smaug@eneevax.UUCP (Kurt J. Lidl) (06/20/88)
Well, I finally broke down and bought myself a 68010 for my 4.2 motherboard B2000. After carefully prying out the old 68000, and installing the shiney, new & improved CPU, I cranked up the Amiga. Everything booted properly, vd0:, faccii, dmouse v1.04, and blitz. Now for the real test! Will MSE (Mandelbrot Set Explorer) perform like greased lightning, at a whopping 3-5% greater speed? Alas no - it gives the dreaded Task Held error, and then drops thru to a Guru... So I guess it was the 68010... Now on the the question - Is there a patch to cause it to non-guru? A ftp'able version of decigel for AmiDOS v1.2? Anything I can do to get it working correctly? BTW, the chip (a real Motorola ceramic case one) only cost me $10. Not too bad. Let's hear it for Hamfests! Thanks, Kurt Lidl -- ================================================================== == Kurt J. Lidl (smaug@eneevax.umd.edu) (301)454-6849 == == UUCP: [seismo,allegra]!umcp-cs!eneevax!smaug == ========"It's after 3am, no point in going to sleep now..."=======
thad@cup.portal.com (06/22/88)
My articles and Scott Turner's DeciGEL (asm source and executable) are on Fish Disk #18 in the MC68010 directory. I'll send you a shar'd uuencoded copy (approx. 20K); look for it in a mailbox near you! :-) Am surprised you had a problem with MSE, though. Since early 1986 I've been using CSA's 68020/68881 boards in my Amigas and have had very few problems during the past 18 months. Most the problems are due to privilege violations or self-modifying code (the latter a real no-no on the '020). The only commercial titles of which I'm aware that are troublesome on the '020 Amigas are SINBAD, Macro Modem, Barbarian (though Ken Hodel pestered Psygnosis DAILY (long distance to England) 'til we received fixed copies), HARRIER Combat Simulator, and SubLOGIC's JET (though their FS-2 works fine). Nearly EVERYTHING (PD and commercial) compiled with early versions of the Lattice compiler (e.g. 3.02 and 3.03) will fuckup an '010 or '020 Amiga. And a lot of new stuff compiled with Lattice is causing other perturbations (witness the spurious requestors "User Abort Requested"). These problems can be fixed by re-compiling with Manx (and you get smaller and faster executables as a side benefit (Lattice ads notwithstanding)).
thad@cup.portal.com (06/23/88)
BTW, just encountered another "major" product that does NOT function properly on a 68020-equipped Amiga: New-Tek's DIGI-VIEW version 3.0 (the latest version, just released) when the DIGI-DROID is connected. The 68020's cache MUST be disabled in order for the motor controlled color-wheel to function; totally UNACCEPTABLE. :-( Difficult to believe they did this, especially when Tim Jenison stood up in front of 1000+ people at the FAUG meeting earlier this year (the meeting at which Max Toy was the featured speaker) and promised the Digi-Droid self-modifying code bug (first appearing in DIGI-VIEW V2.0) would be fixed in version 3.0 (which also, BTW, was supposed to be released "in several weeks", and that was months ago).
toebes@sas.UUCP (John Toebes) (06/25/88)
In article <6749@cup.portal.com> thad@cup.portal.com writes: >Nearly EVERYTHING (PD and commercial) compiled with early versions of the >Lattice compiler (e.g. 3.02 and 3.03) will fuckup an '010 or '020 Amiga. > >And a lot of new stuff compiled with Lattice is causing other perturbations >(witness the spurious requestors "User Abort Requested"). These problems >can be fixed by re-compiling with Manx (and you get smaller and faster >executables as a side benefit (Lattice ads notwithstanding)). I would suggest getting some facts before making such a broad sweeping statement. I also have an 020 machine and have found less of a problem than you. First of all, all the PD stuff I found that didn't run on a 010 or 020 was compiled with GREENHILLS. It tends to output a move sr instruction to collect the condition codes in complex expressions. After hearing this problem attributed to the Lattice compiler, I spent several weeks going over the code to find where we produced such illegal code and found none. Only after talking to Andy Finkel did I find that the culprit was the GREENHILLS compiler. As for comparing 3.02 code and 3.02 code to current Manx code, I would agree with you there is no comparison. However, you clearly haven't looked at the 4.0 code (and have no interest in doing so) which has many innovations specifically for the Amiga. A lot has happened to the compiler in 3 years, anyone still using that old version should consider moving on to something else. /*---------------------All standard Disclaimers apply---------------------*/ /*----Working for but not officially representing SAS or Lattice Inc.-----*/ /*----John A. Toebes, VIII usenet:...!mcnc!rti!sas!toebes-----*/ /*------------------------------------------------------------------------*/
thad@cup.portal.com (07/01/88)
John, not to start any "war", but ... A lot of the PD sources I compiled back in 1985 and early 1986 using the 3.02 or 3.03 would go belly up first on the 010 and later on the 020 systems until I switched to Manx. I can still locate hardcopy (from CBM, I presume) documenting the problems in the LIBRARY code of those early Lattice versions. That's why Scott Turner and I got "involved" with DeciGEL. Re: the "User Abort Requested" requestors: a lot of stuff uploaded to my BBS systems has been (recently) compiled using Lattice, and causes that problem. After I recompile the sources using Manx, the executable size is ALWAYS smaller, and the "User Abort Requested" never again surfaces. Make what you will of that. And several months ago, in this newsgroup, that topic of the "User Abort..." came up with respect to AmigaARC and other programs (compiled using Lattice). Problems with BOTH Manx and Lattice have stymied some of my projects, so I've been recently "investigating" other alternatives (for which I have source). And I've been writing commercial compilers for over 20 years, so I know what I need, and what it takes to satisfy that need.