5,5 @@ RUN apt-get update
 
 COPY install/ubuntu_*.sh /install/
 RUN bash /install/ubuntu_install_core.sh
-RUN apt-get update && apt-get install -y python-pip
 RUN bash /install/ubuntu_install_python.sh
 RUN bash /install/ubuntu_install_iverilog.sh

tests/ci_build/Dockerfile.gpu

-FROM nvidia/cuda:cudnn
+FROM nvidia/cuda:7.5-cudnn5-devel
 
 # Base scripts
 RUN apt-get update

tests/ci_build/install/ubuntu_install_python.sh

 # install libraries for python package on ubuntu
-apt-get update && apt-get install -y python-dev python3-dev
+apt-get update && apt-get install -y python-pip python-dev python3-dev
 
 # the version of the pip shipped with ubuntu may be too lower, install a recent version here
 cd /tmp && wget https://bootstrap.pypa.io/get-pip.py && python3 get-pip.py && python2 get-pip.py

tests/python/integration/test_ewise.py

@@ -59,13 +59,6 @@ def test_log_llvm():
     np.testing.assert_allclose(
         b.asnumpy(), np.log(a.asnumpy()), rtol=1e-5)
 
-from tvm.contrib import nvcc_compiler
-
-@tvm.register_func
-def tvm_callback_cuda_compile(code):
-    print(code)
-    ptx =  nvcc_compiler.compile_source(code, target="ptx", options=["-arch=sm_35"])
-    return ptx
 
 def test_add():
     # graph

tests/python/integration/test_gemm.py

 import tvm
-from tvm.contrib import nvcc_compiler
-from tvm.contrib import metal_compiler
 import numpy as np
 import time
 
-#@tvm.register_func
-def tvm_callback_metal_compile(code):
-    lib = metal_compiler.compile_source(code)
-    return lib
-
 def test_gemm():
     # graph
     nn = 1024

tests/scripts/task_python_docs.sh

 #!/bin/bash
 cd docs
 PYTHONPATH=../python make html
+cd _build/html
+tar czf docs.tgz *
+mv docs.tgz ../../../