On 3/28/2011 1:12 PM, bearophile wrote:
Walter:
By fundamental technical issue, I mean things like Python's numeric types
which require runtime testing for every operation, and are very resistant
to known techniques of optimization.
Life is a bit more complex than that: - The Lua JIT has shown once and for
all that dynamic typing is not as bad (performance-wise) as you think.
There's a lot of money and manpower behind Python. If this were true, why hasn't
this technology been done for Python?
Secondly, even Lua's proponents (like ilikecakes) says he uses Lua in hybrid
configurations with C and C++, so there is clearly some sort of deficit with Lua.
I have
floating-point heavy benchmarks that run faster in jitted Lua than in D
compiled with DMD.
Come on. As has been discussed to death here, dmd's back end does not do
floating point well, in particular, it does not make use of XMM instructions.
Any code generator that does will perform better.
And, Python does not have the dynamic typing issue with floats that it does with
integers. Jitted floating point Python code *should* match performance with
statically compiled floating point code.