Vectorized Emulation - Putting It All Together
Explore the cutting-edge technique of vectorized emulation in this conference talk from Recon 2019. Dive into how AVX-512 is leveraged to run multiple virtual machines in parallel per core, enabling efficient determination of input-driven VM divergence. Learn about the advantages of this approach, including near-zero cost tracking of input effects on system state. Discover how this method goes beyond traditional code coverage, incorporating register, memory, and stack state monitoring. Understand the performance benefits of running multiple VMs per core, allowing for fuzzing at speeds surpassing native x86 execution. Delve into the potential for hardware-accelerated taint tracking and feedback, which enables solver-like capabilities in fuzzing. Focus on these advanced aspects rather than basic performance metrics and design principles.