Jacob Indursky and Alex Salvatore
Homework 7 -- profiling

Help: Justin Shiiba helped with our compile script

Hours spent analyzing: 3
Hours spent solving: 2


Analysis of assembly code: 

        While doing this profiling, we incrementally static inlined all of
our modules. This took away function call overhead from the main. From our 
Labnotes file, it is easy to see the speed increase we achieved at each stage of
the inlining process. We found that inlining all of our functions significantly
optimized our um performance. Now, however, There is no process that takes up
an amount of time as according to the assembly code, all functions have been 
put into the main. 

Before we started the process of inlining all of our functions to main, 
we looked at which process was taking the longest. Our emulator was calling
get_word from the memory interface and it was taking 21.28% of our program.

Analysis of assembly code of get_word(Seg_id index, int offset, T mem)
in memory.c:

 mov    0x28(%rdx), %rax
 mov    %edi, %edi
 movslq %esi, %rsi
 mov    (%rax, %rdi, 8), %rax
 mov    (%rax, %rsi, 4), %eac
 retq


After careful analysis, we realize we still have to read in each parameter
and then access a part of the array. The minimum amount of instructions for
this is 5 and that is what we have. We considered changing the int to an
unsigned, however, it would continue to be the same amount of actual
instructions in our assembly code. Without changing our data structure, we
do not see a way to improve this assembly code. Even with a chnage in 
data structure, we are not confident it would help as this code is already 
very clean. Inlining this function, however, eliminates three lines of asembly
code which is over half of the instructions for the function. Our labnotes show
the significance time-wise of inlining this particular function.