Questions and Answers of Computer Architecture

Consider the following loop that adds a constant to a vector (we discussed this earlier). There's quite a lot of overhead associated with the solitary SIMD instruction. Suppose you were designing a
What is the effect of the PACKssdw MM0, MMl instruction if initially MMO contains OxE000001200000611 and MMl contains Ox00102222FFFFFFFF?
You decide to add a new architectural feature to a processor by creating some new instructions; that is, you are extending its ISA. What consequences could these additions have for the existing ISA?
Consider the waveform in Figure P.5.26. If it were applied to a simple DSP with the transfer function y; = 0.7x; + 0.3x;_1, what would the output look like? Assume the data is 0.0, 0.0, 0.20, 0.5,
What lessons did Intel's MMX and AMD's 3DNow! extensions teach us about ISAs?
You are asked to design a new processor with a 64-bit word. Taking advantage of advances in technology, you decide that you can allocate an extra 5 bits to each word. That is, data words in registers
Explain what the following fragment of code achieves. the data is signed and that the packed shift right arithmetic instruction operates on word (16-bit) operands. MOVQ MMO, MM1 PSRAW MMO, 15 PXOR
Consider the following block of operations that might be found inside a loop. Explain what the instructions do and what operation is being performed on the data. MOVQ MOVQ MM1, A MM2, B MOVO MM3,
What is performance in the context of computer systems and why is it so difficult to define?
A system consists of a CPU, cache memory, main store, and hard disk drive. Where are time and effort best spent improving the system's performance? What factors affect your answer?
Should metrics for computer performance be linear or non-linear? For example, if a linear metric has a value X, the metric 2X would imply twice the performance , whereas if the metric were
Should metrics for computer performance be linear or non-linear? For example, if a linear metric has a value X, the metric 2X would imply twice the performance , whereas if the metric were
The time taken by machines A, B, and C to execute a given task is What is the performance of each of these machines relative to machine A? A B C 16 m, 9 s 14 m, 12 s 12 m, 47 s
Why is clock rate a poor metric of computer performance? What are the relative strengths and weaknesses of clock speed as a performance metric?
The timing diagram in Figure P6.7 illustrates a system in which ope rations occur as three consecutive clock cycles. Actions taking place in clock cycle 1 are scalable; that is, if the clock cycle
What are the relative strengths and weaknesses of the MIPS as a metric of computer performance?
Can you think of a better metric than MIPS or clock speeds that gives a good impression of the power of a processor (without having to use benchmarks).
How is it possible for one computer with a low MIPS rating to have a better performance in practice than a computer with a high MIPS rating?
Overclocking a computer means operating it at a higher clock rate th an that specified by its manufacture r; for example, a 2 GHz chip might be clocked at 2.1 G Hz to squeeze more performance out of
The following figures define the typical operating parameters of a processor. If the clock rate could be reduced by 15% , it would require only 2 cycles to perform a registe r load. Would that be a
A computer has the following parameters. If the average performance of the computer (in terms of its CPI) is to be increased by 20% while executing the same instruction mix, what target must be
A program is run on a computer with the following parameters. What is the MIPS rating of this computer? Clock cycle time Instructions with 1 cycle Instructions with 2 cycles Instructions with 3
For the following data, what is the average number of cycles per instruction? Operation Arithmetic/logical instructions Register load operations Register store operations Unconditional branch
In a particular system, a CPU is used for 78% of the time and a disk drive for 22 % of the time. A designer has two options: a. improve the disc performance by 40% and the CPU performance by
For the following systems that have both serial and parallel activities, calculate the speedup ratio. a. 10 processors b. 100 processors c. 5 processors d. 100 processors fs = 0.1 f, = 0.1 f = 0.4
A computer employed in arithmetic processing uses a software division routine. A program runs for two minutes on this machine with division taking 60% of the total time. If we wish to add a dedicated
A system containing several operational units may have multiple enhancements. In Figure P6.20, a system consists of a process S followed by processes P1 and P2• Both processes P1 and P2 can be
A program is executed in 200 ms during which 250 million instructions are executed. What is the average MIPS for this program?
A coprocessor is added to a computer to speed the execution time of string-processing instructions by a factor of 3.5. What fraction of the execution time must use these string-processing
Consider the current high-performance desktop computer and the laptop ( or notebook). Suppose you wish to increase the performance of both machines. Do you think that the same elements of the system
Amdahl or Gustafson? Consider the following example. A physical process involving a flat area 100 units by 100 units is being simulated. Processing of the units can take place in parallel. However,
Someone decided to use the following C code as part of a benchmark to determine the performance of a computer including its memory. It has two potential faults. What are they? for (i = 0; i < 100;
You are redesigning a system. You can replace the existing single processor by two P processors or by four Q processors. However, the P processors are able to run 80% of the code in parallel, whereas
An operation can be speeded up by applying two different optimizations, 01 and 02• These optimizations operate on different parts of the process and there is no overlap. If 01 speeds up fraction f1
You manufacture a computer that executes a program in 50 minutes whereas your competitor's takes 45 minutes. How are you going to sell (advertise) your processor?
What are the relative advantages and disadvantages of arithmetic, geometric, and harmonic means as methods of averaging benchmarks?
For two benchmarks, x and y, show that their arithmetic mean is always higher than, or the same as, the geometric mean.
The SPEC benchmarks present results with respect to a standard machine by normalizing the benchmarks. That is, a set of benchmarks is run on a reference machine and the times obtained for each of the
Two computers and a reference machine produce the following results. Present the results in a normalized form and provide benchmarks for machines A and B. Machine Reference A B Benchmark 1
In 2013, a woman with a small business at home is going to buy a desktop computer to handle her correspondence and diary/calendar, to allow her to email colleagues and to deal with her tax. Being a
This question asks you to implement register indirect addressing. For the architecture of Figure P7.1 , write the sequence of signals and control actions necessary to execute the instruction ADD
This question asks you to implement memory indirect addressing. For the architecture of Figure P7.1, write the sequence of signals and control actions necessary to execute the instruction ADD [M] ,
For the microprogrammed architecture of Figure P7.l, define the sequence of actions (i.e., microoperations) necessary to implement the instruction TXPl (DO)+' Dl that is defined as:Explain the
Why was microprogramming such a popular means of implementing control units in the 1980s?
Why is microprogramming so unpopular today?
Figure P7.12 from the text demonstrates the execution of a conditional branch instruction in a flow-through computer. The grayed out sections of the computer are not required by a conditional branch
What modifications would have to be made to the architecture of the computer in Figure P7.12 to implement predicated execution like the ARM? FIGURE P7.12 PC_MPLX 00 014 10 11 BRA Target where the
What modifications would have to be made to the architecture of the computer in Figure P7.12 to implement operand shifting ( as part of a normal instruction) like the ARM? FIGURE P7.12 PC_MPLX 00
Derive an expression for the speedup ratio (i.e., the ratio of the execution time without pipelining to the execution time with pipelining) of a pipelined processor in terms of the number of stages
A processor executes an instruction in the following six stages. The time required by each stage in picoseconds (1,000 ps = 1 ns) is given for each stage. a. What is the time to execute an
Both RISC and CISC processors have registers. Answer the following questions about registers. a. Is it true that a larger number of registers in any architecture is always better than a smaller
Someone once said, "RISC is to hardware what UNIX is to software'? What do you think this statement means and is it true?
What are the characteristics of a RISC processor that distinguish it from a CISC processor? Does it matter whether this question is asked in 2015 or 1990?
What, in the context of pipelined processors, is a bubble and why is it detrimental to the performance of a pipelined processor?
To say that the RISC philosophy was all about reducing the size of instruction sets would be wrong and entirely miss the point. What enduring trends or insights did the so-called RISC revolution
There are RAW, WAR, and WAW data hazards. What about RAR (read-after-read)? Can a RAR operation cause problems in a pipelined machine?
Consider the instruction sequence in a five-stage pipeline IF, OF, E, M, OS: Instructions 1 and 2 will create a RAW hazard. What about instructions 3 and 4? Will they also create a hazard? 1. ADD
A RISC processor has a three-address instruction format and typical arithmetic instructions (i.e., ADD, SUB, MUL, DIV etc). Write a suitable sequence of instructions to evaluate the following
Figure P7.27 gives a partial skeleton diagram of a pipelined processor. What is the purpose of the flipflops (registers) in the information paths? FIGURE P7.27 +4 MPLX PC CLK Structure of a
Explain why branch operations reduce the efficiency of a pipelined architecture. Describe how branch prediction improves the performance of a RISC processor and minimizes the effect of branches?
Assume that a RISC processor uses branch prediction to improve its performance. The following table gives the number of cycles taken for predicted and actual branch outcomes. These figures include
RISC processors rely (to some extent) on on-chip registers for their performance increase. A cache memory can provide a similar level of performance increase without restricting the programmer to a
RISC processors best illustrate the difference between architecture and implementation. To what extent is this statement true ( or not true) ?
Consider the following code:The processor has a five-stage pipeline F O E M S; that is, instruction fetch, operand fetch, operand execute, memory, and operand write back to register file. a. How
Why do conditional branches have a greater effect on a pipelined processor than unconditional branches?
Describe the various types of change of flow-ofcontrol operations that modify the normal sequence in which a processor executes instructions. How frequently do these operations occur in typical
What is branchless computing?
What is a delayed branch and how does it contribute to minimizing the effect of pipeline bubbles? Why are delayed branch mechanisms less popular then they were?
A pipelined processor has the following characteristics: Estimate the average cycles per instruction for this processor. Loads Load stall (load penalty) Branches Probability a branch is taken
What is the difference between static and dynamic branch prediction?
A processor uses a 2-bit saturation-counter dynamic branch predictor with the states strongly taken, weakly taken, weakly not taken, and strongly not taken. The symbol T indicates a branch that is
The following sequence of branch outcomes is applied to a saturating counter branch predictor TTTNTTNNNTNNNTTTTTNTTTNNTTTTNT. If the branch penalty is two cycles for an incorrectly predicted branch,
Consider the 4-bit saturating counter as a branch predictor with 16 states from 1111 to 0000? Describe in words the circumstances where such a counter might be effective.
Draw the state diagram of a branch predictor using a 3-bit saturating counter? Under what circumstances do you think such a predictor might prove effective?
The following code is executed by an ARM processor: Assume that a 1-bit branch predictor is used for both branch 1 and branch 2 and that both predictors are initially set to N. Complete the
A processor executes all non-branch instructions in one cycle. This processor implements branch prediction, which incurs an additional penalty of 2 cycles if the prediction is correct and 4 cycles if
A computer has a branch target buffer, BTB. Derive an expression for the average branch penalty if the following apply. • A branch not in the BTB that is not taken incurs a penalty of O
A RISC processor implements a subroutine call using a link register (i.e., the return address is saved in the link register).The cost of a call is 2 cycles and the return costs 1 cycle. If a
Why is the literal in the op-code sign-extended 9 before use (in most computer architectures)?
Why is the address offset shifted two places left in branch/jump operations in 32-bit RISC-like processors?
Assume a five-stage pipeline (instruction fetch, operand fetch, execute, memory, write-back). For the following code show any stalls and indicate where operand forwarding would be needed. ADD R9,
Superscalar processing could be added to any existing processor without having to recompile source code. Why?
The performance of a computer can be expressed as the time taken to execute a task. That time can is given by Many factors affect the design of a processor; for example, technology, ISA, the
If a VLIW form of an existing processor were to be produced, why would the source code have to be recompiled?
Mainstream techniques for accelerating the performance of the computer are the RISC instruction set philosophy, superpipelining, superscalar technology, VLIW technology and code morphing. Briefly
In 1991, Nick Tredennick said: "Superpip elining is a new and special term meaning pipelining. The prefix is attached to increase the probability of funding for research proposals. There is no
You are redesigning an existing RISC-style pipelined processor to make it a three-way superscalar processor that permits three instructions to be executed at the same time. a. Do you expect the new
Is the IA64 Itanium a VLIW processor? Give your reasons.
What is register renaming and why is it used in superscalar processors?
Why does register renaming prevent write-after write hazards (WA W) but not read-after-write (RAW) hazards?
What is the theoretical upper bound on the CPI metric for a RISC processor without superscalar facilities?
What is the limitation on the number of registers (i. e., the user-visible register file) in a processor?
Describe the difference between superscalar and VLIW architectures.
What is software loop unrolling and why does it help to speedup execution?
What is predication and how can it be used to speed up program execution?
The superscalar and VLIW processors share fundamentally different relationships with their compilers. Why?
Consider the following fragment of assembly language that is to be executed on a generic three way superscalar processor that can accommodate one memory access per clock. Rewrite the code to improve
Both the ARM processor and the IA64 architecture have predicated instruction execution. In what way is the IA64's predication mechanism superior to the ARM processor's?
In the context of the IA64, what is the difference between instruction and data speculation?
Demonstrate how a compiler might translate the following fragment of code into a form that can run on a VLIW processor that specifies three operations per instruction word (i.e., three instructions

Showing 500 - 600 of 1374