› HISTORY› -------›› Version 2.03› ------------›› - Hardware interrupts info extended›   according to Western Design Center›   documentation.›› - Bugfix: benchmarks now work›   correctly on 6502 machines.›› - Bugfix: no more messages like›   "System error 1.0".›› - Bugfix: no more troubles with›   screen buffers after a system›   error while capturing a snapshot›   while a menu is open (uh!). Both›   bugs found by WSZ.››› Version 2.02› ------------›› Known bugs: it does NOT detect› te 65c816 high RAM (past the› $FFFF).›› Look below to read discussion of› the screen drawing speed.›› - Code optimizations have saved›   178 bytes.›› - The background picture has›   been packed better (with the›   same algorithm, RLE); 822›   bytes saved.›› - 256 bytes less are allocated›   from the free RAM.›› - Now you can move to another menu›   closing the previous one›   automatically. Use left/right›   arrow. Other windowing routines›   are also changed, but you will›   not see any difference.›› - Benchmark progress indicator›   added.›› - "Quit" has moved from the menu›   bar to the "Desk" menu.›› - Snapshot added. Now you can›   capture the SI screen at any›   time the program is waiting for›   your command. Snapshots are saved›   as pure bitmaps named SI*.GR8.›   Up to 255 bitmaps can be captured›   per one session.›› - Bugfix: due to a misunderstanding›   between me and Gumi/TIGHT, the›   SID was checked wrong. Correct›   now, see SI.DOC for details.›› - Bugfix in high RAM test. Problem›   reported by Mario Trams.›› - Bugfix: on machines with the›   400/800 OS characters !! or !"›   appeared in the field "OS serial›   number", instead of "AA" or "AB"›   respectively. This bug's cause›   was that I converted text›   encoding from ICODE to ASCII and›   omitted a string.›› - Bugfix: "Sector I/O" option was›   impossible to change, if set›   to "OS" and saved to the SI.DEF.›   Found by WSZ.›› - SpartaDOS 3.2 crashes, when the›   SI is attempting to get confi-›   guration from a SIO device with›   code greater than $38. The SI›   will no longer check drives›   D9:-D15: for this DOS. Problem›   reported by Matthias Belitz.›› - "Disk summary" window no longer›   displays all tested devices,›   but only from D1: to the last›   detected disk.›› - New OS types should be recog-›   nized: Tight OS and DracOS. ›› - Text fields to store benchmark›   results are now refreshed before›   each test. ›› - Bars, which indicate the›   benchmark results, will now wrap›   to the line below, if they don't›   fit in the window.›› - PERCOM[4] (number of heads)›   is now ignored for hard drives›   other than connected with IDE›   Interface. Problem reported›   by Matthias Belitz.›› - "Parallel devices" have been›   moved to a separate window.›   IDE, MSC and Black Box detection.›› - New information in the "System›   summary" window: "Installed DOS".›› - When the SDMA is off, benchmarks›   will no longer run with a›   completely black screen.›› - Screen routines are now a bit›   improved. Average improvement›   for both 6502 and 65c816 rou-›   tines is about 10 percent.›› Since the version 2.0 has been› released, many people argue, that› screen output is SLOW. The problem› is, that the SI works on a virtual› screen of 64x24 characters. Because› the ANTIC does not support such› resolution, everything has to be› done by the CPU.›› Unfortunately, almost all critical› operations are 16-bit (because› two bytes of screen memory have› to be accessed at once), so, on the› 65c816 with its 16-bit registers› the speed is satisfying; the 6502› however does all the stuff much› more slowly.›› For example, to fill the whole› virtual screen with characters,› the 6502 needs about 1.5 mln cycles.› Regarding the ANTIC sitting on the› bus from time to time, it is over› 1 sec. With the algorithm I've› used, there's no way to improve› the speed to a big extent.›› Some people have noticed, that the› windows are drawn in two steps:› first an empty window is created,› then it is filled with a text.› The people argued, that it would› be a bit faster to draw the window› filled with the text at once.›› Yes, it would speedup window› drawing, but it would also need to› keep whole windows (i.e. with› the border and all empty fields)› in the memory. Since the SI has› about 30 windows and menus, it› would demand a lot of additional› memory. Because there's no very› great amount of memory left, I› will rather stay at the actual› solution.›   › Another question is background› redraw speed. The problem is, that› the background image is packed› to save some RAM space, and it› remains packed all the time the› SI is running. It means, that› the SI must unpack it (not just› copy) to the screen memory each› time the screen need to be redrawn.›› Again, the 65c816 routine can› do it very quickly - the background› is redrawn just like it would be› a bitmap copy - but with the 6502› unpacking is clearly visible.››› Changes made in 2.01› --------------------›› 1) Bugfix: in the field "OS type/›    '816 oriented" nothing appeared›    (instead of "no") on non-816›    machines.›› 2) Percent-fields are now 4-byte›    long (up to 9999% possible).›    It has been changed to prevent›    crashes on HyperSpeed Ataris.›› 3) SID (a sound chip from the C-64)›    detection added.›› 4) Sector I/O operations were done›    through the SpartaDOS LSIO›    vector, when the SpartaDOS was›    present. Now it can be forced›    back to OS SIO.›› I hear, that the CPU speed bench-› mark crashes on some emulators. I› am far from going to fix it,› because real Ataris do not crash.› So get a real computer, before you› report a bug in the SI!››› Changes made in 2.00 (a big rewrite)› ------------------------------------›› 1) 65c816 support added: the SI›    now recognizes this CPU.››    DOES ANYBODY KNOW, HOW TO DISTIN-›    GUISH THE 65C802 FROM THE 65C816?›    ANY INFO WILL BE WELCOME.›› 2) CPU and OS interrupt vectors info›    added.›› 3) Disk and memory info added.›› 4) Benchmarks added. The MIPS.COM,›    FLOPS.COM and RWTEST.COM have›    been incorporated to the SI›    with some changes and improve-›    ments.›› 5) A bug in 'max.scanlines' test›    has been fixed.›› 6) Some routines have been›    optimized.›› 7) Background image has been added›› 8) VM-support has been removed for›    now. It will possibly be back›    in a future version.›› 9) The RAM test now can detect›    the 65c816 high RAM (i.e. past›    the first 64k).›› 10) Quick help added.›› 11) Any window may be closed now›     with the [Esc] key.›› 12) Options added.›› 13) OS type test has been rewrit-›     ten. Now should recognize the›     MTI OS and ARGS-OS.›› 14) IDE Interface support added.›› 15) Fast redraw added: on '816›     machines the screen elements›     are drawn faster.›› 16) 400/800 machines are now always›     assumed as NTSC ones.›› 17) Unfortunately, all mentioned›     changes, improvements and›     additions made the SI now›     require at least 48k machine›     to run...›› 18) Due to this reason, the SI›     is now *relocatable* to use›     as much free memory as›     possible. The main binary›     segment (SI.PRG) is executed›     by the SI.COM, which is ›     actually a small loader.›     › The SI.PRG format description is› reachable on the WWW at:›› http://www.orient.uw.edu.pl/› 
conradus/›› where "
" sign stands for the› tilde.›› NOTE: memory usage summaries are› based on the OS variables; since› the XL OS does not test the high› RAM nor bank-select, values shown› in the 'Memory usage' window may› differ from the reality. These› values are correct under DracOS,› a 16-bit, 816-aware operating› system.››› Changes made in 1.92 (not released)› -----------------------------------›› 1) Menu bar control has been added.›› 2) Maximum scanlines test added.›› 3) Design change.›› Changes in 1.91 (later release)› -------------------------------›› 1) The speed indicator bug, men-›    tioned below as fixed in 1.91,›    in fact still remained in the›    object code (it has been fixed›    in the source code only - I›    forgot to compile it again,›    sorry). Really fixed now.›› Changes in 1.91 (first release)› -------------------------------›› 1) The code has been recompiled to›    higher addresses ($2600 now) for›    better compatibility with some›    systems. Problem reported by›    Tomasz Tatar.›› 2) MEMLO and MEMTOP values are›    checked before the main segment›    is loaded.›› 3) Two stupid mistakes in the›    Axlon-test have been fixed›    (no bug reports?!?)›› 4) A bug in the speed rate indica-›    tor has been fixed. Due to this›    bug, the speed rate indicator›    worked wrong on 65c02/65c816›    machines. Reported by David›    Paterson and John Harris.›› 5) POKEY test has been added to›    detect stereo circuits.›› 6) One more routine has been added›    to clearly determinate machine›    type (400/800 or XL/XE).›› 7) Screen design has been changed›    a bit.›› 8) The code has been shortened.››› Changes in 1.9› --------------›› 1) A bug, that caused garbage on›    the screen, if tested computer›    hadn't extended RAM, has been›    fixed.›› 2) The SI didn't start neither un-›    der MyDOS nor AtariDOS. Fixed.›› 3) Some other small bugs have been›    fixed.›› 4) Some routines have been optimi-›    zed.›› 5) Axlon-test has been added. My›    computer hasn't the Axlon-RAM,›    so if something goes wrong,›    please report it.›› 6) Virtual Memory test. Now SI is›    a VM-application (and it sounds›    much better). See VM.DOC›› 7) 'XL RAM' is now called 'Linear'›    (due to hope of the 65c816) and›    'XE RAM' is now called 'Exten-›    ded' (due to no particular rea-›    son).›