旼컴컴컴컴컴쩡컴컴컴컴컴컴컴컴컫컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴커 냥袴袴袴袴� 쿑ile:TMi0SDGL.DOC 쿞ubject: TMi0SDGL R2 Library Documentation. � 넋袴袴袴袴軻쳐컴컴컴컴컴컴컴쩡컨컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴캑 낳靈靈爐念 볐Date:25.06.2000� Copyright(c) 1996-2000 by B-coolWare. � 낳聃懃懃쌈캤쿣ersion: 2.15 � This document should not be changed in any way � 냠袴袴袴袴暠쿝evision: 02/16쿪nd should be distributed with related software.� 읕컴컴컴컴컴좔컴컴컴컴컴컴컴좔컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴컴켸 Document no.: LM-0020.96.00M02.16 Too-Much-in-0ne-So-Don't-Get-Lost(tm) ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ CPU/FPU detection library for Intel 80x86 and compatibles Revision 2. TMi0SDGL Revision 2 and this document are Copyright(c) 1996-2000 by B-coolWare. Other brands and product names mentioned herein are trademarks or registered trademarks of their respective holders. The contents of this document is subject to change without prior notices. Printed in Russian Federation (or wherever you are, heh) Printed on recycled paper (at least I hope so...) Product is not tested on animals and is ozone-friendly. ..i Contents................................................................ Page 1. License Agreement . . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.1 Distribution Conditions . . . . . . . . . . . . . . . . . . . . . . 1 2. Library Features . . . . . . . . . . . . . . . . . . . . . . . . . . 2 3. Notes For Programmers . . . . . . . . . . . . . . . . . . . . . . . . 3 3.1 General Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 3.2 Known Problems And Incompatibilities . . . . . . . . . . . . . . . 4 4. Library Reference . . . . . . . . . . . . . . . . . . . . . . . . . . 6 4.1 Function Reference . . . . . . . . . . . . . . . . . . . . . . . . 6 4.2 Variable Reference . . . . . . . . . . . . . . . . . . . . . . . . 6 4.3 Tips On Using DLLs . . . . . . . . . . . . . . . . . . . . . . . . 7 5. Credits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 5.1 Hall Of Fame . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 6. How To Contact The Author . . . . . . . . . . . . . . . . . . . . . . 9 6.1 Author's Reward . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Appendices................................................................... Appendix A. List Of Files In Original Distribution Archive. Appendix B. Useful Resources On Internet/WWW. ..1 1. License Agreement. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ The TMi0SDGL Revision 2 CPU/FPU detection library is free copyrighted software and is copyright(c) 1996-2000 by B-coolWare. Parts of this license agreement may resemble with GNU General Public License Agreement, but THIS IS NOT A GNU product. Author reserves some specific rights for this product which GNU GPL forbid to reserve. Please read this license in its entirety and do not use this product if you do not accept the license. a) GRANT OF LICENSE. i. You, the end user, are hereby granted the permission to use the TMi0SDGL 2 library ("the code") in your own non-commercial applications of any kind without any royalty to its author. You are also granted the right to copy or otherwise distribute the code provided that you charge no fee for it except a reasonable fee for the physical act of copying but not exceeding US $5 in total. You are also granted the right to modify the code to adapt it for your purposes or eliminate a bug. In case of modifying the code, you should add a notice that the code was modified and contact author in order to provide him with copy of all modified files so that changes you made can be incorporated in subsequent releases. Regardless of the extent of such modifications, the code will still copyright(c) by B-coolWare and you should keep all copyright messages intact. However you are not allowed to distribute the modified code and/or change its version number. ii. You are allowed to use the code in commercial software (including shareware) without prior author's written permission provided that none of the key features of the software are based on this code (like they are in System Information tools for example). In this case you should only add the copyright notice somewhere in your software that says that portions of your software are copyright(c) by B-coolWare. If some or all of the key features of your commercial product are based on this code, you should inquire author's written permission to use this code before you publish your software. b) YOUR OBLIGATIONS AND LIABILITIES. You are not allowed to change copyright for this code, sell or lease it, or charge fee for distributing it that exceeds your expences from doing it or exceeds US $5 whichever amount is less. You are not allowed to use the code in commercial software (including shareware) if the key features of such software are based on the use of this code without prior author's written permission. If you will not comply with the terms above you will be liable for violation of Russian Copyright Law for Programs and Databases as well as other national and international copyright laws, treaties and provisions. Parties that aquire this code from you will still have their rights for as long as they will comply with the terms of this license. c) WARRANTIES AND AUTHOR'S LIABILITIES. Author hereby disclaims all warranties regarding to this code, its quality, errorlessness and fitness for a particular purpose. I can only guarantee that the code will occupy disk space. In case of any data losses, physical damage to your hardware or loss of profits author shall not be liable for it even if he or his representatives were advised of the possibility of such damages. This code is provided on "AS IS" basis with no technical support. The whole risk is with you. d) AUTHOR'S SPECIFIC RIGHTS. Author reserves the right to use this code in any software of his own, or the company he is employed at, whether commercial or not. He also reserves the right to change software legal status in future releases as he sees fit (to shareware, for example). Last word to all computer freaks: PLEASE do not sell or lease this software - it took much time and effort of several people to produce it and if you think someone should be paid for it - you're not the right person anyway. 1.1 Distribution Conditions. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ This library is copyrighted free software. This means that it can be freely distributed and used in any non-commercial software provided that no fee charged from recipients and that copyright notices are not changed. Thus, you are free to distribute this library by any means provided that you make no profit from it and do not modify the copyright notices. Upload it to your favorite BBSes and ftp sites, give it to your friends - the wider it will spread - the better. * Archive sites: this software may be included in public non-commercial ftp/http archives and/or CD-ROMs published by public non-commercial archive sites provided any charge is solely to recover the cost of distribution. All such distributions must contain all files that builds the distribution, including this file. Please contact author for the most recent version available before publishing the software. Remember: if you distribute this library, include ALL files that came in original archive (or better distribute the original archive itself). This is especially true for documentation files including the one you read now. The most recent version of the library can always be downloaded via World-Wide-Web at http://come.to/b-coolware This library is also uploaded to SimTel archives shortly after release. It is put into /pub/simtelnet/msdos/sysinfo directory under name cpuXXX.zip, where XXX is the version number (215 for this release), and DLL source code in cpuspXXX.zip. ..2 2. Library Features. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ TMi0SDGL Revision 2 comes in source code for three different languages, namely 80x86 Assembler, C/C++ and Borland Turbo/Object Pascal. It detects CPU brand and model, FPU brand and model, CPU clock speed of the system it is ran on as well as some other features for several CPUs. Currently detected processors: Intel : 8088, 8086, 80C88, 80C86, 80188, 80186, 80286, 80386SX, 80386DX, 386SL, i376, RapidCAD, i486SX, i486DX or i487SX, Pentium, P24T (Pentium Overdrive), P54C, P55C, Pentium Pro (P6), Pentium II, Celeron(tm) and Celeron A (128k cache), Pentium II Xeon(tm) and Pentium IIe, Pentium III brands, any succeeding Intel CPUs when they become available. For newer i486s also determines exact model (SX,SX2,DX,DX2,DX2WB, DX4,Overdrive). NEC : V20, V30 V40 and V50 detection method is known but not implemented. C&T : 38600 and 38600SX AMD : Am386SX-40, Am386DX-40, Enhanced Am486DX2, DX4, DX2+, DX4+, X5 (Am5x86 - 486DX4 class CPU), Elan SC400 (486-class microcontroller), 5k86 and SSA5 (K5 - P5 class CPU), K6, K6-2, K6-III, Athlon(tm), Duron(tm). VIA Cyrix: Cx486SLC, Cx486DLC, all other Cx486 models, M1sc (5x86), M1 (6x86), M2/M-II (6x86MX, P5MMX-like CPU), VIA Cyrix III. TI : Ti486SXL (Potomac), new Ti486DX, DX2, DX4 IBM : 386SLC, 486SLC, 486SLC2, 486BL3 (Blue Lightning) UMC : U5S, U5SD, U486DX2, U486SX2 NexGen : Nx586, Nx686 Centaur/ IDT Tech.: WinChip C6, WinChip 2 (aka C6+), WinChip 2A (100MHz bus version), WinChip 3 Rise : iDragon(tm) mP6, mP6 II Currently detected math co-processors: Intel : 8087, 80287, 80287XL, 80387, RapidCAD, internal where applicable Cyrix : 82x87, 83x87, 84x87, EMC87, internal where applicable AMD : internal, where applicable ULSI : 82x87, 83x87 C&T : 38700 IIT : 2C87, 3C87, 4C87 NexGen : Nx587, internal where applicable Weitek : Abacus (all models simply reported as Weitek) Centaur/ IDT Tech.: internal Rise : internal Also detects emulated FPUs on 386 and up. Also detects L1 cache size for processors with L1 cache. Also determines if CPU is in real or protected (V86) mode (386+). Calculates clock speed for all listed processors, even under Windows and DESQview (OS/2 requires HW_TIMER set to ON in DOS Box Settings for correct calculation). Works fine in plain DOS, Windows (including Windows 95), DESQview, can be ran in OS/2 DOS Box though the results may be disappointing. 16-bit version works fine under Windows NT (32-bit doesn't) :( Even works in Sun's SunPC for Solaris x86 emulation box - reports 286 at speeds of 100MHz or more! :) Includes compiled libraries for Pascal, C/C++, DLLs for Delphi, C/C++, Visual Basic, etc. Applications can target plain DOS, protected mode DOS, Windows 3.1x and Windows 95 (sorry, no Windows NT | OS/2 support for now...) Full source code allows for easy adaptation and expansion of the library. Comments wherever possible tells what is done and why - the code is extremely easy to read and understand. ..3 3. Notes For Programmers. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ This section includes information for users of previous versions of TMi0SDGL (1.1x) as well as information on known problems with the library and what is under the hood. 3.1 General Notes. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Some previous versions of TMi0SDGL were spread wide enough and many people already made use of them. There are significant changes made in Revision 2 and following is detailed information on what was changed and how to deal with this new revision. Note that the code was rewritten from scratch and is not compatible with previous versions. If you made use of any 1.x version, you'll have to rewrite your software in parts that use the library. a) MAKEFILEs do not include rules to make C protected mode executables - there's too many C compilers able to produce such .EXEs, each using its different unique set of command line parameters. I assume that you are using Borland C++ as C compiler, Turbo Librarian as object librarian and Turbo Assembler 4.0 or later as assembler. If you using another compiler - well, try to change command line options used to make the library and examples... You SHOULD use the assembler with similar functionality and language/macro features as Borland Turbo Assembler 4.0 (MASM 5.1 will do ok, for example)! I'm not sure that PharLap 386|ASM will compile this code but you can use -op switch with Turbo Assembler to produce PharLap-style object file to use with PharLap family of DOS Extenders and their 386|LINK. Things you'll have to change in MAKEFILEs are: CC = name of C compiler executable LIB = name of object librarian executable ASM = name of assembler executable INCLUDE = path to include files ASMINC = path to assembler include files CCOPTC = options to produce .OBJ files from .C files CCOPTL = options to link .OBJ files into .EXE file ASMOPTN = options to produce real mode .OBJ files from .ASM files ASMOPTX = options to produce protected mode .OBP files from .ASM files (in fact, these are the same as ASMOPTN but with added __DPMI__ define) It is doubtly that you'll have to change anything else, except maybe the name of the library (I suggest you name it CPUTYPEX.LIB, where X denotes protected mode). If you modify MAKEFILEs to accomodate them for any C compiler other than Borland C++, please send me a copy so that I'll be able to include it in next releases. Microsoft C/C++ 7.0 command line for 16-bit targets is included commented in the makefile._c (tested and working). Note, that starting with v2.02 you can use DLLs in your Windows apps. Some compilers do allow to use DLLs in DOS Protected Mode, not only in Windows, so this might be a way for you to include TMi0SDGL features in your DOS PM apps. b) There are several limitations you should know of when writing a protected- mode program in Borland Pascal that uses TMi0SDGL: 1. If no memory manager is installed on the system where the program is ran (except maybe HIMEM.SYS) it is likely that you will get Runtime Error 216 (General Protection Fault Exception) on 386 and up. This is due to Borland DPMI host(RTM)'s implementation of V86 monitor: it does not allow to write to special control registers of 386+. Most advanced memory managers like Quarterdeck's QEMM or Qualitas's 386Max do allow this operation under certain conditions (they check which flags are subject to be changed) and the TMi0SDGL code works fine under them. Regrettably, it is too complicated or even impossible to get to PL0 if there's no VCPI-compliant memory manager installed and the system is in protected mode, so I do not know how to workaround this problem to this date. Under VCPI though, it is possible to get to privilege level 0 where all special instructions are allowed and are executed with no exceptions. Known VCPI services providers are QEMM, Microsoft's EMM386 and some other memory managers. ..4 2. It is not possible to detect IBM 386SLC, 486SLC, 486SLC2 and 486BL3 CPUs in protected mode. The detection method for these CPUs involves RDMSR instruction which may generate either Exception 6 (Invalid Opcode) or Exception 13 (General Protection Violation) if the CPU is not one of above. These exceptions must be trapped but most V86 monitors do not allow to do it. Thus detection of these CPUs is limited only to real mode or protected mode with DPMI-compliant memory manager (with some ill-behaved DPMI hosts this won't work too). c) Assembler code relies on automatic jump optimization provided in TASM via JUMPS directive. If you use assembler that do not support JUMPS, you will probably get some "jump is out of range" errors during compilation and will have to fix those jumps manually (replace Jcc location JCXZ location with with JNcc temploc and TEST CX,CX JMP location JNZ temploc temploc: JMP location temploc: - this is done automatically by TASM with JUMPS enabled, really useful feature). 3.2 Known Problems And Incompatibilities. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ a) Some of detected processors require CPU to be in real mode or a DPMI- compliant memory manager installed for correct detection. Namely, these are IBM's 386SLC, 486SLC, 486SLC2 and 486BL3 chips. These chips contain MSRs (Machine-Specific Registers) that are accessible via RDMSR and WRMSR instructions, officially introduced in Intel Pentium processor. These instructions can only be executed in real mode or on CPL0. Another issue with these instructions is that when they are given an invalid register index, they raise Exception 13 which should be intercepted again in real mode or on CPL0 to be properly handled. b) Under some memory managers (namely, Soft-ICE 2.6x from Nu-Mega Technologies and Borland's DPMI16BI.OVL from their DOS Extender) the 386SX and 386DX distinguishing do not work properly. This is due to limitations in V86 monitor design of these managers: they do not allow free access to CR0 register even if no dangerous flags are affected. Again, this problem goes away only in real mode or under a VCPI-compliant memory manager. Also, this problem do not appear under Quarterdeck's QEMM386 memory manager - it handles CR0 access much more flexibly and allows to change some flags in it even from V86 task (presumably to allow Manifest to distinguish SX from DX :). I tried to workaround this problem in protected mode code using DPMI's exception interception mechanism but some DPMI hosts may not properly reflect these exceptions to my handlers anyway. c) I have no information on how to distinguish different models of Weitek numeric coprocessors, thus all models are reported simply as Weitek. If anyone know reliable method for distinguishing Weiteks, please let me know. d) Lot of problems appear in OS/2 DOS Box. Some of them: 1. You have to set HW_TIMER=ON in DOS Box settings for correct clock speed calculation, otherwise it gets divided by 2 (well, the OS name says it all :). 2. OS/2's V86 monitor is very rough and don't allow to play with protected instructions and registers. Almost nothing can be done about it - perhaps I should write an OS/2 protected mode driver to do all the dirty work... Or maybe somebody with good experience in writing OS/2 drivers will do it for me - me myself is not experienced in this at all, I never tried to write an OS/2 program yet. If someone is willing to contribute to TMi0SDGL this way - please contact me and we'll discuss details. ..5 e) CPUID-related code assumes that every CPU that support it implements CPUID at least to level 1 (thus software is able to get processor feature flags and Family/Model/Step info). As far as I know this is true for current Cyrix, AMD, UMC, NexGen and Centaur processors, BUT this may be false for other upcoming CPUs, resulting in senceless info retrieved from such CPU. f) Because Intel 487SX is actually a 486DX designed for coprocessor socket, it is impossible to distinguish 487SX from 486DX unless this is a new chip with CPUID support (marked & on chip), thus these older versions are reported as "486DX or 487SX". It is also impossible to distinguish older 486DX2 chips - they are also detected as 486DX or 487SX. g) Windows NT is currently not supported with 32-bit DLL. This is because the OS applies its strict I/O protection to all nonprivileged apps, and TMi0SDGL's DLL is such, causing access violations when trying to access certain I/O ports, which is crucial for detection of some CPUs, like Cyrix ones. I suggest I will have to write a device driver for NT that will execute at ring 0 or 1 and thus will have access to needed ports. But I do not have the DDK to implement one... For now, the DLL detects if it is ran under Windows NT and ceases to function if so. Windows 95 version of the DLL does not perform any checks though and will almost surely crash under Windows NT. h) WATCOM C/C++ development suite is currently not supported. The code will not be linked with WATCOM C/C++. I'll try to add WATCOM-specific defines or source files in subsequent releases. Until then - sorry... i) Some systems equipped with Intel HX chipset and running Windows 95 with EMM386 active cause TSC-based CPU speed detection to crash with GPE on RDTSC instruction. Reason for this behavior is being investigated. For now it is known that the source of the problem is EMM386 and that CR4.TSD bit does not affect execution (even if it is cleared, the GPE still occurs). The only workaround to date is to disable EMM386 (who needs it anyway..?) j) Tomas Navratil reported that Celeron systems with some particular mother- boards caused TSC-based speed detection to hang. Problem is being investigated. Reported troublemaker is Acer ALI M1621 chipset. After changing to VIA Apollo PRO the problem gone. Also, TSC code runs ok on other VIA chipsets as reported. k) Seems that AMD implemented RDTSC differently from Intel's specs: executing it from V86 *always* generates GPE not honoring CR4.TSD state while Intel specifies that CR4.TSD controls GPE generation on RDTSC in *all* processor modes (real, V86, protected). Also, some people reported that this behavior is also exposed by newer Intel processors. Check for CR4.TSD state is added but may cause problems itself because reading control registers is a privileged instruction... ..6 4. Library Reference. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ This chapter gives summary on library functions and variables. 4.1 Function Reference. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ cpu_Type ~~~~~~~~ prototype: Pascal C/C++ Static: function cpu_Type : String; char * cpu_Type(); DLL: function cpu_Type : PChar; char * cpu_Type(); Purpose: returns verbose name of current central processor, including vendor name and model if determined. fpu_Type ~~~~~~~~ prototype: Pascal C/C++ Static: function fpu_Type : String; char * fpu_Type(); DLL: function fpu_Type : PChar; char * fpu_Type(); Purpose: returns verbose name of current numeric extension unit, including vendor name and model if determined. cpu_Speed ~~~~~~~~~ prototype: Pascal C/C++ Static: function cpu_Speed : Word; unsigned int cpu_Speed(); DLL: function cpu_Speed : Integer; int cpu_Speed(); Purpose: returns integer value of raw current processor clock speed in MHz. ncpu_Speed ~~~~~~~~~ prototype: Pascal C/C++ Static: function ncpu_Speed : Word; unsigned int ncpu_Speed(); DLL: function ncpu_Speed : Integer; int ncpu_Speed(); Purpose: returns integer value of normalized processor clock speed in MHz. Speed normalization uses internal tables of known official processor clock speeds and selects value closest to raw speed. Since new clock speeds are frequently announced by chip vendors, speed returned by this function may be inappropriate (too low) for new high-frequency chips. Note: this function does NOT return PR-rating for non-Intel chips, but rather its real (or close to real) speed. fcpu_Speed ~~~~~~~~~~ prototype: Pascal C/C++ Static: function fcpu_Speed : Real; float fcpu_Speed(); DLL: function fcpu_Speed : Double; double fcpu_Speed(); Purpose: returns raw floating point value of current processor clock speed in MHz. getCacheSize ~~~~~~~~~~~~ prototype: Pascal C/C++ Static: function getCacheSize : Word; unsigned int getCacheSize(); DLL: not available not available Purpose: returns integer size of L1 data cache in KBs (not reliable). CxCPUIDEnable ~~~~~~~~~~~~~ prototype: Pascal C/C++ Static: procedure CxCPUIDEnable; void CxCPUIDEnable(); DLL: the same the same Purpose: used internally. Enables CPUID instruction on Cyrix 5x86 and 6x86 CPUs by setting CPUIDEN bit in CCR4 internal register. getCPUID ~~~~~~~~ prototype: Pascal C/C++ Static: procedure getCPUID(Level : LongInt, void getCPUID(long Level, Result: Pointer); void * Result); DLL: the same the same Purpose: returns CPUID response for given Level in structure pointed to by Result. Layout of response structure can be found in interface sections of corresponding source files. getVersion ~~~~~~~~~~ prototype: Pascal C/C++ Static: function getVersion : Word; word getVersion( void ); DLL: the same the same Purpose: returns BCD-encoded library version, major version in high byte, and minor version in low byte. For example, version 2.15 will return 0x0215 ($0215 in Pascal notation), but not 0x020F. 4.2 Variable Reference. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ cpu ~~~ Size: Byte; Purpose: contains numerical ID of current processor, of 0xFF if processor was not yet detected. Note: available from getCPU function in DLL. fpu ~~~ Size: Byte; Purpose: contains numerical ID of current NPX, of 0xFF if NPX was not yet detected. Note: available from getFPU function in DLL. extFlags ~~~~~~~~ Size: Word; Purpose: contains extra efXXX bit flags for current processor. For values and meaning of efXXX flags refer to interface sections of corresponding source files. Note: available from getExtraFlags function in DLL. cpuid0 ~~~~~~ Size: 12 bytes; Purpose: contains vendor ID string returned by CPUID instruction, or all zeros if CPUID is not supported by current processor. Note: available as ASCIIZ string from getVendorString function in DLL. Known vendor strings are: "GenuineIntel" - Intel "NexGenDriven" - NexGen (now part of AMD, doubtly can be seen) "CyrixInstead" - Cyrix "AuthenticAMD" - AMD ("AMD ISBETTER" in engineering samples) "UMC UMC UMC " - UMC "CentaurHauls" - Centaur/IDT Tech. "RiseRiseRise" - Rise Technology cpuid1 ~~~~~~ Size: Double Word (4 Bytes); Purpose: contains CPU family/model/stepping info as returned in EAX by CPUID instruction. Note: available from getSteppingInfo function in DLL. This value is bitmapped as follows (with 31 being MSB and 0 - LSB): [31 . . . 27 . . . 23 . . . 19 . . . 15 . . . 11 . . . 7 . . . 3 . . 0] [reserved ] D O [family=4,5,6,..] [ chip model ] [chip stepping] O (bit 24)= OverDrive processor D (bit 25)= Dual-capable processor (secondary CPU on a dual-processor system) (both bits cannot be set at the same time - reserved combination) Refer to Intel's application note AP-485 for detailed description of bitfields. cpuFeatures ~~~~~~~~~~~ Size: Double Word (4 Bytes, or 32 bitfields); Purpose: contains CPU feature flags as returned in EDX by CPUID instruction. Note: available from getCPUFeatures function in DLL. For bitfields descriptions refer to P5InfoX.PAS or .C, or ask Intel and others :) ..7 4.3 Tips On Using DLLs. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Two DLLs (Windows Dynamic Link Libraries) are provided with this release of TMi0SDGL to allow Windows/DOS Protected Mode programmers to use the library with minimum effort required from them. Both 16-bit and 32-bit DLLs export the same functions, though I provided both 16-bit and 32-bit header files and import units for convenience. For descriptions of exported functions, refer to paragraphs 4.1 and and 4.2 of this manual. These DLLs can be freely included with your applications with no royalty paid to the author (as with other parts of the library). You can even skip "Portions copyright(c) B-coolWare" notice in your product's About box if you use these DLLs, because they have the copyright information within, though I prefer you to add this notice anyway (just to honor my pride, y'know :). * How to use these DLLs in Delphi/Pascal. Simply add CPUDetXX (where XX is 16 or 32 for 16- and 32-bit version respectively) into USES clause of your application. That should do it. See CPU16D.DPR and CPU32D.DPR for examples. Note that there's no wrapper classes/objects for Delphi/Pascal - it's trivial to implement them yourself if you need them. * How to use these DLLs in C/C++. First thing to check before you can try the DLL, is the format of import library. C/C++ programmers should keep in mind that Borland and Microsoft (and some other vendors) use different object file formats, Intel OMF and COFF respectively, which are not compatible. Microsoft linker automatically converts standalone .OBJ object files from OMF to COFF, but this is not true for .LIB library files. Thus, giving an OMF library to Microsoft linker will produce error on invalid object file. To solve this problem, you must cast the IMPLIB spell using IMPLIB.EXE provided with your compiler to generate proper import library. The magic words are IMPLIB .LIB .DLL This will produce libname.LIB static import library from dllname.DLL dynamic link library. Then you can replace import libraries I provided, which are in Intel OMF format, with the ones you created yourself. After that, refer to CPU16C.CPP or CPU32C.CPP for example on using these DLLs. I also provided two .HPPs with useful classes that will simplify your life even further. * How to use these DLLs in Visual Basic. Add CPUDET32.BAS module to your project and put CPUDET32.DLL in Windows\System directory or in directory where your application is - that should do it. * Other development systems. Other development systems, like PowerBuilder or Visual FoxPro, are not currently supported. But if you are experienced in these devsuits, you can create import modules for them yourself, based on supplied import modules in Pascal and C. If you do, please send me a copy of your modules so that I'll be able to incorporate them in next releases. Your copyright will be reserved. ..8 5. Credits. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Too-Much-in-0ne-So-Don't-Get-Lost(tm) Revision 2 CPU/FPU Detection Library is copyright(c) 1996-2000 by B-coolWare. Accompanying documentation is also copyright(c) 1996-2000 by B-coolWare. Portions copyright(c) 1997 by Vladimir S. Lokhov Portions copyright(c) 1995 by Lautenaus Portions copyright(c) 1992 by 2B Programmers Portions copyright(c) 1994-99 by Christian Ludloff Portions copyright(c) 1995 by NexGen Corporation Portions donated to public domain by Compaq Computer Corporation Portions copyright(c) 1995 by Cyrix Corporation Portions copyright(c) 1996-98 by Intel Corporation Runtime libraries are copyright(c) by their respective vendors. Portions I forgot to mention are copyright(c) by those I don't remember. (this last joke is copyright(c) 1996 by Bobby Z. :) This is free copyrighted software! Use it free, give away to your friends and colleagues, upload to your favorite BBSes and public FTP sites, but do not sell or lease it! Please read the license agreement for details. ...huh...Hope this will keep the lawyers happy... 5.1 Hall Of Fame ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Thanks goes to Almighty God for inspiration and everything. In the memory of my father Michael M. Zakharychev (1943-1997) who died from cancer. Please read CHARITY.TXT. This is a short list of contributors to this library, either with code or with information. I cannot mention all people who took part in development of this library for it will take too much space, only the notest of them. 1. Christian Ludloff aka ludloff@sandpile.org This guy is dedicated to x86 family research and maintains the Sandpile site, the most comprehensive set of information regarding undocumented and advanced features of x86 family of processors. I retrieved lots of information that was used to produce TMi0SDGL from Sandpile (and still do it!) Check it out at http://www.sandpile.org. 2. Vladimir A. Prunsky aka VAPorizer He was my boss and I'd like to thank him for his patience while I was writing this library in its early days instead of sticking to my primary job tasks. 3. Sergey Frolov aka Lautenaus aka 2:5030/219.9@fidonet who provided me with lots of useful code regarding Cyrix family of processors and did much testing. 4. Aleksandr Konosevich aka 2:5004/9.7@fidonet who is the Cyrix guru. He knows almost everything about these chips and he didn't charge me a penny for his knowledge! 5. Vadim Selivanow aka 2:5015/17@fidonet who provided a lot of NexGen-related information and code and also did much testing. 6. Alex Bachin aka bachin@garant.ru who provided i386SL detection code and 80C88/C86 detection code and kindly agreed to be TMi0SDGL beta-tester. 7. Dmitry Orloff aka do@kpi.kharkov.ua aka 2:461/27.1@fidonet who helped me a lot in my great hunt for bugs, especially with NEC Vx0 related ones (yeah, he still have access to a PC based on it...!). 8. Alexander Alferowich aka 2:5031/7.3@fidonet who really helped me alot with hunting bugs and discrepancies found in final v2.00 release which was widely spread thru FidoNet network. Geez, I didn't think I missed THAT much of 'em... 9. Michael Radchenko aka 2:2025.25@fidonet who helped me with adjusting new AMD K5 processor speed constant. 10. Vladimir Lokhov aka vladimir@lokhov.lug.tsu.tula.ru aka 2:5022/18.14@fidonet who suggested another method of detecting CPU speed and provided some source code, which was gladly accepted. 11. Vasiliy Sorokin aka vasiliys@hotmail.com who reminded me of my old AMD detection code and checked it. 12. Phil Randal aka philr@hwcces.demon.co.uk BIG thanks for your very valuable comments on TMi0SDGL (and for magic values for AMDs and P6)! Please come back anytime with more. :) 13. Leave devil's for himself... :) 14. Tomas Navratil aka navsoft@post.cz who found and corrected numerous discrepancies in v2.03 and also reported some bugs I missed in v2.06 and v2.08. Still he finds more! 15. Jeff Miko for his invaluable comments and bug reports. Also among contributors are Igor Dral 2:50/321@fidonet Alex B. Shamarokov 2:5020/205.8@fidonet Janis Smits 2:5100/25@fidonet Ilya Tumanov 2:5030/82.6@fidonet Igor Krasikov 2:463/59@fidonet Roman Gasteff 2:5020/177.8@fidonet Levan Natroshvili 2:446/1@fidonet Vsevolod Fedotov 2:5005/2@fidonet Dmitry Dmitrienko 2:5020/375.8@fidonet Thank you: All the participants of v2.00 beta testing - your feedback helped me a lot with cleaning this code of bugs. And, of course, all of you, the users of TMi0SDGL, who wrote me back and reported my faults or suggested some improvements or simply made use of this product - thank you. Without you, this library would never be what it is. ..9 6. How To Contact The Author. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ In case you have spotted a bug or have some suggestions or code to donate, please use form in file BUGREP.TXT (you can also submit bug reports from the Web) and following addresses: e-mail: bobby_z@cyberdude.com (primary, hopefully won't change) zak@dzbjaro.bertelsmann.de (my current business e-mail) Web: http://come.to/b-coolware (redirector to actual site) Surface mail: Vladimir M. Zakharychev 150031, 10-4-13, Dobrynina Str. Yaroslavl, Russia Letters both in english and in russian are welcome. Please do not use any other language if you want to be answered. Letter-bombs will be returned to sender (shhh-boom-BANG!). When writing in Russian, please use Windows CP1251 or KOI-8R charset. 6.1 Author's reward. ~~~~~~~~~~~~~~~~~~~~ If you made use of this code and think it's worth something, please send me a postcard with picture of country you live in or picture of yourself to address above. OR purchase a gift certificate for me on Music Boulevard (http://www.musicblvd.com) where I shop sometimes. It can be of any amount, but not less that US $10 (that's Music Boulevard restriction, not mine). Just drop in to the site, register and buy a gift certificate for Vladimir Zakharychev with e-mail of bobby_z@cyberdude.com for any amount you wish and I will redeem it someday. OR you may just drop me an e-mail with your opinion on this library. Anyway, my best reward is the fact that my work didn't go in vain and is used by people around the globe (hopefully with gratitude). ---- Appendix A. List Of Files In Original Distribution Archive. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ The original distribution SFX archive CPU215.EXE should contain (sorted in several subdirectories): .\ - root TMi0SDGL directory README. Read me file. README.1ST Important information that requires to be read first. README.COM README display engine. freeware by Bobby Z. FILE_ID.DIZ BBS archive description file - for easy archive handling. VMZ.ASC Author's PGP keys. (also at ldap://certserver.pgp.com) TMi0SDGL.REG Windows Registry file with version information for UpdateMe! .\SOURCE\ - source code and makefiles MAKEFILE._P Pascal makefile. MAKEFILE._C C makefile. MAKEFILE._CP C++ makefile. CPUTYPE.C Main library module - C version. CPUTYPE.PAS Main library module - Pascal version. REALCODE.ASM Real mode low-level code. DPMICODE.ASM Protected mode DPMI-compliant low-level code. CPUSPEED.ASM CPU clock speed calculation code. CPUSPD2.ASM Alternate CPU clock speed calculation code. CACHETST.ASM L1 cache size detection code. CPU.C Demonstration program - C version. CPUCPP.CPP Demonstration program - C++ version. CPU.PAS Demonstration program - Pascal version. P5INFO.C Demonstration program - C version. P5INFO.PAS Demonstration program - Pascal version. CPUDET*.PAS DLL import units. CPUDET32.BAS Visual Basic 5.x 32-bit DLL import module. MAKEALL.BAT Quick make batch - makes all examples. .\SOURCE\INCLUDE\ - include files TMi0SDGL.H C header file. TMi0SDGL.HPP C++ header file. TMi0SDGL.INC Assembler include file. HEADER.ASH Common assembler header file. DPMI.ASH Assembler header file for DPMI-compliant code. CPUDET*.H* C/C++ DLL import header files. .\EXE\ - sample executables CPU*.COM compiled sample programs CPU*.EXE compiled sample programs P5INFO*.EXE compiled sample programs GETDIRS.EXE run it and mail results to help me add unlisted/undetected Cyrix-like CPUs to library. Only do this if TMi0SDGL is unable to detect your CPU correctly and names it a Cyrix/TI/SGS CPU. .\LIB\ - compiled libraries CPUTYPE*.LIB compiled C/C++ static libraries CPUTYPE.TP? compiled Borland Pascal 7 units CPUDET*.D?? 16- and 32-bit dynamic link libraries CPUDET*.LIB 16- and 32-bit DLL import libraries (OMF object format). CPUDET*.DCU Delphi import units. CPUDET*.TP? Borland Pascal import units, Windows and DPMI. .\DOC\ - documentation files TMi0SDGL.DOC This file. The TMi0SDGL documentation and license. BUGREP.TXT Bug report/suggestion form. Please use it if you found a bug or have something to suggest about TMi0SDGL. WHATSNEW.215 What's New document for current TMi0SDGL release. HELPME.PLS A call for help - please read this file! CHARITY.TXT Simply read this file... VDBASE55.TXT Tips on using TMi0SDGL with Visual dBase 5.5a TODO.TXT To Do List. This archive may also contain other files, like BBS or group addies. Simply note that these files, if present, are not from B-coolWare and can be deleted. DLL source code is available in separate archive - CPUSP215.EXE. Appendix B. Useful Resources On Internet/WWW. ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ http://come.to/b-coolware This is a redirector to my home page, where you can always obtain the latest version of the TMi0SDGL (including unpublished alphas and betas), as well as some other of my programming stuff. As I'm not sure about future location of my site, it is the safest way to use this redirector to find my homepage wherever it will go in the future. http://www.x86.org R. Collins' Intel Secrets(tm) (what intel doesn't want you to know...) Page. Vast of information about x86 processor family, bugs, undocumented features, source code and shareware/freeware tools for professionals and dummies. Just drop in and see it all yourself. A "must visit" site if you're interested in details on x86 family of CPUs. http://www.sandpile.org Christian Ludloff is back on the net and with lots of new information well worth opening a new site, sandpile.org. Another "must visit" site, even more "must" than Intel Secrets! No comparisons, no affiliations with any chips vendors, no "my stuff is better than others'" crap - only plain technical data, and that's what makes this site cool. http://www.pobox.com/~ralf Ralf Brown's Home page. Ralf Brown is most known as compiler and publisher of the Interrupt List, the biggest and the most precise reference to IBM PC Interrupts. IntList also includes lots of information on CPUs in file OPCODES.LST, donated to IntList by Potemkin's Hackers Group. http://www.intel.com Guess who are these... :) http://www.amd.com ...and these... http://www.cyrix.com ...and these... http://www.???.com - fill in ??? yourself, most vendors now appear on the Web with their commercial blah-blah they call "customer services". Though I suggest you'll have to squeeze the REAL information out of them, drop by drop, slowly increasing pressure up to ten, for they often call it trade secrets or propietary info or kinda like that and won't give it away... When you give up, acquire the 4P and IntList and you'll find there the most of it, all in one piece, from independent researchers that got nothing to hide (almost) :). << end of TMi0SDGL.DOC >>