[理工] 102交大 計組
1.
MIPS is not a unfair metric to compare the performance of two computers because MIPS metric only counts the number of executed instructions but does not consider the CPU idle time for stalling CPU requested by cache miss
2.
programming language and compilers heavily influence the number of executed machine instructions per operation
3.
Algorithms determine the number of executed operations
知道這三段敘述是錯的,可是不太懂他錯在哪邊
4.the number with all bits in exponent and fraction field being 1 is referred to as infinity in IEEE standard 754.
這個敘述是對的,但無限大不是exponent 欄位是1 然後 Fraction 全0?
Fraction 非0的話是用來表示NaN 應該是這樣吧?!
5.
Consider the code sequence below, where we predict beq as taken, but actually it is not taken
For pipelined MIPS CPU which has data forwarding mechanism and the branch outcome is determined in MEM stage it takes 10 cycles to complete the execution of the above code sequence
解答給的算式是 (5-1)+4+1+1=10
但Beq在Mem階段決定猜錯的話,懲罰是3個cycle吧
我算出(5-1)+4+1+3=12cycles
不知道哪裡出錯了
謝謝大家了
--
※ 發信站: 批踢踢實業坊(ptt.cc)
◆ From: 1.164.17.252
→
01/03 17:22, , 1F
01/03 17:22, 1F
推
01/03 17:30, , 2F
01/03 17:30, 2F
→
01/03 17:30, , 3F
01/03 17:30, 3F
→
01/03 17:31, , 4F
01/03 17:31, 4F
→
01/03 17:31, , 5F
01/03 17:31, 5F
→
01/03 17:31, , 6F
01/03 17:31, 6F
→
01/03 17:31, , 7F
01/03 17:31, 7F
→
01/03 17:32, , 8F
01/03 17:32, 8F
→
01/03 17:32, , 9F
01/03 17:32, 9F
→
01/03 17:32, , 10F
01/03 17:32, 10F
→
01/03 17:33, , 11F
01/03 17:33, 11F
→
01/03 17:33, , 12F
01/03 17:33, 12F
→
01/03 17:33, , 13F
01/03 17:33, 13F
→
01/03 17:40, , 14F
01/03 17:40, 14F
→
01/03 17:40, , 15F
01/03 17:40, 15F
推
01/03 17:44, , 16F
01/03 17:44, 16F
→
01/03 17:45, , 17F
01/03 17:45, 17F
→
01/03 17:48, , 18F
01/03 17:48, 18F
→
01/03 17:50, , 19F
01/03 17:50, 19F
→
01/04 09:31, , 20F
01/04 09:31, 20F
推
01/09 17:47, , 21F
01/09 17:47, 21F
討論串 (同標題文章)