Luke Gorrie

@corsix No, I'm not confident of that. This is the loop from the hottest trace in the program but the impact could be elsewhere. The perf data we have is...

Here are direct links to the two files in that Dropbox folder in case of any permissions mixup: - https://www.dropbox.com/s/sh0ntir7ny9drni/gc32-vs-gc64.tar.bz2?dl=0 - https://www.dropbox.com/s/ucxni7ove5pi5qz/perf-maps.tar?dl=0

Good question. This is complicated because the application is running in a live environment (monitoring traffic at an ISP) and the workload can be different depending on the time of...

@eugeneia has also found a similar performance difference for some (not all) workloads in a controlled CI test environment. In that case the same test is run multiple times and...

I am amazed at how *complicated* this whole discussion is. I am really grateful that people are willing to spend this time teasing out the relationships between the projects. I...

Thank you everybody for the $0.02. This is really helpful! @agladysh wrote: > The [semi?]official position of the Lua Team, as I understand it, is that LuaJIT is not an...

@agladysh Seems like it would be very hard for RaptorJIT to commit to being a strict Lua implementation. Just now there are fundamental differences: the way to write efficient code...

> What is exactly the "C API" you want to drop ? To be honest this issue is a bit rhetorical. The C-API is actually used internally in the RaptorJIT...

(Or... I am not quite sure actually how much of the C-API is used internally in the virtual machine itself. The basic calling mechanism is used but I'd have to...

> wouldn't most serious Lua implementations have this problem [of not knowing what incompatibility the future holds]? Maybe this is a feature not a bug? Would the PUC Team prefer...