Soner Onder
Michigan Technological University, MI·Computer Science
How fast a computer processor runs most programs still depends heavily on branch instructions, the points where a program chooses between paths, such as an if-statement or a loop test. When the processor cannot predict which way a branch will go, it wastes time, which limits how many instructions it can finish per clock cycle.
- Fields
Computer Science
Engineering
- Topics
computer architecture
microarchitecture
compilers
- Grants
- 1 known grant, through Sep 2027
What they work on
- Develop Vectorized Instruction Space (VIS), which converts control dependences into data dependences using an executable dynamic single-assignment form at the machine-instruction level, with lower overhead than traditional predication.
- Design microarchitecture support for VIS that combines controlled multi-path execution, running several program paths at once, with speculation, to remove hard-to-predict branch instructions.
- Build compiler techniques that use VIS to unroll loops with unknown iteration counts, targeting both superscalar and very long instruction word (VLIW) processors.
- Release the VIS toolsets and design infrastructure as open source for research and teaching.
More about the research
This project argues that small improvements to existing techniques are not enough and proposes a new way to design processors and compilers, with the goal of speeding up software on everything from phones to large data centers.
Why it matters
Programs on mobile devices and in data centers are often slowed by branch instructions the processor cannot predict. If VIS removes many of these branches and handles loops of unknown length efficiently, it could make computer systems faster at every scale. The open-source tools could also help other researchers and students who study processor and compiler design.
Known grants
From NIH and NSF research grants. Labs often have other funding we can't see.
Closest labs
Nathaniel Bleier
University of Michigan, MI·Computer Science
Combine learning-guided program synthesis with automated logical reasoning to generate instruction sequences, accelerator control codes and microcoded expansions that are formally proven equivalent to the original program.
Yuhao Zhu
University of Rochester, NY·Computer Science·Early-career
Develop the foundations for in-sensor machine vision that combines computational metasurfaces (flat nanostructures that process light) with analog encoders inside CMOS image sensors, so images are partly processed before they are digitized.
Peng Jiang
University of Iowa, IA·Computer Science·Early-career
Develop a runtime system that gives a virtualized view of data partitioned across heterogeneous devices, plus a compiler tool that places data using a statistical performance model of random memory accesses.
The school
Michigan Technological University: 9 labs on PI Pages, $107M a year on research, 86 research doctorates a year.
Reach out
A start for your email. Fill in the brackets.