English version
Accueil     À propos     Téléchargement     Ressources     Contactez-nous    

Ce site est rarement mis à jour. Pour les informations les plus récentes, rendez-vous sur le nouveau site OCaml à l'adresse ocaml.org.

Browse thread
[ Home ] [ Index: by date | by threads ]
[ Search: ]

[ Message by date: previous | next ] [ Message in thread: previous | next ] [ Thread: previous | next ]
Date: 2009-03-05 (19:34)
From: Jon Harrop <jon@f...>
Subject: Re: [Caml-list] stl?
On Thursday 05 March 2009 06:22:28 yoann padioleau wrote:
> Come on, can you stop all those stuff about LLVM. The guy works in a game
> company with people knowing C/C++ for decades, with quite a lot of legacy
> code I guess, and you arrive with your "hey you should use LLVM" that
> almost nobody knows about.

Then they should learn about it. LLVM is already capable of generating SSE 
instructions from higher-level code more effectively than GCC and, 
consequently, several of my benchmarks run 2-4x faster than GCC-compiled C.

> Oh, and by the way, in which programming language is written LLVM ? :)

Sure. That doesn't mean we shouldn't build upon LLVM.

> > It doesn't need to be a JIT and, actually, HLVM already supports both JIT
> > and standalone compilation.
> So what you propose to his company is to switch from C++ to HLVM ? :)
> Be serious.

I suggest they consider using LLVM, most likely from their C++ to begin with. 
This is a low barrier to entry: they can just compile their C++ using 
llvm-gcc and write custom passes that perform the optimizations they want. 
That is ideal for applications that are willing to sacrifice numerical 
robustness for performance, for example.

Dr Jon Harrop, Flying Frog Consultancy Ltd.