Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Processor speed

Author: Eugene Nalimov

Date: 10:21:38 03/10/00

Go up one level in this thread


On March 10, 2000 at 02:58:24, David Blackman wrote:

>On March 09, 2000 at 23:14:52, Robert Hyatt wrote:
>
>>One thing is for sure, the Motorola "architecture" is so far ahead of Intel,
>>with _real_ registers, etc...  it was a shame IBM went the wrong way when they
>>decided to use the x86.  I would _much_ rather be programming 680x0 processors,
>>and had their speed driven by the market pressure that has driven Intel.
>>
>>Programming the 680x0's feels just like programming any well-done architecture
>>of the early 80's...  lots of instructions, lots of registers, sane instruction
>>formats, sane memory addressing modes, etc.  None of that early segment horse-
>>hockey.  :)
>
>The 680x0 was difficult to implement fast. Especially the 68020 compatible ones.
>68020 is one of the reasons people got excited about RISC, whereas the 386
>architecture can be made to go fast more easily (but still not as easy as RISC).
>
>Of course the 386 is more ugly than a 68020 from a user mode programming point
>of view, and the 16 bit Intels were much, much worse.

Here I disagree. I wrote compilers for both of them (actually, it was the same
portable optimizing backend), and 68k was much more pain because of 2 register
classes (addresses and data). Registers on 386 are more or less symmetric (there
are 3 exceptions - shift on non-constant amount, divide and 32x32->64 multiply,
memory movement instructions; fortunately, those instructions are either slow or
not frequent, so extra pushs/moves/pops around them do not slow program a bit).

Eugene



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.