pkimpel / retro-b5500 Goto Github PK

The color and positioning of some buttons on the B5500 Operator Console were 
not correct. In particular, Card Load Select and Memory Check were exchanged, 
the Power On button should be green, and the Halt and Memory Check buttons 
should be red. Card Load Select and Halt should light when activated.

These corrections were taken from a picture of a real B5500 Console, courtesy 
of Sid McHarg. We had to guess about the Memory Check button color, since it 
was missing from the console in the picture.

To be released in emulator version 0.13.

Reported by Fausto Saporito on 2013-06-28:

Hello,

an interesting update.

I fixed most of the syntax errors in the fortran source (i'm still modifying 
the sequence numbers).
From 170 errors I got 7 errors... then I fixed other two... and trying to 
compile I got this :

5:ALGOL/FORTRAN= 1 BOJ 2026
 CRA IN CARD A 1:ALGOL/FORTRAN= 1
 PBD0143 OUT LINE:ALGOL/FORTRAN= 1
-FLAG BIT ALGOL/FORTRAN= 1, S= 3, A= 209
 CRA REL CARD A 1:ALGOL/FORTRAN= 1
 0:PRNPBT/DISK= 2 BOJ 2115
 ALGOL/FORTRAN= 1 DS-ED 2116

I reduced the stack to 1000 (from 3000), and I don't know if this could be a 
problem.

The compiling aborted before reaching the end of it.. the listing is stopping 
in this way:

                 DDT111:  EMITDDT(1,1,1) ;                                                      04088000    0130
                          END;                                                                  04089000    0131 
                          E0(LQV) ; % @14: LOGICAL EQUIVALENCE.                                 04090000    0131 
                          ; % @15: FLOAT.                                                       04091000    0133 
                          GO DDT111 ; % @16: ISIGN, SIGN.                                       04092000    0133 
                          BEGIN E0(RDV); GO AINT END ; % @17: MOD.                              04093000    0133 
                          E0(LOR) ; % @20: LOGICAL OR.                                          04094000    0135
                          BEGIN E0(XCH); GO AIMAG END ; % @21: REAL.                            04095000    0136        
########  -FLAG BIT ALGOL  /FORTRAN= 1, S =   3, A = 209  
######################################################

Attached there's the new job.

Regards,

Fausto

SYMBOL/COOL at 02711000:

is it "((30-I) DIV 2)" or "((80-I) DIV 2)" ?

The coldstart loader script, when doing a cold start, attempts to create the 
SYSTEM/LOG file for the MCP's use. The MCP either removes the file or refuses 
to use it.

The behavior of webUI/B5500SPOUnit occasionally becomes erratic:

  * Appears to allow input while output is being typed.
  * Gets into a state where input cannot be terminated. Clicking End of Message or pressing Enter appears to act like the Error button. Input Request goes out, but immediately comes back on again.

These may be separate, independent problems or symptoms of a common one. The 
sequence of events to cause this is unknown. It's possibly and 
emulator/processor problem, but is more likely in the SPO logic.

Last character on Algol compiler line 10880000 was originally a "-", but may be 
"~" (:=) instead. Very difficult to determine from the scanned listing. Changed 
in both ALGOL and ALGOLXEM sources.

From Tim Sirianni:
2)  Also, has anyone suggested to do the same thing you're talking about with 
disks [support multiple indexedDB databases for separate disk subsystems and 
and system configurations] except for tapes? Set up a sort of a virtual tape 
library. The tape drives could be directly unloaded into the IDB-backed 
library, indexed by the VSN. As soon as you hit UNLOAD, the data on the tape 
just goes to the library. Later we could type a VSN into a UI that pops up when 
a tape drive's LOAD is pressed and have that tape load. Tapes could be exported 
from and imported to the library in the same way the tape drives currently 
handle it. Also fun would be if tape drives could be restored to a saved state 
(including online/offline, write-protect, and "which VSN/scratch tape is 
loaded?" status) at power on time.

The B5500ColdLoader.html utility does not properly initialize the date in the 
DIRECTORYTOP segment. After initialization it comes out as 1/0/0.

There may be a problem initializing time as well, but not sure if the 
ColdLoader is the source of that problem.

Unit busy for DKA and DKB when doing a read-check operation is currently reset 
when the disk-finished interrupt occurs, not at I/O completion. This may not be 
correct. We had to change the way the printer worked so that busy was reset at 
I/O complete, and the disk may require the same change to CentralControl and 
IOUnit.

What steps will reproduce the problem?
1. CE on SPO to start cande
2. login and make a algol
3. try to type '700  WRITE (LINE,<"X = ",F8.5>,X);~'
4. type LIST 700

What is the expected output? What do you see instead?
a line '700  WRITE (LINE,<"X = ",F8.5>,X);' is expected
a line '700  WRITE (LINE"X = ",F8.5,X);' is seen

What version of the product are you using? On what operating system?
0.15

Please provide any additional information below.

Attempted to run the XREF/JONES program from the SPO, equating the CARD file to 
SYMBOL/XREF. Program appeared to start fine, then got open messages for a 
printer-backup file. After several seconds, I/O activity in the emulator 
ceased, and XREF/JONES appeared to be idle. Program could be DS-ed 
successfully, and system seemed to operate normally otherwise.

Line number at which the program was DS-ed is immediately after a write to the 
printer file. Suspect the problem is something in I/O, or possibly a dropped 
interrupt.

Attempting to run with Processor 2 enabled typically results in a system-level 
invalid address "punt". The second processor's Normal light blips briefly, but 
then the system halts.

The B5500 Reference Manual states that the field used for comparing link words 
is [9:24] (Algol bit numbering), but the B5500 Processor Technical Manual 
states it is [18:15] (Algol numbering -- it actually states the field is 
30=>16, which is hardware numbering).

Assume for now that the Reference Manual is correct, as [9:24] appears to be 
used in several places in the MCP by LLL operations. This means that the 
comparison field would be the mantissa portion of the word, which makes sense.

We have reports of invalid address and flag-bit errors, especially when the 
system is under heavy load. We also see occasional disk errors, which look like 
they may be caused by bad memory addresses.

From Paul Cumberworth email of 2013-11-18:

<quote>
I had also experienced the DKA failure a few weeks ago and took a copy for 
interest sake.  I put it down to a corrupted directory at the time.
Inline image 1

I've have also notice the increased frequency of Invalid Addresses is TSSMCP, 
and there also appears some strange behavior in DCMCP under load. Occasionally 
on large compile for instance, I will get a flag bit error, on rerun it will 
run satisfactorily to completion.  A good example is two concurrent compiles of 
the cobol compiler (seems to be load sensitive), reasonably often one of the 
compiles will get a flag bit error.  I did get the odd flag bit error 
previously but hadn't really worried about it too much, but the frequency is 
definitely higher.
</quote>

Code to format the code segment address at 02182760 uses L.[45:2] instead of 
L.[46:2]. Original bug in patch #114 for Mark XVI.0.122 Algol compiler.

*** DO NOT FIX ***

This is an original error in the bitsavers.org listing that should be preserved.

Reported by Terry Heidelberg on 2012-12-15.

            -------- Original Message --------
            Subject: Re: Emulator version 0.07
            From: Fausto Saporito <[email protected]>
            To: Paul Kimpel <[email protected]>
            Cc: Nigel Williams <[email protected]>
            Date: 6/24/2013 2:09 PM
>             Hello,
>
>             ok... after several tries, I got the FATAL:
>
>             5:ALGOL/FORTRAN= 1 BOJ 0911
>              CRA IN CARD A 1:ALGOL/FORTRAN= 1
>              PBD0121 OUT LINE:ALGOL/FORTRAN= 1
>             1 D
>
>              SINVALD LINK
>             trying to abort the job, after the usual CRA stop.
>
>             Maybe this is related to the Algol stack size ... I'll try with 
larger value.
>
>             BFN,
>             Fausto
>
>             2013/6/24 Fausto Saporito <[email protected]>
>
>                 Hi Paul,
>
>                 it's the FORTRAN compiler deck, I sent you.
>                 By the way, the emulator is working, here is the output of 
SPO:
>
>                  5:ALGOL/FORTRAN= 1 BOJ 0607
>                  CRA IN CARD A 1:ALGOL/FORTRAN= 1
>                  PBD0120 OUT LINE:ALGOL/FORTRAN= 1
>                 1 TS
>                 INV KBD 1 TS
>                 MX
>                  5:ALGOL/FORTRAN= 1
>
>                 1 TL
>                  TIME LIMITS FOR ALGOL/FORTRAN= 1 ARE: PRT=NO LIMIT; IOT=NO 
LIMIT.
>                 1 TI
>                  TIME FOR ALGOL/FORTRAN= 1 IS: CP= 14:20, IO= 6:36 IN 17:11
>                 OL CRA
>                  CRA IN USE BY ALGOL/FORTRAN= 1:0000000 CARD A 001 00000 01
>
>                 At the moment, the CRA is stil freezed...
>
>                 Regards,
>                 Fausto
>
>                 2013/6/24 Paul Kimpel <[email protected]>
>
>                     That certainly sounds like some sort of emulator bug. If 
you'll send me that big deck, I'll try to reproduce it.
>
>                     -------- Original Message --------
>                     Subject: Re: Emulator version 0.07
>                     From: Fausto Saporito <[email protected]>
>                     To: Paul Kimpel <[email protected]>
>                     CC: Nigel Williams <[email protected]>
>                     Date: Mon, 24 Jun 2013, 11:03 am
>>                     Hello,
>>
>>                     with the big FORTRAN deck, I'm experiencing very often 
the CRA block... it seems to happen without any valid reason.
>>                     The emulator seems to halt, and on the console (in this 
case) I have just TIMR flashing.
>>
>>                     PA Slack freezed at 81% and PA rate freezed at 4.4%
>>
>>                     I'll try to wait...
>>
>>                     BFN,
>>                     Fausto

From Nigel Williams:
Implement support for at least two datacom stations in the web-based emulator, 
so that the multi-user aspect of the B5500 TSMCP can be demonstrated.

Time of day (e.g., in response to a WT SPO command) frequently does not format 
properly, e.g., TIME IS 2L3M.

As observed thus far, the letters in the time are always in the range J-R, 
which may indicate that the internal time counter from which the time is being 
formatted (XCLOCK) has gone negative, as character mode in the processor 
converts negative binary numbers to characters with a "J-sign" in the low-order 
position. Thus, since the hours and minutes fields of the time are formatted 
separately, 2L3M could indicate a time of -2334.

The Mark XVI ESPOL source contains an original typo at line 02558000. The 
mnemonic for Bit Set, syllable code 52 (64 octal), should be "BIS", not "BIT". 
The latter is the mnemonic for Bit Test.

** DO NOT CORRECT **

This is an original error in the source that was correctly transcribed from the 
listings on bitsavers.org, and should be preserved.

At the end of the CMN (Enter Character Mode In Line) instruction, it executes a 
portion of the RDA (Recall Destination Address) instruction starting at J=2 to 
load the character-mode destination address. If the TOS word is a descriptor, 
the instruction checks for presence, and if not sets the presence-bit 
interrupt. Neither the Reference Manual or Training Manual state whether this 
interrupt is inhibited in Control State. We are going to assume it is.

It would be desirable to separate the B5500 Console and its controls from the 
main browser window. This would allow the Console to be positioned on the 
screen or brought to the top much more easily without the main browser window 
overlaying several peripheral device windows.

ANSN(14) = 0000014000      ≠   RSTN(14) = 000009900+

015130     IF F1D20 IS POSITIVE THEN MOVE 99 TO RSTN (14)
015140         ELSE MOVE 14 TO RSTN (14).

The double-precision arithmetic operators need to be implemented and tested.

If the emulator running the MCP idles for a very long time (hours) with nothing 
to do, it responds very slowly. It takes several seconds for the Input Request 
light to come on after clicking it, and the reply message comes out at perhaps 
0.5 cps.

This behavior has been observed in Firefox 19. Could it be evidence of some 
sort of memory leak or long chain of closures that induces a lot of garbage 
collection overhead?

it looks like there is one parameter too many:
 case 0x25: // XX45: ISO=Variable Field Isolate op
this.adjustAFull();
t2 = variant >>> 3; // number of whole chars
if (t2) {
t1 = this.G*6 + this.H; // starting source bit position
t2 = t2*6 - (variant & 7) - this.H; // number of bits
if (t1+t2 <= 48) {
this.A = this.cc.fieldIsolate(this.A, t1, t2);
} else { // handle wrap-around in the source value
this.A = this.cc.fieldInsert(
//                           vvvvvvvvvvvvvvvvvvvvv
this.cc.fieldIsolate(this.A, 0, t2-48+t1, t1+t2-48), 48-t2, 48-t1,
//                           ^^^^^^^^^^^^^^^^^^^^^
this.cc.fieldIsolate(this.A, t1, 48-t1));
}
// approximate the shift cycle counts
this.cycleCount += (variant >>> 3) + (variant & 7) + this.G + this.H;
this.G = (this.G + variant >>> 3) & 7;
this.H = 0;
}
break;

Attempting to boot the system using Card Load Select does not work in the 
B5500Console.html UI. It does work using SyllableDebugger, however.

What steps will reproduce the problem?
1.list line 433
2.
3.

What is the expected output? What do you see instead?
this.abnormal=true
this.abnomal=true


What version of the product are you using? On what operating system?
0.19 
Linux Slackware 12.1



Please provide any additional information below.

The emulator formerly ran in Google Chrome just fine, but starting sometime in 
the Fall of 2013, now has problems. It starts out okay, but after a few 
minutes, all of the emulator windows crash with a Chrome "Aw Snap!" error. The 
only recovery is to close all of those windows and open a new one for the 
console.

It appears the crash is due either to a memory leak in the emulator or 
inadequate garbage collection by Chrome. The Chrome Memory Profiler shows 
steadily increasing memory utilization until 300-400MB, at which point Chrome 
aborts the scripts.

IF WAITIO(@500000000,@120,U) ! 0 THEN%
  BEGIN; STREAM(BUFF); DS ~ 11 LIT "WRITE LOCK~";% 
…

IF (R~WAITIO(@500000000,@177,U))!0 THEN
  IF R!@120 THEN %ERROR OTHER THAN WRITE LOCK
    BEGIN;STREAM(U~TINU[U],BUFF);
        BEGIN DS~14 LIT "#CANNOT PURGE ";
           …
        END;                                     
        LABELTABLE[U]~@214;      
        GO TO EXIT;
      END ELSE %NO WRITE RING
          BEGIN; STREAM(BUFF); DS ~ 11 LIT "WRITE LOCK~";%
..
         END

     } else {
        if (this.writeRing) {
//            this.buildErrorMask(0); 
                                } else {
            this.errorMask |= 0x50;     // RD bits 26 & 28 => no write ring, don't return Mod III bits
        }
        finish(this.errorMask, 0);

   DO UNTIL NOT (T1~WAITIO(@500000000,@165,U)) OR    

Floating-point computations do not appear to produce the correct results. The 
results are similar, but different significantly from what is expected in a 
couple of program output listings from 1970. Suspect a rounding or 
normalization problem.

From Tim Sirianni:
1)  I would love to be able to run multiple instances of the emulator, with the 
possibility of running shared disk among them.  I believe indexedDB is capable 
of shared access, isn't it?   Even if shared disk isn't possible it would be 
fun to run multiple emulators, which I guess means being able to set the name 
of the disk subsystem IDB to open, so this might be an easy change to make 
while you're in the code.  Yes, this is a way to get around the just-one-thread 
javascript execution model.  Two one-processor emulated system should run twice 
as fast a two-processor one on a two-core real system.

When selecting and copying text from the LinePrinter device "paper" area using 
Firefox 33.1 under Windows 7 and then pasting it into a text editor, there is 
an extra newline in the pasted text every three lines. 

This behavior does not present using Google Chrome. Other browsers and 
platforms have not yet been tested.

This three-line interval corresponds to the internal way the paper area is 
formatted -- each group of three lines is contained within a separate <pre> 
element so that the green-bar effect can be applied with CSS classes. 
Apparently the adjacent <pre> elements cause the Firefox browser to generate an 
extra newline when converting the copied portion of the DOM to text.

This may be a browser-specific issue that we cannot fix within the emulator.

Have seen some instances of leading zeroes not being suppressed in Algol 
I-format output.

Tim Sirianni reports some leading non-zero digits are being replaced by spaces 
in Algol formatted output. See attached

After compiling a new compiler and doing an MC on the new compiler name, the MC 
does not persist across a halt/load.

? USER=TERRY
? COMPILE 0COB/TERRY WITH COBOL FOR LIBRARY
? DATA CARD
000100 * HELLO WORLD PROGRAM                                            00100
000200 IDENTIFICATION DIVISION.                                         00200
000300 PROGRAM-ID.    HELLO.                                            00300
000400 * ENVIRONMENT DIVISION.                                          00400
000500 * DATA DIVISION.                                                 00500
000700 PROCEDURE DIVISION.                                              00700
000800     DISPLAY "HELLO, WORLD"                                       00800
001000     STOP RUN.                                                    99999999
? END

The ESPOL compiler declares the variable NESTOG twice in the same block: (1) as 
a define at 01001310, and (2) as a Boolean variable at 017723000. From the 
cross-reference, it appears that the B5500 compiler ignores the define and uses 
the Boolean. This variable is never used as the destination of an assignment, 
so it's probably just dead code (from a no-longer implemented $-option). The 
define was commented out in ESPOLXEM, so for now will do the same in the base 
ESPOL source.

Entering the CD command on the SPO appears to crash the emulator. This has been 
tried and observed only using the SyllableDebugger thus far. Reloading the 
SyllableDebugger in the browser with F5 and restarting the MCP appears to work, 
but typically get a DKA(R) DISK FAILURE-D21 (EU not ready) error message 
shortly after that. Completely exiting the browser and starting from scratch 
appears to clear the error, and the disk data does not appear to be damaged.