[lkml]   [1996]   [Dec]   [21]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: Yet another cyrix 6x86 question

On Sat, 21 Dec 1996, Stephen S M Wong wrote:

> Hi,
> > For optimal perfomance on a Cyrix 6x86, should one compile as Pentium or
> > PPro ? I've tried both and they both run fine, but which is better?
> I think the current status is that, there is no difference, as no
> production version of gcc has PPro (nor Pentium) specific optimization
> yet. So, 486/Pentium/PPro flag should not make a difference. Maybe
> there are some conditional defines to make the memory alignment more
> generous for Pentium or PPro.
> My 2 cents.

Can anyone say inline assembly ? We use hand optimized assembly all over
the place where speed is an issue. You can use any instruction as long as
you know what it's opcode, and not just it's symbolic representation.

A basic assembly compiler only does a bit of syntax checking, and the
object of the excersise, symbolic -> opcode translation .... If the
compiler dosn't know about fancy_new_tcp_checksum_in_one_tick_instr, then
you can just tell it that you want opcode XXX insted.

We also use opcodes directly, and the places were we do, are protected
behind the approprate #ifdef CPU=x86.

What we (the x86 based gcc using comunity) need is some one, or a group to
add some basic functions to gcc for the pentium, ppro, and non-intel x86
clones (ie -m586, -m686, -m5x86, -m6x86, -mbluethunder (?), -Cy486,
-mxyzclone ...) that will switch on and off the correct optimizations
like for the i386, and i486. After that the in kernel source switches
that are used should srink to a minimum (ie -pipe -mCPU -DCPU, and
optionally -D__SMP__), and with the CPU specfic code in the kernel, we
should be all set for world domination, well maybe ;).

> Stephen Siu Ming Wong Internet: [ O O ]

PGP key pass phrase forgotten, \ Overload -- core meltdown sequence
again :( and I don't care ;) | initiated.
/ This space is intentionally left
| blank, apart from this text ;-)

 \ /
  Last update: 2005-03-22 13:38    [W:0.038 / U:1.208 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site