Here is the summery of letters that I have received to my question: When recompiling the kernel on .99pl5, should the -m486 be used for use in a 386 w/ FPU linux box? Here are the responses that I received from various people. Thanks for all the help guys! I received excellent, informative responses from linux users. I thank all who responded to my question, You'all are great. Neill Means means@sage.cc.purdue.edu P.S. When recomiling the kerel, what does profiling support for the kernel do? Is it good/bad? Pros/Cons? [SUMMERY]: ================================ You probably do not want to compile with the -m486 if you want to get the most performance out of your kernel. The -m486 flag does some optimizations for the 486 (FPU or not doesn't matter; optimizations help on a 486sx as well). These optimizations are due to the instruction timings and pipeline handling on the 486. The 386 will still execute these instructions, but with reduced performance. I have a 386/40 w/387/40 myself. I compared the latest X distribution, which is compiled with -m486, to the previous one, which is not, and there is a noticeable decrease in performance. (this is especially visible when scroling in an xterm) -- | Anatoly Ivasyuk @ Rochester Institute of Technology | |-----------------------------------------------------| | anatoly@nick.csh.rit.edu | ani0349@cs.rit.edu | | Computer Science House | Computer Science Dept. | |-----------------------------------------------------| ================ Theoretically the 486 is capable of going super-scalar (executing more than one instruction in a clock cycle) for certain combinations of certain instructions. These combinations are not neccesarily the obviously fastest way of generating code to do something. So, the -m486 switch causes the compiler to generate code that will run a little faster on a 486 and a little slower on a 386. You would probably do a bit better to compile with out the switch if you don't have a 486.Not a bad question at all. :) jem. -- jem@sunsite.unc.edu\/sunSITE admin ================================================== All -m486 does is cause generated code to be placed so as to improve the performance of the 486's built-in cache. On a 386 it wastes a *little* memory with no speedup... but also no slowdown. ++Brandon -- Brandon S. Allbery bsa@kf8nh.wariat.org ======================================================= -- Sorry, our mailer only posts first 4 lines of the .signature file. Here is the rest of it:
| Greg Kroah-Hartman | [PATCH 004/196] Chinese: add translation of SubmittingPatches |
| James Bottomley | Re: Integration of SCST in the mainstream Linux kernel |
| Jeff Garzik | Re: [Patch v2] Make PCI extended config space (MMCONFIG) a driver opt-in |
| Chodorenko Michail | PROBLEM: Celeron Core |
git: | |
| Linus Torvalds | People unaware of the importance of "git gc"? |
| Johannes Schindelin | Re: Empty directories... |
| Jakub Narebski | Re: VCS comparison table |
| Sam Song | Re: Fwd: [OT] Re: Git via a proxy server? |
| J.W. Zondag | Dell PE1950 III - Perc 6i |
| Richard Stallman | Real men don't attack straw men |
| GVG GVG | ssh_exchange_identification: Connection closed by remote host |
| Anselm R. Garbe | OpenBSD 4.0 / Xorg -> vesa 1920x1200 widescreen resolution |
| Jim Winstead Jr. | Re: Root Disk/Book Disk Compatibility |
| Anselm Lingnau | File creation date in UNIX (was: Re: VMS) |
| Rafal Kustra (summer student) | mount |
| Nicholas Yue | Re: more on 486/33 weirdness |
