Memory and Compilers
summary
Three concrete systems and their numbers: iNVMFS is a lightweight NVRAM file system that checkpoints both program and storage state, uses byte‑level/append‑friendly writes and metadata logging, and reports ~74% execution‑time and ~73% energy reductions versus copy‑on‑write and LFS on sensor workloads. Mercury reorders Simulink actor translation with pipeline‑sensitive profiling, topological candidate selection and penalty updates, cutting runtime 10–33% on x86 and 9–59% on ARM versus baseline generators and improving pipeline stalls. An I/O virtualization stack uses weight‑free ring buffers and programmable QoS regulators to bound DMA memory contention and restore network latency/bandwidth; LLVM changes employ PACGA MACs on register spills, adding ~9% code size and ~17% runtime for integrity (104% with encryption) and blocking many exploit attempts (23/67 Juliet tests prevented).
AI-written summary · about
chapters (16)
- 0:00Start
- 5:03Architecture of Imv and Fs
- 5:35How Is the Writing Policy Designed
- 7:20File Creation
- 9:05Checkpoint Commit and Restore Commits
- 13:26Limitations of Flash Memory
- 22:36Software Device Virtualization
- 38:51Pipeline Friendly Code Synthesis
- 44:23Instruction Scheduling
- 45:54Toward Register Spilling Security Using Llvm and Arm Pointer Authentication
- 47:04Memory Security
- 47:49Register Spilling
- 52:43Matrix Generation
- 53:57Analysis
- 58:44Dynamic Analysis
- 1:01:22Poster Sessions
related talks
01h 11mProgramming language memory models: Problems, solutions, and directions
21mWeak Memory Demands Model-based Compiler Testing
17mRecent Developments in Optimizing Compilers for RISC-V
17mCrossMapping: Harmonizing Memory Consistency in Cross-ISA Binary Translation
32mBit-Stealing Made Legal: Compilation for Custom Memory Representations of Algebr...
23mVerified Compilation of C Programs with a Nominal Memory Model
17mUnlocking 15% More Performance: A Case Study in LLVM Optimization for RISC-V
15mAutomatically Reasoning About How Systems Code Uses the CPU Cache