+-----------+------------------+---------------------------------------------+
|+--------+ |File: CPUTYPE.DOC |Subject: CPU/FPU feature detection library. |
|+--------+++---------------+--+---------------------------------------------+
||Њ-Њ-+-+ ||Date: 28/11/94 | Copyright(c) 1993,94 by B-coolWare. |
||І-+-+-|І-||Version: 1.14e | This document should not be changed in any way |
|+---------+|Revision:03/01 |and should be distributed with related software.|
+-----------+---------------+------------------------------------------------+
Document no.: LR-0006.94.93M03.01
"...Too much is not enough."
(Deep Purple)
Contents
License Agreement.
Appendix A. Library Features and Notes.
Appendix B. Credits.
Appendix C. Known incompatibilities.
Appendix D. Files in distribution.
Too-Much-In-One-So-Don't-Get-Lost(tm) CPU/FPU feature detection library
License.
This license regards to the Too-Much-In-One-So-Don't-Get-Lost(tm) CPU/FPU
detection library version 1.14e dated 28/11/94, mentioned hereafter as
"source code". It describes your rights, obligations and liabilities as for
this source code. It took much time to gather all information used to
produce this code, so the main purpose of this license is to grant anyone
free access to it. Use it free, give away to anyone, but please do not
sell or lease it - it's my work, and if you think someone should be paid
for it, you're not the rite person anyway.
Here it goes:
COPYRIGHT:
This source code is copyrighted freeware. This means that author reserves
copyright for this code. This doesn't mean however that you cannot use,
distribute or modify the code. This just means that you cannot reserve
copyright for this code or charge money for it. However you may reserve
copyright for any changes/additions you made to it. See details hereafter.
YOUR RIGHTS AND LIABILITIES:
You have legal rights to copy, distribute, modify and use entire code or parts
of it in either commercial or noncommercial software, provided that you do
not charge additional fee for distribution/incorporation of this code or its
parts in your own products. If you do so, your rights automatically void and
you'll be liable for violation of the Russian Copyright Law in Computer
Programs and Databases as well as other laws and international provisions
regarding to intellectual property and copyrights. Parties that acquire this
source code from you will still have their rights for the source code as
long as they comply with terms of this license.
YOUR OBLIGATIONS:
The source code SHOULD be distributed along with this license and you should
document any changes made to it either in source files or in additional
documents. The code cannot be distributed partially, you should include all
files that builds the library plus any additional files needed to compile it
in the distribution archive/set. Regardless of the extent of modifications
made by you to this source code, it should remain free, and you should
provide a way for anyone (and partially the author :) to obtain his own copy
as of original, so as of modified code. Although you can reserve copyright
for modifications, the rest of the code will still be copyrighted by
B-coolWare and you should keep the B-coolWare's copyright notice in the code
unchanged.
WARRANTIES:
This source code is provided "AS IS" with no express or implied warranties,
including any implied warranties of fitness for particular purpose. If you
modify the source code included herein, please add a notification about it
and complete description of what and how you modified, so that subsequent
users of this code will know that they use not original but rather modified
version of the code.
AUTHOR'S LIABILITY:
In case of any damage caused by use or misuse of this code or any other kind
of decease including loss of business profits or valuable information,
author shall not be liable for it. I expressly disclaim any warranties
regarding to the quality of this code, or its errorlessness.
AUTHOR'S EXCLUSIVE RIGHTS:
Author reserves the right to change this code entirely or partially without
notification to its users. He also reserves the right to use this code in
any manner in any commercial or non-commercial software of any kind and
give it away to anyone who ask.
If you have any questions, suggestions or usable information on how to
improve this product, feel free to contact author at
e-mail (preferrable):
2:5028/52.6@fidonet
bob@ymz.yaroslavl.su (Relcom@Internet)
do ye think me should arrive on CompuServe? :)
phones:
+7-(0852)-274-551 (9:00am to 6:00pm, Moscow Time) voice\data 14400/V.32b
+7-(0852)-238-663 (8:00pm to 10:00pm,Moscow Time) voice
please don't call earlier or later - gimme a break...
paper mail:
150031, 10/4/13,
Dobrynina Str.,
Yaroslavl,
Russia
Vladimir M. Zakharychev (Владимир Захарычев)
Calls and letters both in English and in Russian are welcome. Please do not
speak/write in any other language if you want to be answered. Letter bombs
will be returned to sender :) shhhh-boom-BANG!
Huh...! Hope I made every effort to keep the lawyers happy... :)
Appendix A. Library Features and Notes.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
The Too-Much-In-One-So-Don't-Get-Lost(tm) CPU/FPU feature detection library
allows you to distinguish between following 80x86 family CPUs:
Intel 8088,8086,80188,80186,80286,80386DX,80386SX,i486DX,i486SX,Pentium and
any succeeding CPUs (when they become available),
NEC V20 and V30,
Cyrix 486SLC, 486SX/DX/DLC and M1 (586),
UMC U5-S, U5-D (486s) (??? is U5-D real thing?)
FPUs:
Intel 8087,80287,80287XL - not 100% reliable,80387,i487sx and 487 internal,
Cyrix 2C87,3C87 and 4C87,
IIT 2C87 and 3C87,
Weitek 1167 and 1367, both reported simply as Weitek;
determine whether or not the processor is in V86 mode (for DOS programs only)-
generally this one just tests for PM flag and assumes V86 if it is set for DOS
+ PM combination automatically leads to V86;
determine 386dx chip step (B0 or earlier,B1,D0 or later) -
this one tests for known bugs in chips of these steps, it does not reset CPU
to look at EAX value; doesn't work under Windows;
determine CPU clock speed (for all listed processors, was not checked on some
models and may require some adjustments). Works fine under _my_ Windows, but
some people stated it produced incorrect results on their systems.
Notes.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Note 1:
AMD Am386SX/40 and Am386DX/40 - can be identified as such by using CPU_Type
in conjunction with Speed: if CPU is identified as Intel 386 and its clock is
more than 33MHz then this is definitely an AMD chip - Intel doesn't produce
40MHz 386s while AMD does. However this method is not implemented in CPU_Type
routine because speed computing requires some floating point calculations and
I don't want to insert FPU instructions in the code because there may be no
FPU... You may implement this method in your high-level routine reporting
CPU type if you want (and as I did :). Also you may conclude i486DX2 if it is
identified as 486DX and has CPU clock of 66 MHz - afaik Intel do not produce
DX chips with such clock frequencies. However I cannot guarantee this is true
and don't use this method by myself. If anyone knows of reliable method of
identifying DX2, DX4 and other chips with internal clock doubling/tripling/etc.
then please let me know (it better be working ASM sources...).
Note 2:
The source code is commented so I didn't include its description here. I think
there'll be no problem using it. I also included samples in Turbo Pascal and
Turbo C, which may be incorporated into your hi-level programs. Lotta probs
vanished, heh :)
Note 3:
More and more CPUs approaching: DX4, SLC2, SLC3; 486xx series's getting more
and more xx modifiers without differences to software... I feel like I'm
gonna give up adding new CPU types here - I got too much problems with those
I already have. They're all alike, no way to determine which are what. Maybe
if vendors add special instructions like CPUID in all their chips and
program microcode to give different results - well, then I'll try. For now
I'm quit.
Note 4:
I have no access to Cyrix M1 machines and thus can't check and fix CPU timings
in CPUSPEED.ASM for this processor. This may lead to inadequate MHz
reported. If anyone have time and access to an M1-based machine I'd like
you to check this and fix CPUFix value if needed. And don't forget to contact
me and send me adjusted value so I'll be able to fix it by myself! Also
please check if M1 is correctly identified - I'm not quite sure it will.
Note 5:
Take a look at the batches I provided to simplify compile process and
correct paths and parameters where needed prior to running them. This is
especially true for MAKEC.BAT and MLIB.BAT, for they contain paths I used on
my PC to compile C version of code.
Note 6: -- for those having problems in certain environments
This code haven't been extensively tested on all CPUs it theoretically should
recognize. I performed extensive tests only on systems available to me which
include:
1. AMD Am386DX-40/Intel 80387DX
running MS-DOS 5.0, QEMM 7.04 + DESQview 2.63, Windows 3.11
2. Intel i486DX2-66
running MS-DOS 6.2, QEMM 7.04, Personal NW, Stacker 4.0,
Windows 3.10 Russian
3. GenuineIntel i486SX-40/None - new generation (supporting CPUID)
running MS-DOS 6.2, QEMM 7.01, Windows 3.10
4. Intel 80286-10/Sun SPARCStation 1+ emulating 80286
5. Intel 80386SX-25/Intel 80387SX
running MS-DOS 6.2/MemMaker, Windows 3.10 Russian
6. AMD Am386DX-40/Cyrix Cx87SLC
running Novell DOS 7, Personal NW, Windows 3.10
7. Intel i486DX-50
running MS-DOS 6.2, Windows for Workgroups 3.11
8. Intel Pentium-S/60
running MS-DOS 6.2, Windows for Workgroups 3.11
[2] also were running OS/2 2.1 few months ago, but it doesn't now and I can't
check if the code still work under it.
On all listed systems the code appeared to work fine giving correct results
under all listed environments except [1,6] when running Soft-ICE 2.62 - code
incorrectly stated that CPU is 386SX. I cannot guarantee that the code will
run o.k. on systems with other configurations/operating environments, though
I hope it will. Please don't bomb me with letters like "wow! this thingie
crashes my PC with such and such CPU/FPU under such and such env." - you got
full source code, and you should know how to use a debugger if you know how to
program, so try to find out the origin of your problem on YOUR system (and
maybe a workaround) before reporting it.
Appendix B. Credits.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
This appendix contains names and netmail addresses of people who contributed
to this product in any way, either by adding any methods or by fixing
existing methods given incorrect results. If you do such additions or fixes,
please contact me and send modified version of the code so that I'll be able
to incorporate it in next releases.
My thanx to
1. Igor Dral of 2:50/321@fidonet
who helped me lot in cleaning the code of bugs, testing, etc. Thanx, pal!
2. VAP of 2:5028/52.6@fidonet (he's my BOSS),
who helped me with C code - I don't like this language myself :)
3. Alex B. Shamarokov of 2:5020/205.8@fidonet
who awared me of strange behavior of the code on new Intel' chips with
CPUID support, so I did an attempt to fix it in v1.14c.
4. Janis Smits of 2:5100/25@fidonet
who awared me of OS/2 problems and pointed out the origin of them.
5. Ilya Tumanov of 2:5030/82.6@fidonet
who advised me to add integer CPU MHz calculations for those who want only
rough approximation of clock speed. He also advised me not to give up with
this library - well, I'll try...
6. YOUR NAME MIGHT BE HERE...
And, of course, thanks to Almighty God for the inspiration and everything,
and all of you, who paid attention to this product, spent your invaluable
time testing it and maybe even made use of it!
Appendix C. Known incompatibilities.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1. Under Soft-Ice 2.62 and below incorrectly states 386DX to be 386SX. This
is because Soft-Ice's V86 monitor do not allow to toggle ANY flags in
CR0 from V86 task. All other CPU identifiers does just such a mistake
under Soft-Ice. Let's hope Nu-Mega will fix their monitor in future
versions of this nice debugger so we'll be able to get correct results. :)
2. Under Windows 3.1 386 chip stepping determination routine crashes DOS
session due to "invalid instruction". This routine was originally taken
from PharLap TellMe utility. BUT this program just do not perform chip
step detection if running under Windows Enhanced Mode! Seems like there's
no other way to correct this problem but do not use this routine under
Windows... so I finally got rid of it if running under Windows.
3. I was awared that on certain systems this code completely craches the OS.
The origin of the problem is not figured out yet, so keep an eye on it and
if the code causes system crash on your PC, please try to find offending
source and/or environment that forces it to crash.
4. Under OS/2 2.x running on 486 systems sometimes causes VDM session crashes.
Regrettably I couldn't check it by myself but Janis Smits who awared me
of the problem also suggested a workaround for it which I included in
current release. I still don't know if it helped.
5. People say that under OS/2 clock frequency gets divided by 2. I didn't
find a workaround yet. Those who can't wait may try this: use isInOSZwei
routine to determine if we're under OS/2 and multiply clock frequency by
2 (shl 1) if so. This is very rough approach though and I hope to find
more elegant way to solve the problem. I tried to find a "begin/end criti-
cal section" call to OS/2's DOS CompBox API and failed. If anyone knows of
correct and reliable way to enter/leave critical section under OS/2 please
let me know.
Appendix D. Files in distribution.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Original issue of the source code as of version 1.14e include:
File Description
------------ ----------------------------------------------------------
CPU_TYPE.ASH Main library module for ASM programs
DOSINOUT.ASH DOS Input/Output routines
LSTRING.ASH Lstring macro to define Pascal-style length,data strings
HEADER.ASH Assembly language header file - memory model definition, etc.
UNIDEF.INC Assembly language macros
CPU.ASM Sample program demonstrating usage of CPU_TYPE.ASH
CPUSPEED.ASM CPU clock speed determination routine
CPU_HL.ASM Main library module for use with hi-level languages
CPU.C Sample program in C
CPUTYPE.C C library implementation
CPUTYPE.H C library interface
CPU.PAS Sample program in Pascal
CPUTYPE.PAS Pascal library interface & implementation
P5INFO.ASM Low level routines for P5INFO.C and P5INFO.PAS
P5INFO.C Sample program providing basic info on P5 in C
P5INFO.PAS Sample program providing basic info on P5 in Pascal
MAKEA.BAT Make batch for assembly-language version
MAKEP.BAT Make batch for pascal version
MAKEC.BAT Make batch for C version
MLIB.BAT Make batch to build C/C++ .lib library file
YESNO.COM Batch service program, displays prompt and waits for Y/N.
HISTORY.DOC History of changes
----------------------------
!!!! ПРИМЕЧАНИЕ СОСТАВИТЕЛЯ:
С разрешения автора здесь публикуются только те файлы,
которые имеют непосредственное отношение к Паскалю.
----------------------------
<< end of CPUTYPE.DOC >>
|
