On Nov 26, 2007, at 5:44 PM, Jon M. Hanson wrote:
> Chris Gehlker wrote:
>> In the Intel world, the data path isn't wider between i386 and
>> ix86-64. There are more registers, but compilers haven't really been
>> optimized to use them yet. But even when they are, one cache miss
>> will
>> wipe out the gains from running hundreds of instructions with more
>> registers.
>>
> This isn't true at all. If you're running in x86-64 and something has
> been compiled as x86-64 you will absolutely use the additional
> registers. Compilers (at least GCC and Intel's compiler, I can't speak
> for Microsoft's) have supported x86-64 for quite a while now.
By "but compilers haven't really been optimized to use them yet" I
didn't mean to imply that the compilers would ignore the registers. I
meant that many of the specific optimization strategies that
compilers run are the same when generating 64-bit code. The benchmarks
bear me out.
--
No matter how far you have gone on the wrong road, turn back.
-Turkish proverb
---------------------------------------------------
PLUG-discuss mailing list -
PLUG-discuss@lists.plug.phoenix.az.us
To subscribe, unsubscribe, or to change your mail settings:
http://lists.PLUG.phoenix.az.us/mailman/listinfo/plug-discuss
(text/plain)