|
Software archaeology with Excel 68000 versions (E2DA,F806 and DB53)
Hi,
When the Excel Mach III arrived in France in november 1988, Pierre Nolot informed Rexton (the importer for France) that there were bugs in the mate finder level (i.e., Problem N°12, http://www.schach-computer.info/wiki...tle=Nolot_Test). Fidelity considered these to be exceptional cases that should be ignored. Rexton threatened to return the quota of 100 Mach III units for France, but the Spracklens intervened quickly and the DB53 version corrected the F806 version. A few decades later, a E2DA version appeared! Its late arrival led it to be considered the ultimate version of the Mach III. As such, it is often used in current tests and tournaments. I tested this ROM under MAME and noticed that the problem n°12 was not found again(checkmate in 6 moves stops the search after 42 secondes!). Is it the return of the F806 bug, or a prototype/pre-release version of Mach III coming out of nowhere? After disassembling the three versions of Mach III, two versions of Mach II, and Designer 2265 (10000 lines each), I created this little executable code comparison: Code:
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------the E2DA version of Mach III is identical to Mach II A (notes #1 and #2), “move.b #$0, D1” and “addq.b #1, D1” in common. The DB53 version is similar to Mach II LA (notes #1 and #2), with “move.b D0, D1” and “eori.b #$10, D1” in common. The E2DA version therefore predates the F806 version. The E2DA and F806 versions predate the Excel Mach II LA (note #3), because “neg.b D6” appearing on the latter, then on DB53 and the designer. Problem N°12 is not solved by the E2DA and F806 versions, and Problem n°13 requires 20% more time. On all Nolot's tests others than checkmat problems, find the solution for F806 version requires 1680 secondes (+1.74%) more time than DB53 version. All these elements have convinced me that the E2DA version is the first version (prototype?) of the Mach III. Therefore, only the DB53 version should be used for tournaments, unless you want to skew the results ;-) For CB-Emu, it would be wise to rename fex68Km3b, for example with fex68km3p, p for prototype (if it has never been sold). In https://www.schachcomputer.info/foru...?t=6986&page=5 , messages #47, Kamoj had remarkably well detected and reported the problem with the Excel Mach III E2DA version : congratulations! Some response times for P. Nolot's tests with mame. Code:
| E2DA | F806 | DB53 |Designer 2265 |Cordialy, mychess |
Re: Software archaeology with Excel 68000 versions (E2DA,F806 and DB53)
Hello,
Happy new year MMXXVI and good health to all ! All the Nolot test for the best of Excel family. With a little riddle ... Code:
Excel Mach II L.A. Excel Mach II,V ;-) Excel Mach III DB53 Designer 2265 Elite A/G V2 Excel Mach IVCordialy, Mychess ps : E455B34A |
Re: Software archaeology with Excel 68000 versions (E2DA,F806 and DB53)
Hello,
The response time for “Mach II,V” is under emulation (Mame). All others response times are taken from Nolot published in "Europe echecs" and available at https://www.blitzchess.fr/fr/tests/europe-echecs/88-90/ . However, I checked 2 positions (Cb7 and Cb10) with emulations (Mame, CB-emu) with the 2 sets available for LA version (fex68km2,fex68km2a), so it seems that Nolot has received from Fidelity/Rexton a Mach II at 13,78 MHz (74' vs 85' for the last). If I remember correctly, it is not the first time Fidelity/Rexton gave it an overclocked chess computer for testing. Cordialy, mychess |
| Alle Zeitangaben in WEZ +1. Es ist jetzt 03:18 Uhr. |
Powered by vBulletin (Deutsch)
Copyright ©2000 - 2026, Jelsoft Enterprises Ltd.
©Schachcomputer.info