Re: CPU options during config

Chuck Lever (chucklever@netscape.net)
8 Jan 99 15:31:36 EST


alan@lxorguk.ukuu.org.uk wrote:
> Unknown instructions raise illegal instruction. There are also a set
> of "official" ways to check chips. The problem is the run time cost
> of testing often outweights the time saved. Thats why Linus stuff on
> the "with TSC" compile makes a lot of sense.

i see where you are coming from. however, i sense that there's a
trade-off between high performance, and how complicated is system
configuration, especially for someone who doesn't have time to
do research on the capability of his/her processor. i don't even
see that everyone here is in agreement about whether the K6 should
be considered a pentium+ or a pentium-pro, for the sake of selecting
a kernel configuration *and* compiler optimizations for the kernel
build. choosing compiler optimization and kernel configuration
may not even be the same issue, as Helge Hafting points out.

someone else recently suggested running a program before the
configuration menu starts to set up reasonable default configuration
settings based on the processor type, and that might be an excellent
compromise between maintaining high-performance and kernel simplicity,
and ease of configuration.

by the way, i have tried compiling a kernel with gcc 2.8.1 and
-march=pentiumpro, on a K6. it hangs very soon after it boots.
recompiling with -march=pentium fixes the problem. i'm not sure
if this falls under "compiler weirdness," or whether it just means
the K6 kernels should be built with I686 and -march=pentium.
anyway, i thought i should report this here, sample size 1.

--
	Chuck

____________________________________________________________________ More than just email--Get your FREE Netscape WebMail account today at http://home.netscape.com/netcenter/mail

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/