Explain your answers clearly. Have fun.

1. (10 points) Fooling around with the PDP8/E.

   There's a PDP8/E instruction I forgot to tell you about. It's an op-code 7 instruction called BSW (for "byte swap"), and it swaps the upper 6 bits of the AC with the lower 6 bits. Its machine code is 7002.

   • With what other op-code 7 instructions is BSW impossible to use? Why?
   • What number is in the AC after you execute CLA IAC BSW? (Go ahead and use the emulator for this.)
   • Where in the order of execution of Group I instructions does BSW fall?
   • What does the following PDP8/E code fragment do, and how? What are the disadvantages of doing this sort of thing?
     
     

     
     *0000
         A,     0000
         B,     0000
     
     *0200
         START, CLA IAC BSW
                CMA IAC
                DCA A
         C,     DCA B
                ISZ C
                ISZ A
                JMP C
                ...
     

   
   
   
2. (10 points) The table below lists execution times for all the PDP8/E instructions we have used. Please answer the following questions.
   • Does the notion of CPI make sense in the context of the PDP8/E? Why or why not?
   • Even if you said CPI does not make sense, you can still use the timing table below to compute the native MIPS for the PDP8/E if you first make some assumptions. What sort of assumptions do you need to make?
   • For any conceivable programs, what are the maximum and minimum native MIPS for the PDP8/E?
   • Using assumptions of the kind you described two questions ago, compute the native MIPS for the PDP8/E. Make sure to explain exactly what your assumptions are.
   • Based on what you know about modern workstations, how does the MIPS rating of the PDP8/E compare to modern machines?
   
   
   

   
   Execution times in micro-seconds for PDP8/E instructions
   --------------------------------------------------------
                            Addressing mode     
                 ------------------------------------------
   Instruction                            Indirect with
      Type       Direct      Indirect     auto-increment
   --------------------------------------------------------
   AND            2.6          3.8             4.0
   TAD            2.6          3.8             4.0
   ISZ            2.6          3.8             4.0
   DCA            2.6          3.8             4.0
   JMS            2.6          3.8             4.0
   JMP            1.2          2.4             2.6
   
   Opcode 7          1.2 (addressing mode irrelevant)
   --------------------------------------------------------
   

   
   
   
3. (15 points) When we implemented function-calling on the PDP8/E, we assumed each function would have only one parameter, and each stack frame would consist of two words--the parameter on top of the return address. The following are my implementations of CALL and RETURN, including the pre-conditions and post-conditions for each. Note that these routines use straight-forward implementations of PUSH and POP subroutines.
   
   

   
   / CALL -- calls a function.
   /
   /  Pre-conditions: The address of the function to be called
   /    is in the page-0 location ADDR, and the parameter of
   /    the function is in the AC.
   /
   /  Post-conditions: The return address to the caller and the
   /    parameter have been pushed (in that order) onto the
   /    system stack.  Control has been transferred to the
   /    function being called.
   
   CALL,       0000
               DCA CALLTMP
               TAD CALL
               JMS PUSH
               TAD CALLTMP
               JMS PUSH
               JMP I ADDR
   CALLTMP,    0000
   
   
   /  RETURN -- returns from a function
   /
   /  Pre-conditions: the return value of the function is in the AC.
   /
   /  Post-conditions: the stack frame has been popped, the return
   /    value is in the AC, and control has been returned to the
   /    original routine that called CALL (via the return address
   /    that was in the stack frame).
   
   RETURN,     DCA RETTMP
               JMS POP
               JMS POP
               DCA ADDR
               TAD RETTMP
               JMP I ADDR
   RETTMP,     0000
   

   
   
   

   How would you modify our function-calling mechanism to allow for a variable number of parameters? How would the stack frame be structured? Show new versions of CALL and RETURN including pre- and post-conditions that implement your extended function-calling mechanism.
   
   

4. (3 points) Direct me to an interesting Web site.
   
   
5. (12 points)
   • Draw a 4-bit two's complement subtractor. This circuit should have eight input lines corresponding to the two 4-bit numbers we want to subtract, and output lines, corresponding to the four digits of the difference.
   • Suppose every gate in your 4-bit subtractor has a propagation delay of 8 nanoseconds (ns). What is the propagation delay of the entire circuit? That is, how long after all eight input values are stable do you have to wait before you can count on all four output values being correct?
   • If you extend your 4-bit subtractor to an N-bit subtractor, what is the propagation delay of the resulting circuit?
   
   
   
6. (10 points) Consider the fictional machine Mbase from problem 14 on page 84 of P&H. The compiler team from problem 17 has gone on vacation, and the iterim compiler team has cooked up yet another compiler. The new compiler generates code with 90% as many instructions of each class as the base compiler, and then replaces every class D instruction by two class A instructions. Call the machine with the new compiler Mnew.
   • What is the CPI for Mnew?
   • How much faster is Mnew than Mbase?
   • The head of the design project reads the Mnew report, and immediately starts firing people. Who, and why?