Luke Gorrie
Luke Gorrie
This branch explores idea https://github.com/raptorjit/raptorjit/issues/251#issuecomment-482614962: to allow `cnewi` instructions to sink even when they are transferred to another root trace via the last snapshot. - [x] Allow `cnewi` to sink...
The feature that I am most excited about supporting in raptorjit is unboxed FFI pointers and 64-bit integers (#174.) I especially want to always be able to load these typed...
Here is a fun little program that demonstrates the problem of code doing 64-bit arithmetic (e.g. on pointers) being vulnerable to surprising performance problems. In this case adding a single...
We should write down how the virtual machine is bootstrapped and built. The VM bootstraps in a complex way including multiple "little language" code generators. This process is not currently...
How should we share relevant changes with LuaJIT? Is there anybody who wants to take responsibility for this? (I feel that my own hands are full already.) For example #246...
Here is an idea that I think comes from @fsfod: Suppose the JIT would always use a separate snapshot for each frame on the call stack. Then each IR instruction...
RaptorJIT needs to support unboxed FFI pointers. This means, above all else, that assigning pointers to local variables and doing arithmetic on them would never allocate on the heap or...
This issue is inspired by a discussion partly [on the LuaJIT mailing list](https://www.freelists.org/post/luajit/ANN-dumpanalyze-tool-for-working-with-LuaJIT-dumps,13). Our experience with Snabb has been that bytecode interpreter performance is much too low to ever use...
Here is a radical idea - something we should consider and discuss seriously - not necessarily something we will do: How about if we remove the Lua C-API from RaptorJIT?...
Question: does RaptorJIT need a new garbage collector? Why? Why not? LuaJIT has a [design for a new garbage collector](http://wiki.luajit.org/New-Garbage-Collector). @corsix and @fsfod both have partial implementations of the new...