AGA Toolkit '97/The AGA Toolkit '97.iso/c/fastexec.doc

Text File | 1996-09-07 | 17KB | 359 lines

                                                                26-Feb-1996
Documentation for

                        ==========================
                        ====  FastExec 39.16  ====
                        ==========================
                Copyright © 1994-1996 PitPlane Productions

             IMPORTANT! Please read copyrights, disclaimer and
                license parts of this document before use.



    *** What is FastExec? ***

    o FastExec is a reset proof patch to move Execbase and supervisor stack
      from chipram into 32bit fastram on Blizzard 12x0 based A1200s, or
      other turbo card based systems having the same problem. FastExec
      requires Kickstart 3.0 or better.


    *** Why FastExec? ***

    o With Blizzard turbos (at least 1230-III) fast memory is added to
      system memory pool after execbase is initialized. This leads to the
      fact that execbase and system supervisor stack are stuffed into chip
      memory! As you may have heard, chipram is slow compared against
      fastram, in fact fastram can be more than ten times faster! It's a
      pity that so vital parts of system get into chip memory, as there
      usually is loads of lighting speed fastram available. FastExec is
      program to fix this problem.

    o There is another program called FastExec (author Torbjörn Andersson)
      but it isn't as good as mine. It's over 4K long. It uses HACKS. It
      doesn't survive resets. It's ShareWare. First version of my FastExec
      was (internally) released at the end of year 1994, so it's an
      independent product.

    o Taken from the document of Andersson's FastExec:

     "FastExec  is  not programmed in a recommended way, because I
      don't know any clean way to solve the problem.  FastExec may
      not work on your system."

      As a side effect Andersson's FastExec disturbs AutoConfig system.
      Just try to examine expansion boards in your system after using it.
      My machine claimed that there weren't any present!? I think there
      isn't any good reason to move expansion.library because it causes
      these severe troubles.

      Well, I know the clean way to move execbase and I implemented it.


    o This tool is FREEWARE!! (see license)
    o FastExec is 100% assembler!
    o FastExec executable is less than 1024 bytes short!
    o Uses only 424 bytes of memory! (Yes, that's true!!)
    o Doesn't conflict with bootmenu! (Andersson's FastExec does!)
    o Can be used to fix similar troubles with other turbos too!
    o I dunno about the speedup, as there aren't many multitasking
      benchmark programs around. Should be worth of it, though.
    o FastExec should be compatible with all future kickstarts.
    o Enforcer hit free (tested)


    *** Notes ***

    o ATTENTION kickstart release 2 users!!
      Would you like to see FastExec running on your system? If you would,
      contact me (author) and ask. This could be _very_ easily implemented,
      but I really didn't see any reason to support this anchient kickstart
      release... ;-) The fact is that most of the powerful systems (that
      need FastExec) run with kickstart 3.0 or better. (Except maybe some
      upgraded A3000s?) Anyway, if you need FastExec for KS 2.x contact me.

    o FastExec has been tested on following system:

      A1200 10/540 MB, Blizzard 1230-III 68030 @ 50MHz
      A1200T 10/820 MB, Blizzard 1260 68060 @ 50MHz

      FastExec works (should work!) with all other turbo cards having
      similar troubles. FastExec doesn't rely on Blizzard products, but
      others aren't tested. If you're having troubles with FastExec on
      your system and you're sure that trouble is hardware (or maybe even
      software) related, contact me. I'll (try to) fix it. I cannot
      guarantine FastExec to be 100% software/hardware compatible!

    o Note that FastExec can *not* move execbase to fast memory which is
      added with AddMem or similar tool in startup-sequence. Credit Card
      (PCMCIA) memory cannot be used, but it's even slower than chipram
      anyway!

    o Some of virus killers and checkers may report that system's kickmem
      and kicktag pointers are in use. That really is the case, but you
      should calmly ignore these warnings. (Most virus killers/checkers
      will show the name of the kicktag, in this case it is 'Fast_Exec'.)

    o FastExec doesn't use Cold- or CoolCapture. This means it'll work fine
      with, for example, ShapeShifter's PrepareEmul tool, in any
      circumstances.

    o FastExec should be compatible with all future Kickstarts. Kickstart
      3.0 is currently the only tested one. If FastExec refuses to
      co-operate with KS 3.1 or maybe with 4.0 beta(?) then contact me.

    o To contact the author (for any reason?) try this:

      Harry Sintonen (Delta)          (Gifts, cli-scripts, ASM Sources
      Kotipolku 30                     etc. are welcome. No illegal trade.)
      48710  Karhula
      FINLAND


    *** Technique ***

    o If you're not interested how this program works you should skip this
      part of the document. Maybe you should skip it anyway, because first
      of all, this part is quite technical and secondly: I'm not very good
      in explaining complicated things in english as this. (Forgive me!!
      English isn't my native language after all. ;-)

    o The very early rom startup hasn't changed much since Kickstart
      release 1.2. All the internals, including hardware initialization,
      execbase and internal variable building are almost identical in all
      kickstart versions. Of course, in newer kickstarts things are done
      much better than in earlier versions, and some structures has been
      expanded, but the main idea is the same.

      When the system has been initialized, temporary execbase has been
      created. Then the system will search ROM memory for resident modules
      and create ResModules list. If any KickTags are defined they'll
      be inserted (or they'll replace existing modules) to ResModules list.
      After this, all the resident modules will be initialized. This really
      gets things running. The ResModules list will be initialized in
      priority order, that is, expansion.library will be run (read:
      initialized) first.

      Expansion.library initialises all the expansions in system and for
      example, memory of SupraRAM 500RX expansion for Amiga 500 will be
      added to system's MemList. Apparently the memory of Blizzard 12x0
      Turbo Board is NOT added in this phase. All memory above 16 megabyte
      area (8MB autoconfig memory area) will be initialized afterwards. The
      task is done by resident called 'diag init'. Its priority is 105
      while exec.library's is 104.

      Exec.library will be the next resident module to initialize. It tries
      to allocate fast memory for execbase, but if it doesn't succeed it
      will use the temporary execbase, which is always in chipram
      (actually, it can be in pseudo fast memory). If it does get the
      memory then the temporary execbase is cloned to new one and the
      memory taken by temporary execbase is free'ed. As the memory of
      Blizzard Turbo Board isn't available, the execbase and supervisor
      stackpointer will be in chipram (The memory is added to system
      afterwards in 'diag init' resident).

      The trick is, that we'll allocate known fastram when the system is
      running with that memory. Then we'll relocate execbase and supervisor
      stack to this newly allocated memory array. The hard part is that
      execbase can NOT be cloned when it is available in multitasking
      enviroiment. Programs might have got execbase pointer from address 4
      and use it happily, when we would free the memory used by that
      execbase and use the new, totally different memory array as new
      execbase. This would cause a total crash. The solution is to clone
      execbase early enough, BUT after fastram is added to system memory
      pool, which means use of kicktag pointer at priority 104. (Just after
      'diag init' resident!)

      Now our kicktag code is called after the fastram is available so we
      can simply AllocMem() the execbase and stack area, clone execbase,
      create new supervisor stack, set new execbase pointer to address 4,
      set new supervisor stack pointer, free old execbase and stack memory,
      and viola!, things are running fine.

      The memory of kicktag code (code handling all the above) is
      allocated with KickMemPtr service of execbase. As this memory is
      allocated by execbase initialization, it must exists in MemList at
      that time (this code is in chipram with A1200). Now, all the memory
      has been allocated and the system is running from fastram.

      This technique will work with future kickstarts. The reason is that
      most of the execbase's structures aren't initialized at the moment
      the cloning is done. Actually the only potential problem is that
      system will crash if some program gets execbase pointer before
      FastExec and uses it afterwards.


      Now, that was a cutted version of _The Technique_!! [grin]


    *** FastExec's Usage  ***

    o FastExec's template is

      FORCE24BIT/S, RESET=REBOOT/S

      FORCE24BIT/S              With this switch you can force FastExec to
                                accept also fast memory below 24-bit
                                address range. Normally only fast memory
                                above 16MB (24-bit) is accepted. Usually
                                this switch isn't needed, I added it just
                                for compability reasons.

      RESET=REBOOT/S            If this switch is specified FastExec will
                                reboot the machine immeditially after
                                FastExec is installed to memory. BEWARE!!
                                You will lose all the information in ram!
                                Of course, no reboot is done if execbase is
                                already in fastram.


    o RC (return code) is set:
      OK (0)                    FastExec installed ok.
      WARN (5)                  FastExec isn't needed. Exec is already in
                                fastram. This is also the result if
                                FastExec has already moved exec. Of course
                                RESET=REBOOT/S has no effect if execbase is
                                already moved!
      ERROR (10)                Things went wrong. Couldn't get memory?
                                This is the error you get if you do not
                                have 'real' fastram at all (see FORCE24BIT
                                switch).
      FAIL  (20)                Something fatal happened! Dos didn't open?


    o If you aren't aware of the fact whether you really need FastExec or
      not (YES, you don't necessarily need this program! :) and you want to
      install FastExec easily you should follow these instructions:

      Write in CLI or Shell:

      1.> FastExec
      1.> Set
      process           1
      RC                5
      Result2           0
      1.>

      If you got result as above (RC=5) you do not need FastExec as
      execbase is already in fastram. If you got RC=10 then you do not have
      enough memory (or you don't have 'real' 32-bit fastram [See
      FORCE24BIT switch]).

      process           1
      RC                0
      Result2           0

      If you got Set's output as above (RC=0), you NEED FastExec. It will
      now activate after next reboot, but if you want to make FastExec's
      effect permanent, you need to copy FastExec executable into your C:
      drawer (if not yet done) and insert the following line to the
      beginning of your S:StartUp-Sequence file:

      C:FastExec REBOOT

      Now, after every coldreboot (total crash, powering up) your machine
      will boot twice, but that isn't too disturbing, I think!

      You can leave REBOOT switch out, the only difference is that execbase
      will not be moved into fastram first time the system is powered up
      (or rebooted after fatal crash), as the system will not be rebooted
      by FastExec. This makes more sense if you have several similary
      acting reset proof patches in your startup. Because of this feature
      you can install, for example, following lines to your
      startup-sequence:

      C:FastExec                ; FastExec will not reboot machine,
      C:PrepareEmul             ; but PrepareEmul will!

      Now, your machine will be extra rebooted only once, instead of two
      reboots. (I really hate triple reboots!! ;^)


                                Simple, ha?

                   And remember... Enjoy the FAST Exec!



    *** COPYRIGHTS ***

      All Amiga technology is Copyright © Amiga Technologies GmbH.

      Blizzard Turbo Boards are © 1994 phase 5 digital products. All rights
      reserved.

      ShapeShifter is © 1993-1995 Christian Bauer. PrepareEmul is a part of
      the ShapeShifter package.

      SupraRAM is © 1991 Supra Corporation. SupraRAM 500RX is a trademark
      of Supra Corporation.

      FastExec is Copyright © 1994-1996 PitPlane Productions.


    *** LICENSE ***

1.   This license applies to the product called "FastExec", a program for
     the Amiga computer, published by PitPlane Productions under the
     concepts of FreeWare, and the accompanying documentation. The terms
     "Program" and "FastExec" below, refer to this product. The licensee is
     addressed as "you".

2.   You may copy and distribute verbatim copies of the program's
     executable code and documentation as you receive it, in any medium,
     provided that you conspicuously and appropriately publish only the
     original, unmodified program, with all copyright notices and
     disclaimers of warranty intact and including all the accompanying
     documentation, example files and anything else that came with the
     original.

3.   You may not copy and/or distribute this program without the
     accompanying documentation and other additional files that came with
     the original. You may not copy and/or distribute modified versions of
     this program.

4.   You may not copy, modify, sublicense, distribute or transfer the
     program except as expressly provided under this license. Any attempt
     otherwise to copy, modify, sublicense, distribute or transfer the
     program is void, and will automatically terminate your rights to use
     the program under this license. However, parties who have received
     copies, or rights to use copies, from you under this license will not
     have their licenses terminated so long as such parties remain in full
     compliance.

5.   By copying, distributing and/or using the program you indicate your
     acceptance of this license to do so, and all its terms and conditions.

6.   Each time you redistribute the program, the recipient automatically
     receives a license from the original licensor to copy, distribute
     and/or use the program subject to these terms and conditions. You may
     not impose any further restrictions on the recipients' exercise of the
     rights granted herein.

7.   You may not disassemble, decompile, re-source or otherwise reverse
     engineer the program.

8.   If you wish to incorporate parts of the program into other programs,
     write to the author to ask for permission.

9.   You agree to cease distributing the program and data involved if
     requested to do so by the author.

10.  You may charge a fee to recover distribution costs. The fee for
     diskette distribution may not be more than the cost to obtain a public
     domain diskette from Fred Fish.


    *** DISCLAIMER ***

     NO GUARANTEE IS GIVEN NOR IMPLIED THAT THIS PROGRAM IS FIT FOR ANY
     USE. THIS PROGRAM AND THE INFORMATION WITHIN THIS TEXT IS PROVIDED ON
     "AS IS" BASIS. THE ENTIRE RISK AS TO ITS QUALITY AND PERFORMANCE IS
     WITH THE USER. SHOULD THE PROGRAM PROVE DEFECTIVE, THE USER (AND NOT
     THE AUTHOR) ASSUMES THE ENTIRE COST OF ALL NECCESSARY DAMAGES. IN NO
     EVENT WILL THE AUTHOR BE LIABLE FOR DIRECT, INDIRECT, INCIDENTAL OR
     CONSEQUENTIAL DAMAGES RESULTING FROM ANY DEFECT IN THE PROGRAM. THE
     AUTHOR RESERVES THE RIGHT TO MAKE CHANGES TO THE SOFTWARE OR
     DOCUMENTATION WITHOUT NOTICE.