Not solved by any means, but there is typically only one GC-related paper per top-tier PL conference such as PLDI or ICFP these days. Many more GC results, particularly for more actively developed systems where the changes tend to be incremental, are published at either ISMM (International Symposium on Memory Management) or MSPC (Memory Systems Performance and Correctness).
thats moreso a reflection that aside from PLDI, none of those conferences really touch GC. its also a tricky domain to analyze wrt theory, and requires huge engineering effort to evaluate in practice. But theres also lots of real opportunities for innovation in the theory & practice of GCs.
"In particular, when garbage collection has five times
as much memory as required, its runtime performance
matches or slightly exceeds that of explicit memory
management."
Yep. Like most problems in life, it's 'solved' if you have the money.
That's talking about throughput. Getting good pause times and good throughput is still hard. Even if you have 100x as much memory, you still get the same pause times (albeit less frequently).
I think that for a conventional GC it's mostly dependent on how much time you have and what properties you want, that's mostly an engineering problem. Remember that academia worked on GCs way before mainstream languages. From what I understand the "experimental" languages are playing with region-based allocation for now.
