*start* 01104 00024 US Date: 25 MAY 1977 1557-PDT From: MCDANIEL Subject: moving to the d1 directories To: d1docs Date: 24 MAY 1977 1551-PDT From: MCDANIEL Subject: moving to the d1 directories To: D1Users: Wednesday, 25 May in the afternoon I will move files from the Aspen directory to the d1 series of directories. I will destroy the files on Aspen. I will maintain the same number of versions, but I will not guarantee the numbers to be the same unless specifically requested (ie., foo;28,27,26 will become foo;3,2,1). I will move d1 files only. I will cause archived d1 files to be unarchived, moved to the appropriate d1 directory and then archived. If D0 people can be convinced to move to another directory, I will cause the Aspen directory to be destroyed, after appropriate D0 actions to save d0 files. Please do not use the Aspen or d1* directories wednesday afternoon (25 may) EXCEPT for READING purposes. The D1 directories are: D1DOCS -- documentation: text & ears files D1SOURCE -- program sources D1 -- running programs and binaries D1LOGIC -- logic diagrams. ------- ------- *start* 00511 00024 US Date: 25 MAY 1977 1610-PDT From: MCDANIEL Subject: <d1docs>message.txt To: D1Users: I have constructed a message.txt file on <d1docs>, and I have added d1docs to <d1docs>d1users.msg. This means that messages you send to d1users will be automatically sent to <d1docs>message.txt. This facility will leave a message history and (hopefully) be of use when you can't remember the contents of the last message(s) you have received. I will periodically remove old messages. Gene ------- *start* 00836 00024 US Date: 25 MAY 1977 1702-PDT From: MCDANIEL Subject: <aspen>switchover is done To: d1docs Date: 25 MAY 1977 1304-PDT From: MCDANIEL Subject: <aspen> switchover is done To: D1Users: cc: chang, charnley D1LOGIC D1DOCS D1 D1SOURCE have been created and <ASPEN> files have been moved to those directories. See <McDaniel>*.runfil for the actions that were taken if you have any question. NOte that memgnd.sil and memgnd.gerb were moved to d1logic manually. PLEASE EXAMINE THE <ASPEN> DIRECTORY: REMOVE YOUR PERSONAL FILES FROM THAT DIRECTORY. IF YOU HAVE PERSONAL FILES ARCHIVED THERE, PLEASE DEARCHIVE THEM AND ARCHIVE THEM ON ANOTHER DIRECTORY. THE ASPEN DIRECTORY WILL BE DESTROYED AS SOON AS POSSIBLE. I WILL DEARCHIVE AND REARCHIVE THE APPROPRIATE D1 ORIENTED FILES ON ASPEN. Gene ------- ------- *start* 00598 00024 US Date: 25 MAY 1977 1703-PDT From: MCDANIEL Subject: <d1docs>message.txt To: d1docs Date: 25 MAY 1977 1610-PDT From: MCDANIEL Subject: <d1docs>message.txt To: D1Users: I have constructed a message.txt file on <d1docs>, and I have added d1docs to <d1docs>d1users.msg. This means that messages you send to d1users will be automatically sent to <d1docs>message.txt. This facility will leave a message history and (hopefully) be of use when you can't remember the contents of the last message(s) you have received. I will periodically remove old messages. Gene - ------- *start* 00540 00024 US Date: 26 MAY 1977 0649-PDT Sender: FIALA From: Edward Fiala Subject: The D1 Microassembler To: D1Users I.D.: <D1DOCS>D1ASSEM.EARS;3 Key words: CSL, D1, processor research, microprocessors Pages: 59 This document describes version 1.0 of the D1 microassembler, compatible with version 1.0 of the hardware manual and such recent changes as I am aware of. The assembler has been fairly well checked out with a test program, and I believe that it is ready for general use. Please report all problems to me. ------- *start* 00300 00024 US Date: 26 MAY 1977 0925-PDT From: FIALA Subject: D1Assem To: D1Users: Sorry that you got two messages about the assembler. I accidentally distributed an obsolete version the first time. You should only print <D1Docs>D1Assem.Ears;3. I deleted the obsolete version. ------- *start* 00290 00024 US Date: 1 JUN 1977 1848-PDT From: FIALA Subject: D1Lang To: D1Users: A new version of <D1Source>D1lang.mc is available. It fixes two bugs reported by D. Charnley and the hardware change reported by K. Pier. There are no external or documentation changes. ------- *start* 00193 00024 US Date: 9 JUN 1977 0758-PDT From: PIER Subject: dict.analyze To: SIL Interest Group:, D1Users: new <sil>dict.analyze;1 corrects SG10139 ECL PROM pinouts. Pier ------- *start* 01808 00024 US Date: 15 JUN 1977 1945-PDT From: FIALA Subject: D1 Microassembler To: D1Users: I have just released a new D1 microassembler (<D1SOURCE>D1LANG.MC and <D1SOURCE>D1ALU.MC) and new documentation (<D1DOCS>D1ASSEM.EARS). This message describes all of the changes to the assembler, so you need not print the assembler manual again unless you want to. The new release fixes one bug and adjusts to the following three hardware changes: 1. The ORSAVEDCARRY function is now USESAVEDCARRY--the function causes the ALU carry-in to be the carry-out of the previous function for the current task. 2. The "insert field" and "extract field" methods of loading SHC are now initiated by BSEL=6 or BSEL=7 in the instruction which does a shift and specify the control for that shift, rather than being initiated by ALUF decodes 0 and 1 in a previous instruction. 3. The 8 FF-decodes for AMux sources are now in the first 64 FF decodes, so any AMux source can be specified for a memory operation. The changes in the assembler are as follows: 1. The macros for building parameters, constants, etc. used to accept either integers, parameters, or addresses as arguments. They now accept only integers and parameters. 2. The D1ALU.MC file no longer contains special stuff for defining the "insert field" and "extract field" ALU operations. ALUFM 0 and ALUFM 1 are no longer special. 3. The ALU operation assembled into ALUFM 16 should be "NOT A" rather than "A" because the shifter output is complemented on the AMux. 4. The macros for constructing "insert field" and "extract field" constants used to be written as (ALUOP)&RSH T[3]. They are now written as standalone clauses RSH T[3], etc. and are only legal in instructions which are doing a shift operation (SHIFTNOMASK, etc.). ------- *start* 00397 00024 US Date: 29 JUN 1977 1701-PDT From: SPROULL Subject: New Gobble To: D1Users: There is a new Gobble on <D1> (no room left on <SIL> until I get an easement from some local god), which fixes a bug reported by Brian Rosen: if you are doing re-work and have REMOVED an IC (that is, all of its uses have disappeared), the AD file got hopelessly mangled. Now fixed. Bob ------- *start* 00584 00024 US Date: 14 JUL 1977 1846-PDT From: SPROULL Subject: New NPPR, ANALYZE To: D1Users: There are new test versions of these on <sproull> (because I can't fit them anywhere else). NPPR fixes a long-standing bug about fonts used in printing. ANALYZE should remove all of those nasty extraneous error messages (warnings), and should generate some brand new ones if you should forget to ire up some pin in a section you have used. The reason I am sending this notice to D1 users only is that I hope wyou will try it out and report problems to me asap. Bob ------- *start* 00521 00024 US Date: 16 JUL 1977 0515-PDT From: FIALA Subject: New Release of Hardware Manual To: D1Users: I have just released the first revision of the hardware manual. It is on <D1DOCS>D1MANUAL.EARS (94 pages). There are extensive revisions to the Processor and Debugging sections of the manual and fewer revisions elsewhere. I do not expect to make another release of the manual for several months, but I would appreciate being notified of any errors, omissions, or unclear parts in the manual. ------- *start* 00861 00024 US Date: 18 JUL 1977 0845-PDT From: SPROULL Subject: More on DA message To: D1Users: OOPs -- forgot some things: GOBBLE has Pier's proposed fix for re-working nets with terminators assigned: if net has cable or edge pins, or if either new name or old name ends in !, compare terminator assignments as well. Please thry this out. I also forgot to mention in the description of BUILD that the password in the SIL file will be changed to indicate the file has not been modified since being BUILT, and SIL will change the password back if the file is ever modified. In addition, some "marks" will be placed in the file that will be plainly visible when printed, so you know the drawing is correct given the REV level cited. ANALYZE will clearly mark in its comments those files that are unchanged since last BUILD operation. Bob ------- *start* 00522 00024 US Date: 21 JUL 1977 2159-PDT From: SPROULL Subject: New SIL manual (preliminary release) To: D1Users: You will find on <SIL>SilManual.Press a preliminary version of the new SIL manual (print it on Dover!). Please offer (modest) suggestions for improvements (chiefly in accuracy, not in beauty of exposition). The intent was to start wtih CHuck's manual, add documentation of new stuff, work in Ken's conventions memo (or most of it), and offer a design of BUILD for you to consider. Bob ------- *start* 00254 00024 US Date: 1 AUG 1977 1357-PDT From: FIALA Subject: Midas To: D1Users: Version 1 of the Dorado Midas manual is hereby released. A copy will be kept near the micoprocessor, so you probably won't need to print your own copy. ------- *start* 00477 00024 US Date: 3 AUG 1977 1227-PDT From: ORNSTEIN Subject: Interim Dorado Display Review To: D1Users: Any or all interested in going over the rough design of the display controller with me tomorrow morning at 10AM - let us meet in the Alcove. (If that's busy, then the Juniper/Dorado room). The idea is to see if it makes sense. I am sending around a brief informal writeup and sketches this afternoon for you to look over in advance. Thanks, Severo ------- *start* 00591 00024 US Date: 3 AUG 1977 1755-PDT From: WILLIE-SUE Subject: ALUFMEM To: D1Users: In writing the microcode for BitBlt, I have discovered that I want another ALU fn that is variable, just as ALUF[17] is. Since ALUF[16] ia already special, I propose to make ALUF[15] an emulator specific ALU fn. BitBlt will save and restore its value; all other tasks will have to be content with 14 (decimal) ALU fns (values to be decided sometime). If anyone has violent (or other) objections, come talk to me, or we can all discuss them at the Friday meeting. Willie Sue ------- *start* 00148 00024 US Date: 9 AUG 1977 1729-PDT From: FIALA Subject: D1Lang To: D1Users: A new release of D1Lang.mc is on <D1SOURCE>. ------- *start* 00191 00024 US Date: 10 AUG 1977 1617-PDT From: FIALA Subject: D1Lang.mc To: D1Users: A new D1Lang.mc is released on <D1Source> fixing bugs in branch condition numbering. ------- *start* 00369 00024 US Date: 21 AUG 1977 0605-PDT From: FIALA Subject: Midas To: D1Users: I have just finished revising <D1DOCS>D1MIDAS.BRAVO and <D1DOCS>D1MIDAS.EARS. This is the first relatively complete documentation on Midas. I am going to place a copy of the documentation next to Dorado, and I recommend that you read through it before using Midas. ------- *start* 00693 00024 US Date: 22 AUG 1977 1040-PDT From: FIALA Subject: D1 Microassembler To: D1Users: I have released a new version of D1Lang.mc on <D1SOURCE> and new documentation on <D1DOCS>D1Assem.Ears. The only change in D1Lang is that the branch stuff is changed to allow you to write the branch clause anywhere in the instruction statement. Formerly, the branch clause had to be the left-most clause. The documentation is also unchanged except a few examples have been added to sections describing parts of the assembler that were a source of questions. I do not recommend that you print a new copy of the assembler document, but do pick up the new release of D1Lang.mc. ------- *start* 00601 00024 US Date: 1 SEP 1977 1618-PDT From: PIER Subject: MULTIWIRE To: D1Users: cc: chang, charnley, rosen Next Thursday, Ken Tyler and Paul Burdsall of MULTIWIRE West will give a presentation on their wiring technology. We are investigating this technology as a possible alternative to stitchweld for the D1 system. Interested parties should remain after this talk to give Ken and Paul a picture of our specific needs, and whatever they require to make an estimate of the feasibility and cost of MULTIWIRE for D1. That's THURSDAY, SEPTEMBER 8, at 1:30, CSL BEANBAG ROOM ------- *start* 00273 00024 US Date: 3 SEP 1977 1412-PDT From: MCDANIEL Subject: new d1lang To: D1Users: i have moved a new version of d1lang to <d1source>. This version incorporates the change to the fastbranch code that makes d1lang compatible w/ the hardware. Gene ------- *start* 01128 00024 US Date: 3 SEP 1977 1542-PDT From: MCDANIEL Subject: [ifs]<d1drawings> To: D1Users: I have arranged a d1drawings directory on ifs. Because ifs is not a stable system only .press or .ears files should be stored on this directory (as opposed to the .sil files stored on the d1logic directory on maxc). I have moved all the .press files on <d1logic> to [ifs]<d1drawings>. I will delete the .press files on <d1logic> after a backup has been performed on ifs. I have arranged for all of you to have accounts on ifs if you don't have one already (this was necessary to have access to the "group" d1drawings account). The d1drawings password is the same as the password for d1logic, d1source, etc. For those of you not familiar w/ ifs, the maxc program, pupftp, and the alto program, ftp, can be used to retrieve or store files on ifs. The familiar retrieve and store commands work as expected. The ftp name for ifs is "ifs". I encourage you to move future .press files to ifs. Gene ps. If you didn't have an account on ifs prior to this, your password is the same as the one for d1drawings. G. ------- *start* 00753 00024 US Date: 7 SEP 1977 0221-PDT From: FIALA Subject: D1Lang.mc To: D1Users: I have released a new D1Lang.mc on <D1Source> with extensive changes, so go carefully. The changes are as follows: 1) Branch condition numbering changes 2) Bug fixes to memory-reference stuff relating to overwrite of FF field because of failure to set USESFF flag for MicroD. 3) Failure to set USESFF for RBASE←. 4) Elimination of ONPAGE macro. 5) Significant speed improvement by rewrite of branching and IM post-macro stuff--this is only loadable by LPD's newest undebugged MicroD. 6) Bug fixes for reversal if IMLH← and IMRH←. 7) Fix RSH and RCY to accept shift count of 0. Please re-report any bugs I failed to fix and any new bugs. ------- *start* 00350 00024 US Date: 8 SEP 1977 1141-PDT From: MCCREIGHT Subject: Storage Board Interface Spec. To: D1Users: The first edition of the Dorado Storage board interface specification is available as <D1DOCS>storinterface.press. Please read and feed back if this (admittedly somewhat arcane) subject is of interest to you. Thanks, Ed ------- *start* 00578 00024 US Date: 15 SEP 1977 1034-PDT From: PIER Subject: NEW HARDWARE To: D1Users: Per suggestion from Willie-Sue, the primary branch test for the alu has been changed from ALU GE Zero to ALU LT ZERO. D1LANG must be changed to reflect this. The result of this change is that all of the following tests may then be specified in a single instruction: ALU EQ 0, ALU NEQ 0, ALU LT 0, ALU GE 0, ALU GT 0, ALU LE 0, by the appropriate use of JCN and FF fastbranches. This modification is effective immediately. Let a thousand recompilations bloom. (Sorry). ------- *start* 00183 00024 US Date: 15 SEP 1977 1041-PDT From: FIALA Subject: D1Lang To: D1Users: There is a new D1Lang.mc on <D1SOURCE> with the branch condition changes in it. ------- *start* 00592 00024 US Date: 17 SEP 1977 1706-PDT From: MCDANIEL Subject: KERNEL.MB To: D1Users: The four kernel diagnostics have been coalesced into one diagnostic: KERNEL.MB This diagnostic noramlly operates in an infinite loop. Any breakpoints indicate a hardware (software?) failure. KERNEL.MC is the driver program, KERNEL.DLS is a listing of the instruction placements. THERE IS A NEW D1LANG.MC on <D1SOURCE>. It redefines the IM size to 10000B. To start the diagnostic, Enter midas Reset the machine Set task to 0. Load kernel Read commands from SETP3 GO at IM 0 Gene ------- *start* 00464 00024 US Date: 20 SEP 1977 0005-PDT From: FIALA Subject: D1Lang To: D1Users: I have released a new <D1SOURCE>D1Lang.mc with the following changes: (1) With shifts, RSH and RCY now accept an argument of 0. (2) BRANCH and DBLBRANCH are now controlled by the TOP LEVEL and SUBROUTINE declarations, so BRANCH inside a subroutine is equivalent to GOTO, while at the top level, BRANCH does not restrict the placement of the target instruction. ------- *start* 00181 00024 US Date: 20 SEP 1977 1432-PDT From: FIALA Subject: Midas To: D1Users: New release of Midas on <D1>d1midasrun.dm. Fixes symbol table bug, hopefully. ------- *start* 00298 00024 US Date: 23 SEP 1977 1012-PDT From: MCDANIEL Subject: new poked1.run, poked2.run on <d1> To: D1Users: I have modified pokeD to allow comments in the parameter file. The characters between an asterisk and carriage return are ignored when poke is reading numbers. G. ------- *start* 00318 00024 US Date: 24 SEP 1977 1334-PDT From: MCDANIEL Subject: <d1>poked1.run, <d1>poked2.run To: D1Users: There is a new version of poke. This version allows larger parameter files when executing test 30. In particular the imtest that contains about 500 instructions can be read by poke. Gene ------- *start* 00943 00024 US Date: 30 SEP 1977 1743-PDT From: FIALA Subject: D1Lang.mc and D1alu.mc To: D1Users: There is a new release of both of these files on <D1Source>. You must pick up new versions of both simultaneously. The changes are as follows: 1. Addition of McrD←B and DummyMRef←A in memory section. 2. Error checks for use of B←Pipex with memory references. 3. Error checks for use of IMLH←, IMRH←, READIM, LDTPC←, or RDTPC← inside subroutines. 4. Error checks for use of arithmetic ALU operations with external BMux sources that are too slow for this. Don't forget that you will have to modify D1ALU.mc to produce the ALU op defn file that you need for assemblies. The only changes in D1ALU.mc are the addition of the letter "S" in front of the four ALUOP statements for "A+B", "A+B+1", "A-B", and "A-B-1". Hence, you might find it easier to directly edit your defn file rather than recreating it from D1ALU.mc. ------- *start* 00382 00024 US Date: 1 OCT 1977 1520-PDT From: MCDANIEL Subject: [ifs]<d1drawings>*.results To: D1Users: The above named files are a new version of the delay times that I constructed: the new values are some what larger -- they reflect a better equation. Also, the information is presented twice: once sorted alphabetically and once sorted by delay time. G. ------- *start* 02040 00024 US Date: 5 OCT 1977 0157-PDT From: FIALA Subject: Midas Release To: D1Users: I have just released a new <D1>d1midasrun.dm. It fixes most of the bugs reported and incorporates most of the new features suggested. YOU MUST RELOAD ALL MICROCODE WITH LPD'S LATEST MICROD RELEASE BEFORE USING THIS MIDAS. The changes are: 1. All bugs in DMux simulator are believed to be fixed. 2. When you are displaying the DMux errors (middle-button over "CkDMux") the prettyprint procedure for DMux locations is changed to show exactly the signal names that have failed. 3. Symbolic input for instruction memory is changed to accept "RETURN", "GO[33]","LONGGO[FOO+3]", and "GCALL[1000]" as legal clauses. These can be used in conjunction with the previous "RSTK←3" type clauses. The argument to GO must be in the range 0 to 77 and is equivalent to JCN←arg. The arg to LONGGO is a virtual address in Virtual mode or an absolute address in absolute mode. Midas translates this to an appropriate Long goto or long call in the FF and JCN fields. GCALL translates its absolute argument into an appropriate global call. Note for these that you write GO or LONGGO even when you mean CALL. 4. The bug causing symbols with value 0 not to be prettyprinted has been fixed. 5. The DMux signals in the control section have been rearranged for the purposes of viewing these from Midas. The deviations from the hardware DMux addressing will be documented in the Midas manual very shortly. 6. An OLINK memory (read-only) has been added. This memory shows the value in corresponding TLINK -1. This is useless in absolute mode. In virtual mode it will show the location of the last instruction that caused LINK to be loaded. This is provided for convenience in locating the diagnostic instructions which do BRANCH[ERROR], where ERROR is located at a global call location. Please rereport any bugs/features I missed except the Repeat-Go problem which I haven't got to yet (It is possible that Repeat-Go actually works now.). ------- *start* 00372 00024 US Date: 5 OCT 1977 1722-PDT From: FIALA Subject: Midas Documentation To: D1Users: The Midas user documentation, <D1DOCS>D1MIDAS.EARS, has been revised. This release is not very different from the last release and serves only to document some internal changes. The new document will be put next to the hardware debugging console shortly. ------- *start* 00600 00024 US Date: 6 OCT 1977 0833-PDT From: MCDANIEL Subject: <d1source>kernel.dls, kernelsources.dm; <d1>kernel.mb To: D1Users: I have changed the way the kernel diagnostics are stored on maxc: The dump file on <d1source> contains the sources and command files sufficient to generate a new kernel.mb given a disk with micro, microd and d1lang. As before <d1source>kernel.dls is the listing file and <d1>kernel.mb is the object file. There is a new version of the kernel diagnostics on maxc. It is compatible with the latest round of changes in d1lang and midas. Gene ------- *start* 00283 00024 US Date: 6 OCT 1977 2342-PDT From: FIALA Subject: Midas Release To: D1Users: I have just released a new <D1>D1midasrun.dm. It fixes several bugs. You should delete showvhist.midas, dmuxa.mc, and dmuxa.mb from your disk before loading this release. ------- *start* 00183 00024 US Date: 20 OCT 1977 1310-PDT From: WILLIE-SUE Subject: ifs directories To: D1Users: There are now directories on ifs for: D1, D1Source, and D1Logic. ------- *start* 00406 00024 US Date: 20 OCT 1977 1442-PDT From: MCDANIEL Subject: <d1source>control1.dm To: D1Users: The above file contains the controlboard diagnostic that I recently wrote. The .mb file is in the .dm file, and I will separate them soon. In the mean time, the package is available. There is a listing w/ the d1 and instructions regarding the use of the test are in the listing. Gene ------- *start* 00783 00024 US Date: 21 OCT 1977 1652-PDT From: FIALA Subject: Dorado Hardware Manual To: D1Users: The new release of the hardware manual is available as <D1DOCS>D1Manual. Both Press and Ears forms are there at the moment. The major changes since the 16 July 77 release are as follows: 1. Major revisions and additions to the Memory section. 2. Major editting to Alto Debugging Interface section (probably not interesting to anyone except me and maybe K.P.) 3. Addition of compressor and hold simulator stuff in Processor section. 4. Assorted small changes and clarifications in the Processor section. The IFU, Fast I/O, Slow I/O, Introduction, and Overview sections are approximately unchanged. Some of the figures have changed a little. Size = 108 pp. ------- *start* 00356 00024 US Date: 23 OCT 1977 1641-PDT From: MCDANIEL Subject: new kernel diagnostics release To: D1Users: I have discovered that the old diagnostics had a bug that would cause a spurious failure in the arithmetic shift test code, depending upon the implementation of d1lang. This has been changed. There are no other changes. Gene ------- *start* 00382 00024 US Date: 24 OCT 1977 1740-PDT From: MCCREIGHT Subject: New BUILD.RUN To: D1Users: ... is located on <SIL>. Now, for the first time, it includes the .BP files in the backup dump files. Somehow these were lost in the original shuffle... Mike Overton would appreciate everyone's using the new BUILD so he can find the .BP files in some orderly way. Ed ------- *start* 01146 00024 US Date: 31 OCT 1977 1406-PST From: FIALA Subject: Midas Release To: D1Users: A new release of <D1>D1MidasRun.dm is ready. This release fixes the following problems: 1. Abort after ShowError now aborts all nested command files, not just the one containing ShowError. 2. Reset, Go, and Step now reset the NoTasking condition correctly. 3. A ShowMem command file is added and other stuff for displaying the state of stuff in the memory section. Prettyprint routines exist for muffler signals in the memory section (though these are still being changed). The printout of simulator-detected muffler errors has been changed; there will be automatic printout of errors in a subsequent Midas release. 4. The sequential data pattern stuff has been fixed. However, note that the two parity bits are the low-order bits counted when testing IMX, so the sequential pattern isn't as pretty for this as one would like. 5. The reported problem with TimeOut being *10 too large is not fixed; I am suspicious that this is due to the Timer subr in the OS not delivering time in units of msec on Alto II, or something. ------- *start* 00920 00024 US Date: 31 OCT 1977 1759-PST From: FIALA Subject: D1LANG.MC To: D1Users: I have just released a new <D1SOURCE>D1LANG.MC with the following changes: 1. PREDA, PREDB, TREE, CARRY400, PROCSRN←A, MCR←A, CHECKFAULT deimplemented. 2. IOFETCH← and IOSTORE← made into uncommon memory references (non-RM source uses up FF) and legal only outside emulator. 3. MAP← and CACHEFLUSH← now legal only in emulator. 4. Error checks for IFUDATA outside emulator. 5. PROCSRN←B, LOADMCR[A,B] added. 6. The BSEL encodings of WF← and CNT← are deimplemented (the FF encodings for these are still defined). The CARRY400 was a left-over old deimplemented feature from way back, the PROCSRN←B and LOADMCR[A,B] stuff and the IOFETCH← and IOSTORE← encoding changes will be described in the next release of the hardware manual. These changes have also been documented in the new <D1DOCS>D1ASSEM.EARS. ------- *start* 00574 00024 US Date: 2 NOV 1977 1024-PST From: MCDANIEL Subject: new kernel To: D1Users: The debugging disk and the d1 directories have a new version of kernel. The hold register is now used during the diagnostics. There is a counter kept in IM that records how many times the diagnostics have been executed. Both the value of hold and the number of iterations are displayed on the midas screen (ie., there is a new kernel.midas, too). The value for hold begins at zero and is incremented by one (modulo 7) each time thru the main diagnostic loop. Gene ------- *start* 00683 00024 US Date: 2 NOV 1977 1706-PST From: PIER Subject: Dorado debugging environment To: D1Users: I have just been reminded of a known problem in the Dorado area, basement of bldg. 34. When the machine shop is running the heliarc welder, Altos (and presumably Dorados) do unpredictable things. The most reliable indicator of the heliarc running is the generation of spurious characters on an Alto II, the kind we have for Midas. So, when things go awry and you start to curse the keyboard, check in the shop first. This problem has been worked on in the past, and has not been successfully eliminated. Fortunately, heliarc welding is a rare occurance. K ------- *start* 03215 00024 US Date: 7 NOV 1977 0439-PST From: FIALA Subject: Midas To: D1Users: I have just dumped a new <D1>D1MidasRun.dm and expect to be dumping new versions daily for about the next week. The new Midas has the following revisions: 1. Display controller reset done when you do a "Reset". 2. Nomenclature changes for print routines as described in last message about D1Lang changes. 3. Complete implementation of latest revisions of MemC stuff on DMux and in Figure 9. If you have the 10/21/77 hardware manual, that is completely obsolete--everything has been changed around since then. You should get a copy of the latest hardware manual Figure 9, DMux signals in the "Debugging" chapter, and "Memory" chapter from me. 4. MCR and CFLAGS are implemented as registers. The value read for MCR is obtained from the DMux; all of the MCR bits on MEMC can be tested with "Test". CFLAGS is implemented using CFLAGS←A and B←CFLAGS; only the 4 bits written by CFLAGS←A are testable. 5. The prettyprint routines are all current with the latest Figure 9 and DMux signal list in the "Debugging" chapter. The following are not yet implemented: 1. BR readout (You can put BR on the display and write into it, but the value won't be read out.) 2. CACHE read/write; if you have any ancient documentation, it is wrong; none of the stuff for reading and writing the cache is implemented and it is being done as separate CACHEA and CACHED memories. The stuff for writing ProcSRN and reading out the Pipe registers is implemented, although this may not be too useful because ProcSRN is not on the MemC board--assuming that the lack of ProcSRN is neutralized somehow, then you can look at the Pipe stuff with Midas. Midas rearranges the B←Pipe0 to B←Pipe4 stuff into 4 separate memories called PVA, PCACHE, PMAP, and PERRS. You will probably be able to figure out how these are rearranged from the way they printout. I will document this pretty soon. I recommend that fairly early in the debugging, you attempt to get MCR and CFLAGS working via the "Test" command in Midas. You will have to manually set the mask for these (Mask is 177600 for MCR, 360 for CFLAGS) before testing. The MemC DMux signals are rearranged as detailed in the Table in the "Alto Debugging Interface" chapter. The changes from the way the hardware delivers these signals are essentially as follows: 1. The Cinxx bits are extracted and stored in two (unused) DMux words called CINH and CINL. 2. Aad[15:19]b is extracted and stored in an (unused) DMux word arranged so that there are four 0's to the right. 3. MemB[0:4] are extracted and stored right-justified in an (unused) DMux word. 4. All of the signals for MCR are extracted and stored in an (unused) DMux word. 5. All of the positions from which items were extracted are zeroed in the DMux readout. The "ShowMem" command file, which you can execute using "Read-Cmds" displays all of the memory stuff (except MCR, which is on the display anyway). I am planning to do BR, CACHE, and MemC DMux simulation, in that order; if that isn't appropriate, let me know. Please report any new bugs; I still have several old ones on my list so don't rereport them. ------- *start* 00282 00024 US Date: 8 NOV 1977 1555-PST From: FIALA Subject: Midas To: D1Users: New <D1>D1MidasRun.dm is released. This release implements the CACHEA and BR memories. However, there may be bugs in this code, particularly the code for reading/writing CACHEA. ------- *start* 01208 00024 US Date: 9 NOV 1977 0533-PST From: FIALA Subject: Midas To: D1Users: New Midas release with following changes: 1. "Passive" and "Active" commands implemented (The opposite of the current mode is in the menu.). Normal mode is "Active",which works as before. When "Passive", only registers whose values can be read/written without giving Dorado any clocks are updated/can-be-modified on the display. t1, t2, t3, go, step are supposed to work properly while passive. Other commands such as "Test" won't work and aren't error-checked yet. 2. Termination for a "Go" now monitors the "Running" bit in ESTAT rather than the DMux bit; the DWATCH item on the display allows you to select a particular DMux address to monitor with the scope while running. You should use the hardware address not the rearranged Midas address for this purpose. Midas will strobe the address in DWATCH into the hardware before "Go", "Continue", etc. 3. Some changes to "Reset" for the memory system (nothing important). It is moderately likely that the above changes to determine "Stopped" based upon "Running" won't work correctly. If so, please suffer with the old Midas until after Dealer. ------- *start* 02364 00024 US Date: 10 NOV 1977 0713-PST From: FIALA Subject: Midas To: D1Users: There is a new release of Midas on <D1>D1Midasrun.dm (usual place). This new release has extensive changes to implement the long-awaited "Poke" feature and also has some other modifications. A summary of the changes follows; new Midas documentation will be forthcoming. 1. Display now stays white when overlays swap in, rather than having the displaced display lines turn black. 2. Two new registers are added called STROBE and D1OUT. Storing into these registers will result, respectively, in a three-step strobing sequence or in a sta @DoradoOut operation as discussed in the "Alto Debugging Interface" chapter of the hardware manual. 3. The Passive/Active stuff is now fully implemented--the actions which only make sense in one mode now appear only when that mode is current. This resulted in a rearrangement of the command menu which will take awhile to get used to. 4. When you enter Passive mode, Midas restores the machine, as nearly as possible, to its state at the last breakpoint. During Passive mode, the only clocks given to Dorado are those you explicitly cause to be issued. You can freely write into MIR and ABMUX (which can be loaded without giving any clocks), but all other register writes are illegal. Midas freely reads the DMux whenever it suspects anything has changed. Registers not derived from DMux readout aren't read; they will retain whatever value they had when you entered Passive mode. 5. While in Passive mode, you can update all the registers by doing an "Update" command (or you can do "Active" then "Passive" again). 6. Continue-step and Continue-Go can be observed easily on the scope because there are no extraneous clocks issued. 7. A new command called "Rep-t2" will repeatedly clock the instruction in MIR and then reload MIR (for scope loops). 8. A new command file called "Poke" puts ABMUX, STROBE, and D1OUT on the display. The expected use of the STROBE and D1OUT stuff is that you will directly modify ABMUX and MIR, so that STROBE and D1OUT will only be used on the Control and Clock registers. There is more information that you will need to use the STROBE and D1OUT stuff, so don't use these yet unless you talk to me. (In 4 above, read "they will display ..." not "they will retain ...") ------- *start* 00305 00024 US Date: 11 NOV 1977 0709-PST From: FIALA Subject: Midas To: D1Users: This release of Midas has been pretty well checked out and bugs in previous versions are supposed to be fixed. Also added virtual and absolute printout for registers containing control store addresses. ------- *start* 00319 00024 US Date: 14 NOV 1977 1728-PST From: FIALA Subject: Midas To: D1Users: New Midas documentation is available as <D1DOCS>D1Midas.Ears or .Press. A copy of this is being put at the debugging console. Changes in the documentation reflect the various features which have been added recently. ------- *start* 00276 00024 US Date: 15 NOV 1977 0910-PST From: FIALA Subject: D1LANG.MC To: D1Users: I have just released a new <D1SOURCE>D1LANG.MC which adds a B←PIPE5 source and renames the B←CFLAGS function to be B←CONFIG. These conform to a hardware change on MemC. ------- *start* 00595 00024 US Date: 16 NOV 1977 1454-PST From: FIALA Subject: Midas Release To: D1Users: <d1>d1midasrun.dm;38 is released with the following changes: 1. Bug in TimeOut is fixed 2. Bug in "Test" of MCR is fixed 3. Bug reported in obscure case of "Go" is fixed 4. VNV memory is implemented (read-only) 5. Exit action is eliminated (use shift-swat) 6. Command menu is rearranged 7. Test and TestAll are idiot-proofed against execution in virtual mode 8. Numerous internal changes The hardware changes in the Pipe and the DMux signals are not in this release of Midas. ------- *start* 00798 00024 US Date: 17 NOV 1977 1959-PST From: FIALA Subject: Dorado Hardware Manual To: D1Users: The 17 November 77 release of the Dorado Hardware Manual is on <D1DOCS>d1manual.ears and d1manual.press (120 pp.). The significant changes in this release are: 1. Extensive revisions to the memory chapter, to the memory DMux signals in the debugging chapter, and to Figure 9. 2. Some additions to the control section about task switching. 3. Removal of the compressor section from the processor chapter (If you want to save the compressor stuff for historical reasons, you should retain the pages from the previous manual release.). 4. A few small changes in the slow io chapter. 5. Revisions to Figure 1A and incorporation of new figures 15 and 16, formerly omitted. ------- *start* 00742 00024 US Date: 18 NOV 1977 1931-PST From: FIALA Subject: Midas Release To: D1Users: <D1>D1MidasRun.dm is released. This release has been updated to reflect the changes in Figure 9 and the DMux signals and is supposed to agree with the hardware. Before loading this new Midas, you should delete SHOW*.MIDAS. I have renamed all of the "Show" command files so that they won't use up as much room in the menu. The new files are the same names as the old ones but omit the "Show" at the beginning of the name. There are also some other command files for showing rbase, membase, and br. There is a new memory called PREF which shows the new B←Pipe5 stuff. The CFLAGS register has been replaced by a CONFIG register. ------- *start* 00788 00024 US Date: 22 NOV 1977 2139-PST From: FIALA Subject: Midas Release To: D1Users: A new Midas release is on <D1>D1midasrun.dm;41. This new release has the following bug fixes and new features: 1. TestAll printout of memories is augmented to include address range. 2. TestAll is modified to include the tests of shmv, ef, if, rf, wf, br, and mcr. 3. The initialization has been speeded up via an improvement in MDI. 4. Bugs in BR, CACHEA, and VNV read-write have been fixed; BR is believed to work but CACHEA and VNV are untested. 5. A bug in the control section reset subroutine called by the Reset action and by Step and Go at new addresses has been fixed. 6. A bug in passive-mode Update is fixed. 7. The command files have been rearranged a little. ------- *start* 00673 00024 US Date: 1 DEC 1977 1921-PST From: FIALA Subject: Midas Release To: D1Users: A new version of midas is on <D1>d1midasrun.dm. This version has the following changes: 1. SimTest fixed to not flash when run after loading a program. 2. Initialization speeded up by reducing the size of the Midas.State file. 3. Bug in TimeOut action fixed. 4. <esc> made to operate as a "Continue" when "Test" is the last action. 5. New code added to read/write CacheA (still doesn't work). 6. CURRENT-ADDR added to test code and display to show the address of the memory word suffering a test error. 7. A number of internal changes to make more space. ------- *start* 01012 00024 US Date: 2 DEC 1977 0854-PST From: FIALA Subject: Midas Release To: D1Users: A new <D1>D1midasrun.dm is released. Changes are: 1. Addition of ADDR-INCREMENT which allows the address increment used in memory tests to be controlled (formerly this was always 1). If you set (32-bit) ADDR-INC negative, it will count backwards from the high to the low test address. 2. TRESET command file added. This is used to restore the right-hand display column the way Midas testing (as opposed to diagnostic testing) wants to see it. TRESET is also useful for resetting the address range and bits-checked stuff which gets modified while testing something else. In addition, this release (actually last release but I forgot to announce it) has a modification to the Skip and BackSkip (command file) commands allowing specification of the destination command number by a tag beginning with ".", e.g.: L X Skip .FOO L X ShowError Go at BEGIN timed out L X Skip .END .FOO ... ... .END ------- *start* 00206 00024 US Date: 5 DEC 1977 1523-PST From: FIALA Subject: D1LANG Release To: D1Users: I have just stored a new <D1SOURCE>D1Lang.mc. It fixes bugs in LDF T and LDF R with POS eq 0. ------- *start* 00260 00024 US Date: 7 DEC 1977 1823-PST From: FIALA Subject: Midas Release To: D1Users: <D1>D1MidasRun.dm;45 is released. This release contains the DMux signals for the MemX board and a new command file (MemX.midas) for displaying them. ------- *start* 00251 00024 US Date: 9 DEC 1977 1646-PST From: FIALA Subject: Midas Release To: D1Users: A new release of <D1>D1MidasRun.dm exists. It has the DMux signals for both MemX and MemD and fixes two problems with the previous release. ------- *start* 00389 00024 US Date: 12 DEC 1977 0804-PST From: MCDANIEL Subject: more files moved to ifs To: D1Users: I have moved more files to [ifs]<d1> and <d1source>. In particular, kernelsources.dm, contrl1.dm, kernel.mb, and control1.mb have been copied to ifs. Archive and deletion is pending on the [maxc] files. The pier and lampson changes to memc have been coalesced. Gene ------- *start* 00383 00024 US Date: 12 DEC 1977 2104-PST From: FIALA Subject: Midas Release To: D1Users: <D1>D1MidasRun.dm;47 is released. This release fixes the removal of WF←B and CNT←B encodings in BSEL and the addition of HOLD←B encoding in BSEL. It fixes several bugs in DMux printout. <esc> works for a few more situations in the Test code. HOLD←0 is done at brkp's. ------- *start* 00243 00024 US Date: 14 DEC 1977 1840-PST From: FIALA Subject: <D1>D1Midasrun.dm;48 To: D1Users: Is released. This one adds 11 new DMux signals for the MemX boards and fixes 4 other DMux printing bugs and a bug in Rep-T2. ------- *start* 00361 00024 US Date: 14 DEC 1977 2234-PST From: MCDANIEL Subject: new kernel on ifs To: D1Users: This version of kernel contains fixes to the code affected by holdSim. It appears to work correctly, but the absence of the memory boards makes it impossible to checkout. Set the flags w/ caution. There is a new listing in the notebook. Gene ------- *start* 00312 00024 US Date: 15 DEC 1977 1045-PST From: MCDANIEL Subject: new kernel, memc, memx To: D1Users: I have placed a new version of kernel, memc, memx on [ifs]. This release reflects software reorganization w/ no functional changes. Get the new sources if you want to modify the code. Gene ------- *start* 00321 00024 US Date: 19 DEC 1977 1113-PST From: FIALA Subject: <d1>d1midasrun.dm;49 To: D1Users: Is released. This one has numerous internal changes and fixes several problems with the "Reset" action; also rearranged some DMux printing on MemX board and nth round of revisions on CacheA read-write. ------- *start* 00426 00024 US Date: 20 DEC 1977 2145-PST From: FIALA Subject: <D1>D1midasrun.dm;50 To: D1Users: This new release fixes problems reported with the last one and increases the display size from 16 lines to 20. The various hardware inversions on the D board are also accounted for. The command files for operating the microdiagnostics will have to be revised slightly to handle the new display arrangement. ------- *start* 00251 00024 US Date: 30 DEC 1977 2151-PST From: FIALA Subject: <D1Docs>D1Midas.Press To: D1Users: A new release of the midas manual is on <D1Docs>. This release improves the discussion of command files and has some other changes. -------