Merged in documents/sima.md (pull request apache#1)

Updated SIMA.md to make sure we could build and link TVM correctly on

SIMA.md

@@ -6,79 +6,98 @@ The original instructions to install TVM are here: https://docs.tvm.ai/install/f
 Here, we will simplify the steps so to get you started quickly on your Macs.
 
 # Pre-requisites
-Install Homebrew, miniconda, create a tvm environment.
+## Prepare for conda environment
+Install `Homebrew`, `miniconda`, and create a tvm environment under conda.
 
-You can use pyenv too, just make sure to create a tvm environment.
+You can use `pyenv` too, just make sure to create a tvm environment.
 
-```shell script
-# install Homebrew
+```shell
+# Install Homebrew
 /bin/bash -c "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install.sh)"
-brew update; brew upgrade
+brew update && brew upgrade
 
 # Install miniconda
 brew cask install miniconda
 
-# open a new terminal with Command-N
+# Source conda.sh whenever opening a new terminal(Path to conda.sh may differ)
+echo ". /usr/local/Caskroom/miniconda/base/etc/profile.d/conda.sh" >> ~/.bash_profile
 
-# create tvm environment
-conda create -n tvm
+# Open a new terminal with Command-N or make sure you source ~/.bash_profile
+
+# Create tvm environment, make sure we use python3.7 (for now)
+conda create -n tvm python=3.7
 conda activate tvm
 ```
 
+Now we are in tvm environment!
+
+## Install required libraries
 Since Macs don't have Clang/LLVM with OpenMP, we install it
-```shell script
-conda install -y -c conda-forge llvmdev llvm-openmp llvm-tools clang clang-tools clang-dev
+```shell
+conda install -y -c conda-forge llvmdev llvm-openmp llvm-tools clang clang-tools clangdev
 ```
 
 Install ANTLR parser for TensorFlow parsing
-```shell script
+```shell
 conda install -y -c conda-forge antlr antlr-python-runtime
 ```
 
 Install build tools (can also be done with brew for system-wide)
-```shell script
+```shell
 conda install -y ninja cmake 
 ```
 
 Install misc packages
-```shell script
+```shell
 conda install -y -c conda-forge numpy decorator attrs tornado xgboost psutil cython
 ```
+
+Install glfw3
+```shell
+conda install -y -c menpo glfw3
+```
+
 Not required by useful
-```shell script
+```shell
 conda install -y -c conda-forge matplotlib opencv py-opencv yaml yapf python.app h5py hdf5 cffi
 ```
 
+Install tensorflow and tflite
+```shell
+# Make sure the pip command is pointing to conda
+which pip
+pip install tensorflow==1.14.0 tflite==2.1.0
+```
+
 # Setup
-```shell script
-git clone --recursive https://github.com/apache/incubator-tvm tvm
+```shell
+# Get into sima's TVM and update the repo
 cd tvm
 git submodule init
 git submodule update
 ```
 Now you are in ```$TVM_HOME```
 
 # Build
-```shell script
+```shell
 cd $TVM_HOME
 mkdir build
 cd build
 cp ../sima-config.cmake config.cmake
-cmake ..
+cmake -G Ninja ..
 ninja
 cd ..
 make cython
 ```
 It takes about 10mn to build and 2s for cython.
 
 # Add TVM python paths to system-wide paths
-```shell script
+```shell
 # at the end of ~/.zshrc, add
 export TVM_HOME=/path/to/tvm
 export PYTHONPATH=$TVM_HOME/python:$TVM_HOME/topi/python:${PYTHONPATH}
 ```
 
 # PyCharm
 - Point PyCharm to use your conda tvm environment.
-- Add those 2 paths ```$TVM_HOME/python:$TVM_HOME/topi/python``` to your project.
-
+- Add those 2 paths ```$TVM_HOME/python:$TVM_HOME/topi/python``` to your project.