Addressing modes in assembler (1)

Addressing 24-bit storage and 31-bit storage

System 370 (predecessor to System z) could only address 16MB of storage when I started to learn assembler and we did not have words for addressing mode(s). Therefor creative people used the high order byte of an address as a sort of parameter to the address that was in the low order three bytes. One example is an address in the Data Control Block (DCB):

0001E0 00         269+         DC    BL1'00000000'  BFTEK, BFALN, DCBE INDICATORS 0001E1 0000AE     270+         DC    AL3(CLOSE)     EODAD (END OF DATA ROUTINE ADDRESS)

The address of the End Of Data routine is only 24 bit long. That is why the DCB is still located below the line.

Bear in mind that the general registers have always been four bytes each.

I have made a small program that just reads records and writes them out again. This is a part of it which I will play with in the following articles in this series.

BAL   R14,OPENIN                           BAL   R14,OPENOUT                  READ     DS    0H                                   BAL   R14,READIN                           L     R2,RECADR                            MVC   OUTREC,0(R2)                     BAL   R14,WRITE                            B     READ                         CLOSE    DS    0H                                   DROP  R9                                   BAL   R14,CLOSEIN                          BAL   R14,CLOSEOUT                         B     RETURN

24-bit addressing mode was sufficient for a long time. In the 70’s came eXtended Architecture (XA) though which meant you could now address 2GB of storage. It was a game changer in assembler programming. IBM has always protected the customer's investment in their software (Cobol, PL1, and Assembler) so the XA-system had to be backward compatible. One single bit in the PSW did that. The high order bit in the instruction address in the PSW told the CPU whether, when zero, it run in addressing mode 24 or in 31-bit addressing mode (addressing mode 31) when one.

Addressing mode 24 PSW: 078D0000   00007E1A - The high order bit is off

Addressing mode 31 PSW: 078D1000   80007DFE - The high order bit is on

In addressing mode 24 the instructions work as usual but in addressing mode 31 some of the instructions work a little bit different. They now have to take the longer storage address into account. An example

Addressing mode 24

LA R1,BELOW16

R1 will after this instruction contain the address of BELOW16 in the three rightmost bytes and the leftmost byte will be zero.

Addressing mode 31

LA R1,ABOVE16

R1 will now contain the address in all four bytes except the high order bit that will be zero. Sometimes referred to as an XA-address.

Change mode

We suddenly encountered a problem in our legacy addressing mode 24 programs if we had to address storage “above the line” as the 16MB limit was called. We couldn’t just change the addressing mode for the whole program because we might be using the leftmost byte as parameter bits. In that case we could use some new instructions that could flip the high order bit of the PSW i.e. go to addressing mode 31 and back to addressing mode 24.

BAL   R14,OPENIN                           BAL   R14,OPENOUT                          GETMAIN RC,LV=WS31_LEN,LOC=(31,64)         LR    R9,1                                 USING WS31,R9                      READ     DS    0H                                   BAL   R14,READIN                           L     R2,RECADR                            LA    R3,LABEL31                           O     R3,=X'80000000'                      BSM   0,R3                         LABEL31  DS    0H                                   MVC   RECORD31+1(80),0(R2)                     MVC   OUTREC,RECORD31                      LA    R3,LABEL24                           N     R3,=X'7FFFFFFF'                      BSM   0,R3                         LABEL24  DS    0H                                   BAL   R14,WRITE                            B     READ

Partial addressing mode 31 program.

I admit this is a stupid program but I have made it to show how you can change mode to addressing mode 31 in a program that otherwise is addressing mode 24. I Getmain some storage “above the line”. R9 will contain an XA-address. Ex:

R9: 1F700F78

I intend to move the read record to RECORD31. Right after I move it back “below the line” to OUTREC. I then must go back to addressing mode 24 in the running program in order to be able to write the record.

LA    R3,LABEL31                           O     R3,=X'80000000'                      BSM   0,R3                         LABEL31  DS    0H

Instructions to go from addressing mode 24 to addressing mode 31

• I start to load the address of the first addressing mode 31 instruction at LABEL31
• Then I set the high order bit on that address to indicate that from this address I can reach above the line
• BSM is Branch and Set Mode. The first argument is set to zero because I just need to branch to the address in second argument
• LABEL31 is the location counter of the first addressing mode 31 instruction

All following instructions executed will be in addressing mode 31 until I do the following instructions. Actually the opposite of the previous.

LA    R3,LABEL24                           N     R3,=X'7FFFFFFF'                      BSM   0,R3                         LABEL24  DS    0H

Instructions to go from addressing mode 31 to addressing mode 24

• I start to load the address of the first addressing mode 24 instruction at LABEL24
• Then I set the high order bit off that address to indicate that I am now unable to reach above the line
• BSM is Branch and Set Mode. The first argument is set to zero because I just need to branch to the address in second argument
• LABEL24 is the location counter of the first addressing mode 24 instruction

New addressing mode 31 program

Although 24-bit addressing mode is the default if you do not specify anything neither in the assembler source nor the linkage editor, you will probably wish to write addressing mode 31 programs. You will probably also wish to locate your program “above the line”, i.e. in 31-bit storage area. In order to do so you may either code the RMODE 31 in the assembler source or as input to the linkage editor

Specified in the source or in linkage editor	Result
AMODE               UNSPECIFIED   RMODE               UNSPECIFIED	AMODE                24 RMODE                24
AMODE               UNSPECIFIED RMODE               ANY or 31	AMODE                31 RMODE                31
AMODE               31 RMODE               UNSPECIFIED	AMODE                31 RMODE                24
AMODE               31 RMODE               ANY	AMODE                31 RMODE                31

Just specify RMODE 31 and you are are also in addressing mode 31.

If the program is located in 31-bit storage it must also be in addressing mode 31 because the base register(s) will have to address the instructions. So just make the RMODE-instruction 31 or ANY.

PGM5     CSECT      PGM5     RMODE ANY         ENTRY PGM5

How to start your CSECT if running AMODE 31 and RMODE 31

The best way to make sure that your program is in the right mode is to take a look at the  directory entry for the loadmodule. You can find the modes on the very right of the program entry.

BROWSE            JEE.LOADLIB                           Row 0000001 of 0000003 Command ===>                                                  Scroll ===> CSR           Name     Prompt        Alias-of     Size      TTR     AC   AM   RM _________ PGM1                               000001B8   00000C   00    31  ANY _________ PGM5                               00001140   00042F   00    31  ANY _________ PGM6                               000002C8   000404   00    24   24

Next time I will explore the possibilities of the 64-bit programming. That will be fun!