| Title: | Alpha Support Conference |
| Notice: | This is a new Alphanotes, please read note 2.2 |
| Moderator: | VAXAXP::BERNARDO |
| Created: | Thu Jan 02 1997 |
| Last Modified: | Fri Jun 06 1997 |
| Last Successful Update: | Fri Jun 06 1997 |
| Number of topics: | 128 |
| Total number of notes: | 617 |
A customer is attempting to use EDIT/SUM on OpenVMS/Alpha. Although it
mostly works, there is at least one case where it consumes virtual memory
until it hits the wall whereupon an ACCVIO occurs. The same files "work"
on OpenVMS/VAX.
$ edit/sum/out=new.dml/list=temp.lis -
Q30758.dml -
/update=(q307581.upd,q307582.upd)
%SUM-W-EDITSCLSH, edits clash
- 361 USER$TSC:[GILLINGS.Q30758]Q307582.UPD;1
- 361 USER$TSC:[GILLINGS.Q30758]Q307581.UPD;1
%SUM-W-EDITSCLSH, edits clash
- 701 USER$TSC:[GILLINGS.Q30758]Q307582.UPD;1
- 701 USER$TSC:[GILLINGS.Q30758]Q307581.UPD;1
%SUM-W-EDITSCLSH, edits clash
- 746 USER$TSC:[GILLINGS.Q30758]Q307582.UPD;1
- 746 USER$TSC:[GILLINGS.Q30758]Q307581.UPD;1
%SUM-W-EDITSCLSH, edits clash
- 765 USER$TSC:[GILLINGS.Q30758]Q307582.UPD;1
- 765, 769 USER$TSC:[GILLINGS.Q30758]Q307581.UPD;1
%SUM-W-EDITSCLSH, edits clash
- 779 USER$TSC:[GILLINGS.Q30758]Q307582.UPD;1
- 779, 783 USER$TSC:[GILLINGS.Q30758]Q307581.UPD;1
%SUM-W-EDITSCLSH, edits clash
- 795 USER$TSC:[GILLINGS.Q30758]Q307582.UPD;1
- 795, 799 USER$TSC:[GILLINGS.Q30758]Q307581.UPD;1
%SUM-W-EDITSCLSH, edits clash
- 1003 USER$TSC:[GILLINGS.Q30758]Q307582.UPD;1
- 1003 USER$TSC:[GILLINGS.Q30758]Q307581.UPD;1
%SYSTEM-F-ACCVIO, access violation, reason mask=00,
virtual address=FFFFFFFF84F4BF94, PC=FFFFFFFF82887180, PS=0000001B
Improperly handled condition, image exit forced.
Signal arguments: Number = 0000000000000005
Name = 000000000000000C
0000000000010000
FFFFFFFF84F4BF94
FFFFFFFF82887180
000000000000001B
Register dump:
R0 = 0000000000000041 R1 = FFFFFFFF82892044 R2 = 000000007AFF7F78
R3 = 0000000084F4BF90 R4 = 000000007AFF7E00 R5 = 000000007AFF7F78
R6 = 000000007AFF7FC0 R7 = 000000000000001B R8 = 0000000000000021
R9 = FFFFFFFF84F4BF90 R10 = 000000007FFAD238 R11 = 0000000000020074
R12 = 0000000000000000 R13 = FFFFFFFF82892720 R14 = 0000000000000000
R15 = 000000007AFF79F0 R16 = FFFFFFFF84F4BF90 R17 = 0000000000000000
R18 = 0000000000000000 R19 = FFFFFFFF8143E690 R20 = 0000000000000001
R21 = 0000000000000001 R22 = 0000000000000003 R23 = 0000000000000003
R24 = FFFFFFFF8143E690 R25 = 0000000000000003 R26 = 0000000000000000
R27 = 0000000000002604 R28 = FFFFFFFF84F4BF90 R29 = 000000007AFF7C50
SP = 000000007AFF79F0 PC = FFFFFFFF82887180 PS = 300000000000001B
Process GILLINGS_1 logged out at 7-MAR-1997 12:10:30.26
Accounting information:
Buffered I/O count: 131 Peak working set size: 25984
Direct I/O count: 237 Peak virtual size: 260880
Page faults: 6230 Mounted volumes: 0
Charged CPU time: 0 00:00:02.19 Elapsed time: 0 00:00:27.83
Note the peak working set size and peak virtual size. This was a subprocess
which only executed the above command. The file sizes are:
Q30758.DML;1 380/380
Q307581.UPD;1 37/40
Q307582.UPD;1 20/20
The OpenVMS/VAX gives 3 additional warnings and completes with much less
resource consumption:
%SUM-W-EDITSCLSH, edits clash
- 1003 DISK$USER1:[GILLINGS.Q30758]Q307582.UPD;1
- 1003 DISK$USER1:[GILLINGS.Q30758]Q307581.UPD;1
%SUM-W-EDITSCLSH, edits clash
- 1757 DISK$USER1:[GILLINGS.Q30758]Q307582.UPD;1
- 1757 DISK$USER1:[GILLINGS.Q30758]Q307581.UPD;1
%SUM-W-EDITSCLSH, edits clash
- 1777 DISK$USER1:[GILLINGS.Q30758]Q307582.UPD;1
- 1777, 1781 DISK$USER1:[GILLINGS.Q30758]Q307581.UPD;1
%SUM-W-EDITSCLSH, edits clash
- 6734 DISK$USER1:[GILLINGS.Q30758]Q307582.UPD;1
- 6734 DISK$USER1:[GILLINGS.Q30758]Q307581.UPD;1
Accounting information:
Buffered I/O count: 274 Peak working set size: 575
Direct I/O count: 183 Peak virtual size: 3372
Page faults: 681 Mounted volumes: 0
Checked on OpenVMS/Alpha V6.1 and V7.1, same result (V7.1 trace posted)
Obviously a bug, but perhaps there's an OpenVMS Archeologist out there
somewhere who can read the heiroglyphics inside the UPD files and
figure out what's going wrong. Files are at XARA(59.516)::Q30758*.*.
John Gillings, Sydney CSC
| T.R | Title | User | Personal Name | Date | Lines |
|---|---|---|---|---|---|
| 56.1 | (?wrong conf) source bug: your IPMT :) | COMEUP::SIMMONDS | lock (M); while (not *SOMETHING) { Wait(C,M); } unlock(M) | Thu Mar 06 1997 23:12 | 17 |
Simple code bug in SUMSLP.EXE, module UPDATE, routine DEC_LTOR:
DEC_LTOR: .JSB_ENTRY INPUT=<R2,R7>, OUTPUT=<R2,R3,R7>
CLRL R3 ; Clear high register
10$: .JSB_ENTRY INPUT=<R2,R7>, OUTPUT=<R2,R3,R7>
EDIV #10,R2,R2,R0 ; Get least significant digit
[...]
should be (for example) :
DEC_LTOR: .JSB_ENTRY INPUT=<R2,R7>, OUTPUT=<R2,R3,R7>
CLRL R3 ; Clear high register
10$: .JSB_ENTRY INPUT=<R2,R7>, OUTPUT=<R2,R3,R7>
CLRL R3 ; Clear high register
EDIV #10,R2,R2,R0 ; Get least significant digit
[...]
John.
| |||||
| 56.2 | Mr Carter, I presume? | AUSS::GARSON | DECcharity Program Office | Fri Mar 07 1997 00:42 | 3 |
re .1
Well, I'm impressed. Pity there's no PATCH for Alpha.
| |||||
| 56.3 | COMEUP::SIMMONDS | lock (M); while (not *SOMETHING) { Wait(C,M); } unlock(M) | Fri Mar 07 1997 19:50 | 11 | |
Re: .1
-< (?wrong conf) source bug: your IPMT :) >-
^^^^^^^^^^^
I withdraw that remark, as the code error is clearly a porting 'slip'..
(there appears to have been a goal during our VAX MACRO code port to use
more registers and to eliminate stack references for result temporaries
where possible; this particular ported MACRO module doesn't seem to have
been picked up and used in OpenVMS VAX V7.1, btw...)
..and this Conference seems to be our only choice for Alpha porting topics.
| |||||
| 56.4 | ZIMBRA::BERNARDO | Dave Bernardo, VMS Engineering | Fri Mar 07 1997 20:01 | 3 | |
>>and this Conference seems to be our only choice for Alpha porting topics.
nah, you've also got VMSNOTES.
| |||||