Brain Image Data Mining
Design your datapath for a pipelined machine. You decide how many stages.
Provide a complete drawing of your datapath indicating all stages, functional
units, multiplexors, registers, etc( you may indicate information passed between
pipeline stages as a “black box” as done in the notes). Indicate all features
that you include to improve the performance of your machine. Is your machine
hardwired or microprogrammed?
Project had to deal with:
Control hazards: Assume a branch instruction is encountered every 8 instructions
on average and that 90% of the branches are taken. Deal with this as
efficiently as possible with hardware and/or nops…no delayed branching.
Data Hazards: Assume that 1 instruction in four involves a data hazard with the
instruction immediately preceding it and that 20% of these hazards
involve a load word instruction where the previous instruction was
a load word and the next instruction requires the value that was to
Structural Hazard: Include in your evaluation iff your machine design has a
structural hazard. Explain why it does or does not have such a hazard.
Assumption: 40 % of all instructions are load/store.
R. Benton, S. Choubey, D. Clark, T. Johnsten, V.V. Raghavan. “Diagnosis and Grading of Alzheimer’s Disease Via Automatic Classification of FDG-PET Scans”, Journal of Neurology, (Under Review)