Hi all, I am proud to announce a double-update:





PCISLEEP - scans your PCI bus and shows device information like:

bus.device(.function) [vendor:model] classcode(/iface) vendor class [details]
BusDevF  vend:type  class   vendor  description...
00.0b   [10ec:8139] 0200    Realtek LAN / Ethernet

... in the LIST mode (also shows some ACPI port base guess, please
let me know if it is correct for your system). There are also two
other modes: SLEEP mode, which puts all devices which support that
into "idle" or "halted" mode, and VGASLEEP mode, which also activates
a DPMS screen saver and puts your VGA into "suspend" mode, if it
supports that (PCI and AGP work, even dual-VGA should work). To wake
up, press a key. After VGASLEEP, waking up involves re-running the
VGA BIOS (looks like rebooting). Let me know if it works for you. You
may want to try without EMM386 if you get VGA wakeup problems.

FDAPM - collection of energy saving related DOS functions, a big time
superset of MS POWER functionality. New feature since the 18jan2005
"clock update after suspend" (and standby) update: ACPI support!

Actually, the ACPI support is very basic, but still very cool...
Plus there is a new ZAPSTATS option which resets the busy / idle time
statistics, as suggested by Alain. The statistics are now updated more
often, hopefully improving accuracy. Supported ACPI functions are:

- CPU freeze: Use SPEED0 to stop your CPU clock as in ACPI S1 standby
  mode, press the power button to continue. This is also used if you
  tried to use STANDBY / SUSPEND but no BIOS APM support is found.

- CPU speed control: Use SPEED1 ... SPEED8 to make your system run on
  1/8 ... 8/8 of the maximum speed (some mainboards support steps of
  1/16, but FDAPM does not use this). While FDAPM used the ACPI tables
  of the BIOS, and the assumption that P_BLK base port is at PM1a_EVT
  plus 10h, the THROTTLE tool ( http://www.oldskool.org/pc/throttle/ ,
  open source) uses PCI bus scan and a list of known chipsets instead
  of asking the BIOS. Let me know if either of them fails for you.
  Throttle also triggers C2 CPU state to enter throttled state, which
  FDAPM does not do (because ACPI specs do not seem to suggest it).

- System power off: The POWEROFF function can now use the ACPI tables
  to turn off your computer, if no BIOS APM support is present :-).
  You can use the ACPIOFF function to explicitly test the ACPI style
  even if your BIOS does support APM. Let me know if it works.

Note that those FDAPM functions are not meant to be used in DOS boxes.
THROTTLE can be used with a driver which enables it to run in DOS boxes
of several Windows versions, so you may want to try that. FDAPM ACPI
support is working for me, but if it sets YOUR computer on fire, then
I cannot help you with it, sorry.

Technical information: ACPI tables are memory structures at the end
of your main RAM, linked by a pointer in BIOS or EBDA. The FDAPM way
to access them is to use int 15.87 memory copy - works in plain DOS,
even with HIMEM / EMM386 present, as we are accessing normal RAM. If
we would access a framebuffer, you would need MEMCHECK option for the
EMM386 case, or alternatively a 32bit C library for FDAPM ;-). FDAPM
is about 3200 lines of Assembly code at the moment...
The ACPI tables contain information about the right port to use for
power off / freeze, and about the right values to use for speed / duty
cycle encoding. But (yuck) you have to parse a complex data stream,
the DSDT, to know the right VALUES to use for power off / freeze, or
the right PORT to use for speed setting. I ignore the second problem,
as the port is (for all chipsets known by THROTTLE, except one exotic
old one) always at a fixed offset from another known port.
For the first problem, however, I scan the DSDT to find some byte
constants for the power off / freeze values, and ignore everything
else (fully analyzing DSDT would involve building a tree of named
entities and even running some functions stored in a virtual machine
which processes specific byte code - Intel tells that a normal full
ACPI driver implementation will be around 70 kB big, uncompressed...).
So of course I am very interested in hearing from you whether FDAPM
"just pick the name and the constant" works for you, as I ignore the
whole "name tree" and "functions" stuff, including "prepare to sleep".

Enjoy the updates - learning about ACPI kept me quite busy during
the last several weeks. Feedback is welcome :-).

PS: compressed binary sizes are 4k for PCISLEEP and 6k for FDAPM. This
is possible because PCISLEEP only knows 15 widespread vendor names ;-).


[Would be nice if somebody could think of a nice SHORT summary of the
above for a news item...]

This SF.Net email is sponsored by: IntelliVIEW -- Interactive Reporting
Tool for open source databases. Create drag-&-drop reports. Save time
by over 75%! Publish reports on the web. Export to DOC, XLS, RTF, etc.
Download a FREE copy at http://www.intelliview.com/go/osdn_nl
Freedos-user mailing list

Reply via email to