llvmparse user manual

Title llvmparse
Description LLVM assembly parsing standalone utilities
Author Nikolaos Kavvadias 2012-2013
Copyright Nikolaos Kavvadias
Contact nikos@nkavvadias.com
Website http://www.nkavvadias.com
Release Date 25 February 2013
Version 1.0.0
Rev. history  


First commercial release.

1. Introduction

llvmparse is a set of standalone parsers for the popular LLVM textual assembly format. This format is used as a human-readable representation of the LLVM IR (intermediate representation) which is used by the LLVM compiler (http://www.llvm.org).

This version of llvmparse includes two standalone parsers:

  1. a flex/bison-based parser (compiled) and
  2. a TXL-based parser (interpreted)

that are compatible with LLVM 3.1 and 3.2. TXL is a functional programming language that is commonly used for source-to-source transformation tasks (http://www.txl.ca).

llvmparse will always be kept up-to-date with the latest official LLVM release.

2. File listing

The llvmparse distribution includes the files listed in the following table. There is a single release for llvmparse containing all source code, Windows (MinGW) binaries as well as Linux static binaries.

/llvmparse Top-level directory






Automated build script.

Clean the generated debris files in /tests.

Script for setting up the environment on Linux.

Script for setting up the environment on Windows.

Exercise all test cases found in /tests.

/bin/lin Linux executable directory
llvmparse.exe Static executable of the flex/bison-based parser for Linux platforms.
/bin/win Windows executable directory
llvmparse.exe Static executable of the flex/bison-based parser for Windows platforms.
/doc Documentation directory








List of authors.

End-user license agreement.

The LLVM source code open-source license.

This file.

HTML version of README.

PDF version of README.

Current version of mprfgen.

/src Source code directory







Makefile for building the flex/bison-based parser.

TXL grammar for the LLVM assembly format.

Flex/lex scanner for the LLVM assembly format.

Bison/yacc grammar for the LLVM assembly format.

Driver TXL program for the TXL-based parser.

Driver C program for the flex/bison-based parser.

/tests Directory for testing llvmparse



ANSI/ISO C source for a test case.

LLVM textual assembly files generated by clang-3.2 (not included).

3. llvmparse setup

For using llvmparse, this distribution comes with ready-made native executables for Windows and Linux, which can be found in the corresponding /bin/exedir directory, where exedir is lin for Linux or win for Windows.

In order to build llvmparse from source on your platform, just run the build.sh bash script from the command prompt. Cygwin or MinGW are suggested as POSIX emulation environments on Windows.

First, change to the appropriate directory, for instance:

cd /home/user/llvmparse/

assuming llvmparse was installed inside /home/user. A few environmental variables should be set and for this purpose the env-*.sh scripts can be used. For instance, a Windows installation requires the following:

$ ./env-win.sh

The llvmparse distribution can be built from sources by using:

$ ./build.sh

4. llvmparse usage

The llvmparse parsers can be invoked as shown in the test-all.sh script. However, for exercising a single test case, this can be accomplished using the flex/bison-based parser as follows:

$ ./bin/llvmparse.exe fibo.ll >& fibo.txt

which parses the test program fibo.ll with llvmparse.exe and stores the parsed program to fibo.txt.

For using the TXL-based parser, it is suggested that the txl interpreter is visible. Usually txl is installed in the /usr/local/bin path in Linux and Cygwin/Windows systems. For MinGW/Windows systems, it is suggested that the Windows port of TXL is used. For instance, TXL 10.6 can be installed to /c/Txl106/.

The env-*.sh scripts can be used to set the appropriate value for the TXL_PATH environment variable. An example invocation of the TXL-based parser is as follows:

$ $TXL_PATH/bin/txl -q -raw -o $fibo.txt $LLVMPARSE_PATH/src/llvmparse.Txl fibo.ll

5. Running the test suite

The test suite that accompanies llvmparse can be found in the /test subdirectory. It consists of C source code and the generated LLVM textual assembly files, which use the .ll suffix. These LLVM IR files can be exercised by running the top-level test-all.sh script:

$ ./test-all.sh

The generated debris files (using the *.bisonparsed and *.txlparsed patterns) can be deleted using the corresponding clean-all.sh script:

$ ./clean-all.sh

6. Prerequisities