# I need to analyze assembly code...

## Recommended Posts

lride    674
I must analyze some assembly code..

Is there some kind of "assembly code analyzer"?
Or do i need to learn assembly? [img]http://public.gamedev.net//public/style_emoticons/default/sad.png[/img] Edited by lride

##### Share on other sites
Erik Rufelt    5901
Kinda need to learn it.. what is the code and what is the purpose of the analysis?

##### Share on other sites
Bacterius    13165
Depends when you mean by "analyze", if you mean find out what it does, then yes, you probably need to understand the code and the computer won't help you do that. If you mean how fast it'll be, how to optimize it, etc, then there are tools available for this kind of stuff (but it doesn't hurt to understand what the assembly is doing, either).

##### Share on other sites
lride    674
I need to know how conditional statements, loops and etc translate into assembly.
Where can I learn some assembly?

##### Share on other sites
SiCrane    11839
If you just want to see how code in one language turns into assembly, most compilers give you an option to see generated assembly. For gcc you can use the -S switch and with MSVC you can use the /FA switch.

##### Share on other sites
alvaro    21246
That part is not too hard. The first hit on Google for `assembly x86 tutorial' was [url="http://www.cs.virginia.edu/~evans/cs216/guides/x86.html"]this page[/url], which seems reasonable. You can then ask your C or C++ compiler to generate assembly output for some simple programs and try to analyze what it's doing. If you have optimizations turned off, it should be pretty straight forward most of the time.

If you get far enough to understand how conditional statements and loops are implemented, you should make a little extra effort and understand how function calling and local variables work, since that's very informative for any programmer.

##### Share on other sites
lride    674
[quote name='alvaro' timestamp='1349571343' post='4987542']
If you have optimizations turned off, it should be pretty straight forward most of the time.
[/quote]

Well, I must anaylze a code with full optimization especially in a loop, so I know what kind of loop optimization(such as interchanging, unrolling) is happening.
Is optimized assembly harder to understand?

alvaro    21246

##### Share on other sites
kuramayoko10    390
I found the following free e-book:[url="http://savannah.nongnu.org/projects/pgubook/"] Programming From the Ground-Up[/url]
I believe it will cover the intel syntax, which is easier to learn.
Try to find resources on the intel developer zone as well.

I like to use objdump + gcc on Linux to have the assembly code with the C code commented in between.
[CODE]
> gcc -g test.c -o test.o
> objdump -dS test.o > test.asm
[/CODE]

On Windows, Visual Studio gives you assembly code pretty easily as well: right-click on your code and select "Show Dissasembly" Edited by kuramayoko10

##### Share on other sites
lride    674
[quote name='kuramayoko10' timestamp='1349572149' post='4987545']
I like to use objdump + gcc on Linux to have the assembly code with the C code commented in between.
> gcc -g test.c -o test.o> objdump -dS test.o > test.asm
On Windows, Visual Studio gives you assembly code pretty easily as well: right-click on your code and select "Show Dissasembly"
[/quote]

I did

g++ -O3 main.cpp -o main.o
objdump -dS main.o >main.asm

but I don't get c++ code commented in between.

(excerpt)
[source lang="plain"]main.o: file format pei-i386
main.o: file format pei-i386

Disassembly of section .text:

00401000 <___mingw_CRTStartup>:
401000: 53 push %ebx
401001: 83 ec 38 sub $0x38,%esp 401004: a1 70 40 40 00 mov 0x404070,%eax 401009: 85 c0 test %eax,%eax 40100b: 74 1c je 401029 <___mingw_CRTStartup+0x29> 40100d: c7 44 24 08 00 00 00 movl$0x0,0x8(%esp)
401014: 00
401015: c7 44 24 04 02 00 00 movl \$0x2,0x4(%esp)[/source]

by the way "go to disassembly" works excellently in VC++ Edited by lride

##### Share on other sites
Bacterius    13165
1 C++ line is [u]not[/u] 1 assembly line, especially with optimizations enabled. There is no simple one-to-one translation between a lower level and a higher level language. You will just need to learn basic assembly, there is no way around that. You can't learn how to ride a bike by driving a car!

##### Share on other sites
alvaro    21246
You can get the original code interleaved with the generated assembly with this command: g++ test.cpp -c -g -O3 -Wa,-ahl=test.lst

Give it a try!

##### Share on other sites
lride    674
Thanks, it works.

I was wondering if assembly code could show if branch prediction is taking place.

##### Share on other sites
alvaro    21246
[quote name='lride' timestamp='1349576724' post='4987559']
I was wondering if assembly code could show if branch prediction is taking place.
[/quote]

No, that doesn't make any sense. Branch prediction is a feature of the CPU, which tries to execute the code as fast as possible, but the assembly is not instrumented in any way to enable it: The CPU will do it automatically everywhere.

##### Share on other sites
lride    674
Thanks to everyone who answered my question!

so much to know for a science fair project...

##### Share on other sites
alvaro    21246
The science fair project is so you learn a lot of things, not the other way around.

##### Share on other sites
Ohforf sake    2052
[quote name='alvaro' timestamp='1349576874' post='4987560']
[quote name='lride' timestamp='1349576724' post='4987559']
I was wondering if assembly code could show if branch prediction is taking place.
[/quote]

No, that doesn't make any sense. Branch prediction is a feature of the CPU, which tries to execute the code as fast as possible, but the assembly is not instrumented in any way to enable it: The CPU will do it automatically everywhere.
[/quote]

Actually, this is not true for all instruction sets. Some don't have complex branch prediction mechanisms but rely on branch hinting where a special branch hint instruction has to be issued a few cycles before the branch. In those cases you can actually check in the assembly if the branch hint instruction is present and located in the right spot.

For x86 (or x86_64) however, this is not the case, as alvaro already pointed out. But all modern intel and amd CPUs have hardware counters that can be used by profilers to tell you, where branch mispredictions occure. See oProfile (Linux) or vTune and CodeAnalyst (Windows).

##### Share on other sites
deftware    1778
My two cents: grab a disassembler - a program which shows you the assembly of a compiled EXE or DLL file.. I used to modify all sorts of programs (while simultaneously gaining a bare-minimum awareness of assembly itself) by disassembling them, and editing their code directly using a hex editor - overwriting small bits of existing code by surmising the actual byte opcodes of the asm instructions I desired and writing them in by hand.. Also, writing programs that would do it all in memory (WriteProcessMemoryEx) so that the original file could be left alone while effecting the equivalent change in functionality once the target was running. That may or may not help you.

##### Share on other sites
Heath    357
I know it's not the same architecture (6502 is not x86), but I just wanted to throw this out there: [url="http://www.atariarchives.org/mlb/"]Machine Language for Beginners[/url]. Edited by Heath

##### Share on other sites
slmgc    168
[quote name='lride' timestamp='1349570746' post='4987539']
I need to know how conditional statements, loops and etc translate into assembly.
Where can I learn some assembly?
[/quote]

You should definitely read [url="http://www.amazon.com/Code-Optimization-Effective-Memory-Usage/dp/1931769249"]Code Optimization: Effective Memory Usage[/url] by well-known code-hacker Kris Kaspersky. One of the best books on the subject you want to know.

##### Share on other sites
LorenzoGatti    4442
What do you need to analyze assembly code [i]for[/i]? There must be a specific question to answer. For example, I once tried to look at the relevant code for a rather inexplicable C++ bug, and I only had to look at a few symbol names and trivial push, mov and call instructions to get evidence that a class constructor was wrong enough to call itself recursively; there was absolutely no need to make modifications, predict branches, understand every line of the program, etc.

Science fair project suggestion: illustrating different ways to do something in assembly, to show which libraries and compilers are more clever and which approaches fit specific processors and use cases.
You should analyze a task that is:[list]
[*]easy to understand (to avoid losing public)
[*]non-obvious to implement, with some difficulties and tradeoffs (to find interesting differences between implementations)
[*]simple to test (because you'll have to run performance measurements)
[/list]
BLAS routines, for example dense matrix multiplication, should be good choices.