Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: GCC beats MSVC by 50% on crafty according to german magazine!

Author: Alessandro Damiani

Date: 08:06:59 03/14/02

Go up one level in this thread


On March 14, 2002 at 07:34:59, James T. Walker wrote:

>On March 14, 2002 at 05:56:02, Alessandro Damiani wrote:
>
>>On March 13, 2002 at 22:49:35, Robert Hyatt wrote:
>>
>>>On March 13, 2002 at 21:50:56, James T. Walker wrote:
>>>
>>>>On March 13, 2002 at 14:12:29, Robert Hyatt wrote:
>>>>
>>>>>On March 13, 2002 at 00:29:21, martin fierz wrote:
>>>>>
>>>>>>aloha!
>>>>>>
>>>>>>here's something i found on a german computer magazine website:
>>>>>>(http://www.heise.de/ct/english/02/05/182/)
>>>>>>
>>>>>>
>>>>>>"Under Windows we made use of Visual Studio 6 (with Service Pack 5), with which
>>>>>>in all probability most Windows applications have been created. The SPEC results
>>>>>>obtained with the new compilers such as the current GCC 3.0 or Intel's in-house
>>>>>>compiler are better by between ten and more than twenty percent.
>>>>>>[snip]
>>>>>>With a SPECint_base value of 306 Apple's 1 GHz machine under Mac OS X ran almost
>>>>>>head to head with the equally clocked Pentium III, combined with Linux and GCC,
>>>>>>with a SPECint_base value of 309. Under Windows, the bad quality of Microsoft's
>>>>>>run-of-the-mill compiler, which pushed the system down to a SPECint_base value
>>>>>>of 236, below the 242 value of the PowerMac running at a clock cycle of 800 MHz,
>>>>>>came back to haunt the Intel processor."
>>>>>>
>>>>>>and then there is the link http://www.heise.de/ct/english/02/05/182/qpic01.jpg
>>>>>>which shows the specint crafty result which is a whopping 444 for GCC and
>>>>>>only 293 for MSVC.
>>>>>>
>>>>>>is this really possible?? i remember i once tried GCC for my checkers program,
>>>>>>and of course it's long ago, but it was clearly worse than MSVC at the time. i
>>>>>>just can't remember anybody posting anything like this here, GCC being 50%
>>>>>>faster than MSVC... but usually, this magazine is good...
>>>>>>
>>>>>>cheers
>>>>>>  martin
>>>>>>
>>>>>>PS: just another question: is linux 32-bit or 64-bit? can i use more than 2-4GB
>>>>>>ram under linux?
>>>>>
>>>>>
>>>>>1.  I've never seen GCC within 10% of the speed of MSVC.  I doubt it has
>>>>>suddenly happened.
>>>>>
>>>>>2.  Linux is _both_.  On intel (non-IA64 machines) it is a 32 bit operating
>>>>>system.  On 64 bit processors like that Alpha or IA64 it is a 64 bit operating
>>>>>system.  The RAM limit is not an OS issue, it is an architectural issue.  Except
>>>>>for a bizarre hack Intel added a couple of years back, the 32 bit machines are
>>>>>limited to 4 gigs (2^32).  With a kludge they added, this goes to 32 gigs I
>>>>>believe, but only for (at the time) the Xeons...
>>>>
>>>>Hello Bob,
>>>>Can you confirm that the P3/P4 and AMD Athlons have only 32 address lines?  My
>>>>understanding of microprocessors is that the memory limit is due to the number
>>>>of address lines and not the number of data bit lines.  Meaning that a 32 bit
>>>>processor can pass 32 bits in parallel but the amount of memory that it can
>>>>address is 2^X  where "X" is dependent on the number of address lines the cpu
>>>>has.  I remember the Motorola 68000 had something like 21/23 address lines even
>>>>though it was a 16 bit processor(Don't tie me to that exactly).  I have tried to
>>>>look up the Pentium and have been unable to get a pin-out of it.
>>>>Jim
>>>
>>>I'm not sure either.  I only know that the xeons have 4 extra bits that are
>>>usable, although it might be that _all_ the processors have 36 bits but they
>>>don't function on anything but xeons...  no idea...  I'll try to find out...
>>
>>The 68000 was 32bit (defending it :). And yes, the address space was not 2^32,
>>as Jim wrote. IIRC something like 2^24? But I never met the limit at that
>>time...
>>
>>Alessandro
>
>I believe the 68000 was a 16 bit processor (externally) but handled 32 bits
>internally.  This made it a kind of bastard processor since it had a 16 bit data
>bus but operated on 32 bit words inside.  The 68020 was a real 32 bit processor
>(in/out).  I studied/worked with cpu's many years ago and I rember that the data
>bus being either 16 bits or 32 bits did not determine the amount of memory it
>could address.  That is dependent on the number of address lines it had which
>could be any number the designers decided were required.  And of course in the
>binary system the amount of memory it could address was 2^X with "X" being the
>number of address lines available.
>Jim

You are right, but if you say "it was a 16bit processor" that means to me you
also mean internally, which is - as you say now - not true. I looked at it from
a programmers point of view. And then I had the internal 32bit to work with.
Yes, the data bus was 16bit. I did not say something against not being 32bit
externally. I thought my text was not misleading...still thinking it. But don't
worry, Jim, it was not an affront, just a detail.

Alessandro



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.