Re: 64 bit data types

From: "John Herbster" <herb-sci1_AT_sbcglobal.net>
Date: Tue, 31 Jan 2006 07:18:38 -0600

"Henrick Hellström [StreamSec]" <henrick@xxxxxxxxxxxx> wrote

> > According definition Integer and Cardinal [are] generic
> > datatypes which should become 64-bit.

> I thought so too, but according to Danny Thorpe, Integer
> has only been guaranteed to be the most effective (not
> largest) signed integer type supported by the CPU.
> On x64 that is still a 32 bit integer, because 32-bit
> operations are generally faster that 64 bit operations
> even on these CPUs.

So maybe we need both "Integer" and "FastInt". --JohnH

.

References:
- 64 bit data types
  - From: Florent Ouchet
- Re: 64 bit data types
  - From: Ender
- Re: 64 bit data types
  - From: "Henrick Hellström [StreamSec]"

Prev by Date: Re: 64 bit data types
Next by Date: Re: Runtime design forms support
Previous by thread: Re: 64 bit data types
Next by thread: Re: 64 bit data types
Index(es):
- Date
- Thread

Relevant Pages

Re: Borland Delphi + 64Bits
... Danny Thorpe wrote: ... It contradicts what happened between D1 and D2, ... that takes full advantage of the CPU performance characteristics. ... registers each containing a 32 bit number, than to multiply two 32 bit ...
(borland.public.delphi.non-technical)
Re: Borland Delphi + 64Bits
... Danny Thorpe wrote: ... My point was just that in some cases there are valid performance reasons to have Integer and Cardinal bumped up to 64 bit. ... It won't be possible to write single source big num code that both compiles on both x86-32 and x86-64 and takes maximum advantage of the CPU, without having an unsigned long type that is 32 bit on the former and 64 bit on the latter, and an unsigned long long type that is 64 bit on the former and 128 bit on the latter. ...
(borland.public.delphi.non-technical)