'Booksim'에 해당되는 글 25건

PROGRESS REPORT

Lab.work 2009. 7. 21. 10:39
■ THIS WEEK  (7/15~7/20)

□ Phoenix 2.0

about degradation after 32 threads:  
first, I conjectured that the reason was skewed distribution of words.
so I randomly generated input file using a-z and white space.
but there still exists slight degradation.

now, I divided the start_worker function into 7 parts and find the degradation part.
here I attached my work. 

□ Booksim 2.0

the simulation script is still running.
the result file using uniform traffic is attached.
I will analysis the result when other data come out.


■ NEXT WEEK  (7/21~7/27)

□ Phoenix 
Test more with huge data set and re-plot the speedup graph.
Find way to determine working set size and L2.

□ Simics & Gems
a little later..

□ Booksim 2.0
Plot all graph using other traffics and analysis the result.

'Lab.work' 카테고리의 다른 글

PROGRESS REPORT (8/3)  (0) 2009.08.04
PROGRESS REPORT  (0) 2009.07.28
PROGRESS REPORT  (0) 2009.07.07
6월 18일 발표논문  (0) 2009.07.01
6월 29일 월요일  (0) 2009.06.29
블로그 이미지

민둥

,

Virtual-Channel Flow Control

Architecture 2009. 7. 8. 17:03
Interconnection network is characterized by its topology, routing, and flow control.

Topology: arrangement of nodes and channels into a graph
Routing: how a packet chooses a path in this graph
Flow control: allocation of channel and buffer resources to a packet as it traveses the path

2 types of resources: buffers and channels.
Typically, a single buffer is associated with each channel.
Virtual channels decouple resource allocation by providing multiple buffers for each channel in the network.
In addition to incresing throughput, wirtual channelss provide an additional degree of freedom in allocating resources to packets in the network. 

Virtual Channels:
vc for deadlock avoidance
output queueing or split input queues for partial resource decoupling.

Flow control protocol:
1) how resources (buffers and channel bandwidth) are allocated
2) how packet collisions over resources are resolved

Collisions:
1) blocking P (a packet) in place
2) buffering P in a node prior to where the collision occurs.
3) dropping P
4) misrouting P to a channel other than the one it requires.

Flits.
flits have no routing or sequencing information, the allocation must be  done in a manner that keeps the flits associated with a particular packet together.

Wormhole Routing:
advance each flit of a packet as soon as it arrives at a node (pipelining) and blocks packets in place when required resources are unavailable.
1) reduces the latency of message delivery compared to store and forward routing.
2) requires only a few flit buffers per node.


'Architecture' 카테고리의 다른 글

GARNET  (0) 2009.09.15
Network traffic patterns  (0) 2009.08.02
Interconnection Network Topologies  (0) 2009.07.01
PARSEC vs. SPLASH-2  (0) 2009.06.16
The PARSEC Benchmark Suite  (0) 2009.06.05
블로그 이미지

민둥

,

PROGRESS REPORT

Lab.work 2009. 7. 7. 11:38
■ THIS WEEK  (6/29~7/6)

□ Phoenix 2.0
I re-ploted the result graph with the x-axis.
There is no degredation after 32 nodes when data set is huge.
Optimal chunk size is dependent on the application.
□ Booksim 2.0
I tested 2D mesh network with different number of virtual channel and buffer size.

■ NEXT WEEK  (7/7~7/13)

□ Phoenix 
Test more with huge data set and re-plot the speedup graph.
Find way to determine working set size and L2.

□ Simics & Gems
run and test PARSEC on Simics and Gems.

□ SigMP
SigMP seminar will be on Thursday. 

'Lab.work' 카테고리의 다른 글

PROGRESS REPORT  (0) 2009.07.28
PROGRESS REPORT  (0) 2009.07.21
6월 18일 발표논문  (0) 2009.07.01
6월 29일 월요일  (0) 2009.06.29
PROGRESS REPORT (6/29)  (0) 2009.06.29
블로그 이미지

민둥

,

Interconnection Network Topologies

Architecture 2009. 7. 1. 14:45
□ Butterfly

(2-ary 4-fly)

k-ary n-fly Butterfly Network:
An n stage of radix k switches butterfly network. 
Such a network is composed of N=kn source/destination terminal nodes.
n stages of kn-1 switching nodes each with k inputs and k outputs.
All (n+1)N channels are unidirectional. 

□ Mesh
A k-ary n-mesh (mesh) topology


□ Cmesh (Concentrated Mesh)
Concentrated mesh topology. the 'c' determines the concentration



□ Fat tree
The links in a fat-tree become "fatter" as one moves up the tree towards the root.

□ Torus
A k-ary n-cube (torus) topology
A closed surface defined as the product of two circles


□ Quad Tree
A quadtree is a tree data structure in which each internal node has up to four children. 
Quadtrees are most often used to partition a two dimensional space by recursively subdividing it into four quadrants or regions.


□ Flattened Butterfly
"Flattened Butterfly : A Cost-Efficient Topology for High-Radix Networks" ISCA 2007

□ CMO (Concentrated Multi-dimensional Octagon)
Created by Sang Kyun Kim and Wook-Jin Chung from EE382C

□ MECS (Multidrop Express Channels)
"Scalable On-Chip Interconnect Topologies"


'Architecture' 카테고리의 다른 글

Network traffic patterns  (0) 2009.08.02
Virtual-Channel Flow Control  (0) 2009.07.08
PARSEC vs. SPLASH-2  (0) 2009.06.16
The PARSEC Benchmark Suite  (0) 2009.06.05
The SPLASH-2 Programs  (0) 2009.06.05
블로그 이미지

민둥

,

6월 29일 월요일

Lab.work 2009. 6. 29. 22:21
■ Sun Machine 

케이블 가격 알아보기
혹시 자주 다운된다면 다시 OS 깔아야하나.

■ Phoenix

check ** L2가 working set size를 포함하기 못할때..!!

Phoenix on Simics!ㅠ_ㅠ
일단 실제 기계에서 다 돌려보고 Simics에서 해보삼

wordcount가 왜 초반에는 super linear speedup 인데 다음에는 떨어지는걸까
wordcount working set이랑 L2 size 확인
다른 application에서도 chuck size가 어떤 영향을 미치는지

Speedup 그래프 다시 그리기
chunk size가 정확히 어떤 영향을 미치는지 + best chunk size적용

linux x86랑 sun sparc이랑 무슨 차이? 왜??
hyper threading? super 어쩌고 threading? throughput이 더 중요한지 연산을 하는데 최적화 되어있는지 등등
+ thread끼리 공유하는 cache등등의 문제
sun 같은 경우는 모두 on chip communication이라서!?
linux같은 경우는 하나의 board에서... 그래서?

sun system config 찾아보기

■ Booksim

목요일날 다시 얘기!
Booksim에 대해서 다시 설명해주심
topology에 대해서 간단히 찾아보고, code도 대충 보기!

networks.cpp
iq_router.cpp
가장 highlevel은 trafficmanager.cpp <= 여기서 모든게 실행

"wait_for_tail_credit" config 살펴보기
booksim_config.cpp

■ Gems

목요일전에 gems돌려보기

-------------------------------------

목요일 한국시간 8시 다시 미팅


'Lab.work' 카테고리의 다른 글

PROGRESS REPORT  (0) 2009.07.07
6월 18일 발표논문  (0) 2009.07.01
PROGRESS REPORT (6/29)  (0) 2009.06.29
PROGRESS REPORT (6/23)  (0) 2009.06.23
PROGRESS REPORT (6/15)  (0) 2009.06.16
블로그 이미지

민둥

,

공지사항

최근에 받은 트랙백

글 보관함

달력

«   2024/05   »
일 월 화 수 목 금 토
1 2 3 4
5 6 7 8 9 10 11
12 13 14 15 16 17 18
19 20 21 22 23 24 25
26 27 28 29 30 31

티스토리툴바