Computer Laboratory > Teaching > Course material 2009–10 > All courses > System on Chip Design and Modelling

 

System on Chip Design and Modelling
2009–10

Principal lecturer: Dr David Greaves
Taken by: MPhil ACS
Syllabus

Page will be updated as the course develops...


Course Material

The taught part of this course takes place over 16 sesions of one hour in SW02. The expected time allocation is that the Monday sessions are nominally for lectures and the Wednesday sessions are practical classes and group seminars, but there will be some weeks with Wednesday lectures.


The course has been arranged in four sections ... details still in flux while I determine the interests of the class.

SECTION 1: Low-level Modelling and Design Refactoring Problems

Section 1: Topics: Verilog RTL Design with examples. Basic RTL to gates synthesis algorithm. Event-driven simulation cycle. Using signals, variables and transactions for component inter-communication. SystemC overview. Hazards, retiming, refactoring.

Slides 1.1: (RTL): Pack 1.1 Not all will be used.

Slides 1.2: (SystemC): Pack 1.2, Examples.

Section 1: Slides III (Design Flow): Pack 1.3 (short)

Section 1: Assessed work:

  • Exercise: Tick 1 (due 26th Jan, 5 credits).
  • Exercise: Tick 2 (due 2nd Feb, 5 credits).

SECTION 2: Design Partition and Hybrid Modelling.

Section 2: Topics: Bus and network structures. Architecture, design partition and design database. ESL. IP-XACT.

Slides 2.0: (Soc Components) Please see next year's course.

Slides 2.1: (Technology and Design Partition) Pack 2.1

Slides 2.2: (ESL modelling/Architecture Exploration) Pack 2.2, ESL toy PRACTICALS.

Section 2: Reading List:

Section 2: Assessed work:


SECTION 3: Assertions, Higher-level Entry & Hands-on SOC Design

Section 3: Topics: Assertion-based testing. Temporal Logic, PSL, SVA. Refinement, Synthesis from higher forms (SysML, OneSpin, Kiwi, Bluespec).

Slides 3.1: (Assertion-Based Design) Pack 3.1

Slides 3.2: (High-Level Synthesis) Pack 3.2

Section 3: Reading List:

Assertion-Based Design

NB: You can run your own PSL experiments using Modelsim. There is medium-sized example in /usr/groups/ecad/mentor/modelsim/current/modeltech/examples/psl/verilog/modeling/dram_controller.

High-Level Synthesis

Section 3: Practical work: We will start with an ORP 1000 core that is already set up to run raw C code and also single-core linux. This is available in Verilog RTL and as a cycle-accurate C model and as a high-level ISS (for use with GDB). The C models can be connected to a SystemC TLM bus/system model or to a cycle accurate SystemC model. We will insert assertions and detect errors.

Section 3: Assessed work: To be added.

  • Exercise: Tick 4 (due 23rd Feb, 5 credits).
  • Exercise: Tick 5 (due 9th Mar, 5 credits).

SECTION 4:

Section 4: Topics: Real-world bus structures (AHB, AXI, BVCI), DRAM. Some examples (transactor synthesis, network synthesis, glue logic synthesis).

Slides 4.1: (Bus/NoC Structures) Pack 4.1

Slides 4.2: (SoC Topics (Scale, Power Control etc)) Pack 4.2

Slides 4.3: (Examples) Work In Progress!

Section 4: Practical work: Design and testing a cache consistency system and message passing or transactional memory for a multi-processor SoC using TLM modelling in SystemC with the ORP 1K cores.

Section 4: Assessed work:


Primary Materials

Primary Material: Tools

Tool setup for Computer Laboratory machines is described here toolsetup/README.txt.

Primary Material: SystemC

Download and install SystemC on your own system (requires registration on systemc.org) or use the provided distribution on the lab machines toolsetup/README.txt.

There are a number of on-line tutorials and examples, such as ASIC-WORLD.

Primary Material: RTL (Verilog/VHDL)

Much of the course uses SystemC, but some examples and exercises are in RTL of VHDL/Verilog style. A teach-yourself Verilog course is available from the Computer Laboratory. Please register using crsid@cam.ac.uk email address on page: https://www.ivc.cl.cam.ac.uk/ivc/login.pl.

Another on-line tutorial for Verilog is here ASIC-WORLD.

See also the part IB undergraduate course notes for ECAD and ARCH, VerilogWiki Tutorial and DJ Greaves Tutorial 1985.

PSL Property Specification Language

There are plenty of introductory and advanced on-line materials for PSL, just google for them...

Here is the standardised manual: PSL Manual

Primary Material: ORP 1K Soft Core

The ORP open risc processor can be downloaded from OpenCores or from ovpworld.

The pre-built version, based on the Embecosm downloads will be/is installed on the lab machines here: toolsetup/README.txt.


Coursework Assessments

The coursework for this ACS module is assessed using six ticks and two mini-projects. The ticks count in total for 30 percent and the mini-projects each count for 35 percent.

All coursework must be submitted both in corner-stapled hardcopy and by zip file email by given deadlines. The email will contain source code and a brief report in pdf format. This form of cover sheet (available from receiption or print your own) must be used on the top of all submissions and a member of Student Admin staff must sign and date each hardcopy submission as it is received.

A ten percent per day penalty will be applied to late submissions. The penalty is applied until the earlier of the hardcopy or email is received.

Six Proficiency Ticks

Proficiency Ticks account for 30 percent of the module credit in total.

Proficiency Ticks are marked on a scale of 1 to 5 and the full mark will be awarded for simply meeting the tick requirements without extra credit for presentation etc.. Ticks might sometimes be completed during the practical classes. The ticks must be handed in to student admin in hard-copy form and email by 4:00 pm on the Tuesday following the week where the tick was set.

It is expected that most candidates will be awarded full credit for their tickable work on first submission, but tickable work can be resubmitted later in the term for re-assessment as desired.

Mini-project I: Structured Research Essay (30 Credits)

Structured Research Essay: System-Level Design Techniques.

The deadline for the Structured Research Essay is Monday March 15th 2009 (end of Lent full term).

Mini-project II / Essay II (40 Credits).

Mini-project II involves the design and testing of either a message passing or transactional memory for a multi-processor SoC. An accompanying report and research essay is needed.

DETAILS TO BE ADDED. The deadline for Mini_project II / Essay II is the first day of Easter full term.

External moderation and assessment of the essays: Mr Peter Flake (Elda Technology Ltd).

Nominal time budget: 16 SW02 sessions, 4 post session supervisions, 20 private study, 10 proficiency ticks (excluding additional time during classes), 10 Mini-project I, 20 Mini-project II. This is a total of 80 hours, mostly in the Lent Term but also (for mini-project II) during the easter break.


Supplementary Materials


Easter Term Research Project: Suggestion

Subject: Power estimation of various memory communication and consistency models while using high-level modelling.

In this project, a rough model of the power consumption for the various IPC (inter-processor communication) configurations developed during the taught and practical parts of the course will first be developed. Then the research will be to explore how well the same results can be derived when using TLM modelling for various parts of the system.

Taxonomy of power estimation methods:
1. Proper power estimates come from post-layout simulation,
2. Rough estimates come from post-synthesis, pre-layout simulation,
3. Very rough estimates come from pre-synthesis simulation of RTL,
4. Cycle-accurate SystemC should give the same as 3,
5. TLM models can be gradually moved away from cycle accuracy by increasing the simulation quantum: how does this degrade the power estimate accuracy? Do all plots of power versus cache/size, bus width and IPC mechanisms preserve their qualitative shape, even if they are no longer quantitatively accurate ?

Supervision: I am happy to serve as supervisor for this project. More than one person may be intending to do it. I suggest they collaborate on infrastructure and reading lists, but it is a condition of my supervision that the specific research questions investigated and experimental goals do not overlap between candidates.


(C) 2010 David J Greaves.