Re: [建議] Final project grading 注意事項

[ric2k1 (Ric)]

※ 引述《ric2k1 (Ric)》之銘言：
(新增/修訂的部分用水藍色標示)

說明一下我們對於期末專題大致的評分原則，同學們可以自行斟酌參考。


我會選擇性的挑一些(或全部，或 combine)項目來測試，每個 dofile 隨著重要性

有一定的占分。


請注意，spec 有說明，分數分配為:

Report          15% (Please do spend time to write a good report!!)

Implementation  85%

- Circuit parsing and reporting: 40%
  (correctness: 75%, error detection: 15%, performance: 10%)

- Strash:                       10%  (correctness: 90%, performance: 10%)

- Simulation:                   20%  (correctness: 70%, performance: 30%)

- Fraig:                        25%  (correctness: 50%, performance: 50%)

Bonus Up to extra 10%


1. CIRRead <(string fileName)> [-Replace]

Major tests:

- Can read in correct aag files correctly (tested with "cirp")

- Able to handle large circuit parsing

- Able to re-read circuits (-replace)

Minor tests:

- Able to detect if there's error in aag file

- Runtime and memory efficiency


2. CIRPrint [-Summary | -Netlist | -PI | -PO | -FLoating | -FECpairs]

Major tests:

- Able to correctly report "-summary", "-PI" and "-PO" BEFORE optimization

- Able to report "-Netlist" anytime without crashes.

- Given a circuit with no structurally equivalent gates, able to report

  "-netlist" and "-floating" consistently with ref prog BEFORE performing

  "strash" and "fraig".

- Given a small circuit (#gates < 100), able to report ALL the FEC pairs.

  (Tested with "CIRSimulate" command)

Minor tests:

- Able to report "-floating" and "-fec" anytime without crashes.


3. CIRGate <<(int gateId)> [<-FANIn | -FANOut><(int level)>]>

Major tests:

- Given a circuit with no structurally equivalent gates, able to report

  fanins consistently with ref prog BEFORE "strash" and "fraig".

- Able to report fanouts without crashes.

- Able to do "CIRG" (no option) anytime without crashes.

Minor tests:

- Given a large circuit, able to report deep fanins/fanouts without crashes.


4. CIRSTRash

Major tests:

- Given some very small circuits (#gates < 10), able to perform ALL the

  trivial strash operations.

- Given a circuit composed by AND(f, !f) of two structurally identical

  circuit, able to simplify it into a constant 0.

- Should be reasonably fast (linear time complexity)

- myHash.cpp is implemented correctly (tested by a separate program).

Minor tests:

- Able to perform strash on large circuit.


5. CIRSimulate <-Random | -File <string patternFile>>
               [-Output (string logFile)]

Major tests:

- Given a small circuit (#gates < 100), able to identify ALL the FEC pairs

  by random simulation. (Tested with "CIRGate" command)

- Given a large file, able to simulate from pattern files and write the

  output to file correctly.

- Given a medium size (~1000 gates) file, able to perform simulation

  efficiently.

Minor tests:

- Able to repeat the simulation commands.


6. CIRFraig

Major tests:

- Given a circuit composed by two functionally equivalent circuits (e.g.

  sim10.aag), able to EFFICIENTLY simplify it into a constant WITHOUT

  performing strash.

- Given a large circuit, able to preform fraig efficiently.

Minor tests:

- Given a large circuit, able to repeatedly perform "cirstrash" -->

  "cirsim -random" --> "cirfraig" at least 3 times to simplify the circuit.


--
※ 發信站: 批踢踢實業坊(ptt.cc)
◆ From: 114.36.54.188
→ ric2k1:等一下來說明一下 "report" 的要求
※ 編輯: ric2k1          來自: 114.36.62.203        (01/16 02:37)

--
※ 發信站: 批踢踢實業坊(ptt.cc) 
◆ From: 114.36.62.203
推 ckmarkoh:fraig指令執行之後不能有strash的動作嗎?
→ ckmarkoh:可是若為了增加效率 每化簡一個就會做strash和optimize
→ ckmarkoh:是不被允許的?