Solved–Lab Assignment 5 –Solution

Description

Purpose:

Generate an understanding of performance, energy, and area tradeoffs. The specific context of this assignment is the design of a cache, while the some of the basic lessons do extend to compute logic. There is no code to be written, you will exercise the Cacti tool for this assignment.

(Note: The simulator does not evaluate miss rate! Keep in mind that even if a configuration takes up more resources, it may produce a lower miss rate. Try to combine your knowledge from Assignment 2 – Cache Simulator and this assignment to understand design tradeoffs)

Part I: Downloading and Installing Cacti.

For this assignment, you will use the Cacti tool – an integrated cache access time, cycle time, area, leakage, and dynamic power model for uniform and non-uniform cache architectures. The download link for the current version and installation instructions is as follows.

1. Download link: http://www.cs.utah.edu/~rajeev/cacti6/

2. Under the “Latest Update” section, click on “CACTI 6.5” to start downloading the cacti65.tgz file

3. Untar the file (Linux):

tar -xvf cacti65.tgz

4. Execute the following commands cd cacti65

make

You may need to install the correct dependencies on your host machine, please come to office hours if you are having trouble setting up.

4. Leave all parameters as default expect for the ones identified below.

4. Comment out -cache type “main memory” from the cache.cfg file and uncomment – cache type “cache”

4. Comment out -Data array cell type – “comm-dram” from the cache.cfg file and uncomment -Data array cell type – “itrs-hp”

4. Have “UCA Bank Count” as 1

4. Technology node 0.032

5. Access mode is set to “sequential”

6. Optimize tag is set to “ED”

7. For future reuse, make a copy of the cache.cfg file, this will be your “default” cache.cfg file

8. run the executable:

./cacti -infile ./cache.cfg to generate simulation results

It may be easier to generate a script to execute the configurations required below.

_{ECE 3056}

Part II: Analysis (80 pts)

First quantify certain relationships between performance and energy. Then find an optimized cache configuration as described below. Consider the design of a 256 Kbyte data cache. Provide the following.

1. Energy Analysis:

1. 1. [10 points] Provide a bar chart of the dynamic read access energy as a function of associativity (2,4,8,16) with a block size of 64 bytes. See example figure below. Provide a brief explanation in architectural terms, i.e., look at the structure of the cache from the text/class. What aspect of the cache is responsible for this behavior.

b.

1. 3. [10 points] Provide a bar chart of the dynamic read access energy as a function of block size (32, 64, 128, 256) with an associativity of 4. Provide a brief explanation in architectural terms, i.e., look at the structure of the cache from the text/class. What aspect of the cache is responsible for this behavior?

2. Time Analysis:

2. 1. [10 points] Provide a bar chart of the access time as a function of associativity (2,4,8, 16) with a 64-byte block size. Provide a brief explanation in architectural terms, i.e., look at the structure of the cache from the text/class. What aspect of the cache is responsible for this behavior?

2. 2. [10 points] Provide bar chart of the access time as a function of block size (32,64, 128, 256) for an associativity of 4. Provide a brief explanation in architectural terms, i.e., look at the structure of the cache from the text/class. What aspect of the cache is responsible for this behavior.

2. Configuration Analysis:

2. 1. [40 points] Find a set associative cache configuration that fits as closely as possible in 1mm² (configuration parameters must be a power of 2), and has minimum energy-delay product (product of dynamic read access energy and

2

_{ECE 3056}

access time). Provide a brief and clear explanation as to the steps you followed (perhaps influenced by 1 and 2 above) to find this configuration. This is more important than the answer you find.

Grading Guidelines

1. Complete plots and clear explanation of 1 and 2: 40 pts

2. Methodology and explanations for item 3: 40 points.

The overall report is expected to be on the order of 2-3 pages or so.

Submissions Instructions

Brevity and precision is valued over volume in your writing. Your submission should be a single PDF file called Assignment-5-cacti.pdf submitted to Canvas.

3