Computer Architecture Lab/SS2015
|Dropbox is not an appropriate source for file links. Please upload files you'd like to share to Wikiversity instead.|
Marek Dejak s136075
Bartek Gasior (s107181)
Luca Pezzarossa s121273
Jesper Lønbæk s094726
Russell Barnes s146105
Jakob Toft s113012
Design and implementation of an Ethernet connection, based on the UDP protocol, between the Patmos processor and a PC (or another Patmos processor) for the Terasic DE2-115 development board.
Jens Joachim Krogsløkke Pedersen s103023
Thomas Onstrup Risager s103040
Stephan A. Nielsen s103035
Project: Roll-your-own CPU, based on a 16-bit RISC'ish architecture, with compact home made 16-bit instruction set.
Joakim Sindholt s093442
Martin Wiingaard s103639
Daniel Kenji Pedersen s103636
- Project Definition
- Develop a NoC that uses distributed routing. Investigate the opportunity to remove the routing part of the header to enable a lager word pointer and increase the size of the payload in order to increase throughput.
Alexandros Andreou, s145883
Linas Kaminskas, s082936
Mohamed Elmi, s146899
Our current ideas for a project will be one of the following:
- Floating Point Unit (FPU) for Patmos.
- While in general FPU's can consume a lot of power and processing time, performing DSP on embedded systems is still very useful and is performed in some scenarios. For example when one is working with a data collection systems that includes a multitude of sensors (ADC's, DAC's, actuators, accelerometers etc.) and one wants to do real-time processing. The goal is to implement an IEEE 784-2008 compliant FPU unit and instructions which the Patmos project can make use of. The IEEE 784-2008 standard can be found at IEEE Xplore Digital Library. The older standard IEEE 784-1985 can also be found at IEEE Xplore Digital Library. Further material at IEEE-754 Reference Material.
- Direct Memory Access (DMA) for Patmos.
- For a SoC or a modern computer architecture, Direct Memory Access (DMA) is undoubt one of the most useful features to have when moving data from peripherals to peripherals, memory to memory, or memory to peripherals (and vice versa). This reliefs the processing unit of performing tedious looping in order to copy or move data between two points. By having an arbitration mechanism or a secondary bus for data movement, and a fast implementation of a DMA unit, performance and power reductions can be gained, as parts of the main processing unit are not needed. With an interrupt mechanism, the processor can either go into a sleep state or continue by performing some other useful work. This project will concern the implementation of a simple DMA unit that is capable of moving data for the Patmos processor.
Paulo Antonio Ferreira Da Silva Botelho, s146338
Christian Monrad Larsen, s097077
Géza Husi, s146146
NoC programmed in Chisel.
- Wormhole flow control
- XY routing
- Virtual Channels
- 4x4 mesh topology