Date: Mon, 31 Oct 2005 22:48:20 +0100 From: "Poul-Henning Kamp" <phk@phk.freebsd.dk> Cc: Alexander Leidinger <Alexander@Leidinger.net>, freebsd-current@freebsd.org, Robert Watson <rwatson@FreeBSD.org>, David Xu <davidxu@freebsd.org> Subject: Re: TSC instead of ACPI: powerd doesn't work anymore (to be expected?) Message-ID: <84718.1130795300@critter.freebsd.dk> In-Reply-To: Your message of "Mon, 31 Oct 2005 12:03:48 %2B0100." <81484.1130756628@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
In message <81484.1130756628@critter.freebsd.dk>, "Poul-Henning Kamp" writes:
>And finally, WinTel know that the ACPI timer sucks, thats why more
>recent hardware has HPET timers, I'm working on support for that.
OK, I've added a rudimentary HPET support that gives us a timecounter.
It has quality -200, so it will not be used unless manually configured
with the sysctl.
Here is the short summary of the two machines I could test it on:
TSC ACPI-fast HPET i8254
dual Xeon: 580 nsec 1429 nsec 1120 nsec 3980 nsec
dual Opteron: 212 nsec 1270 nsec 1760 nsec 4420 nsec
So while AMD is better at TSC and ACPI-fast, Intel is better at
HPET and i8254.
And yes, this probably depends on just about everything on the motherboard
so your mileage will vary...
Fortunately we can choose timecounter based on its speed, once we have
convinced ourselves that the quality is OK. We don't currently do this,
but the code can be resurrected.
Poul-Henning
Full results (HEX = Opteron, Y = Xeon)
critter phk> ministat HEX/*.[AHT]* Y/*.[AHT]*
x HEX/_.tc.ACPI-fast
+ HEX/_.tc.HPET
* HEX/_.tc.TSC
% Y/_.tc.ACPI-fast
# Y/_.tc.HPET
@ Y/_.tc.TSC
+--------------------------------------------------------------------------+
| @ # % |
| @ # % |
| @ # % |
|* @ # % |
|* @ # x % + |
|* @ # x % + |
|* @ # x % + |
|* @ # x % ++ |
|** @ # xxx % ++ |
|*** @ # xxx % +++|
|MA A A |A| A MA |
+--------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 10 1.257e-06 1.299e-06 1.265e-06 1.2705e-06 1.5042532e-08
+ 10 1.752e-06 1.794e-06 1.76e-06 1.7637e-06 1.2543701e-08
Difference at 95.0% confidence
4.932e-07 +/- 1.3013e-08
38.8194% +/- 1.02424%
(Student's t, pooled s = 1.38496e-08)
* 10 2.11e-07 2.45e-07 2.12e-07 2.189e-07 1.1551335e-08
Difference at 95.0% confidence
-1.0516e-06 +/- 1.26009e-08
-82.7706% +/- 0.99181%
(Student's t, pooled s = 1.3411e-08)
% 10 1.429e-06 1.431e-06 1.429e-06 1.4295e-06 8.4983659e-10
Difference at 95.0% confidence
1.59e-07 +/- 1.00101e-08
12.5148% +/- 0.787888%
(Student's t, pooled s = 1.06536e-08)
# 10 1.12e-06 1.121e-06 1.12e-06 1.1201e-06 3.1622777e-10
Difference at 95.0% confidence
-1.504e-07 +/- 9.99638e-09
-11.8379% +/- 0.786807%
(Student's t, pooled s = 1.0639e-08)
@ 10 5.79e-07 5.8e-07 5.8e-07 5.796e-07 5.1639778e-10
Difference at 95.0% confidence
-6.909e-07 +/- 1.00001e-08
-54.3802% +/- 0.787097%
(Student's t, pooled s = 1.06429e-08)
critter phk> ministat HEX/*
x HEX/_.tc.ACPI-fast
+ HEX/_.tc.HPET
* HEX/_.tc.TSC
% HEX/_.tc.i8254
+--------------------------------------------------------------------------+
| %|
|* + %|
|* x + %|
|* x + %|
|* x + %|
|* x + %|
|* x + %|
|* x + %|
|* xx + %|
|** xx + %|
|A A| A A|
+--------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 10 1.257e-06 1.299e-06 1.265e-06 1.2705e-06 1.5042532e-08
+ 10 1.752e-06 1.794e-06 1.76e-06 1.7637e-06 1.2543701e-08
Difference at 95.0% confidence
4.932e-07 +/- 1.3013e-08
38.8194% +/- 1.02424%
(Student's t, pooled s = 1.38496e-08)
* 10 2.11e-07 2.45e-07 2.12e-07 2.189e-07 1.1551335e-08
Difference at 95.0% confidence
-1.0516e-06 +/- 1.26009e-08
-82.7706% +/- 0.99181%
(Student's t, pooled s = 1.3411e-08)
% 10 4.418e-06 4.439e-06 4.424e-06 4.4264e-06 7.5894664e-09
Difference at 95.0% confidence
3.1559e-06 +/- 1.11942e-08
248.398% +/- 0.881084%
(Student's t, pooled s = 1.19138e-08)
critter phk> ministat Y/*
x Y/_.tc.ACPI-fast
+ Y/_.tc.HPET
* Y/_.tc.TSC
% Y/_.tc.i8254
+--------------------------------------------------------------------------+
|* + x %|
|* + x %|
|* + x %|
|* + x %|
|* + x %|
|* + x %|
|* + x %|
|* + x %|
|* + x %|
|* + x %|
|A A A A|
+--------------------------------------------------------------------------+
N Min Max Median Avg Stddev
x 10 1.429e-06 1.431e-06 1.429e-06 1.4295e-06 8.4983659e-10
+ 10 1.12e-06 1.121e-06 1.12e-06 1.1201e-06 3.1622777e-10
Difference at 95.0% confidence
-3.094e-07 +/- 6.0245e-10
-21.6439% +/- 0.0421441%
(Student's t, pooled s = 6.41179e-10)
* 10 5.79e-07 5.8e-07 5.8e-07 5.796e-07 5.1639778e-10
Difference at 95.0% confidence
-8.499e-07 +/- 6.60693e-10
-59.4544% +/- 0.0462185%
(Student's t, pooled s = 7.03167e-10)
% 10 3.973e-06 3.986e-06 3.98e-06 3.9791e-06 3.4140234e-09
Difference at 95.0% confidence
2.5496e-06 +/- 2.33748e-09
178.356% +/- 0.163517%
(Student's t, pooled s = 2.48775e-09)
--
Poul-Henning Kamp | UNIX since Zilog Zeus 3.20
phk@FreeBSD.ORG | TCP/IP since RFC 956
FreeBSD committer | BSD since 4.3-tahoe
Never attribute to malice what can adequately be explained by incompetence.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?84718.1130795300>