[LLVMdev] Improving Garbage Collection

[Nate Fries]

On 7/6/2011 6:24 PM, Talin wrote:
> The LLVM code generators and analysis passes have a much more thorough 
> knowledge of SSA value lifetimes than frontends do, and therefore 
> could avoid spilling and reloading of values when it wasn't needed.
Although this would indeed be nice, it is not done by similar platforms 
in practice. I have investigated [very] briefly into whether the CLR or 
JVM implement garbage collection in their IR, and it does not seem that 
they do (meaning, the CLR/JVM implementation itself is responsible for 
garbage collection, not the code generated by the CLR/Java language 
compilers).

However, I have left remaining the only strong point in taking an IR 
approach to GC that you pointed out. I have to say that it makes a 
convincing argument, since the generators and analysis passes would know 
even more about the lifetime of variables in code than any plug-in GC 
library would. I'd be curious as to what benefits could be gained from 
this approach over the current standard way of doing it; but not 
terribly hopeful.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20110707/b8612dd2/attachment.html>