From: | charles.marslett@tempe.vlsi.com (Charles Marslett) |
Newsgroups: | comp.compilers,comp.lang.asm.x86 |
Date: | 8 Jul 1997 00:35:03 -0400 |
Organization: | VLSI Technology, Inc. |
References: | 97-06-071 97-06-081 97-06-101 97-06-123 97-07-008 |
Keywords: | optimize, practice |
On 4 Jul 1997 14:35:25 -0400, als@tusc.com.au (Anthony Shipman) wrote:
>I think that part of the gap in the discussion is that some people are
>talking about asm vs C compiles whereas others are talking about asm
>vs Fortran compilers. It is easier to optimise Fortran than C and
>because Fortran is used more for number crunching much greater
>attention is paid to optimising it.
>
>I expect that it is unlikely that a human can write better asm than a
>good Fortran compiler. However given the quality of C compilers,
>especially, from what I've heard about VC++, I expect that a human
>could often write better asm, in small bursts.
Actually, that was part of what I was trying to say -- even in Fortran
optimizing a small routine better than the compiler is far easier for
a human being than optimizing a large routine (even one that is only a
few hundred lines long).
And I agree with you about PC C compilers. The quality of Microsoft
compilers' code has clearly declined over the past 4-5 years, and I
suspect it has been a result of pressure to add features to the tools
at the expense of maintaining (or improving) the code generator
quality. And lest I be criticized for Microsoft bashing, I have to
add that Borland and Watcom have lost even more ground.
--Charles
--
Return to the
comp.compilers page.
Search the
comp.compilers archives again.