Tape Sort

[halfmeg]

Jay J, does tape backspace work for the TOS system IPL with Hercules
version 3.0 ?

Have been digging around for more info on the tape sort. An older
SYNCSORT version (3.4 circa 1989) specifies that a tape sort may only
use the 24-bit addressing mode, cannot be invoked by PGM=SYNCSORT but
by PGM=SORT, SORTLIB containing the special tape sort program be
specified instead of the regular SYNCSORT library. All this leads me
to believe SYNCSORT folks used the OS/360 sort software almost
unchanged from the distribution we have. Earlier posts concerning
similarity in messages indicated the original source may have been
the OS/360 sort as well.

The sortwork DD statement as coded in an example in the SYNCSORT
manual looks like:

//SORTWK01 DD UNIT=TAPE

which is identical to my test job's DD statements. In the details
however, it states the UNIT parameter must be the installation
specific name chosen to define a tape device. The MVS Turnkey system
is genned with 3420s. OS/360 Sort knows about 2400s and 3400s
because it estimates sort capacity by type and density of the drive.

This from OS/360 Sort source:

THE FOLLOWING ASSUMPTIONS ABOUT DENSITY MUST BE MADE :
* A 7 TRACK UNITS WITHOUT SPECIF. DENSITY ARE ASSUMED 800 BPI.
* B 9 TRACK UNITS WITHOUT SPECIFIED DENSITY ARE ASSUMMED = 1600 BPI
<snip>
STH R6,CPINWKU PUT NO. OF WORK UNITS IN CPI
CLC CIWKFCT+2(2),CITP2400 ARE WORK UNITS 2400 TAPE
BE CI060A IF TRUE TAKE BRANCH FIR
CLC CIWKFCT+2(2),CITP3400 IS WORKUNITS 3400 TAPE FIR
BNE CI012 IF NOT TAKE BRANCH FIR
TM CIWKFCT+1,X'C0' ARE WORKTAPES 9-TRACKS FIR
BE CI012 IF TRUE TAKE BRANCH

IIRC Jay Moseley changed the MVT Sort distribution he provides after
my tinkering with the COBOL sort to include tape by adding the parm
2400 to the gen of SORT. His distibution is what Volker has included
with the Turnkey MVS system.

I am now wondering if it should/can be 3400 instead. Attempting to
attach a 2420 didn't work with Hercules 3.0 version.

Phil - think SYNCSORT may have added 3480s and 6250bpi to tape sort


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[halfmeg]

Post by halfmeg
Jay J, does tape backspace work for the TOS system IPL with
Hercules version 3.0 ?

Does previous problem with TOS IPL still occur with Hercules version
3.0 ? ( Might have been read backward instead of backspace. )

Post by halfmeg
<snip>
I am now wondering if it should/can be 3400 instead. Attempting to
attach a 2420 didn't work with Hercules 3.0 version.

Looks like 2400 is only choice in SYSGEN for SORTMERG GEN MACRO.

Have attempted the Tape Sort with Volker's Turnkey MVS system on the
P/390 and have gotten a different error.

19.31.55 JOB 48 *IEF233A M 480,SCRTCH,SL,SORTRCP,HELOWRLD
19.31.55 JOB 48 *IEF233A M 481,SCRTCH,SL,SORTRCP,HELOWRLD
19.31.55 JOB 48 *IEF233A M 482,SCRTCH,SL,SORTRCP,HELOWRLD
19.31.55 JOB 48 *IEF233A M 483,SCRTCH,SL,SORTRCP,HELOWRLD
19.34.26 JOB 48 IEC705I TAPE ON 480,111111,SL,6250
BPI,SORTRCP,HELOWRLD
19.34.26 JOB 48 IEC705I TAPE ON 481,222222,SL,6250
BPI,SORTRCP,HELOWRLD
19.34.31 JOB 48 IEC705I TAPE ON 482,333333,SL,6250
BPI,SORTRCP,HELOWRLD
19.34.31 JOB 48 IEC705I TAPE ON 483,444444,SL,6250
BPI,SORTRCP,HELOWRLD
19.34.31 JOB 48 IEA000I 480,IOE,0C,0E20,,,111111,SORTRCP ,19.34.31
19.34.31 JOB 48 IEA000I
480,,,,0040000200402D000008000000000000000000000000001A
19.34.31 JOB 48 IEA000I 481,IOE,0C,0E20,,,222222,SORTRCP ,19.34.31
19.34.31 JOB 48 IEA000I
481,,,,0040000200402D000008000000000000000000000000001A
19.34.32 JOB 48 IEF450I SORTRCP HELOWRLD - ABEND S0C1 U0000 -
TIME=19.34.32

These IOEs do not appear with Hercules test. The SORT error message
is different in description also:

IER061A - I/O ERR SORTRCP ,HELOWRLD,480,TA,SORTWK01,0C- OP,PROGRAM
CHECK ,0000043,EXCP

Have noticed OS/360 Sort OC1s a lot as it seems to fill a word with
zeros then branches to execute there after the real error has already
been encountered.

Another difference between Hercules and P/390 executions:

The two other SORTWORK tapes are empty on the P/390, with Hercules
they both have some valid data on them.

Makes me think P/390 doesn't support readbackwards for tapes at all,
while Hercules may just have a minor glitch.

Phil - anyone know if P/390 can read an AWS tape backwards ??


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Rent DVDs from home.
Over 14,500 titles. Free Shipping
& No Late Fees. Try Netflix for FREE!
http://us.click.yahoo.com/mk9osC/hP.FAA/3jkFAA/W4wwlB/TM
---------------------------------------------------------------------~->

[cube1us]

Yes, the problem was with read backward (not backspace).

Jay Jaeger

Post by halfmeg

Post by halfmeg
Jay J, does tape backspace work for the TOS system IPL with
Hercules version 3.0 ?

Does previous problem with TOS IPL still occur with Hercules

version

Post by halfmeg
3.0 ? ( Might have been read backward instead of backspace. )

Post by halfmeg
<snip>

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[cube1us]

I don't know -- I haven't tried it. I have been extremely busy since
I got it working under 2.17b9. I saw a reference to the fixed code
in the release announcement that Jay Maynard sent out.

Jay Jaeger

Post by halfmeg
Jay J, does tape backspace work for the TOS system IPL with

Hercules

Post by halfmeg
version 3.0 ?

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[halfmeg]

Post by cube1us
I don't know -- I haven't tried it. I have been extremely busy
since I got it working under 2.17b9. I saw a reference to the
fixed code in the release announcement that Jay Maynard sent out.
Jay Jaeger

Post by halfmeg
Jay J, does tape backspace work for the TOS system IPL with

Hercules

Post by halfmeg
version 3.0 ?

Understand, no rush. In fact it may work if block size is small and
data is not split across pages in host (?) . At least that is what I
think I see in CHANNEL.C or TAPEDEV.C code. The error that is
present now is for large blocks and the index/pointer for the
remainder of the block after current read is done is not
incremented/decremented. In other words it keeps reading in the
beginning of the block and never reads in the end unless the block is
small and fits within the page (?) .
/* Copy data between main storage and channel buffer */
if (IS_CCW_RDBACK(code))
{
idadata = (idadata - idalen) + 1;
memcpy (dev->mainstor + idadata,
iobuf + dev->curblkrem, idalen);
}
else
{
if (readcmd)
memcpy (dev->mainstor + idadata, iobuf, idalen);
else
memcpy (iobuf, dev->mainstor + idadata, idalen);

/* Increment buffer pointer */
iobuf += idalen;
}

Should this last line concerning buffer pointer be between 1st IF and
ELSE also ?

Phil - remember fellas C newbie here :-)


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[cube1us]

No, I don't think so. I doubt it would be quite that simple. My
recollection of that code for read backwards, generally, was that one
was always copying from the *end*, so one didn't need to increment
iobuf. I think the critical variable is *probably* curblkrem, which
is probably handled OK. It may well be that *something* needs to be
done, but I doubt it would be to increment iobuf.

If there really is a problem, I suspect it will take a little bit of
careful thought to fix it correctly.

I do vaguely recall some special code having to do with page
boundaries and/or IDA mode which I think I did *not* modify, or that
I was suspicious of -- that looked "funny". Because I wasn't at all
sure what to modify, I left it alone, figuring that no virtual memory
system was likely to actually use read backwards, and having no test
case anyway. You might want to search thru the archives and see
where I posted the code, to see if there was any discussion of that
code.

Jay Jaeger

Post by halfmeg

Post by cube1us
I don't know -- I haven't tried it. I have been extremely busy
since I got it working under 2.17b9. I saw a reference to the
fixed code in the release announcement that Jay Maynard sent out.
Jay Jaeger

Post by halfmeg
Jay J, does tape backspace work for the TOS system IPL with

Hercules

Post by halfmeg
version 3.0 ?

Understand, no rush. In fact it may work if block size is small and
data is not split across pages in host (?) . At least that is what I
think I see in CHANNEL.C or TAPEDEV.C code. The error that is
present now is for large blocks and the index/pointer for the
remainder of the block after current read is done is not
incremented/decremented. In other words it keeps reading in the
beginning of the block and never reads in the end unless the block is
small and fits within the page (?) .
/* Copy data between main storage and channel buffer */
if (IS_CCW_RDBACK(code))
{
idadata = (idadata - idalen) + 1;
memcpy (dev->mainstor + idadata,
iobuf + dev->curblkrem, idalen);
}
else
{
if (readcmd)
memcpy (dev->mainstor + idadata, iobuf, idalen);
else
memcpy (iobuf, dev->mainstor + idadata, idalen);
/* Increment buffer pointer */
iobuf += idalen;
}
Should this last line concerning buffer pointer be between 1st IF and
ELSE also ?
Phil - remember fellas C newbie here :-)

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[halfmeg]

Post by cube1us
No, I don't think so. I doubt it would be quite that simple. My
recollection of that code for read backwards, generally, was that
one was always copying from the *end*, so one didn't need to
increment iobuf. I think the critical variable is *probably*
curblkrem, which is probably handled OK. It may well be that
*something* needs to be done, but I doubt it would be to increment
iobuf.
If there really is a problem, I suspect it will take a little bit
of careful thought to fix it correctly.
I do vaguely recall some special code having to do with page
boundaries and/or IDA mode which I think I did *not* modify, or
that I was suspicious of -- that looked "funny". Because I wasn't
at all sure what to modify, I left it alone, figuring that no
virtual memory system was likely to actually use read backwards,
and having no test case anyway. You might want to search thru the
archives and see where I posted the code, to see if there was any
discussion of that code.
Jay Jaeger
<snip>

Have taken your advice and researched a little on your and others
earlier postings. Was curious about what all the data movement
(memmove) was with read forward. Saw you comment about sliding the
data forward to the beginning of IOBUF. The same thing probably
needs to happen for read backward only the slide needs to be toward
CURBLKREM. This would allow all the current coding to remain
unchanged after adding the slide in the IS_CCW_RDBACK(code) section
in TAPDEV.C where the current comment is "We don't need to move
anything in this case - just set length".

We have created a tape for testing this stuff. LRECL is 2048 and
blksize is 20480 (we were going to try to test 64K blocksize also but
got a JCL error on invalid number in DCB parms). We put together a
small program to read it in and traced the tape drive while reading.
The result of this is much clearer than with a sort job.

CCW=02FF5420 44005000=> <snip> .\0..\8.......y.
IDAW=00E0F000 Len=800=> <snip> 01 01 01 01
IDAW=00E0F800 Len=800=> <snip> 02 02 02 02
IDAW=00DCA000 Len=800=> <snip> 03 03 03 03
IDAW=00DCA800 Len=800=> <snip> 04 04 04 04
IDAW=00DC9000 Len=800=> <snip> 05 05 05 05
IDAW=00DC9800 Len=800=> <snip> 06 06 06 06
IDAW=00DC8000 Len=800=> <snip> 07 07 07 07
IDAW=00DC8800 Len=800=> <snip> 08 08 08 08
IDAW=00DC7000 Len=800=> <snip> 09 09 09 09
IDAW=00DC7800 Len=800=> <snip> 10 10 10 10
Stat=0C00 Count=0000 => <snip> .\0..\8.......y.
CCW=03000000 60400001=> <snip> ......ok........
Stat=0C00 Count=0000 <snip>
CCW=08FF5B90 0009CF30=> <snip> ......&.....- ..
CCW=02FF5448 44005000=> <snip> ..}...Q...-.....
IDAW=00DCD000 Len=800=> <snip> 11 11 11 11
IDAW=00DCD800 Len=800=> <snip> 12 12 12 12
IDAW=00DC6000 Len=800=> <snip> 13 13 13 13
IDAW=00DC6800 Len=800=> <snip> 14 14 14 14
IDAW=00DC5000 Len=800=> <snip> 15 15 15 15
IDAW=00DC5800 Len=800=> <snip> 16 16 16 16
IDAW=00DC4000 Len=800=> <snip> 17 17 17 17
IDAW=00DC4800 Len=800=> <snip> 18 18 18 18
IDAW=00DC3000 Len=800=> <snip> 19 19 19 19
IDAW=00DC3800 Len=800=> <snip> 20 20 20 20
Stat=0C00 Count=0000 => <snip> ..}...Q...-.....
CCW=03000000 60400001=> <snip> ......ok........
Stat=0C00 Count=0000 <snip>
CCW=08FF5EC0 0009CF68=> <snip> ......&.....- ..
CCW=02FF5470 44005000=> <snip> ..\...Y.........
IDAW=00DCE000 Len=800=> <snip> 21 21 21 21
IDAW=00DCE800 Len=800=> <snip> 22 22 22 22
IDAW=00DC2000 Len=800=> <snip> 23 23 23 23
IDAW=00DC2800 Len=800=> <snip> 24 24 24 24
IDAW=00DC1000 Len=800=> <snip> 25 25 25 25
IDAW=00DC1800 Len=800=> <snip> 26 26 26 26
IDAW=00DC0000 Len=800=> <snip> 27 27 27 27
IDAW=00DC0800 Len=800=> <snip> 28 28 28 28
IDAW=00DBF000 Len=800=> <snip> 29 29 29 29
IDAW=00DBF800 Len=800=> <snip> 30 30 30 30

We unfortunately don't have the knowledge to put together the read
backward test. The test above was very simple, open, get, etc...

Our experience with the duplication of records (and truncation of
remainder of the block) during read backwards while sorting still
indicates a problem. The test tape is in the files area for anyone
that is interested in solving this thing one way or the other.

Phil - good luck



------------------------ Yahoo! Groups Sponsor ---------------------~-->
Rent DVDs Online - Over 14,500 titles.
No Late Fees & Free Shipping.
Try Netflix for FREE!
http://us.click.yahoo.com/vhSowB/XP.FAA/3jkFAA/W4wwlB/TM
---------------------------------------------------------------------~->

[halfmeg]

Post by halfmeg
<snip>
We unfortunately don't have the knowledge to put together the read
backward test. The test above was very simple, open, get, etc...
Our experience with the duplication of records (and truncation of
remainder of the block) during read backwards while sorting still
indicates a problem. The test tape is in the files area for anyone
that is interested in solving this thing one way or the other.

But how we can tinker with stuff. Below is a small example of a
modified exeuction of CBT370 contents to read and read backwards. It
can also FSF and other things as well. A well rounded tape testing
utility for Hercules could be created from it for someone who knows
what they are doing. I could only modify it a little (can't get it
to read the whole block in (chained reads?)).

Anyway if I issue 3 read forwards then 8 reads back I should cover
the same territory. Not so yet. See trace below:

0280:CCW=02DD1628 20000028=> <snip> 01 01 01 01
0280:Stat=0C00 Count=0000 => <snip> 11 11 11 11
0280:CCW=02DD1628 20000028=> <snip> 11 11 11 11
0280:Stat=0C00 Count=0000 => <snip> 21 21 21 21
0280:CCW=02DD1628 20000028=> <snip> 21 21 21 21
0280:Stat=0D00 Count=0028 => <snip> 21 21 21 21


0280:CCW=0CDD1628 20000028=> <snip> 21 21 21 21
0280:Stat=0D00 Count=0000 => <snip> 1 21 21 21
0280:CCW=0CDD1628 20000028=> <snip> 1 21 21 21
0280:Stat=0C00 Count=0000 => <snip> 1 21 21 21
0280:CCW=0CDD1628 20000028=> <snip> 1 21 21 21
0280:Stat=0C00 Count=0000 => <snip> 1 21 21 21
0280:CCW=0CDD1628 20000028=> <snip> 1 21 21 21
0280:Stat=0C00 Count=0000 => <snip> 1 21 21 21
0280:CCW=0CDD1628 20000028=> <snip> 1 21 21 21
0280:Stat=0D00 Count=0000 => <snip> 1 21 21 21
0280:CCW=0CDD1628 20000028=> <snip> 1 21 21 21
0280:Stat=0C00 Count=0000 => <snip> 1 21 21 21
0280:CCW=0CDD1628 20000028=> <snip> 1 21 21 21
0280:Stat=0C00 Count=0000 => <snip> 1 21 21 21
0280:CCW=0CDD1628 20000028=> <snip> 1 21 21 21
0280:Stat=0C00 Count=0000 => <snip> 1 21 21 21
0280:CCW=0CDD1628 20000028=> <snip> 1 21 21 21
0280:Stat=0E00 Count=0000 => <snip> 1 21 21 21

This also shows up the missing byte I thought I saw earlier in sort.
It also shows a need to detect load point as EOF on Read Backward
(that is if Read Backward worked correctly).

Phil - alright, I'll be quiet about this for a while :->



------------------------ Yahoo! Groups Sponsor ---------------------~-->
Rent DVDs Online - Over 14,500 titles.
No Late Fees & Free Shipping.
Try Netflix for FREE!
http://us.click.yahoo.com/vhSowB/XP.FAA/3jkFAA/W4wwlB/TM
---------------------------------------------------------------------~->

[cube1us]

Post by halfmeg
Have taken your advice and researched a little on your and others
earlier postings. Was curious about what all the data movement
(memmove) was with read forward. Saw you comment about sliding the
data forward to the beginning of IOBUF. The same thing probably
needs to happen for read backward only the slide needs to be toward
CURBLKREM. This would allow all the current coding to remain
unchanged after adding the slide in the IS_CCW_RDBACK(code) section
in TAPDEV.C where the current comment is "We don't need to move
anything in this case - just set length".

If I remember the code right, when reading forward, it needed to move
the remaining data that in the AWS "buffer" (IOBUF) towards the front
for subsequent transfers to memory. When reading backwards, you do
not need (nor do you want) to do that - you just pick different
points for the move.

Remember, that at least at the point I sent code in, read backwards
of simple records, in non-IDA mode, such as required to IPL TOS,
WORKED FINE.

So, I would suggest that before you start fussing with the code or
worrying too much, trying to IPL TOS under 3.0. If that works, then
I would look at what is different between the reads you are doing and
those used by TOS. If in fact you are using IDA in the channel
program, then I would focus on that *particular* code, and do so by
comparing it to the non-IDA code (which you know works after you get
TOS to IPL).

One other point: The data that the Hercules IO trace shows is pretty
useless for read backwards. It typically shows the wrong part of the
buffer, and I did *not* fix that (as far as I can recall). So you
can't go by what the Hercules trace shows.

Jay Jaeger



------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[halfmeg]

Post by halfmeg
<snip>
If I remember the code right, when reading forward, it needed to
move the remaining data that in the AWS "buffer" (IOBUF) towards
the front for subsequent transfers to memory. When reading
backwards, you do not need (nor do you want) to do that - you just
pick different points for the move.

That may be the way it has been coded. Remember wondering about the
movement of data after it had been read in ( from HOST, to HERCULES,
to GUEST OS, to application program (?) ). Am still not sure why the
movement is necessary instead of using a pointer.

If read forward requires data shuffling, then it would seem read
backward would require the same type of shuffling. If read backward
doesn't need data shuffling, then it would seem read forward could
use the same method as read backward in reverse. Increment or
decrement a pointer instead of moving data around multiple times.

Post by halfmeg
Remember, that at least at the point I sent code in, read backwards
of simple records, in non-IDA mode, such as required to IPL TOS,
WORKED FINE.

As I posted that it might with the modifications which were made.

Post by halfmeg
So, I would suggest that before you start fussing with the code or
worrying too much, trying to IPL TOS under 3.0. If that works,
then I would look at what is different between the reads you are
doing and those used by TOS. If in fact you are using IDA in the
channel program, then I would focus on that *particular* code, and
do so by comparing it to the non-IDA code (which you know works
after you get TOS to IPL).

Quick review:

Two known instances of requirements for tape read backward function.

1. TOS IPL
2. TAPE SORT

Two individuals wanting read backwards to work.

1. Jay - TOS
2. Phil - SORT

Two wildly divergent knowledge levels ( both system internals and C )

1. Jay - Experienced
2. Phil - Tinker - Apprentice level 2

Modifications made by Jay - Successful TOS read backward

Tape Sort performed by Phil - 1 unique record # read in - same unique
record # duplicated 5 times in sortwork file before abend - Wrong
Length Block error. Also remainder of block missing, it is as if the
beginning of the block was read in 5 times instead of incrementing to
obtain the next section of the block. Something really wrong, so
look for simple test program to demo error very plainly. Find CBT
program which I can alter CCWs in to Read Forward then Read Backward
via EXCPs like SORT does, strange output.

Post by halfmeg
One other point: The data that the Hercules IO trace shows is
pretty useless for read backwards. It typically shows the wrong
part of the buffer, and I did *not* fix that (as far as I can
recall). So you can't go by what the Hercules trace shows.

Uggghhhhh!!!!!!!

Ok, at this point I have, with Volker's kind assistance, built a new
CYGWIN area for compiling Hercules in house.

Have a KLUDGE to suggest which might permit all READ BACKWARD
functions to work correctly and the trace. Ready for it!!

Revert CHANNEL.c and TAPEDEV.c back to pre-version 3.00. Add a flag
to switch on or off upon execution of READ BACKWARD. The first time
x'0C' is encountered; backspace 1 block and read forward 1; if two
x'0C's consecutive; backspace 2 blocks and read forward 1; reset flag
if there are not two consecutive x'0C's.

Phil - rigged for silent running on this subject for now -- unless :-)


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Rent DVDs from home.
Over 14,500 titles. Free Shipping
& No Late Fees. Try Netflix for FREE!
http://us.click.yahoo.com/ybSovB/hP.FAA/3jkFAA/W4wwlB/TM
---------------------------------------------------------------------~->

[cube1us]

It might seem so, but that isn't how it works right now.

It isn't so much a matter of what the particular code in question
does as much as it is a matter of what the *rest* of the code
involved is expecting.

Yes, it could be done with just pointers, but that wasn't how the
*rest* of the code was set up.

Generally, the *rest* of the code expected that the buffer began in
one place. When reading forward, that meant moving the AWS data that
had not yet been placed into memory up towards the front of the
buffer.

For read backwards, that meant simply pretending that the AWS data
got shorter (by the amount already transferred from the end of the
buffer to memory (for read backwards, the *end* of the data is the
*beginning* of the AWS buffer).

Have you (or whomever it is that is working on SORT) simply tried to
IPL the TOS image with 3.00 yet, to see if the patches are still
working in the context of 3.00? I haven't yet had time, but might
later tonite and tomorrow. I am expecting a non-trivial effort to
get 3.00 running -- perhaps having to update CYGWIN yet again.

Alternatively, you could to back to 2.19b7 and apply just my tape
patches and see what happens.

In either case, if TOS ipl's OK, but sort does not work, and sort is
in fact using IDA, then that would be the area to focus on.

Jay Jaeger

Post by halfmeg
If read forward requires data shuffling, then it would seem read
backward would require the same type of shuffling. If read

backward

Post by halfmeg
doesn't need data shuffling, then it would seem read forward could
use the same method as read backward in reverse. Increment or
decrement a pointer instead of moving data around multiple times.

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[Kevin Leonard]

Post by cube1us
Have you (or whomever it is that is working on SORT) simply tried to
IPL the TOS image with 3.00 yet, to see if the patches are still
working in the context of 3.00?

TOS IPLs with no problem under 3.00.
--
Kevin


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[Mark Drummond]

I ran into a data corruption problem with read backwards and found
two causes. They both are related to read backwards with IDALs. I
don't know if this is related to the Tape Sort problem and I'm still
experiencing a tape positioning problem myself that I am trying to
debug.

Both data corruption problems are in channel.c.

At line 1465:

/* Pre-decrement first idaw */
idaw -= (idaseq == 0 ? 1 : 0);

This line should be deleted. The address in the first IDAW does not
need to be decremented by 1. This line causes the first block of
data to be moved 1 byte to the left of where it was intended. On a
read backward, data is copied from the specified address to the 2k or
4k LOWER boundary. If the IDAW contains value X'8000', only 1 byte
is supposed to be copied at location X'8000'. If the IDAW contains
X'7FFF', 2kb or 4kb of data is copied to x'6000'-x'7fff' or x'4000'-
x'7fff'.

At line 1563:

/* Copy data between main storage and channel buffer */
if (IS_CCW_RDBACK(code))
{
idadata = (idadata - idalen) + 1;
memcpy (dev->mainstor + idadata,
iobuf + dev->curblkrem, idalen);
}

Replace this with:

/* Copy data between main storage and channel buffer */
if (IS_CCW_RDBACK(code))
{
idadata = (idadata - idalen) + 1;
memcpy (dev->mainstor + idadata,
iobuf + dev->curblkrem + idacount - idalen,
idalen);
}

Without this change, the data area for every IDAW gets the same data
copied into it. dev->curblkrem is the offset in ioibuf where the
data for this CCW is located. idacount is initially equal to the CCW
count. idalen is the amount of data that can fit into the area where
the IDAW points. For read backwards, the first IDAW gets the
rightmost data, the next IDAW gets the data to the left of the first,
and the thirds get the data to the left of the second, etc. This
change implements this type of data movement.

I uploaded a file called ReadBackwardsTest.aws that can be used to
test the read backwards logic. The tape has 3 data blocks: 24-byte
IPL test block, 136-byte CCW/IDAW block and 256-byte data block. The
tape is IPL'able and reads the same 256-byte block of data into
storage in 8 locations.

The reads are as follows:

Read (forward) into x'1000'
Read (backward) into x'2000'
Read (forward) with IDAWs into x'4f80'
Read (backward) with IDAWs into x'5f80' (data split across
page)
Read (forward) with IDAWs into x'6f80'
Read (backward) with IDAWs into x'7f00' (data fits entirely
in first IDAW area)
Read (forward) with data chaining into 3000
Read (backward) with data chaining into 4000

After you IPL the tape, and the block is read the 8 times, a wait
state is entered. Issue these commands to display the storage blocks:

r 1000.100
r 2000.100
r 4f80.100
r 5f80.100
r 6f80.100
r 7f00.100
r 3000.100
r 4000.100

All 8 storage locations must have the same data or else there is a
data corruption problem.

If you IPL this tape under VM, the read backward with data chaining
block @ x'4000' is corrupted by VM (the block contains data from the
3rd and 1st blocks on the tape - very weird).

Mark D.

Post by cube1us
It might seem so, but that isn't how it works right now.
It isn't so much a matter of what the particular code in question
does as much as it is a matter of what the *rest* of the code
involved is expecting.
Yes, it could be done with just pointers, but that wasn't how the
*rest* of the code was set up.
Generally, the *rest* of the code expected that the buffer began in
one place. When reading forward, that meant moving the AWS data that
had not yet been placed into memory up towards the front of the
buffer.
For read backwards, that meant simply pretending that the AWS data
got shorter (by the amount already transferred from the end of the
buffer to memory (for read backwards, the *end* of the data is the
*beginning* of the AWS buffer).
Have you (or whomever it is that is working on SORT) simply tried to
IPL the TOS image with 3.00 yet, to see if the patches are still
working in the context of 3.00? I haven't yet had time, but might
later tonite and tomorrow. I am expecting a non-trivial effort to
get 3.00 running -- perhaps having to update CYGWIN yet again.
Alternatively, you could to back to 2.19b7 and apply just my tape
patches and see what happens.
In either case, if TOS ipl's OK, but sort does not work, and sort is
in fact using IDA, then that would be the area to focus on.
Jay Jaeger

Post by halfmeg
If read forward requires data shuffling, then it would seem read
backward would require the same type of shuffling. If read

backward

Post by halfmeg
doesn't need data shuffling, then it would seem read forward

could

Post by cube1us

Post by halfmeg
use the same method as read backward in reverse. Increment or
decrement a pointer instead of moving data around multiple times.

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[Norman Hollander]

I have completed a customized WLM Goal Mode
Service Policy especially for Hercules. If anyone would
like it (and will install it), I'll be happy to send it
(with related
docs) to you.



------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[Julio - IG]

Hi Norman.

I am interested, could you send it to me?

Thanks
Julio

----- Original Message -----
From: "Norman Hollander" <norman-hollander-***@public.gmane.org>
To: <hercules-390-***@public.gmane.org>
Sent: Wednesday, November 26, 2003 3:55 PM
Subject: [hercules-390] WLM Goal Mode Service Policy for Hercules

Post by Norman Hollander
I have completed a customized WLM Goal Mode
Service Policy especially for Hercules. If anyone would
like it (and will install it), I'll be happy to send it
(with related
docs) to you.
http://groups.yahoo.com/group/hercules-390
http://www.conmicro.cx/hercules
Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[Hailin Wu]

Can you send me to have a look? Thanks.

Post by Norman Hollander
I have completed a customized WLM Goal Mode
Service Policy especially for Hercules. If anyone
would
like it (and will install it), I'll be happy to send
it
(with related
docs) to you.

__________________________________
Do you Yahoo!?
Free Pop-Up Blocker - Get it now
http://companion.yahoo.com/

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[sjb390]

Hello Norman !

I am very interested on your work. Please send me a copy !

so long.

Joerg Burdorf

Post by Norman Hollander
I have completed a customized WLM Goal Mode
Service Policy especially for Hercules. If anyone would
like it (and will install it), I'll be happy to send it
(with related
docs) to you.

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[halfmeg]

Post by Mark Drummond
I ran into a data corruption problem with read backwards and found
two causes. They both are related to read backwards with IDALs. I
don't know if this is related to the Tape Sort problem and I'm
still experiencing a tape positioning problem myself that I am
trying to debug.
<snip>

Hi Mark,

Thanks for picking up the baton, I am a little too weak in capacity
to do much other than tinker with the code here and there. While I
was making a few changes however, it occurred to me that the 1st
movement of a IDA block might not be handled by the IDA code section
(True - False). If True, then perhaps some of the lengths/position
pointers/currblkrem variables may be getting altered to where IDA
can't work correctly without resetting of correct variable values.

Phil - Tape Sort test with your mods coming soon to a post near you



------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[Fish]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

(Apologies for not replying sooner; been down with the flu since
Thanksgiving...)

Post by Mark Drummond
I ran into a data corruption problem with read backwards
and found two causes. They both are related to read
backwards with IDALs. [...]

<snip>

Great work, Mark! I've saved your post so we can eventually get this
fixed for you in a future release. Sorry this issue has been causing
you can Phil and Jay (and others?) so much grief. We'll eventually
get it straightened out I'm sure.

- --
"Fish" (David B. Trout)
fish-6N/dkqvhA+***@public.gmane.org

Fight Spam! Join CAUCE!
http://www.cauce.org/

-----BEGIN PGP SIGNATURE-----
Version: PGP 7.0.4

iQA/AwUBP9PJVEj11/TE7j4qEQKwJwCfQYvBJyq3hL6fwBBYjchugkubAKsAoORh
oyhgv2zX0RVxjwO+WfP76Yrr
=AmmZ
-----END PGP SIGNATURE-----


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[Gregg C Levine]

Hello (again) from Gregg C Levine
My sympathies, Fish. That's the worst thing to happen to anyone. That
being said, regarding Mark Drummond's discovery, I also remember going
over it with him, and I vaguely remember the issue also concerning
itself, with the channel command structure.

How that happened, I can't recall, but I do remember the discussions
both on, and off the H390-MVS list, concerning itself with this IBM
Pub GA22-6974-10. If you want my copy, its in Bookie format, I'd be
glad to send you it, otherwise I don't know what else to do, because
IBM didn't generate a copy in PDF form. The site does have it there,
that is readable in their HTML format, though. Perhaps it’s due to the
way I think things through.
-------------------
Gregg C Levine hansolofalcon-XfrvlLN1Pqtfpb/***@public.gmane.org
------------------------------------------------------------
"The Force will be with you...Always." Obi-Wan Kenobi
"Use the Force, Luke."  Obi-Wan Kenobi
(This company dedicates this E-Mail to General Obi-Wan Kenobi )
(This company dedicates this E-Mail to Master Yoda )

-----Original Message-----
Sent: Sunday, December 07, 2003 7:44 PM
Subject: RE: [hercules-390] Re: Read Backward - was -Tape Sort
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
(Apologies for not replying sooner; been down with the flu since
Thanksgiving...)

Post by Mark Drummond
I ran into a data corruption problem with read backwards
and found two causes. They both are related to read
backwards with IDALs. [...]

<snip>
Great work, Mark! I've saved your post so we can eventually get this
fixed for you in a future release. Sorry this issue has been causing
you can Phil and Jay (and others?) so much grief. We'll eventually
get it straightened out I'm sure.
- --
"Fish" (David B. Trout)
Fight Spam! Join CAUCE!
http://www.cauce.org/
-----BEGIN PGP SIGNATURE-----
Version: PGP 7.0.4
iQA/AwUBP9PJVEj11/TE7j4qEQKwJwCfQYvBJyq3hL6fwBBYjchugkubAKsAo
ORh
oyhgv2zX0RVxjwO+WfP76Yrr
=AmmZ
-----END PGP SIGNATURE-----

( Sign on back of Imperial Star Destroyer, "Kick me!". )


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[A. Philip Dickinson]

Does anyone have any performance data on a 3033?

I would like to put together a PC with performance that is similar to a
3033, and I was wondering if anyone has done any analytical testing of
Hercules performance on an Intel machine versus real hardware.

My recollection is that IBM decided to define the performance of a 158
as 1 MIP and then quote performance of later models relative to that.

i.e. I recall that a 3032 was 3 MIPS (per cpu?), a 3081 was 8 (per
cpu?). This may be totally wrong.

Those were the three systems I worked on as an Operator and then
Sysprog.

I have done a little testing using Greg's Imon, and I came up with about
60 Intel PPRO Mhz per S/370 MIP, about 110 Intel Celeron Mhz per S/370
MIP and about 75 AMD "Mhz" (but not really) per S/370 MIP. (Based on an
IBM MIP being 120% of the Imon Test 2, as per the Imon help
description).

This was very unscientific and just represents the three machines I had
available at the time.

I know this ignores IO capability, but so did IBM when quoting MIPs, and
this would be very configuration dependent anyway (as it would on an
Intel box. My multi bus RAID proliant has better IO than my laptop for
instance).

Finally, would anyone be interested in a configuration option in
Hercules to limit it's performance to a specified MIP rate? I think
this would be interesting as it would allow me to emulate multiple small
systems (like a 3033) on a single fast Intel box, with some predictable
performance.

Thoughts?

Phil.






------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[A. Philip Dickinson]

The 3033 had support for 26 bit real addressing.

How was this supported under MVS (by one of the SE variants, I guess).

Does anyone have a pointer to a manual that might educate me?

Does Hercules have the necessary support for 26 bit real addressing?

Phil.

[Jay Maynard]

Post by A. Philip Dickinson
The 3033 had support for 26 bit real addressing.
How was this supported under MVS (by one of the SE variants, I guess).

It required MVS/SE3, IIRC, or MVS/SP. The extra memory was only accessible
in virtual mode, as the high order two bits were supplied by bits in the
page table entry.

Post by A. Philip Dickinson
Does Hercules have the necessary support for 26 bit real addressing?

It's there, but I can't say that it's been well tested.

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Rent DVDs Online - Over 14,500 titles.
No Late Fees & Free Shipping.
Try Netflix for FREE!
http://us.click.yahoo.com/vhSowB/XP.FAA/3jkFAA/W4wwlB/TM
---------------------------------------------------------------------~->

[Binyamin Dissen]

On Sat, 01 Nov 2003 15:59:36 +1100 "A. Philip Dickinson"
<phil-+***@public.gmane.org> wrote:

:>The 3033 had support for 26 bit real addressing.

:>How was this supported under MVS (by one of the SE variants, I guess).

It was weird.

The support was only for real pages about the 16m line in virtual mode, i.e.,
the page table entries could point to real storage above the line while in
real mode the pages were not accessible.

One of the few times (the only time?) VM CP went into virtual mode was to
access those pages.

:>Does anyone have a pointer to a manual that might educate me?

:>Does Hercules have the necessary support for 26 bit real addressing?

--
Binyamin Dissen <bdissen-***@public.gmane.org>
http://www.dissensoftware.com

Director, Dissen Software, Bar & Grill - Israel

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[zapzap50]

Have a look at
http://www.computer-archiv.de/comp0028.htm#3033
Although it is in german , it is quite easy to read
Bruno
zapzap1(at)wanadoo(dot)fr

Post by A. Philip Dickinson
Does anyone have any performance data on a 3033?

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[halfmeg]

Post by A. Philip Dickinson
Does anyone have any performance data on a 3033?

Buried here and there are some performance "things" concerning older
processors. Check the bottom of the below message:

http://groups.yahoo.com/group/hercules-390/message/17969

Some of the other tapes/files which were compressed may give
indications of measurable data.

Another web link gives a performance column compared to 360/50 or
370/158 along with some other data:

http://www.jcmit.com/cpu-performance.htm

Post by A. Philip Dickinson
<snip>
Finally, would anyone be interested in a configuration option in
Hercules to limit it's performance to a specified MIP rate? I think
this would be interesting as it would allow me to emulate multiple
small systems (like a 3033) on a single fast Intel box, with some
predictable performance.
Thoughts?

IBM might be interested. The only way to "cripple" the speed that I
can think of is to tie it to the HOST clock with sleeps (thinking
sleep is a clock interrupt type of thing instead of a loop till clock
expired thing). This would hopefully permit full cycles of the HOST
CPU(s) to continue other work without emulation spinning through a
CPU consuming loop to slow execution.

Phil - time to go hiking


------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[kyle5091]

This is amazing --- we work hard at trying to improve performance --
we get faster and faster machines -- now someone wants to "Cripple"
the speed.

I know it's out of interest but I still find it amazing.

"Back to the Future" I guess. !!!!!

Post by halfmeg

Post by A. Philip Dickinson
Does anyone have any performance data on a 3033?

IBM might be interested. The only way to "cripple" the speed that I
can think of is to tie it to the HOST clock with sleeps (thinking
sleep is a clock interrupt type of thing instead of a loop till clock
expired thing). This would hopefully permit full cycles of the HOST
CPU(s) to continue other work without emulation spinning through a
CPU consuming loop to slow execution.
Phil - time to go hiking

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[A. Philip Dickinson]

Well actually.....

I'd like to make use of the rather impressive speed of my dual Athlon
system or my quad xeon to emulate large numbers of 3033's, but in a
realistic manner.

What I'd really like is to emulate eight 3033MP's (total sixteen
emulated CPU's) all managed in a loosely coupled environment under JES3.
This would have been the largest system possible in 1980 and the largest
system supported by my favourite operating system MVS 3.8J (unless we
can get the > 2 cpu stuff to work).

What configuration of intel or AMD processors do I need to achieve this?


Phil D. - eagerly awaiting a CTC that will work with JES3.

-----Original Message-----
Sent: Tuesday, 4 November 2003 7:43 PM
Subject: [hercules-390] Re: How fast was a 3033?
This is amazing --- we work hard at trying to improve performance --
we get faster and faster machines -- now someone wants to "Cripple"
the speed.
I know it's out of interest but I still find it amazing.
"Back to the Future" I guess. !!!!!

Post by halfmeg

Post by A. Philip Dickinson
Does anyone have any performance data on a 3033?

IBM might be interested. The only way to "cripple" the speed that

I

Post by halfmeg
can think of is to tie it to the HOST clock with sleeps (thinking
sleep is a clock interrupt type of thing instead of a loop till

clock

Post by halfmeg
expired thing). This would hopefully permit full cycles of the

HOST

Post by halfmeg
CPU(s) to continue other work without emulation spinning through a
CPU consuming loop to slow execution.
Phil - time to go hiking

------------------------ Yahoo! Groups Sponsor
---------------------~--> Buy Ink Cartridges or Refill Kits
for your HP, Epson, Canon or Lexmark Printer at MyInks.com.
Free s/h on orders $50 or more to the US & Canada.

http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[Ronald Tatum]

Phil,
Your remark about "eagerly awaiting a CTC that will work with JES3"
puzzles me...
Are you implying that the CTC adapter that was actually used for LASP
(local attached support processor) that was used by numerous shops on
360/85s, 370/195s, etc., running various releases of OS/360 failed to
operate properly under JES3, the successor to LASP? I believe that TI
upgraded in the middle/late 70s from the LASP to a JES3 OS and I never
*heard* that they had problems.

Regards,
Ron Tatum

P.S., I'm through scanning the -10 rev of Data Management for System
Programmers, into the -11 rev. My *good* ISP folded his tent, so I'm stuck
for the moment with the BAD folk that kill all transmission/receipt of
e-mails to nonUSA internet addresses, 'cause the f...ing idiots still insist
that "furriners" don't have the same laws about spam that we *fine* US types
have... And of course, we DON'T have ANY laws regarding spam ... This ISP
has a purely incompetent sysprog staff and the owner is a stupid redneck...
So if you or anyone else has tried to contact me by direct e-mail: No, I'm
not being rude or inattentive, I just can't get aything from you or send
anything to you if you're not in the USA.
----- Original Message -----
From: "A. Philip Dickinson" <phil-+***@public.gmane.org>
To: <hercules-390-***@public.gmane.org>
Sent: Tuesday, November 04, 2003 3:10 AM
Subject: RE: [hercules-390] Re: How fast was a 3033?

Post by A. Philip Dickinson
Well actually.....
I'd like to make use of the rather impressive speed of my dual Athlon
system or my quad xeon to emulate large numbers of 3033's, but in a
realistic manner.
What I'd really like is to emulate eight 3033MP's (total sixteen
emulated CPU's) all managed in a loosely coupled environment under JES3.
This would have been the largest system possible in 1980 and the largest
system supported by my favourite operating system MVS 3.8J (unless we
can get the > 2 cpu stuff to work).
What configuration of intel or AMD processors do I need to achieve this?
Phil D. - eagerly awaiting a CTC that will work with JES3.

-----Original Message-----
Sent: Tuesday, 4 November 2003 7:43 PM
Subject: [hercules-390] Re: How fast was a 3033?
This is amazing --- we work hard at trying to improve performance --
we get faster and faster machines -- now someone wants to "Cripple"
the speed.
I know it's out of interest but I still find it amazing.
"Back to the Future" I guess. !!!!!

Post by halfmeg

Post by A. Philip Dickinson
Does anyone have any performance data on a 3033?

IBM might be interested. The only way to "cripple" the speed that

I

Post by halfmeg
can think of is to tie it to the HOST clock with sleeps (thinking
sleep is a clock interrupt type of thing instead of a loop till

clock

Post by halfmeg
expired thing). This would hopefully permit full cycles of the

HOST

Post by halfmeg
CPU(s) to continue other work without emulation spinning through a
CPU consuming loop to slow execution.
Phil - time to go hiking

------------------------ Yahoo! Groups Sponsor
---------------------~--> Buy Ink Cartridges or Refill Kits
for your HP, Epson, Canon or Lexmark Printer at MyInks.com.
Free s/h on orders $50 or more to the US & Canada.

http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->
http://groups.yahoo.com/group/hercules-390
http://www.conmicro.cx/hercules
Your use of Yahoo! Groups is subject to
http://docs.yahoo.com/info/terms/
http://groups.yahoo.com/group/hercules-390
http://www.conmicro.cx/hercules
Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

[A. Philip Dickinson]

Ron,

No, what I meant was support in Hercules for an emulated ctc that works
with jes3. At the moment we can only run a jes3 complex (a global and
one or more locals) under VM, because VM does emulate a full ctc.

Yeah, I run into this "not in the USA" problem all the time too. Very
annoying.

Phil.
Not in the USA.

Post by Ronald Tatum
Phil,
Your remark about "eagerly awaiting a CTC that will work
with JES3" puzzles me...
Are you implying that the CTC adapter that was actually
used for LASP (local attached support processor) that was
used by numerous shops on 360/85s, 370/195s, etc., running
various releases of OS/360 failed to operate properly under
JES3, the successor to LASP? I believe that TI upgraded in
the middle/late 70s from the LASP to a JES3 OS and I never
*heard* that they had problems.
Regards,
Ron Tatum
P.S., I'm through scanning the -10 rev of Data Management for
System Programmers, into the -11 rev. My *good* ISP folded
his tent, so I'm stuck for the moment with the BAD folk that
kill all transmission/receipt of e-mails to nonUSA internet
addresses, 'cause the f...ing idiots still insist that
"furriners" don't have the same laws about spam that we
*fine* US types have... And of course, we DON'T have ANY laws
regarding spam ... This ISP has a purely incompetent sysprog
staff and the owner is a stupid redneck... So if you or
anyone else has tried to contact me by direct e-mail: No, I'm
not being rude or inattentive, I just can't get aything from
you or send anything to you if you're not in the USA.

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->

[Richard Higson]

Date: Mon, 10 Nov 2003 17:29:18 -0600
Subject: Re: [hercules-390] Re: How fast was a 3033?
Regards,
Ron Tatum
My *good* ISP folded his tent, so I'm stuck
for the moment with the BAD folk that kill all transmission/receipt of
e-mails to nonUSA internet addresses, 'cause the f...ing idiots still insist
that "furriners" don't have the same laws about spam that we *fine* US types
have... And of course, we DON'T have ANY laws regarding spam ... This ISP
has a purely incompetent sysprog staff and the owner is a stupid redneck...
So if you or anyone else has tried to contact me by direct e-mail: No, I'm
not being rude or inattentive, I just can't get aything from you or send
anything to you if you're not in the USA.

If you (or other herculeans) would like an account on opa.gt.owl.de,
accessable via ssh (putty.exe/winscp.exe) then let me know ... somehow ...

If your self-appointed home land sherrif won't let you write me directly,
try my firstname.lastname at the company I work for (syskoplan) in the .com
zone. (Obviously anything in .com is 'one of us' ;-)

It's a sad day when an European has to offer an American email-asylum.

Richard
--
Cogito, ergo non possum dormire ;-(
Have a nice day ;-) Richard Higson mailto:richard.higson-qoLni4XNXEKzQB+***@public.gmane.org

------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/W4wwlB/TM
---------------------------------------------------------------------~->