Pancho  writes:
> On 03/09/2020 09:47, Richard Kettlewell wrote:
>> Mono’s CIL JIT was empirically beating GCC on computationally intensive
>> code a decade ago (probably by making better register allocation choices
>> although I didn’t delve into it deeply). No idea what the situation is
>> today but I’d expect them to be pretty similar.
>
> Yeah, maybe optimal reordering of instructions and branch prediction
> too, in combination with register allocation. C# was brilliant at
> numerical analysis type performance right from the get-go.
>
> I have a vague memory of .net runtime statistics being stored and used
> to optimise subsequent runs of a program, but after a quick google I'm
> not sure if I just made that up.

I think that would be Managed MPGO in the language used by
https://docs.microsoft.com/en-us/archive/msdn-magazine/2016/march/compilers-man
aged-profile-guided-optimization-using-background-jit

--
https://www.greenend.org.uk/rjk/

--- SoupGate-Win32 v1.05