Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss

TSC and Cyrix

0 views
Skip to first unread message

Trever Adams

unread,
Nov 8, 1998, 3:00:00 AM11/8/98
to
Well, 2.1.127 sound works great for me as well as the rest. I was under
the impression that TSC code that worked on Cyrix chips was included in
2.1.126, but from 2.1.127 /proc/cpuinfo I get

processor : 0
vendor_id : CyrixInstead
cpu family : 6
model : 84
model name : M II 3.5x Core/Bus Clock
stepping : 2.8
fdiv_bug : no
hlt_bug : no
sep_bug : no
f00f_bug : no
fpu : yes
fpu_exception : yes
cpuid level : 1
wp : yes
flags : fpu de msr cx8 pge cmov mmx
bogomips : 231.01

Supposedly, in 2.1.127 it should output Mhz if the CPU supports TSC
(note that bogomips is pretty accurate to the Mhz with Cyrix chips, but
that is not what I am after). I have seen a tremendous speed up in all
apps (multitasking not just the foreground process) with 2.1.127, so I
assumed that the TSC code was active... now I found out it doesn't seem
to be.

Anyone out there care to shed some light?

Trever

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

Kurt Garloff

unread,
Nov 10, 1998, 3:00:00 AM11/10/98
to
On Sun, Nov 08, 1998 at 02:42:03PM -0600, Trever Adams wrote:
> Supposedly, in 2.1.127 it should output Mhz if the CPU supports TSC
> (note that bogomips is pretty accurate to the Mhz with Cyrix chips, but
> that is not what I am after). I have seen a tremendous speed up in all
> apps (multitasking not just the foreground process) with 2.1.127, so I
> assumed that the TSC code was active... now I found out it doesn't seem
> to be.

Why do you expect the TSC code to speed up your apps? It just makes the
timing more accurate, if I didn't get anything completely wrong.

--
Kurt Garloff <K.Ga...@ping.de> (Dortmund, FRG)
PGP key on http://student.physik.uni-dortmund.de/homepages/garloff
Unix IS user friendly - it's just selective about who its friends are!

Trever Adams

unread,
Nov 10, 1998, 3:00:00 AM11/10/98
to
Kurt Garloff wrote:
>
> On Sun, Nov 08, 1998 at 02:42:03PM -0600, Trever Adams wrote:
> > Supposedly, in 2.1.127 it should output Mhz if the CPU supports TSC
> > (note that bogomips is pretty accurate to the Mhz with Cyrix chips, but
> > that is not what I am after). I have seen a tremendous speed up in all
> > apps (multitasking not just the foreground process) with 2.1.127, so I
> > assumed that the TSC code was active... now I found out it doesn't seem
> > to be.
>
> Why do you expect the TSC code to speed up your apps? It just makes the
> timing more accurate, if I didn't get anything completely wrong.
>

This is what I have understood from following the TSC additions etc:
Every time a time type function is called from userspace (or in the
kernel) it has to either use the TSC, or go CPU to the 85xx (I think
that is the number anyway) and fetch the time. It was my understanding
that this was some ridiculously long period (for computers). I also
seem to remember David M. saying that the networking code does work a
bit faster with TSC. My non-perfect tests also would support that...
but that could also be caching issues since I do not have two identical
boards with one TSC usable chip (pre 2.1.127) and one not.

Again, I may be flawed, but that is what I understood.

Trever

Lawrence Walton

unread,
Nov 10, 1998, 3:00:00 AM11/10/98
to
I don't know about you but 2.1.127 JUST seems faster in gereral, can't put
my finger on it, but it's just snappy. :)

Hardware: P233MMX
128 MEGS RAM
SCSI 975
Seagate HD

On Tue, 10 Nov 1998, Kurt Garloff wrote:

> On Sun, Nov 08, 1998 at 02:42:03PM -0600, Trever Adams wrote:
> > Supposedly, in 2.1.127 it should output Mhz if the CPU supports TSC
> > (note that bogomips is pretty accurate to the Mhz with Cyrix chips, but
> > that is not what I am after). I have seen a tremendous speed up in all
> > apps (multitasking not just the foreground process) with 2.1.127, so I
> > assumed that the TSC code was active... now I found out it doesn't seem
> > to be.
>
> Why do you expect the TSC code to speed up your apps? It just makes the
> timing more accurate, if I didn't get anything completely wrong.
>

> --
> Kurt Garloff <K.Ga...@ping.de> (Dortmund, FRG)
> PGP key on http://student.physik.uni-dortmund.de/homepages/garloff
> Unix IS user friendly - it's just selective about who its friends are!
>

Brian Gerst

unread,
Nov 11, 1998, 3:00:00 AM11/11/98
to
Kurt Garloff wrote:
>
> On Sun, Nov 08, 1998 at 02:42:03PM -0600, Trever Adams wrote:
> > Supposedly, in 2.1.127 it should output Mhz if the CPU supports TSC
> > (note that bogomips is pretty accurate to the Mhz with Cyrix chips, but
> > that is not what I am after). I have seen a tremendous speed up in all
> > apps (multitasking not just the foreground process) with 2.1.127, so I
> > assumed that the TSC code was active... now I found out it doesn't seem
> > to be.
>
> Why do you expect the TSC code to speed up your apps? It just makes the
> timing more accurate, if I didn't get anything completely wrong.

Because reading the TSC only takes a few cycles and doesn't involve ISA
port I/O, which is comparatively very slow.

--

Brian Gerst

0 new messages