|
View previous topic :: View next topic
|
| Author |
Message |
Manish Kumar Gupta
New User
Joined: 19 Feb 2008
Posts: 6
Location: Bangalore
|
|
|
|
Hi,
I have been asked the following question:
01 Var-1 Pic 9(4) Value 1234.
01 Var-2 Redefines Var-1 Pic 9(2).
What will happen if we move Var-2 to Var-1 and Move Var-1 to Var-2
IBM Manual says that the results are undefined so i said the same thing, but later i thought to give it a try, here are the results:
Initial Display:
Display Var-1 = 1234
Display Var-2 = 12
------------------------------
Move Var-1 to Var-2 (Larger to Smaller)
Display Var-1 = 3434
Display Var-2 = 34
------------------------------
But when i moved Var-2 to Var-1 (Smaller to Larger) results were
Before move i made the values as
Display Var-1 = 1234
Display Var-2 = 12
After move the values were
Display Var-1 = 0000
Display Var-2 = 00
-------------------------------
I tried it with different values but the results followed the same pattern, i am not sure what happened, i would be grateful if someone can explain. |
|
| Back to top |
|
 |
dbzTHEdinosauer
Global Moderator
Joined: 20 Oct 2006
Posts: 6966
Location: porcelain throne
|
|
|
|
Manish,
I know that the programmers guide has the same type of graph:
var1 pic 9(04).
var2 redefines var1 pic 9(02).
var1 starts at address 0001 and continues thru 0004
var2 starts at address 0001 and continues thru 0002.
if you understand the above,
you should be able to answer your own question. |
|
| Back to top |
|
|
dick scherrer
Moderator Emeritus
Joined: 23 Nov 2006
Posts: 19244
Location: Inside the Matrix
|
|
|
|
Hello,
| Quote: |
| I tried it with different values but the results followed the same pattern, i am not sure what happened |
Things happened as they should - just what the code said to do.
You might also look at the compiler-generated code to see the actual instructions. . . |
|
| Back to top |
|
|
Manish Kumar Gupta
New User
Joined: 19 Feb 2008
Posts: 6
Location: Bangalore
|
|
|
|
Thanks for the reply! I would take a deeper look into this, When i said
"I am not sure what happened" i was wondering about the difference in results (moving from larger to smaller / smaller to larger)
I will go thru the compiler generated instruction to understand whats happening when i am moving Smaller Variable into Larger.
Thanks |
|
| Back to top |
|
|
Manish Kumar Gupta
New User
Joined: 19 Feb 2008
Posts: 6
Location: Bangalore
|
|
|
|
| The results shouldn't be same for the two move statements, but i am not able to understand the zeros in the move from Smaller to Larger. |
|
| Back to top |
|
|
dick scherrer
Moderator Emeritus
Joined: 23 Nov 2006
Posts: 19244
Location: Inside the Matrix
|
|
|
|
Hello,
Look closely at the memory addresses being used as well as the actual instructions . . . |
|
| Back to top |
|
|
dbzTHEdinosauer
Global Moderator
Joined: 20 Oct 2006
Posts: 6966
Location: porcelain throne
|
|
|
|
| Quote: |
| I will go thru the compiler generated instruction to understand |
stop wasting your time.
read about numeric moves and what constitutes truncation (why and how it is done)
and about justification and zero filling to the left.
while you are at it,
read about alphanumeric moves,
what constitutes truncation,
and about justification and space filling to the right
you are aware, that redefining something is
only allocating the same data area to two reference names.
as I have tried to imply and Dick overtly suggested,
look at the addresses....................... |
|
| Back to top |
|
|
CICS Guy
Senior Member
Joined: 18 Jul 2007
Posts: 2146
Location: At my coffee table
|
|
|
|
| Manish Kumar Gupta wrote: |
"I am not sure what happened" i was wondering about the difference in results (moving from larger to smaller / smaller to larger)
I will go thru the compiler generated instruction to understand whats happening when i am moving Smaller Variable into Larger. |
Why not post those generated instructions for both the moves?
I don't have a compiler available and I am curious as to what is generated for the dissimilar size moves. |
|
| Back to top |
|
|
Robert Sample
Global Moderator
Joined: 06 Jun 2008
Posts: 8696
Location: Dubuque, Iowa, USA
|
|
|
|
| Code: |
DISPLAY 'VAR-1 AT START ' VAR-1.
DISPLAY 'VAR-2 AT START ' VAR-2.
MOVE VAR-1 TO VAR-2.
DISPLAY 'VAR-1 AFTER MOVE TO VAR-2 ' VAR-1.
DISPLAY 'VAR-2 AFTER MOVE TO VAR-2 ' VAR-2.
MOVE VAR-2 TO VAR-1.
DISPLAY 'VAR-1 AFTER MOVE TO VAR-1 ' VAR-1.
DISPLAY 'VAR-2 AFTER MOVE TO VAR-1 ' VAR-2. |
produces
| Code: |
000020 DISPLAY
000432 5820 905C L 2,92(0,9) TGTFIXD+92
000436 58F0 202C L 15,44(0,2) V(IGZCDSP )
00043A 4110 A1A6 LA 1,422(0,10) PGMLIT AT +414
00043E 05EF BALR 14,15
000021 DISPLAY
000440 58F0 202C L 15,44(0,2) V(IGZCDSP )
000444 4110 A190 LA 1,400(0,10) PGMLIT AT +392
000448 05EF BALR 14,15
000022 MOVE
00044A D201 8000 8002 MVC 0(2,8),2(8) VAR-2
000450 96F0 8001 OI 1(8),X'F0' VAR-2+1
000023 DISPLAY
000454 58F0 202C L 15,44(0,2) V(IGZCDSP )
000458 4110 A17A LA 1,378(0,10) PGMLIT AT +370
00045C 05EF BALR 14,15
000024 DISPLAY
00045E 58F0 202C L 15,44(0,2) V(IGZCDSP )
000462 4110 A164 LA 1,356(0,10) PGMLIT AT +348
000466 05EF BALR 14,15
000025 MOVE
000468 D201 8000 C022 MVC 0(2,8),34(12) VAR-1
00046E D201 8002 8000 MVC 2(2,8),0(8) VAR-1+2
000474 96F0 8003 OI 3(8),X'F0' VAR-1+3
000026 DISPLAY
000478 58F0 202C L 15,44(0,2) V(IGZCDSP )
00047C 4110 A14E LA 1,334(0,10) PGMLIT AT +326
000480 05EF BALR 14,15
000027 DISPLAY
000482 58F0 202C L 15,44(0,2) V(IGZCDSP )
000486 4110 A138 LA 1,312(0,10) PGMLIT AT +304
00048A 05EF BALR 14,15 |
Register 8 points to BLW=0 according to the prologue while Register 12 points to the PGT |
|
| Back to top |
|
|
William Thompson
Global Moderator
Joined: 18 Nov 2006
Posts: 3156
Location: Tucson AZ
|
|
|
|
Dang.
The move of the large to the small makes so much sense.
But what is in X'34(12)'?
I can only assume the required leading zeros....
Which means that the redefined var-2 now contains those zeros and the following MVC 2(2,8),0(8) puts those zeros into the rear two bytes of var-1.
Thank you "Mr. TANSTAAFL"....giggle...... |
|
| Back to top |
|
|
CICS Guy
Senior Member
Joined: 18 Jul 2007
Posts: 2146
Location: At my coffee table
|
|
|
|
Dang, forget to switch...grumble...
But not too bad for an old VSE systems programmer...grin.... |
|
| Back to top |
|
|
Robert Sample
Global Moderator
Joined: 06 Jun 2008
Posts: 8696
Location: Dubuque, Iowa, USA
|
|
|
|
| X 34(12) contains 'F0F0' -- I went back and did a little dump to find out. |
|
| Back to top |
|
|
CICS Guy
Senior Member
Joined: 18 Jul 2007
Posts: 2146
Location: At my coffee table
|
|
|
|
As I kinda thought.
And, thank you again "Mr. TANSTAAFL"...grin.... |
|
| Back to top |
|
|
anshul_gugnani
New User
Joined: 02 Nov 2009
Posts: 73
Location: Mumbai
|
|
|
|
Hi,
I could understand the move of var-1 to var-2 but move of var-2 to var-1 is not clear.
When var-2 will be moved to var-1 it should move '0012' to var-1 as it will be padded wid leading zeroes.
Considering this value of var-1 will bw '0012' and value of var-1 will be '00'.
Can someone please explain me the how thew value of var-1 is '0000'
Tried to understand the compiler generated instructions but could not.
Please help.
Thanks. |
|
| Back to top |
|
|
Ronald Burr
Active User
Joined: 22 Oct 2009
Posts: 293
Location: U.S.A.
|
|
|
|
The "problem" with the move of Var-2 to Var-1 is in the sequence of instructions generated by the compiler. Apparently, the author(s) of the compiler neglected to take into consideration the possibility of overlapping fields in a straight numeric move of shorter to longer lengths. So, the generated code:
| Code: |
000025 MOVE
000468 D201 8000 C022 MVC 0(2,8),34(12) VAR-1
00046E D201 8002 8000 MVC 2(2,8),0(8) VAR-1+2
000474 96F0 8003 OI 3(8),X'F0' VAR-1+3 |
says, in the sequence generated, to:
1. 468: zero fill the leftmost digits of Var-1 by moving 2 zeros to the first 2 bytes of Var-1, resulting ( because of the redefines ) in values of Var-1 = 0034 and Var-2 = 00
2. 46E: move the 2 bytes of Var-2 to the last 2 bytes of Var-1, resulting in values of Var-1 = 0000 and Var-2 = 00
3. 474: ORing the last byte of Var-1 to insure that it has a proper sign ( x'F' )
The "correct" code SHOULD have been in the sequence:
| Code: |
000025 MOVE
000468 D201 8002 8000 MVC 2(2,8),0(8) VAR-1+2
00046E 96F0 8003 OI 3(8),X'F0' VAR-1+3
000472 D201 8000 C022 MVC 0(2,8),34(12) VAR-1 |
Namely, move Var-2 to Var-1 FIRST, THEN zero fill Var-1 to the left. Perhaps someone should open a problem ticket with IBM related to how the compiler generates code when moving a shorter numeric field to a longer numeric field and the fields overlap.
FWIW, In the case of overlapping fields it "might" be safer to, instead of MOVEing Var-2 to Var-1, to say COMPUTE Var-1 = Var-2. It would be interesting to see the generated code for that. |
|
| Back to top |
|
|
anshul_gugnani
New User
Joined: 02 Nov 2009
Posts: 73
Location: Mumbai
|
|
|
|
Hello Ronald,
Thanks a lot for such detailed explaination. You are awesome. 
Can you please tell how can we see the compiler generated instructions.
we use changeman tool at our shop.
Thanks. |
|
| Back to top |
|
|
Ronald Burr
Active User
Joined: 22 Oct 2009
Posts: 293
Location: U.S.A.
|
|
|
|
To generate a LISTing of the generated Assembler Code, specify the Compiler Option LIST
If you are not permitted to override the Changeman Options, then, if your shop permits it, you can specify the option INside of your source code. To wit, include the statement
CBL LIST
before the IDENTIFICATION DIVISION header. |
|
| Back to top |
|
|
anshul_gugnani
New User
Joined: 02 Nov 2009
Posts: 73
Location: Mumbai
|
|
|
|
Hello Ronald,
When I listed the compiler generated instructions at my shop for the same move statement I found this -
| Code: |
000026 MOVE
000344 5830 912C L 3,300(0,9) BLW=0
000348 D201 3008 C004 MVC 8(2,3),4(12) VAR1
00034E D201 300A 3008 MVC 10(2,3),8(3) VAR1+2
000354 96F0 300B OI 11(3),X'F0' VAR1+3 |
which is diffrent from the instructions mentioned here.
Please suggest. Thanks. |
|
| Back to top |
|
|
CICS Guy
Senior Member
Joined: 18 Jul 2007
Posts: 2146
Location: At my coffee table
|
|
|
|
| Same thing, just the Load to establish addressability to the piece of WS where the vars are. |
|
| Back to top |
|
|
Manish Kumar Gupta
New User
Joined: 19 Feb 2008
Posts: 6
Location: Bangalore
|
|
| Back to top |
|
|
|
|
