Go up to the Labs table of contents page
This lab is one of two labs meant to familiarize you with the process of writing, assembling, and linking assembly language code. The purposes of the in-lab and post-lab activities are to investigate how various C++ language features are implemented at the assembly level.
There are both 32 bit (md) and 64 bit (md) versions of this lab. This is the 32 bit version.
The Intel x86 assembly language is currently one of the most popular assembly languages and runs on many architectures from the x86 line through the Pentium 4. It is a CISC instruction set that has been extended multiple times (e.g. MMX) into a larger instruction set. In 2004 it was extended to allow for a 64 bit memory space.
cin
), not through command-line parameters.make
!
There are four different platforms that students are potentially developing their code on:
Your code must compile and run on the submission server, which is a 64-bit Linux machine!
There are three changes that will have to be made to compile your program (and this to the Makefile) depending on your own development platform:
vecsum
instead of _vecsum
(note the lack of
underscore in the former) in vecsum.s (this file is described more
below). In the final linking step, if you get a message such as,
main.cpp:(.text+0x12): undefined reference to 'vecsum'
,
then you should change the name of the function.CXX
or
CXXFLAGS
macro(s) line in your Makefile-f
) that will need to be specified.The first bullet point highlights a compatibility problem between
Linux and Mac OS X. When calling a subroutine, which in C++ would be
called foo()
, there are two standards as to how to name the
assembly routine: you can name it either _foo
(adding an
underscore is added before the name), or name it just foo
(with no underscore). Unfortunately, Linux uses a different standard
than Mac OS X, so we have to make (minor) code modifications in order to
compile the code on the other system: in Mac OS X, the vecsum.s file
should have the subroutine be called _vecsum
, and under
Linux, it should be called vecsum
(this is twice, on lines
9 and 21).
In an effort to make sure all the files submitted conform to one
standard or the other, all assembly and C/C++ code must be
submitted in Linux form (i.e. will be called foo
and not _foo
). Note that in many programs, such as the
vecsum.s that we provided you, you have to change the name in TWO
places: on the global
line (line 9 of vecsum.s) and on the
label line (line 21 of vecsum.s). If your code does not compile
on the submission system, you will receive zero credit!
Also note that your code must compile with make
. We
provide a sample Makefile that will compile vecsum, so you can just
modify this Makefile to compile your pre-lab program. Please
note that you should NOT specify a -o
flag to clang++ (not
even -o a
), as we want it to be named the default
(a.out). This allows easy porting between the two operating systems.
If you plan to develop it in Mac OS X, we suggest that you develop it
normally (putting in the _
before the subroutine name).
Then, once you have verified everything works, remove the underscores
from all the relevant lines, and test it out on a
32-bit Linux machine, such as the VirtualBox image, before submitting
it.
Each of these different platforms has different compilation lines to allow it to compile. Some of them require changing the assembly files as well. You only need to read the line(s) pertaining to the platform(s) you are developing on.
The provided code works under both 32-bit and 64-bit Linux (although 64-bit Linux usage should read below).
32-bit Linux: The provided code and Makefile works
on 32-bit Linux. All assembly subroutine names must NOT
have a leading underscore (i.e. they should be vecsum
and
not _vecsum
). nasm is invoked with the -f elf
option. We have the -m32
flag on the CXX macro line so that
it will work fine on a 64-bit Linux machine (including our submission
server), but it is technically not necessary (it doesn’t hurt, it just
doesn’t do anything on a 32-bit machine).
64-bit Linux: You have to explicitly tell clang++ to
compile in 32-bit mode by passing in the -m32
parameter
(note that this parameter is fine to pass in on 32-bit systems as well).
You may need to install the g++-multilib package - we realize that we
are not using the g++ compiler, but this installs the correct library in
the correct place (if this differs with your version of Linux, then
please let us know!). The options for nasm (-f elf
) and the
naming of the subroutines (vecsum
, not
_vecsum
) are the same as 32-bit Linux.
32-bit Mac OS X: to run the code, will need to
rename all the assembly function names with a leading underscore
(i.e. _vecsum
not vecsum
). You will also have
to use the -f macho
format for nasm, and tell clang++ to
generate the correct architecture code. In the past, to generate the
correct architecture code WAS done by providing the
-arch i386
parameter to the compiler we used previously; we
are not sure if this is necessary with clang++ (please let us know how
this works for you; we do not have access to a 32-bit Mac machine with
clang/clang++ to test this out). In the provided Makefile, try the
CXX
macro line to be clang++ -arch i386
(instead of the default clang++
), and change the
ASFLAGS
macro line to -f macho
(instead of the
default -f elf
). You will probably want to remove the
-m32
flag on the CXX
macro line, but be sure
to put that back in before you resubmit. Note that you
MUST change all of this back in order for it to compile
via the submission system! Also note that Mac OS X does not support the
format of assembly that we use in this course, which means that you will
be stuck reading the assembly in the other format we discussed in
class.
64-bit Mac OS X: As far as we are aware, this is the same as the 32-bit Mac OS X. If you run into problems with this, please let us know. Note that Mac OS X does not support the format of assembly that we use in this course, which means that you will be stuck reading the assembly in the other format we discussed in class.
Below is a table summarizing the changes
Platform | nasm flag | x86 subroutine name | clang++ flags | Notes |
---|---|---|---|---|
32-bit Linux | -f elf | vecsum | -m32 | This is the platform on the 001 machines and in the VirtualBox image; -m32 is optional, but keep it in there for compatibility with 64-bit Linux |
64-bit Linux | -f elf | vecsum | -m32 | This is what our submission server is running, and what your code must work on. If you have it on your computer, you must install a few packages as well - see above |
32 bit Mac OS X | -f macho | _vecsum | -arch i386 | We are unsure about the -arch i386 flag’s necessity. May not be able to print the assembly in the format discussed in class. |
64 bit Mac OS X | -f macho | _vecsum | -arch i386 | We think this is the same as 32-bit Mac OS X platform. We are unsure about the -arch i386 flag’s necessity. May not be able to print the assembly in the format discussed in class. |
IMPORANT: Just to repeat, when you submit your code, it MUST be in 64-bit Linux format.
For this part, you will need to download three files: vecsum.s (src), main.cpp (src), and Makefile (src).
To compile a program written partly in x86 assembly and partly in C++, we have to build the program in parts. We build the C++ file as we have in the past:
clang++ -m32 -Wall -g -c -o main.o main.cpp
Note that we used the -c flag, which tells the compiler to compile
but not link the program. Linking it will create the final executable –
but as there is not a vecsum()
function defined (yet), the
compiler will report an error stating that it does now know the vecsum()
function. The -o main.o
part tells clang++ to put the
compilation output into the file named main.o. Note that the
-o
flag wasn’t really necessary here (as clang++ will use
main.o by default when compiling main.cpp), but we wanted to include it,
as we are going to use it below. We include the -m32
flag
to force it to be a 32-bit file. We also added a few more flags
(-Wall -g
) to print all warnings and compile debugging
symbols into the program.
Next, we need to compile the assembly file. To do this, we enter the following:
nasm -f elf -g -o vecsum.o vecsum.s
This invokes nasm, which is the assembler that we are using for this
course. We’ll get to the -f elf
part in a moment. The
-o vecsum.o
option is the same as with clang++ – it is
telling the assembler to put the output into a file named vecsum.o. If
you do not specify a filename with the -o
flag, it will
default to vecsum.obj, NOT vecsum.o – this is why we are using the
-o
flag. We also tell it to include debugging symbols via
-g
. The assembly file name is specified by the
vecsum.s
at the end of the command line.
The new flag here is the -f elf
. This tells the
assembler the output format for the final executable. Operating systems
can typically execute a number of different formats. As we are running
under 32 bit Linux, we specify the elf format. Mac OS X uses -f
macho
– see the above section for more details.
Finally, we have to link the two files into the final executable. We do this as before:
clang++ -m32 -Wall -g vecsum.o main.o
This tells clang++ to link both of the .o files created above into an
executable, which it called a.out
. Note that there isn’t
any compiling done at this stage (the compilation was done before) –
this just links the two object files into the final executable. Also
note that for our submitted Makefiles, we will NOT have the
-o
flag present.
Complete the C++/assembly tutorial, which consists of reading two book chapters that are contained in this repository: x86 Assembly and The x86 C Calling Convention.
Examine the vecsum subroutine in vecsum.s (src). Use the slides and readings to help understand what is happening in vecsum.s. Make sure you understand the prologue and epilogue implementation, as well as the instructions used in the subroutine.
Compile and run the vecsum program:
-g
flag, which allows the program to work well with the gdb
debugger.step
function to step into
the next instruction. However, since no debugging information was
included with the assembler (a shortcoming of nasm), we can’t use
step
– it will just move to the next C++ instruction (the
cout
). Instead, we will use stepi
, which will
step exactly one assembly instruction, which is what we
want.disassemble
. This is just like list
, but it
displays the assembly code instead of the C++ code.set disassembly-flavor intel
. Now
enter disassemble
again – the format should look more
familiar. You only have to enter that set command once (unless you exit
and re-enter gdb).disassemble vecsum
.
Note that this only lists the first third (or so) of the routine – up
until the vecsum_loop
label. To see the rest of the code,
enter disassemble vecsum_loop
, disassemble
vecsum_done
, etc.info
registers
command.You will need to write two routines in assembly, one that computes the product of two numbers, and one that computes the power of two numbers.
The first subroutine will compute the product of the two integer
parameters passed in. The restrictions are that it can only use
addition, and thus cannot use a multiplication operation. We
will assume that both of the parameters are positive integers. It must
compute this iteratively, not recursively. The
resulting product is then returned to the calling routine. This
subroutine should be called product
. We will assume that
values will not be provided to the subroutine that will cause an
overflow, nor will negative (or zero) parameters be passed in.
The second subroutine will compute the power of the two integer
parameters passed in. We will assume that the first parameter is the
base, and the second parameter is the exponent. Again, both are
integers. The restrictions on this routine are that it can only
use the multiplication routine described above – it cannot call
any exponentiation routine. Furthermore, it must be defined
recursively, not iteratively. This routine should be
called power
.
You can assume that the numbers passed into both routines will both be positive, so you need not consider negative numbers or zero. Furthermore, as described above, no values will be used on your program that could cause an integer overflow.
Both of these routines should be in a file called mathlib.s, and must
use the proper C-style calling convention. You must also provide a
mathfun.cpp file, which calls both of your subroutines – see the
main.cpp file provided as a template. The program should take in ONLY
two integers (we’ll call them x and y). It should then
print out the output of calling product(x,y)
and
power(x,y)
. Thus, if the input is 3 and 4, it would print
out 12 and 81.
Input is to be read via standard input (i.e., cin
), not
through command-line parameters.
If you are going to have multiple routines in a single assembly file
(as is needed for mathlib.s), you just have to have multiple
global
lines for each subroutine that you plan on calling
from your C/C++ code.
Come to lab with a functioning version of the pre-lab, and be prepared to demonstrate that you understand how to build and run the pre-lab programs. If you are unsure about any part of the pre-lab, talk to a TA. The in-lab will ask you to write C++ code and examine the generated assembly language for a variety of topics.
You should be able to explain and write recursive functions for the final exam, so make sure that you understand how to implement the pre-lab program. Speak to a TA if you have any questions.
The general activity of this in-lab will be to write small snippets of C++ code, compile them so that you can look at the generated assembly code, then make modifications and recompile as needed in order to deduce the representation of a number of C++ constructs (listed below).
For the in-lab, you will need to work on the two topics shown below – note that there will be a different topics to work through on the next lab.
The deliverable for this in-lab is a document named inlab8.pdf. It must be in PDF format! See the How to convert a file to PDF page for details about creating a PDF file.
In the report, you should explain all the items in one of the categories below (either objects or parameter passing). For the post-lab, you will need to have all items from both categories explained. We are looking for significant evidence that you were able to complete some work during the in-lab, and thus are not setting page length requirements.
Recall that using the -S
flag with clang++ will generate
the assembly code. You will also want to use the -mllvm
--x86-asm-syntax=intel
flags.
As discussed in class (specifically, here), clang
will often optimize away many parts of the calling convention, and has
even been known to pass parameters in registers. So if you see code that
is unexpected, trace it by hand to determine what is really happening.
Likely, it is an optimization of the calling convention. To avoid this
in the in-lab, use g++ rather than clang++ to compile. The correct g++
flags are -S -m32 -masm=intel
.
The questions posed below are example questions to answer. Some of them may overlap. Others may not be necessary. And there may be questions not listed that are worth answering. The purpose of posing those questions is to help you think about what topics and ideas you should address in your response - it’s not meant to be a rigid structure that you absolutely must follow. We are going to look at whether you have addressed the general idea of each of the list topics.
Show and explain assembly code generated by the compiler for a simple function and function call that passes parameters by a variety of means. Be sure to show what is happening both in the caller (the function which makes a call to another function) and in the callee (the function which is called by another function, possibly a recursive call). You do not need to describe parts of the C calling convention we described in class (e.g. saving registers, saving the base pointer, how the call instruction works). The focus here is on examining in detail what happens when parameters are passed.
For the post-lab, you should explore, investigate, and understand all of the items from the in-lab list. Be able to answer “how” and possibly “why” for each item. Use test cases and the debugger as resources. Additionally use resources other than yourself (e.g. books, Web, people). Be sure to credit these sources. You must use (and cite!) additional resources for this post-lab!
The idea is that you will take what you started in the in-lab, and flesh it out a bit more for the post-lab.
Prepare a report that explains your findings. Follow the guidelines in the Post-lab Report Guideline section, below. In particular, you should address how the compiler implements the construct at the machine and assembly levels, and what lead you to this conclusion. You must show evidence of this behavior in the form of assembly code, C++, screenshots, memory dumps, manual quotations, output, etc. Where did you find the information that lead to your conclusion (i.e. your sources)?
Your report should be in PDF file called postlab8.pdf. It must be in PDF format! See the How to convert a file to PDF page for details about creating a PDF file.
Think about how best to investigate the issues you choose. A good starting point is to write a small C++ program that illustrates one of the issues. This program should be as simple as possible.
Look at the assembly code associated with your C++ code. To examine
the disassembled code you have two main options: you can step through
the code in the debugger using the disassembly view, or you can have the
C++ code output to an assembly file (using the -S
and
-masm=intel
flags), which you can then browse or edit.
Generating assembly listings: to generate an assembly listing in clang++, use the flags described above, and see the wiki page for details. Probably the most useful listing will include source, and assembly code. For some issues it will be of interest to see the machine code as well.
A couple of things you will notice almost immediately about these assembly files is that they can be surprisingly long, and that they contain a bunch of labels, directives, and instructions that at first glance appear to have little to do with your original source program. Don’t despair, with a little perseverance you will be able to make heads and tails of a good bit of this.
Note that printing out these disassembled files is probably not your most useful option. You will most likely find that it is significantly easier to view the files in a browser of your choice, such as emacs. In this way you can navigate through the file, searching for particular labels or C++ code. Besides, you may want to make a slight modification to your C++ code and recompile often anyway.
Still stuck? Some of these issues are non-trivial to figure out. Remember that you can use basically any resource whatsoever to figure these things out. There will almost certainly still remain some things in the disassembled code that you do not fully understand. Don’t let this paralyze you. Focus on devising experiments that will help you learn more about the particular issues in lists 1 and 2. By tracing though some parts of the code and by modifying your C++ code and comparing the generated assembly code for the two different versions, you should be able to come up with some reasonably good hypotheses about what is happening. Seek out books, manuals, and web pages that explain the issue. Keep in mind that you are required to list your sources in your post-lab report.
You should submit a nicely formatted report that explains your findings. The report should be a PDF file called postlab8.pdf. At a minimum your post-lab report should address the items in the in-lab list. In your report, label the items according to which list item they came from (parameter passing or objects), and their item number within that list.
Note that this is supposed to be a polished report. Code snippets should be embedded into the document, not just printed out on a page by themselves and added in at the end. Similarly, screen shots (which are optional) should be embedded in the document. Highlighting portions of code or drawing arrows between things may help make your explanations clearer. I would expect the explanation of each item to be a page or two long at least, including embedded code snippets and screenshots. Keep in mind that you are only submitting one file for this report: postlab8.pdf. Thus, everything must be included in that one file.
We are being very careful to not specify the length of the report, only what we would guess is the estimated length. As long as you properly answer the questions, the length is up to you. Double versus single spacing is also up to you, but we prefer single spacing.
Other than your own experiments, feel free to use online x86 assembly references, C++ books, and resources you may find on the Internet or elsewhere. Discussing these issues is allowed, however, remember that your code and final report must be your own work and that you must credit ANY resources used.
We want you to investigate the particular topic area from the given list, write code to discover the answers, and learn about this topic on your own. The questions that we pose are just meant to get you thinking about the possible ramifications of a given question. They aren’t meant to be specific questions that necessarily need answering one at a time.
As with the previous lab, I would expect the explanation of each item (you have to do two items) to be a page or two long, including embedded code snippets and screenshots (obviously, we want a reasonable amount of English text here – if you do a lot of screen shots, then your total length will be a bit longer). Did you investigate the topic? Did you write code to discover what you didn’t know? Was this written in a reasonably readable format? This is what we are looking for.
This is somewhat vague, and purposely so – research is often vague. If we told you exactly what to write, then there wouldn’t be much discovery of that on your part, which would defeat the whole point of this lab.
We are not looking for you to spend hours and hours and hours on this! A page or two per list item (and you have to do two of them) - which means your final report needs to be 2-4 pages long. Keep in mind if you have a lot of screenshots, that doesn’t count much towards that page limit.
The grading will be based on a set of points that we would expect you to discover when investigating a given topic. Your grade will be based mostly on how well you hit those points. A small portion of your grade will be based on the overall report presentation and written ability (while we are a computer science class, we expect you to be able to write in English to some extent!).