Browse thread
Ocamlopt code generator question
[
Home
]
[ Index:
by date
|
by threads
]
[ Message by date: previous | next ] [ Message in thread: previous | next ] [ Thread: previous | next ]
[ Message by date: previous | next ] [ Message in thread: previous | next ] [ Thread: previous | next ]
Date: | 2009-05-05 (16:26) |
From: | Dmitry Bely <dmitry.bely@g...> |
Subject: | Re: [Caml-list] Re: Ocamlopt code generator question |
On Tue, May 5, 2009 at 8:06 PM, Jean-Marc Eber <jeanmarc.eber@lexifi.com> wrote: > Hi Dimitry, > > Firstly thanks for looking again at the sse2 stuff! > > A difference may occur, if I'm not wrong, from the intermediate results > precision: > > in sse2, eveything is done on 8 bytes (if you "do" doubles at least), while, > in x87, intermdiate results (kept on the stack) are 10 bytes precision. > > This may result in differences (and the bytecode runtime never uses x87 > intermediate storage, so it behaves _numerically_ like the sse2 code > generator, I guess). I wouldn't be so sure. Bytecode runtime is C compiler-dependent (that does use x87 for floating-point calculations), so rounding errors can lead to different results. Floating point is always approximate... - Dmitry Bely