Athlon testing progress

["M. Edward Borasky" <znmeb at aracnet dot com>]

1. I downloaded the gcc 20010514 snapshot tarball. It turns out to be easier
to deal with that than the RPMs, and I'm not quite ready to interact with
CVS (in other words, my Internet connection is still a Windows 2000
machine -- I don't have my Linux box on the net yet)

2. "configure", "make bootstrap", "make check" and "make install" all ran
fine. As a reminder, the environment is a 1.333 GHz Athlon Thunderbird with
a 266 MHz front-side bus, 512 MB of DDR RAM and more or less vanilla Red Hat
Linux 7.1. I captured the logs for the "make" steps if anyone is interested
and timed the runs with "/usr/bin/time". Here are the times for the two big
ones:

	make bootstrap:
	1384.40user 245.71system 28:33.59elapsed 95%CPU (0avgtext+0avgdata
0maxresident)k
	0inputs+0outputs (10118910major+10805370minor)pagefaults 0swaps

	make -k check:
	1521.45user 631.93system 44:41.18elapsed 80%CPU (0avgtext+0avgdata
0maxresident)k
	0inputs+0outputs (29026939major+24244306minor)pagefaults 0swaps

3. I'm going to run the Atlas compile with the pre-release gcc 3.0
overnight; IIRC it takes about eight hours for it to run all its benchmarks.
Before I do that, I have two questions:

	a. "g77 -v" tells me I am still using the assembler that came with RH 7.1:
	   GNU assembler version 2.10.91 (i386-redhat-linux) using BFD version
2.10.91.0.2

	   Do I need to get a more recent assembler to get the latest Athlon
wisdom?

	b. I see a number of Athlon-related flags in the output of
"gcc -dumpspecs":
	   %{march=athlon:-D__athlon -D__athlon__ %{!mcpu*:-D__tune_athlon__ }}%
	   %{mcpu=athlon:-D__tune_athlon__ }%

	   Which one(s) of these should I use?

Right now I'm working with Atlas version 3.2.1, which was their most recent
stable release about a week or so ago. If I don't have any problems with
this one, I'll try this testing again with a later release (and probably the
gcc 20010521 snapshot as well, since it will be a while before I get to that
point).

4. Also, is there any point in running the "make bootstrap" again, using the
*new* gcc as the compiler with Athlon flags? I can do that very easily if
there's any information to be gained from it. The runs I did last night used
the "i686" assumption throughout, given that the stock Red Hat 7.1 is the
n-times cursed 2.96-81.
--
M. Edward (Ed) Borasky, Chief Scientist, Borasky Research
http://www.borasky-research.net  http://www.aracnet.com/~znmeb
mailto:znmeb@borasky-research.com  mailto:znmeb@aracnet.com

If there's nothing to astrology, how come so many famous men were born on
holidays?