If it were written out like that in the source, I wouldn't call it clever! I'm guessing you disassembled it from something, and it's somewhat ancient, from before the ESA/390 relative and immediate instructions were introduced.
I've encountered various IBM product run-time library routines which systematically use that approach to implement multiple functions in the same module. Jonathan Scott -----Original Message----- From: IBM Mainframe Assembler List <[email protected]> On Behalf Of Martin Ward Sent: 17 February 2026 18:21 To: [email protected] Subject: Clever code to celebrate Some "clever code" for everyone to "celebrate". When I saw it, I have to confess, I *didn't* think "Why didn't I think of that?"(!) The module starts with a block of data, followed by some code (do you think that this will be dead code therefore?): A00000 DC X'000047F0' DC X'F01447F0' DC X'F01047F0' DC X'F00C47F0' DC X'F00847F0' DC X'F004' STM 14,12,12(13) L 0,0(0,1) The entry points, XXXX, YYYY, XXXX1, XXXX2 and YYYY1, are defined as EQUates which are offsets into this data block: XXXX EQU A00000+2 YYYY EQU XXXX+4 XXXX1 EQU YYYY+4 XXXX2 EQU XXXX1+4 YYYY1 EQU XXXX2+4 So when you call "XXXX" you start executing the data two bytes after the address A00000. This happens to be the hex data X'47F0F014' which is, of course, the object code for a branch instruction. The instruction will branch to the address at X'0014' off R15. Since R15 is the address of the entry point, this will branch to address X'00016' which is the STM instruction. It copies R15 to R1, does a BALR on R15 (to get a single base address for the USING), then compares R1 with the address A00028 to get an offset into a jump table, which it loads and branches to. The jump table code blocks just set R3 to 0, 1, 2, 3 or 4 and branch to A00062. I mean, you could just copy the savearea chaining and set R3 for each entry point, but that would be boring! To see if R3 equals 2, we could do CH R3,=H'2' but that is too boring! So we do: CH 3,A001EA where A001EA is defined as: A001EA EQU A001E8+2 and A001E8 is, naturally, a table of two halfwords: A001E8 DC X'00010002' Wasn't that clever! -- Martin Dr Martin Ward | Email: [email protected] | http://www.gkc.org.uk G.K.Chesterton site: http://www.gkc.org.uk/gkc | Erdos number: 4
