※ 引述《a8909132 (shin)》之銘言：
: ※ 引述《privatewind (傷神客)》之銘言：
: : Compulsory misses.
: 所以說上面題目給的Address stream就是要用來判斷3c model囉？
The main factor is the media stream. Because media stream
seldom reads the past data, it usually reads the next block
to play. By descriptions of access adresses from the problem,
the fact is certain again.
: : The miss rate is proportion to the block size.
: : In a 64KB direct-mapped cache, there exists 512K / 32 = 16K misses.
: : In 16byte, 512K/16 = 32K.
: : In 64byte, 512K/64 = 8K.
: : In 128byte, 512K/128= 4K.
: : Spatial locality.
: 這邊懂了，謝謝囉^ ^
: : ^CPU^ ??
: 這邊張凡講義上是clock rate，不過有可能是它講義上有打錯。
: : CPU = 3G/3M = 100 IO/second
^^ User program要用到1M, OS也要用到2M，加起來3M
: : memory bus = 640MB/64KB=10000 IO/second
^^ bandwidth / block size = the number of blocks / seconds,
and a block fetching is equivalent to an IO operation.
: : 1 disk = 1/( 9ms+64KB/64MB) = 1/(9ms+1ms) = 100 IO/second
disk access latency = seek time + rotation time + transfer time+ overhead
1/period = times
: : disk set = 20 * 100 = 2000 IO/second
: : CPU is the bottleneck.
: 這部份我再想一下，有不懂再麻煩了，先在此說一聲謝謝，太感謝這位高手的幫忙^^
--
※ 發信站: 批踢踢實業坊(ptt.cc)
◆ From: 59.126.187.85
※ 編輯: privatewind 來自: 59.126.187.85 (09/18 10:57)
※ 編輯: privatewind 來自: 59.126.187.85 (09/18 10:57)
※ 編輯: privatewind 來自: 59.126.187.85 (09/18 10:57)
※ 編輯: privatewind 來自: 59.126.187.85 (09/18 11:05)