Questions and Answers

Q:

Could you explain in personal email how FOO works?

A:

I (bsy) do not want to type in long explanations in email. I have tendonitis, and would prefer to avoid typing unnecessarily. If you have a question, come by my office and ask it to me face to face; I can get better feedback on what your confused about, and avoid typing explanations on parts that you already know and give more details on what you don't.

Q:

What if I keep the calculator stack in registers? Is that okay?

A:

It's okay to use registers, but you have to use separate functions for the binary operators; that's one of the requirements for this assignment. The standard calling convention do not allow you to assume anything about the contents of the s registers, nor can you change them in the functions -- when functions return, the s registers must have the same values that they had before the function is called. Similarly, the caller is not allowed to make any assumptions about what the called function did with the t registers.

If you had stuck with the standard register usage / calling conventions as required, and you kept the calculator stack in registers, then the only function that could possibly know how the stack is implemented is your main function, which would satisfy the abstraction requirement to limit knowledge about the stacks implementation to a few functions.

Q:

Can the stack be global?

A:

The stack can be global, but it should be abstract: the details of how it works should be isolated to main or to push/pop routines -- the functions implementing the calculator's binary operators should not know the details of how the stack is implemented.

Q:

What sort of illegal commands do we have to detect?

A:

We are pretending that we are implementing a circa 1970 HP calculator. Keystrokes are always single characters. (Numeric input is multicharacter only because it would be too painful to type in numbers with a ``Enter'' after every digit.) You can just look at the first character of the input and ignore the rest for the non-numeric input. For numeric input you must check that every character is valid and generate an error message if an invalid character is detected. See ~/../public/calc for the I/O behavior that I expect your program to have.

Q:

Do I have to have variables named x, y, z, and w, or can I have an explicit stack or a pointer to the top of the stack or ....

A:

You may do it any way you please, except that your program must satisfy the I/O requirements of the assignment (do the operations, be able to output negative numbers, etc), use separate functions for each of the binary operators, and obey the standard calling conventions.

Q:

In assignment 4, do we need to manipulate the stack in the functions or can we do it in the main loop? Likewise, can we detect errors such as division by zero in the main loop?

A:

That can be in the main loop.

Q:

In assignment 4, do we have to turn in our C or C++ version, or is turning in the MIPS version ``good enough''.

A:

You should have C, C++, or at least high level pseudo-code to document your MIPS assembly code. I strongly recommend that you get your code working as C or C++ code first, before translating it to MIPS. This avoids conceptual bugs and makes debugging easier. You should turn in the C or C++ version as part of the documentation for your MIPS program; it does not have to work.

Q:

How are expanded ble, bgtu, mul, la, etc... ?

A:

Write a short asm program using them (it doesn't have to make any sense). Then load it into SPIM and check the code panel. It will show how these pseudo-instructions were expanded.

Q:

In assignment 4, do we have to have a jump table?

A:

No. I talked about jump tables because there was a question about how to translate switch statements. While you should know how to use jump tables -- this was part of the reading assignment -- I do not require that you use them for this assignment. In this assignment, you are given the functional specifications, and you must design the overall software architecture -- with the exception that I require you to use separate functions for each binary operator (+, -, *, /, ^) of the calculator. Your design should be reasonably clean and well documented, so it is easy to understand / modify.

Q:

In assignment 3, when I enter in test cases with large values, e.g., 100¹⁰⁰, the MIPS emulator tells me that exception 9 was caught and ignored. Should I worry about this?

A:

You should worry about the size of the MIPS registers instead. You should answer the following questions yourself: (1) how many bits are in a MIPS R2000 register? (2) How many bits would be required to represent the answer to 100¹⁰⁰? (3) What are overflows anyway? Understanding this would be helpful in many situations, including the midterm/final.

Q:

What's the difference between the branch instructions and the jump instruction?

A:

Read page 148 of Patterson & Hennessey. This was part of the Chapter 3 reading.

Q:

Do I have to handle hexidecimal input for assignment 3?

A:

You are just supposed to write the exp routine. You should not modify the main code at all -- that is where the I/O is handled.

Q:

I see that you are logged on ieng9. I am trying to use talk. Why don't you ever answer?

A:

The fact that you see me logged on with w or other commands doesn't mean that I am actually sitting in front of my computer.

Q:

Can I use spim at home?

A:

Yes. You can use the text-only version (spim) through any connection to ieng9 (even a very slow modem). If you want to use the X-Windows version (xspim), you need an X server on your home machine (there might be freely available X servers for Windows 95 and the Mac). You can't run xspim across the wire without it (it will say that it can't find a DISPLAY). I am told that there is also a native Windows version of spim available through FTP (see the book, page A-40). This is certainly going to be easier, and you won't need to be connected at all. In case you are using it, make sure that your solution runs the same on ieng9, as we won't go to your place to see it work.

NOTE: PCs are little-endian machines, whereas the Suns are big-endian machines. If your code has endian dependencies, then you will definitely have problems getting it to work on the Suns if you first get it to work on a PC.

Q:

Can I use the mult instruction in assignment 3?

A:

Yes.

Q:

How do I include my test cases in assignment 3?

A:

As comments in your .asm file (close to your code). It doesn't matter exactly where, just make sure it's clear.

Q:

How do I run the turnin program?

A:

Go to a shell and type in help turnin for instructions provided by ACS.

Q:

When is the midterm?

A:

Nov 3rd.

Q:

Is this going to be on the (midterm, final)?

A:

Yes.

Q:

Will we need to know how to do related base conversions fast?

A:

Yes. On the test there will be base conversion problems. If you have to convert first to base 10 and then to the target base, you will take much too much time doing those problems. My tests are hard and have too many questions, and you will not be allowed calculators.

Q:

Does our assignment 2 have to handle 0⁰?

A:

Yes. It may return either 0 or 1. (Mathematically, it is not well defined: look at various limits of x^y.)

Q:

In a subroutine call, couldn't we just move the complete return instruction into where it should go rather than generating it in two steps?

A:

Yes, that would be better. The dynamic generation method is illustrative: it can be used to create a switch statement, or to implement a read/write array.

Q:

Should assignment 2 use two subroutines, i.e., one for mult and another for add, or is one enough.

A:

Having a single subroutine for mult is enough.

bsy@cse.ucsd.edu, last updated Mon Nov 17 21:29:47 PST 1997.