Re: Terra : "Lua" can be as fast as C
- From: Alek Paunov <alex@xxxxxxxxxxx>
- To: luajit@xxxxxxxxxxxxx
- Date: Sat, 25 May 2013 22:52:57 +0300
On 25.05.2013 21:40, John Abrahamsen wrote:
In my opinion that falls outside the scope and mission of Lua. If my target
were to compile to mcode...
But their targets are significantly different/specific - GPU offloading
and alike. Please read at least their paper abstract:
http://terralang.org/publications.html
Their choice of Lua as clean base language for the required LLVM
frontend is probably the path, which most of the people here would get
in that direction (macro enhanced runtime specialization). The employing
of the LuaJIT FFI goodness as driving horse for the hard work of the
mixed execution is perfectly natural choice too. All good so far.
What's driving me crazy is the Acknowledgments paragraph in the above paper:
"This work has been supported by the DOE Office of Science ASCR in the
ExMatEx and ExaCT Exascale Co-Design Centers, program manager Karen Pao;
DARPA Contract No. HR0011-11-C-0007; and the Stanford Pervasive
Parallelism Lab (supported by Oracle, AMD, Intel, and NVIDIA)."
I am sure that the budget of this relatively thin layer on top of
LuaJIT, Lua and LLVM exceeds at least several times the year budget of
LuaJIT itself.
And what - no proper credits, not a single message from the authors in
both Lua lists, not even properly spelled link to the LuaJIT site. Sad
world :-(.
As the host of this list already said, for all further curiosity, their
list is here:
https://mailman.stanford.edu/pipermail/terralang/2013-May/thread.html
Other related posts:
