Memory access profiler by Kevin Nam
## INTRODUCTION ##
This program was written for the LegUp HLS Suite. It checks whether a function is memory-independent with other functions or other invocations of itself, to see if the function can be parallelized.
To do this, it takes the instruction trace of the program we want to analyze. Then it builds a hash table of all the accessed memory addresses. For each address, we build a list of all the functions that access it.
Once we have this table, we go through each address and check if there are memory dependencies.
## REQUIREMENTS ##
The program requires glib. Get it with:
apt-get install libglib2.0-dev
Then simply make.
## USAGE ##
First get the instruction trace using GXEMUL. Due to problems with how gxemul's instruction trace was outputted (printfs), GXEMUL code was modified to write the output to a file rather than print to terminal. Replace “main.cc” with the modified one, in /src/old_main/. Then build GXEMUL. Now the instruction trace will be saved to a file called “gxemuldump”
Once you have this file, you can run the profiler:
./mem_access_profiler <instruction trace file> <access threshhold (integer)> <function name>
argv[1] is the instruction trace file
argv[2] is the number of accesses above which the address will be printed out. ex if you choose 10, then all the addresses where more than 10 different functions access it will be printed out. This is useful if you want to find what variable(s) might be causing the memory dependencies. Once you have the address, you can check the .src file of the program to find the variable name.
argv[3] is the name of the function you want to analyze
Once the profiler finished running, the results will be saved to a text file.
The text file might look like this:
------------- Checking for function: logscl ------------- stack conflict with other: 0 stack conflict with self: 0 heap conflict with other: 0 heap conflict with self: 0 heap dependency conflict with self: 0 stack dependency conflict with self: 0 ----------------------- DONE FOR: logscl ----------------------- Invocation count of this function: 100
here, logscl is the function that was checked. This function was invoked 100 times, and has no memory dependencies ie. parallelizable.
Another result might look like:
------------- Checking for function: encode ------------- Heap conflict with other at: 80033274 Heap conflict with self at: 80033274 Heap conflict with other at: 80032eb0 Heap conflict with self at: 80032eb0 stack conflict with other: 0 stack conflict with self: 0 heap conflict with other: 1 heap conflict with self: 1 heap dependency conflict with self: 1 stack dependency conflict with self: 0 ----------------------- DONE FOR: encode ----------------------- Invocation count of this function: 50
Here, there are memory dependencies, ie. the function is not parallelizable. the 1s indicate TRUE not the number of conflicts.
