sigma/sigma_bugs.txt

1. CPU: register pointer was being loaded from PSW1 instead of PSW2.
2. CPU: register pointer was not preserved on an XPSD 0.
3. CPU: some decode points were taking unimplemented rather than non-existent
   instruction traps.
4. CPU: BDR branches on result >=0, instead of > 0.
5. CPU: PSD change instructions were not saving old PC in PC queue.
6. CPU: illegal register pointer does not stop the system on the Sigma 5-7;
   instead, registers read as 0's, and writes are ignored.
7. CPU: traps were setting the Sigma 9 vector field unconditionally.
8. CPU: CH does not SEXT Rn<16:31>.
9. CPU: MTW and MTH overflow traps not implemented.
10. CPU: MTH condition code calculations not implemented correctly.
11. CPU: illegal shift opcodes on the Sigma 5-7 are treated as arithmetic
    single (according to the AUTO diagnostic).
12. FP: SF did not calculate condition codes correctly.
13. FP: SF left did not detect normalized if the count reached zero.
14. FP: SF right was retaining a guard digit incorrectly.
15. FP: SF right test was not taking into account single/double precision.
16. FP: fp_unpack killed the operand sign before testing it.
17. FP: fp_pack did not recomplement a negative operand.
18. CPU: DW overflow test was incorrect.
19. IO: device status bit field was incorrectly defined.
20. IO: AIO was testing for == 0 instead of != 0.
21. TT: input keyboard character mappings were incorrect.
22. SYS: shifts were not displaying or accepting an index register.
23. CPU: mapping from pulse interrupt number to counter interrupt number
    was incorrect.
24. CPU, CIS: handling of instructions aborts was incorrect.
25. CPU: CC2,4 set incorrectly on aborted stack instruction.
26. CPU: CC2,4 set incorrectly on MSP,0 instruction.
27. CPU: PLW not converting operand address to byte address.
28. CPU: PSM, PLM not converting limit operand address to byte address correctly.
29. CPU: PLM wrote registers in inverse order.
30. Definitions: number of virtual, physical pages derived incorrectly.
31. MAP: MMC control set wrap incorrect.
32. MAP: MMC register update of Rn|1 incorrect.
33. CPU: CBS set CC's incorrectly on unequal.
34. CPU: TTBS set CC's incorrectly on early exit.
35. IO: chan_proc_epilog and four other routines extracted device address incorrectly.
36. CPU: DW was fetching halfwords instead of words.
37. CPU: multiples were not setting condition codes on 0 operands.
38. FP: floating add zero test on second operand was inverted.
39. FP: floating add zero cases failed to set the result variable.
40. FP: floating add zero cases failed to go through postnormalization logic.
41. FP: double precision add macro lost high carry out.
42. FP: double precision add macro had an undocumented restriction about operands.
43. FP: all double precision instructions unpacked the wrong low order memory operand.
44. FP: normalization failed to decrement the exponent.
45. FP: in single precision, removing the guard digit created a spurious low-order digit.
46. FP: denormalization failed to clear the low word (single precision).
47. FP: denormalization failed to clear the guard digit (double precision).
48. FP: add/subtract failed to treat abnormal 0's as normal numbers.
49. CIS: WriteDecA set local rather than global condition codes.
50. CIS: Illegal digit check missed byte 0.
51. CIS: Overflow check set wrong condition codes.
52. CIS: NibbleLShift routine overflow check was (also broken in VAX, PDP11).
53. CIS: WordLShift routine overflow check was wrong (also broken in VAX, PDP11).
54. CIS: WordLShift had signed/unsigned variable conflict in compare.
55. CIS: WriteDecA not setting correct condition codes for -0.
55. CIS: WriteDECA not clearing CC3/4 before setting.
56. CIS: DM shift-and-test loop did 14 iterations instead of 15.
57. CIS: DM final shift of 16 done outside non-zero case instead of inside.
58. CIS: DD put quotient and remainder in wrong registers.
59. CIS: DD did not shift remainder to proper place, based on dividend/divisor widths.
60. CIS: DD overflow test missed exact 16 digit quotient case.
61. CIS: algorithm to compute length of operand failed for certain lengths (also broken
    in VAX, PDP11).
62. CIS: DM and DD answer was wrong on restart cases. [NOT FIXED YET]
63. CIS: EBS fetched pattern from wrong address pointer.
64. CIS: EBS field separator wrote wrong byte.
65. MAP: Access codes defined incorrectly for access protection check.
66. CPU: LPSD fetching operands with write check rather than read check.
67. CPU: Illegal instruction stop (and other trap stops) weren't rolling back the PC.
68. CPU: History routine merged CC's into history array incorrectly.
69. CPU: History routine stored incremented rather than actualy PC.
70. CPU: History reporting routine printed wrong operand value, due to colliding declarations.
71. IO: Power up/down interrupts should not be implemented.
72. IO: WD was not recalculating int_hiact.
73. IO: Interrupt chain was not linked at powerup.
74. IO: Interrupt chain link algorithm was wrong.
75. IO: WD processing was spanning [beg,end) rather than [beg,end].
76. RTC: RTC interrupts were happening at the wrong level.
77. PTR: leader skip was testing for channel end.
78. PTR: leader skip was testing per command instead of per reel/file.
79. PTR: end of file was treated as fatal error instead of channel end + length error.
80. PTR, PTP: units were not marked UNIT_SEQ.
81. LPT: unit was not marked UNIT_SEQ.
82. IO: device address set/show updated/displayed wrong field in single unit devices.
83. CPU: address calculation for MTx in interrupt location used incorrect length.
84. LPT: print, format opcode definitions inverted.
85. LPT: wrong index used to count buffer fill loop.
86. IO: CPU/IOP communications through 20/21 were not modelled.
87. CIS: Multiply algorithm incorrect, multiply restart not modelled.
88. CIS: Divide algorithm incorrect, divide restart not modelled.
89. IO: WD .44 not recognized as effective NOP.
90. CPU: Sigma 5 (uniquely) does not implement CVA or CVS.
91. IO: AIO merging status incorrectly.
92. IO: system for returning status byte from IO was muddled; rewrite required.
93. RAD: 7232 track shift offset was incorrectly defined.
94. CPU: MTX for interrupts returned wrong value.
95. CPU: Counter overflow trigger equation was incorrect.
96. CPU: MTX interrupts not recalculating interrupt summary values.
97. RTC: SET/SHOW C1-C4 routines were broken.
98. DP: comparison for sector field overflow was incorrect.
99. DP: cross-cylinder incorrectly set on read to end of cylinder.
100. DP: seek never executed.
101. DP: seek followon state codes set incorrectly.
102. DP: cylinder offset defined incorrectly.
103. DP: SIO not initing unit thread properly for non-zero units.
104. MUX: Line enable must be remembered for lines that are not currently connected.
105. RAD: write check was reading words off disk instead of bytes.
106. RAD: end of transfer routine checking for address error incorrectly.
107. DK: end of transfer routine checking for address error incorrectly.


Diagnostic Notes
----------------

1. PATTERN paper tape (SA = 60, end pass = 1AC), SET CPU RBLKS=32.
   Break at 60 is triggered once during loading process.
2. VERIFY paper tape (SA = 140, end pass = 8E6).
3. AUTO paper tape (SA = 105), SET CPU LASLMS to suppress spurious error message.
   AUTO magtape runs correctly with no messages.
4. SUFFIX magtape (SA = 100, end pass = 115), no printouts.
5. FLOAT magtape.
6. DECIMAL paper tape (SA = EE).
7. PROTECT paper tape (SA = F9), long runtime until printout.
8. MAP paper tape (SA = 7C).
9. MEDIC paper tape (SA = 68), set SSW2 after breakpoint, long runtime.
10. INT magtape, stops after M6, set SSW2 and continue,
   enters pattern generator and loops forever, as expected.
11. RTC paper tape, clocks must be 500Mhz, reports clocks as "too slow" due
   to faster simulation speed.
   Runs correctly with SET THROTTLE 500K.