Predictions for 2021
- There will be processor in memory, and it will be x86 (Click)
- Distributed GC article in Wikipedia will be longer (Blackburn)
- Memory management will be so important that ISMM will dwarf PLDI (Bacon)
- Scripting languages will be a major GC workload (Blackburn)
- 99% of code, by source lines, will be GC’d; 99% of code, by instructions executed, will use malloc (Bacon)
- Secure/reliable systems require arch/PL/MM/OS/security/etc communities to build a cross-disciplinary solution (McKinley)
- Custom HW will be important to memory management (Blackburn)
- There will be hardware GC support in FPGAs and ASICs -
it’s inevitable! (McKinley)
- There won’t be hardware GC support - don’t you ever learn?! (Boehm)
- No one will be developing applications with transactional memory (Bacon)
- Reference counting is back! (Hertz)
- Shared memory will give way to distinct local memories (Richards)
- Memory hierarchy will be "richer" (ie, much more complex), with SSD etc (Click)
- Data placement will have to be exposed at language level for performance, due to increasing complexity (Click)
- Any programming model that exposes the increased complexity of the memory system will fail (O’Neill)
- Jones et al will still be the definitive GC text (Jones)
- Most memory will be managed by dlmalloc (Click)
- Much memory will be managed by HotSpot, using dlmalloc (Click)
- Heterogeneous CMP will be a major target (Blackburn)
- Embedded/mobile device GC will become dominant (Blackburn)
- There will be an automatically verified concurrent, incremental, real-time garbage collector in production (Bacon)