This should take no longer than 30 minutes and is simply intended to probe whether you have understood the material in the prac. Use your textbook if you think it will help. Answer on the question sheet (3 sides).)
1. I have a memory like my friend Eratosqenes and his sief and cannot remember - remind me of your name (surname) and student number [1]
Pat Terry 63T0844
2. Translate the following program into PVM code, making use of the LDL, STL, INC, DEC and CAP opcodes that you should have introduced into the system during the course of the week. [10]
void main () {
// Reads and stores a simple sentence
// Ima Jeenyus, 2008
int n = 0;
char ch;
char[] sentence = new char[1000];
repeat
read(ch);
write(toUpperCase(ch));
sentence[n] = ch;
n++;
until (ch == '.');
}
0 DSP 3
2 LDC 0
4 STL 0 ; n = 0
6 LDC 1000
8 ANEW
9 STL 2 ; char [] sentence = ...
11 LDA 1
13 INPC ; read(ch);
14 LDL 1
16 CAP
17 PRNC ; write(toUpperCase(ch));
18 LDL 2
20 LDL 0
22 LDXA
23 LDL 1
25 STO ; sentence[n] = ch;
26 LDA 0
28 INC ; n++
29 LDL 1
31 LDC 46 ; ch == "."
33 CEQ
34 BZE 11
36 HALT
3. Having a character data type is all very well, but it could get you into trouble. Consider code like
int i = 0;
char c = 'a';
while (i < 1000) {
i++; c++;
}
Sooner or later (but, of course, when you least wanted trouble) this sort of code would cause trouble. What extensions would you suggest be made to the opcode set and/or the interpreter to allow it to catch such errors? [4]
There are various strategies we could employ. The simplest is probably to have a variation on the STO opcode (and its derivatives like STL) and on INC and DEC that will perform range checks:
case PVM.stoc: // character checked store
tos = pop(); adr = pop();
if (inBounds(adr))
if (tos >= 0 && tos <= maxChar) mem[adr] = tos;
else ps = badVal;
break;
case PVM.stlc: // character checked pop to local variable
tos = pop(); adr = cpu.fp - 1 - next();
if (inBounds(adr))
if (tos >= 0 && tos <= maxChar) mem[adr] = tos;
else ps = badVal;
break;
case PVM.incc: // ++ (characters)
adr = pop();
if (inBounds(adr))
if (mem[adr]) < maxChar) mem[adr]++;
else ps = badVal;
break;
case PVM.decc: // -- (characters)
adr = pop();
if (inBounds(adr))
if (mem[adr]) > 0) mem[adr]--;
else ps = badVal;
break;
4. The PVM currently supports the INPI, INPB and INPC operations that expect to find a destination address on the top of the stack, as you should know, favouring the translation of code like
read(x);
read(sentence[i]);
Suppose instead that the system were changed to support a style of programming like
x = readInt() + readInt(); // read and add two integers
sentence[i] = readChar(); // read and store a single character
What implications would this have for the PVM, and what code sequences would be needed to correspond to these fragments? [4]
One would need to modify the INPx instructions so that they did not store, but push the value read onto the stack:
case PVM.inpi: // integer input
push(data.readInt());
if (data.error()) ps = badData;
break;
case PVM.inpb: // boolean input
push(data.readBool() ? 1 : 0);
if (data.error()) ps = badData;
break;
case PVM.inpc: // character input
push(data.readChar());
if (data.error()) ps = badData;
break;
The source above would then have to be translated to
LDA x
INPI
INPI
ADD
STO
and
LDL sentence
LDL i
LDXA
INPC
STOC
5. Suppose the PVM were to be extended to provide a few more "functions", specifically MIN and SQR - the first for selecting the lesser of the two values currently on the top of the stack, and the second for squaring the value on the top of the stack.
(a) Show the additions that would be needed to the switch statement in the emulator (as it appears on pages 62 - 65 of the text) to support these new opcodes. [6]
case PVM.min: // minimum
tos = pop();
sos = pop();
push(tos < sos? tos : sos);
break;
case PVM.sqr: // square
tos = pop();
push(tos * tos);
break;
(b) Show how they could be used in a translation of the statement
write(sqr(min(a, min(b, c)));
(Hint - you should be able to do this in about 7 lines of PVM code). [3]
LDL a
LDL b
LDL c
MIN
MIN
SQR
PRNI
(c) Briefly indicate where, if anywhere, you would have to make other changes to the PVM.java file to handle these new opcodes [2].
We should have to add internal values for min and sqr to the "enumeration" at the start of the file, and we should have to add the corresponding initialisation of the elements of the mnemonic look-up array (at the end of the file). Since these are "zero address" instructions, they do not give rise to further changes in the parts of the file responsible for tracing, assembling or for listing code.