Efficient Memory Tracing for Mobile Architectures

Analyzed compression and memory tracing techniques for mobile architectures.

Duration: July 2020 to June 2021

Mentor: Prof. Biswabandan Panda, CSE Dept. IIT Kanpur

Details

  • Funded by Qualcomm Research to improve the front-end and memory subsystem of Systems-on-Chip.
  • Developed a framework to collect Memory & Data Traces (MDT) through emulation or natively for Android.
  • Modified QEMU, the emulator used by Android Studio, to collect MDT from Android 9.0 API with x86_64 ABI.
  • Extended Valgrind, a memory profiling framework, collected MDT natively from ARMv8-based devices.
  • Extended ChampSim, a trace-driven simulator, utilized MDT and analyzed patterns to improve branch prediction, prefetching, and cache compression at LLC.
Avatar
Anish Saxena
Computer Science PhD Student

Computer Architecture, Systems, and Security