Related articles |
---|
Use of unaligned load/stores by compilers sazal@aol.com (1998-01-24) |
Re: Use of unaligned load/stores by compilers - or doing things by hal dlmoore@ix.netcom.com (David L Moore) (1998-01-25) |
Re: Use of unaligned load/stores by compilers - or doing things by hal tmoog@mcs.net (Tom Moog) (1998-01-25) |
Re: Use of unaligned load/stores by compilers - or doing things by hal henry@zoo.toronto.edu (Henry Spencer) (1998-01-26) |
Re: Use of unaligned load/stores by compilers reid@micro.ti.com (Reid Tatge) (1998-01-30) |
Re: Use of unaligned load/stores by compilers tgl@netcom.com (Tom Lane) (1998-02-01) |
Re: Use of unaligned load/stores by compilers albaugh@agames.com (1998-02-01) |
Re: Use of unaligned load/stores by compilers hrubin@stat.purdue.edu (1998-02-01) |
Re: Use of unaligned load/stores by compilers dlmoore@ix.netcom.com (David L Moore) (1998-02-01) |
From: | Henry Spencer <henry@zoo.toronto.edu> |
Newsgroups: | comp.compilers |
Date: | 26 Jan 1998 00:53:57 -0500 |
Organization: | SP Systems, Toronto |
References: | 98-01-099 98-01-105 |
Keywords: | architecture, debug |
David L Moore <dlmoore@ix.netcom.com> wrote:
>Some processors have implemented unaligned loads as extra-code traps.
>(An extra-code is an instruction that is emulated in software)...
An alternate approach, which MIPS took (at least, at one time) was to have
a compiler option that generated alignment-insensitive code. This let the
people who needed it -- folks porting big ill-behaved applications -- have
it, with relatively low overhead and no hardware cost.
John Mashey said, roughly, "the applications developers are much happier
if they can compile their code and have it work right away, even if it's a
bit slow, and then track down the alignment problems at their leisure".
As I recall, he said the performance penalty was not huge, maybe 10-15%
typically.
--
| Henry Spencer
| henry@zoo.toronto.edu
--
Return to the
comp.compilers page.
Search the
comp.compilers archives again.