Starting from:

$29

HW7 – Fun with Self-Modifying Code in PAL

HW7 – Fun with Self-Modifying Code in PAL
The goal of this assignment is to practice more PAL by solving a tricky problem that will required indepth understanding of many PAL instructions so you can use them in unique ways. Complete this
assignment with your current partner.
Deliverables: Your solution program in a file named user1_user2.pal.
Your Task – Clear All Memory
Write a PAL program no longer than 100 lines that will erase (by erase I mean ``set to zero'') as many
memory locations as possible, including the memory where the program is written. The number of
lines is the number output in terminal before “end of file” when you assemble your program. You will
use loops (and/or the entire memory space as one big loop given PAL's automatic continuing with
instruction 0 after 7777) and so the number of lines executed will be much (much) higher than 100, the
limit is on lines of actual PAL code in your .pal file. FYI, you shouldn’t need anywhere close to 100 lines,
so there is still plenty of room for whitespace and the comments you are required to include in your
code.
Your grade will be determined by how many memory slots are not erased after your program either
halts or has run for a reasonable amount of time, we’ll say 10 seconds running in regular (not trace)
mode on a lab computer using the PDP8 simulator. Your program does NOT need to stop itself with a
HLT, and indeed the best solutions could not do so, you can instead use the timeout after 10 seconds
as your HLT.
Scores:
• It is a reasonable task given the PAL we’ve learned in the class/reading so far to clear all but a
few memory slots; if your program leaves 4 or fewer slots non-zero you will receive 8/10 points.
• It is not too much more difficult using auto-indexing (see below) and understanding of all PAL
instructions to get down to all but 2 memory slots cleared; if your program can do this you will
receive 9/10 points.
• It is tricky but doable to clear all but 1 memory slot; for this you will receive all 10/10 points.
• It is possible to clear all memory slots completely; I can show you some examples of solutions
that do so if you are interested, though I will not post them anywhere for the sake of future 208
classes. If you come up with a solution that does this you will receive my being very impressed
and 12/10 points.
• If your program leaves more than 4 slots uncleared, you will receive a variable amount of points
based on your solution quality.
• You still must comment your program and use good style to receive the maximum possible
points described above! Your program is only correct if it works AND you can explain how it
works!
Hints:
- *** The most common mistake in the past *** was to see that all memory had value 0 after your
program ran and thus conclude that you were done, when in fact one or more memory locations
had not been modified at all by your program. A memory address is not "cleared" just because it
happens to have the value 0; you cannot assume any default initial values in memory. You must
actively somehow set the value at an address to 0 to "clear" it . The most obvious way to do this is
to execute a DCA instruction (after ensuring the acc value is currently 0) to the memory address
you wish to clear. There are multiple other ways to end up setting a value in memory to 0, and it's
worth thinking about those.
- A corollary to the hint above is that executing uninitialized memory addresses as instructions
means carrying out completely random instructions and accepting whatever consequences they
have. If you haven't set the value at an address to 0 (or something else) explicitly, and the PC
reaches that address, anything that can happen (in PAL), will happen (with some probability). I will
test your program by first loading it, then filling all unused memory locations with random values.
- Use the red hexagon stop button to stop a program that is running in an infinite loop. Once you
stop it the CPU window will re-expand so you can see the result.
o *** Even more useful tip from another student: the window never minimizes when running
in trace mode. Execution speed is much much slower, but that can be useful when
debugging, and you can click on the red stop sign button anytime to pause at a specific
place relatively accurately (within a few instructions).
- Double click on the BP column to the left of any instruction to set a breakpoint there. You can then
run in either regular or trace mode and your program will pause at the breakpoint until you click
the run button again, allowing you to see intermediate results.
- A PAL program that is not stopped simply keeps executing with the PC wrapping around from 7777
to 0. Remember that every sequence of 12 bits is some instruction so executing any memory slot
does “something”, even if the result doesn’t actually change anything in the end. For example
executing a cleared memory slot (value is just 0) runs the “AND 0” instruction which sets AC = AC
AND value-at-mem-address-0. If the AC is already 0 or the value at address 0 is 0, this instruction
has no effect.
- Your program will complete thousands of loops per second even if each loop executes all 10000
(octal) instructions, so no matter how complicated your solution, it should not need more than 10
seconds to reach a point where no more memory will be cleared (in standard run mode, trace is
much slower).
- Make sure that if your program does keep running in an infinite loop, whatever non-zero
instructions are left don’t re-set any memory values to be non-zero.
- The PC always starts with the value 200, so if you want your program to begin somewhere else, put
a “JMP X” instruction at memory address 200, where X is the address of the first instruction in your
actual program. This is only important when you first start your program, so if you later clear the
memory and lose the JMP X instruction that’s just fine!
Auto-Indexing
- PAL treats memory slots 10-17 (octal) in a special manner. When you access these memory
addresses using an indirect instruction, the value at that slot is incremented by 1 before taking the
2nd jump to the given address to get the actual value for the instruction. This is explained in the
PDP8 pdf in chapter 3 (3-27 to 3-28). Here is an example:
Program using indirect access to normal memory slots (i.e. all except for 10-17):
*100
1
2
3
4
*176
100
*200
TAD I 176
TAD I 176
TAD I 176
HLT
After this program runs the AC would contain the value 3, because each of the TAD’s does the same
thing - gets the value 100 from mem address 176, then goes and gets the actual value to add from
mem address 100, which is a 1.
However, if we change the program to use an indirect TAD to one of the special auto-index memory
slots, the behavior is very different. The modified program is shown below:
*10
100
*100
1
2
3
4
*200
TAD I 10
TAD I 10
TAD I 10
HLT
After this program stops the AC would contain the value 9 (2+3+4) because each of the indirect TAD’s
does more than just the regular indirect load. The first “TAD I 10” first increments the value in memory
address 10, making it 101, and then it goes and gets the actual value to load from this mem address
101, which is a 2. The second “TAD I 10” does the same thing, except the initial value in memory
address 10 is now 101, so it adds 1 to that making it 102, then goes and gets the value from memory
address 102, which is a 3. Again the last TAD makes the value at address 10 a 103 and adds the value 4
from address 103 to the AC. You can see how if you made the above a loop that ran TAD I 10 over and
over again, the value in 10 would be incremented each time and thus the actual value loaded would
come from the next memory address. This works in the same way for any of the 6 memory access
instructions. Using the auto-index memory in this way is different from using the ISZ instruction as it is
doesn’t perform the skip on zero part; instead the value at memory address 10 just keeps getting
incremented, wrapping around to 0 after it reaches 7777. Like all memory addresses it is treated as an
unsigned integer.
Think First…
- Spend some time just thinking about this problem and possible solutions. Start by writing a
program that clears the first page of memory, or addresses 0-177. A simple algorithm for this
would be to loop 200 (octal) times, each time using DCA (make sure the AC is 0 first) to clear the
next memory slot (starting with the first slot 0). You could do this by using DCA I with an autoindexed memory address, or by using DCA I to a normal memory slot and then ISZ to increment
that normal slot each time in the loop.
- Expand your program to clearing “most” of the memory, even if not all but 4 or fewer slots. This
just requires looping more times, and being careful that if you start to clear instructions from your
program itself you don’t cause cleared memory slots to be re-written with non-zero values.
- Now consider ways to reduce the number of instructions in your program, thus letting you clear
more memory slots without clearing the important part of your program itself.
o You might be able to get rid of your own looping entirely by using the fact that PAL
automatically loops back around to execute the instruction at memory address 0 after it
executes address 7777.
o You might be able to use auto-indexing to increment values instead of ISZ and this may be
simpler and/or use fewer instructions.
- Finally order your instructions and strategically place them in memory so that your actual program
instructions are cleared last, and as many as possible of them are cleared before the program is so
broken that it can’t do anything anymore.

More products