複選題

8. Assume that the frequency of Floating Point (FP) operations is 25% for a computer, and the average CPI (Clock Cycles Per Instruction) of FP operations is 4. The average CPI of other instructions is 1.5. By performing some enhancement on the hardware, we can reduce the average CPI of all FP operations to 2. All other factors are not changed after enhancement. Which of the followings are true?
(A)The speedup for the FP operations is 2
(B) The total speedup can be computed by Amdahl's Law using the following equation:
(C)The total speedup is 1.307
(D)The CPI after enhancement is 1.859
(E) The CPI before enhancement is 2.125

複選題4. Given the execution time of the 5 stages of a CPU: IF 200 ps, ID 100 ps, EXE 200 ps, MEM 200 ps, WB 100 ps, which of the following statements are true? (A) The ideal cycle time of a single cycle CPU is 800 ps (B) The ideal execution time of "sw" for a single cycle CPU is 800 ps (C)The ideal cycle time of a multiple cycle CPU is 200 ps (D)The ideal execution time of "Iw" for a multiple cycle CPU is 800 ps (E) None of the above

複選題5. Given a 16-bit computer with the I-type instruction in the following format, which of the following statements are true? (A)The computer supports at most 16 operations (B) The computer has 8 registers (C) The range of the number in the immediate field (in 2's complement) is from -25 to 25-1 (D)Each register is 8-bits long (E) None of the above

複選題6. Which of the following statements are true for dynamic branch prediction? (A)Branch history table or branch prediction buffer can be implemented as a memory indexed by the lower bits of the program counter address. (B) Branch prediction buffer must have tag and address check. (C)A 2-bit dynamic branch predictor must miss twice before the prediction is changed. (D)Mis-prediction can be caused by getting the branch history of wrong branch when indexing the branch history table. (E) Correlating branch predictors assume that the behaviour of a branch would not be affected by the behaviours of other branches.

複選題7.For RISC (Reduced Instruction Set Computer), which of the following statements are true? (A)For any given level of general performance, a RISC chip will typically have far more transistors dedicated to the core logic compared with CISC (complex instruction set computer). (B) RISC allows designers to increase the size of the register set. (C)RISC usually uses uniform instruction format with with the opcode in the same bit positions in every instruction, which demands less decoding complexity. (D)RISC is not suitable for pipeline design and would decrease internal parallelism. (E) RISC usually supports more addressing modes than CISC. rs: 3 bits

複選題9. About instruction level parallelism, which of the followings are true? (A)To perform loop unrolling, loop iterations must be independent with no loop-carried dependence. (B) True dependencies can be resloved by register renaming. (C)Anti-dependence occurs when instruction j writes a register or memory location that. instruction i read from, and instruction i is executed first. (D)Loop unrolling can reduce control stalls. (E) Issuing multiple instructions per cycle can not reduce ideal CPI.

複選題10. About memory hierarchy, which of the followings are true? (A)For virtual memory, usually write-through strategy is used to update the lower level storage. (B) Implementing higher associativity in cache can reduce conflict miss. (C)For virtual memory, usually fully associative placement strategy is used. (D)If the page size is larger, the size of the page table is also larger. (E) If write-back strategy is used, it is possible that read misses would produce writes. 0.25

複選題11. What are the factors that should be considered when deciding the page size? (A)Size of page table (B) Fragmentation (C)l/O overhead (D)Number of page faults (E) None of the above

複選題12. Choose the correct statements from the multiple choices regarding threads (A)Kernel-level threads are easier and faster to create than user-level threads (B) Different user-level threads of the same process can have different scheduling priorities (C)A blocking kernel-level thread blocks all threads in the process (D)For the Many-to-One multithreading model, one thread blocking causes all to block (E) None of the above

13. Choose the correct statements from the multiple choices regarding deadlocks (A)If the resource allocation graph contains a cycle, then the system is deadlocked (B) If a system is in an unsafe state, then there is no allocation sequence that allows the processes to finish executing (C)Pretending that deadlocks never occur in the system is a method to deal with deadlocks (D)Banker's algorithm is a deadlock prevention algorithm (E) None of the above

複選題14. Choose the correct statements from the multiple choices (A)Each container includes the application, the necessary binaries and libraries and an entire guest operating system (B) Containers running on a single machine all share the same operating system kernel (C)Virtual machine run as an isolated process in user space on the host operating system (D)Containers are lightweight than virtual machines (E) None of the above