Re: Optimization flags
Giuseppe Ghibò <[email protected]> Wed, 30 Mar 2005 23:44:31 +0200
| Newsgroups | gmane.linux.mandrake.cooker.amd64 |
|---|---|
| Message-ID | <[email protected]> |
Gwenole Beauchesne ha scritto:
> If you are refering to IA-64, IA-32 EL does not support the x86-64 ISA
> yet. Besides, IA-32 EL (the ia32 DR software emulator) is far from slow.
> Only the built-in hardware x86 decoder was slow. It's now replaced in
> favor of IA-32 EL (module for Linux and Windows) in new Itanium chips.
so, how it performs (e.g. on a 1.5Ghz Itanium)?
>
>>> Caring for that would be like caring about about some X86 flags which
>>> might affect
>>> some X86 emulator under PPC architecture.
>
>
> { m68k, x86 } -> ppc can be implemented effeciently enough that playing
> with those flags won't affect performance beyond the effect seen
> natively too.
>
>> When I ran the exact same line when I was using Gentoo, I got shot
>> down in
>> flames by several users saying that using the sse unit for math is a bad
>> idea. They could be wrong, but this what I'm basing my knowledge on.
>
>
> On x86-64? They surely didn't know about the ABI, thus were wrong. Note
> that old i387 is only used for "long double" operations by default.
>
long double are at 128bit, but wasn't the i387 precision limited
to 80bit?
Bye.
Giuseppe