From: davidsen@ariel.crd.GE.COM (william E Davidsen) Subject: Re: find 3.7 and fileutils 3.3 uploaded Date: 29 Jul 1992 19:35:43 GMT
In article <dirkst.712335866@messua>, dirkst@messua.informatik.rwth-aachen.de (Dirk Steinberg) writes:
| I think it is a bad practice to upload static binaries. They are totally
| useless for anyone who does not want to waste massive amounts of disk space.
| I have both Linux 0.96cpl2 and 386BSD 0.1 installed on my 486 and I prefer
| to use LINUX because it is more efficient and faster and it has shared
| libraries (and because gdb-4.6 and X386-1.2E work). The shared libraries
| are the single most biggest plus in favor of Linux, especially when it comes
| to X.
|
| We should always use the latest official gcc release shared images for
| uploading binaries. Starting from next release, these will be based on
| jump-tables anyway. If you really want, you can include /lib/libc.2.2.2
But uploading binaries based on shared libraries makes them break as
soon as you upgrade your gcc. Your thought about .a files was better,
because gcc is relatively small, and you can trim the compiler and still
use the linker if you're all that tight for disk.
The statement about "we should always use..." doesn't make sense
unless people are uploading stuff as a paying job. I would encourage
people to upload .a files if possible, or static linked files as an
alternative. I have a nice collection of files I've downloaded and which
don't work because I have gcc 2.2.2. Let's not add to the problem.
BTW: how much os a hit does the library take on using the jump table?
Between the extra level of indirect, dumping the cache and pipeline
twice as often, and paging on small systems, this looks like a potential
5% increase in clock time. Do you have numbers? I see people taking the
compile time to use -O2 and a bunch of hand selected -f options to make
things faster, so performance hits of a percent or more are probably
going to be noticed.
Note, I'm looking for clarification rather than saying "this is bad."
My guess of what it adds to small systems calls could be way off, etc.
I'm well aware of the benefits of jump tables, just concerned about
possible performance costs. Maybe the only hit is the extra pieline
dump, I suspect the table would be memory, if not in cache, all the
time.
--
bill davidsen, GE Corp. R&D Center; Box 8; Schenectady NY 12345
It never ceases to amaze me that otherwise rational people, able to
understand calculus, compound interest, and the income tax form, can
continue to believe that poker is a game of chance.