this tutorial deals with:

• GNU cross-compilers and utilities for low-level program analysis
• examples of use of such tools for various processors and program development systems:
  • for ARM simulator VisUAL v. 1.27 with GCC 3.2 cross-compiler, GNU Binutils 2.13, and ad-hoc script
  • execution with debugger on ARM Cortex-A9 processor (DE1-SoC HPS)
  • execution with debugger on Nios II processor (DE1-SoC FPGA)
  • for ARM Cortex-A9 processor with GCC 4.8.1 cross-compiler and GNU Binutils 2.23
  • for Nios II processor with GCC 5.3.0 cross-compiler and GNU Binutils 2.25
• lab experience: reproduction of the execution of the examples and production of tutorial documentation

a first example showcases the use of the VisUAL v. 1.27 ARM emulator, for the simulation of the execution of a program for GCD computation with Euclid's algorithm (Listing 7.11 in Schaumont, with modified main because of the different simulation target)

• the emulator accepts an ARM source assembly program coded in a subset of the instruction set defined by the UAL (Unified Assembly Language) syntax

in this example, the assembly program is obtained from the C source by means of the arm-linux-gcc cross-compiler, for ISA ARMv4, implemented in the StrongARM processor

• this cross-compiler and its binary utilities can be retrieved from
  rijndael.ece.vt.edu/gezel2repo/pool/main/a/arm-linux-gcc/

however, the assembly program so obtained does not meet the restrictions posed by the emulator, e.g. it contains assembly directives and other syntactic sugar that are not accepted by VisUAL

• this obstacle is overcome by an ad-hoc script that edits a copy of the previous assembly source to produce a version of it conforming to the VisUAL syntax

archive arm_visual_examples.tgz, available in folder crosstools of the reserved lab area, provides the C sources and the required scripts for the cross-compilation and VisUAL emulation of the GCD example and of a second example, that presents a software implementation of the Collatz delay calculation, similar in functionality to the hardware implementations considered in the previous lab experiences

• to run the examples, after unpacking the archive with tar xzpvf, in folder arm_visual_examples run respectively ./c2visual gcd, and ./c2visual delay_collatz, then launch VisUAL and open the respectively generated files gcd/gcd.s, and delay_collatz/delay_collatz.s (warning: with lowercase extension .s !)

the figure shows an execution snapshot of program delay_collatz.c on the DE1-SoC ARM processor

cross-compilation, loading and execution take place by means of the altera-monitor-program, under control of its GNU debugger GDB

• the system allows the loading of a C or assembly source, whereby cross-compilation is done through the monitor program itself

the same C source from the previous example may be cross-compiled and executed by a Nios II softcore on the FPGA of the DE1-SoC system

execution of the delay_collatz example on the ARM processor by means of the monitor program produces two files in the project directory that are relevant to low-level analysis:

• delay_collatz.axf: the ELF executable
• delay_collatz.axf.objdump: its disassembly produced by the objdump utility

in order to carry out low-level analysis even further, it helps one to know where the binary utilities are located, for the gcc cross-compiler used by the monitor program

• in the installation directory of the Quartus Prime software, the cross-compiler and the binutils are located in directory
  University_Program/Monitor_Program/arm_tools/baremetal/bin

N.B. actually, the monitor program makes use of specialized versions of some of these tools, that are placed elsewhere; however, for general use, independent of any execution in the contest of the monitor program, the aforementioned tool collection is the proper one

processing the C sources by the monitor program for execution on the Nios II processor produces the .elf executable in the project directory, but not its disassembly

• in the installation directory of the Quartus Prime software, the cross-compiler and the binutils are found in directory
  nios2eds/bin/gnu/H-x86_64-pc-linux-gnu/bin

hre is a simple example of use of the size utility: comparing the sizes of the executables produced by the three cross-compilers of interest here, for both of the examples proposed and with the same compilation options; in particular, optimization level O2

• the two C sources and the scripts to this purpose are available in the cross-compiled_sizes.tgz archive, within the crosstools directory in the reserved lab area

the following table presents the outcome of this exercise

run the examples provided with this tutorial, examine the files produced by their execution, find a few not well-known aspects of their contents, search the web for information about them and produce some tutorial documentation in form of a list of questions and answers

• each answer should give a reference to its source; if the answer is drawn from a directly accessible web source, insert the specific link; if this points to information that only answers the question, then giving the link as answer is enough

here are three examples of question and answer, one for each of the three types described above:

• Q: What is the meaning of the acronyms VMA and LMA in the output of objdump?
• A: Respectively virtual memory address and load memory address; the former is the address of a program section during execution, the latter is its address when the program is first loaded in memory; the two addresses may differ when the program is first loaded on a different memory than the one which holds it at runtime, for example it is loaded on a Flash memory and then copied into RAM when it executes. (Schaumont, 2012, pp. 216-217)
• Q: What does the ARM instruction rsb do?
• A: It stands for reverse subtract, as it subtracts the first source operand from the second one, that is, the two arguments are in reverse order with respect to sub; please note that arithmetic ARM instructions allow the second source operand to be immediate, whereas the first one and the destination must be registers, so this instruction allows subtraction from a constant. (www.heyrick.co.uk/armwiki/RSB)
• Q: What is the origin of the name bss of the executable program section that contains the global variables?
• A: en.wikipedia.org/wiki/.bss#Origin

recommended readings:

• Schaumont, Ch. 7, Sect. 7.4

readings for further consultation:

• Schaumont, Ch. 7, Sect. 7.5, 7.6
• The ARM Instruction Set, tutorial, ARM University Program, V1.0
• VisUAL - A highly visual ARM emulator
• tutorials from source Intel® FPGA University Program (November 2016):
• Introduction to the ARM® Processor Using Intel FPGA Toolchain - For Quartus Prime 16.1
• Introduction to the the Intel Nios II Soft Processor - For Quartus Prime 16.1
• Intel FPGA Monitor Program Tutorial for ARM - For Quartus Prime 16.1
• Intel FPGA Monitor Program Tutorial for Nios II - For Quartus Prime 16.1

useful materials for the proposed lab experience:

• GCC, the GNU Compiler Collection
• GNU Binary Utilities