Re: >>/<< 32 Bit wrong on 32 architechture
From: Benjamin Riefenstahl (Benjamin.Riefenstahl_at_epost.de)
Date: 12/31/03
- Next message: Bryan Oakley: "Re: Extracting patterned filenames from [glob] without a loop - possible?"
- Previous message: Joe English: "Re: ANTIALIASING ???"
- In reply to: Gerhard Hintermayer: "Re: >>/<< 32 Bit wrong on 32 architechture"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Wed, 31 Dec 2003 19:03:46 +0100
Hi Gerhard,
Gerhard Hintermayer <g.hintermayer@inode.at> writes:
> Well, I was wondering, why "undefined" is obviously implemented as
> "sometimes correct" and "sometimes not".
My understanding is that math expressions in C were supposed to be
implemented in just one assembly instruction each, give or take some
register juggling (C is sometimes defined as "a portable assembler").
So error checks or preconditions were not wanted, instead some
variation was allowed. With those variations, the outcome even often
depends on compiler optimization flags.
> Do you think, that this "undefined" is useful ?
In C, primarily it just is as it is. Even if you were to lobby for a
change in the C language now, portable code would have to wait for the
spec to change and than maybe some 10 years more, until compilers on
all reasonably platforms have the change.
> we would have an implemetation of expr in tcl, which could handle
> this
Tcl doesn't have those constraints, so it can be more precise.
> instead of saying: but look, C can't handle that either.
IIRC, Tcl just took the whole math expression language from C and
said, "we just use what you already know from C with the same
semantics." That made definition and implementation very simple.
benny
- Next message: Bryan Oakley: "Re: Extracting patterned filenames from [glob] without a loop - possible?"
- Previous message: Joe English: "Re: ANTIALIASING ???"
- In reply to: Gerhard Hintermayer: "Re: >>/<< 32 Bit wrong on 32 architechture"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|
