Vis enkel innførsel

dc.contributor.advisorTybell, Thomas
dc.contributor.advisorKnauserud, Øystein
dc.contributor.authorSolomnishvili, Giorgi
dc.date.accessioned2023-10-04T17:20:48Z
dc.date.available2023-10-04T17:20:48Z
dc.date.issued2023
dc.identifierno.ntnu:inspera:143674355:121282616
dc.identifier.urihttps://hdl.handle.net/11250/3094251
dc.description.abstract
dc.description.abstractThe master thesis is carried out in collaboration with Silicon Labs. The main goal of the thesis is to increase the speed of the CV32E40X processor developed by openHW group, without increasing the area usage and power consumption by more than 20%. The metric used for speed evaluation is Instruction Per Clock Cycle (IPC); for the area usage - net area, cell area, and total area; for the power consumption - Watts (W). The CV32E40X is an open-source RISC-V pipelined processor with four pipeline stages. The core can execute compressed (16-bit long) and uncompressed (32- bit long) instructions. The baseline IPC value is 0.46; the baseline net area - 5874.265, the baseline cell area - 16704.66, the total area - 22578.927, the baseline total power - 270 MicroW. The thesis addresses how implementing a 2-bit branch prediction unit can improve the core speed and cost of the improvement in terms of power consumption and area usage. The results revealed that the 2-bit branch predictor with a cache that has 8 cachelines increases IPC by 8.7%, at the cost of 8.9% net area, 14.7% cell area, 13.1% of the total area, and 10.4% total power increase.
dc.languageeng
dc.publisherNTNU
dc.titleBranch Prediction - a Low Power Solution to Performance Optimization
dc.typeMaster thesis


Tilhørende fil(er)

Thumbnail
Thumbnail

Denne innførselen finnes i følgende samling(er)

Vis enkel innførsel