Appeals for Moed B should be sent to Adi by e-mail till 23.3.17
|The scanned exam must be attached to the e-mail |
Appeals should be sent to Lihu by e-mail till 21.2.17
Following a few questions I got, here are some clarifications for foil 10 in the last lecture.
|Each column (1,2,3,…) represents one cycle at allocation.|
The foil refers to a 4-wide machine, in which at most 4 uops can be allocated (the foils says “issued”) per cycle.
In this machine allocation is done all-or-none: if in a given cycle the is a back-end stall (e.g., due to RS full or due to ROB full), allocation is stalled completely.
That is: if the frontend supply n≤4 uops in a given cycle, and the backend does not have room for all n uops (e.g. RS has fewer than n free entries), allocation is stalled, and no uop is allocated.
Allocation slots classification:
Backend bound: for a cycle with a Back-end stall, all 4 allocation slots are backend bound.
Frontend Bound: in a cycle without a back-end stall, if the frontend provides n<4, this cycle is bound by Fronted supply: it has (4 – n) frontend bound allocation slots.
Retiring: the number (≤4) of uops in the cycle which eventually retire.
Bad speculation: the number (≤4) of uops in the cycle which eventually do not retire (due to some bad speculation, e.g., jump misprediction).
HW4 solution was published in the homework section
Please bring a calculator to the exam, you may need it
Reception hour before the exam - Lihu
|Lihu will hold a reception hour on Tuesday at 14:00 in Taub 5.|
Please arrive no later than 14:30 (if no one will come, Lihu will leave at 14:30).
Foil number 10 in the last lecture was updated
|Foil number 10 in the lecture "Perf Analysis in OOO cores" was updated: |
In the previous version some columns in the table were hidden.
In the new version this is fixed, and all the columns are visible.
המבחן הוא עם חומר פתוח
Reception hour before the exam - Itai
|Itai will hold a reception hour on Monday at 16:00 in Taub 5.|
Please arrive no later than 16:30 (if no one will come, Itai will leave at 16:30).
Clarification on HW4 - question 2.ד
|the following sentence:|
הקיבול הדינאמי עבור אפליקציות TDP גדל ב – 20% בצורה קורלטיבית לגידול הכולל של ה- IPC של שני ה – Threads הרצים בו זמנית על כל Core
means that when one core is running 2 threads in SMT, the total IPC will increase only by 20%
(indeed, this doesn't provide the same performance as when we add annother core, which would double the IPC)
As a result, the dynamic capacitance also increases by 20%
|Homework 4 is published. Due date is 26/01/2017.|
|In order to be aligned with the lectures, Franck will do a תרגול השלמה on Out Of Order Execution today at 16h30 in Taub 4. |
|Homework 3 is published. Due date is 12/01/2017.|
Created on 8/12/2016, 20:34:03
|There will we two tirgul sessions for Itai's class this sunday:|
- 11h30: usual class
- 16h30: תרגול השלמה in Taub6
|Homework 2 is published. Due date is 22/12/2016.|
|Homework 1 is published. Due date is 11/12/2016.|
No tutorials on the first week
|The tutorial on Wednesday 26.10.16 is cancelled|
The tutorial on Sunday 30.10.16 is cancelled
The tutorials will start following the first lecture