hqd2009 / ptmc Goto Github PK

//===----------------------------------------------------------------------===//
//                  Steps for building DTMC infrastructure
//===----------------------------------------------------------------------===//


//===----------------------- Contact

Contact information: Martin Nowack ([email protected])
                     Torvald Riegel ([email protected])

//===----------------------- Contents

This directory contains the Dresden Compiler Infrastructure:

   - LLVM_28_R3 : Low Level Virtual Machine 2.8
                  (available as well under
                  http://www.llvm.org/releases/download.html#2.8 )
   - LLVM_GCC_R3: llvm-gcc-4.2 (based on LLVM 2.8) with support for
                  Transactional Memory language extensions 
   - TANGER_R3  : Tanger - Transactification tool
                  Feature Freeze 3


//===----------------------- Building the Dresden TM Compiler

//===----------------------- Prerequisites
LLVM-GCC front end needs at least the version 2.20 of binutils. If your system 
does not have this version, We recommend to replace your current version for a 
newer one or to install the new version in a local folder a reconfigure your 
environment values. For example, if you are using bash and you installed the 
binutils under /opt/binutils-2.20 folder, you can use the following 
configuration:

   export PATH="/opt/binutils-2.20/bin:$PATH"
   export LD_LIBRARY_PATH="/opt/binutils-2.20/lib:$LD_LIBRARY_PATH"
   export LD_RUN_PATH="/opt/binutils-2.20/lib:$LD_RUN_PATH"

//===----------------------- Introduction

The three components described above should be configured, compiled and 
installed in this order:

   1.- LLVM_28_R3
   2.- LLVM_GCC_R3
   3.- TANGER_R3


This file will talk you through the configuration steps necessary to get DTMC 
up and running on Linux. We will not discuss LLVM-GCC/LLVM itself (the LLVM 
projects have its own documentation available) nor will we talk about 
specialized setups that might be very interesting for LLVM setups but are 
not LLVM-related.

//===----------------------- Installation

0.- Unpack the source files (LLVM_28, LLVM_GCC_R3 and TANGER) in a working 
directory. LLVM will need the LLVM-GCC sources location for its installation.
You can unpack the files using "tar xvf filename.tar.gz -C destinationFolder"

1.-   LLVM 2.8 Release 3
1.a.- Configure LLVM 2.8 Release 3

   Run configure from the folder that you unpack LLVM_28.
   This is an example of parameters that can be used:
   ./configure --prefix=/opt/llvm/2.8 --with-llvmgccdir=/opt/llvm-gcc-tud/2.8 --enable-targets=x86,x86_64

   The "prefix" flag specifies the location for the installation. 
   The value "/opt/llvm/2.8/" will be used throughout the following text as an example. Please change
   all occurrences according to your needs.
   The "with-llvmgccdir" flag specifies the location of the installed directory for the GCC frontend,
   which will be installed at the next steps.


   For additional information please refer to the documentation of LLVM
   http://www.llvm.org/releases/2.8/docs/GettingStarted.html

1.b.- Compile and test LLVM 2.8 Release 3

   make
   make check

1.c.- Install LLVM 2.8 Release 3
   
   make install

2.-   LLVM-GCC R3
2.a.- Configure LLVM-GCC R3

   Update the environment values with the location from the previous 
   installation:

   export PATH="/opt/llvm/2.8/bin/:$PATH"
   export LD_LIBRARY_PATH="/opt/llvm/2.8/lib:$LD_LIBRARY_PATH"
   export LD_RUN_PATH="/opt/llvm/2.8/lib:$LD_RUN_PATH"

   Run configure from the folder that you unpack LLVM_GCC_R3.
   This is an example of parameters that can be used:
   ./configure --prefix=/opt/llvm-gcc-tud/2.8 --program-prefix=llvm- --enable-llvm=/opt/llvm/2.8 --disable-shared  --enable-languages=c,c++
   
   The "enable-llvm" flag specifies the location of llvm installed 
   in the previous step.

   If you want just pure 64 bit compiler, configure with --disable-multilib, 
   without this flag, you will get both version (32 and 64 bits).

   For additional information please refer to the documentation of LLVM
   http://www.llvm.org/releases/2.8/docs/GettingStarted.html



2.b.- Compile and test LLVM-GCC R3

   make LLVM_VERSION=2.8
   make check

2.c.- Install LLVM-GCC R3
  
   make install

3.-   TANGER
3.a.- Configure TANGER R3
   Run configure from the folder that you unpack TANGER.
   NOTE: TANGER should be compiled just from the src directory.
   You can not build it from another directory. The compilation will 
   not succeed.

   Update the environment values with the location from the previous 
   installation:
 
   export PATH="/opt/llvm-gcc-tud/2.8/bin/:$PATH"
   export LD_LIBRARY_PATH="/opt/llvm-gcc-tud/2.8/lib:$LD_LIBRARY_PATH"
   export LD_RUN_PATH="/opt/llvm-gcc-tud/2.8/lib:$LD_RUN_PATH"

   Run configure from the folder that you unpack TANGER.
   This is an example of parameters that can be used:
   ./configure --prefix=/opt/tanger/current --with-llvmsrc=/opt/llvm/2.8/src --with-llvmobj=/opt/llvm/2.8/build


3.b.- Compile and test TANGER R3

   make check

3.c.- Install TANGER R3
  
   make install

4.- If you want to use this set of tools, and you installed the infrastructure 
using an explicit folder, you should set your environment variables according 
to these locations. For example:

   export PATH="/opt/binutils-2.20/bin:$PATH"
   export LD_LIBRARY_PATH="/opt/binutils-2.20/lib:$LD_LIBRARY_PATH"
   export LD_RUN_PATH="/opt/binutils-2.20/lib:$LD_RUN_PATH"

   export PATH="/opt/llvm/2.8/bin/:$PATH"
   export LD_LIBRARY_PATH="/opt/llvm/2.8/lib:$LD_LIBRARY_PATH"
   export LD_RUN_PATH="/opt/llvm/2.8/lib:$LD_RUN_PATH"

   export PATH="/opt/llvm-gcc-tud/2.8/bin/:$PATH"
   export LD_LIBRARY_PATH="/opt/llvm-gcc-tud/2.8/lib:$LD_LIBRARY_PATH"
   export LD_RUN_PATH="/opt/llvm-gcc-tud/2.8/lib:$LD_RUN_PATH"

   export PATH="/opt/tanger/current/bin/:$PATH"
   export LD_LIBRARY_PATH="/opt/tanger/current/lib:$LD_LIBRARY_PATH"
   export LD_RUN_PATH="/opt/tanger/current/lib:$LD_RUN_PATH"

//===----------------------- VELOX project

European research consortium VELOX supports programming of  multi-core systems

Research project will aim to make parallel programming easier for the masses 
by developing integrated Transactional Memory systems for multi-core computers.

This document was prepared by Javier Arias. If you have any comment or 
suggestion feel free to send an email to [email protected].