21 June SS VLIW Pentium Bug (pp 30-35)Scope Instruction level parallelism
Current Microprocessors (including IA64)
28 June Evolution of architecture
C. Kozyrakis and D. Patterson, "A new direction in computre architecture
research", IEEE Computer, November 1998, pp.24-32. (pdf)
1B transistors processor
29 June Project reports (7 July adds more reports) Off-line until 9th August 1999
Approach The lectures will give direction for students to probe further. Students have to read assigned papers to broaden and deepen their knowledge. One project will be issued. Students will do the architectural exploration how the instruction level parallelism work.
Syllabus 3 weeks lecture followed by presentation of individual projects. Topics of lecture : Instruction level parallelism, Performance metrics, Pipelining, Classical method : Scoreboard, Tomasulo, Modern processors.
Lecture
Report Off line until 9th August
ToolsS1 with branch prediction : heuristic tuning of branch prediction scheme, Wiwat Vatanawood. 2 Stages pipeline of S1, Pichayothai Mahatthanapiwat Reduction of CPI in a VLIW processor, Worasait Suwannik Dynamic Scheduling with Tomasulo technique, Radchaporn Kienprasit Two-level branch prediction, Dechanuchit Katunyataweetip Minimize data hazard stalls by forwarding technique, Nuanwan Soonthornpaisaj Improvement the S1p with branch prediction and 2-bit history branch target buffer, Nuttakorn Tubthong Scoreboard technique with 2 ALUs, Sukree Sinthupinyo
VLIW (the first two can be obtained from me)