IBM Mainframe Forum

by **xboss** » Sat Feb 04, 2012 2:00 am

Hello all,
I am trying to run a simple assembler program that do basic arithmetic function (ADD, MULTIPLY, DELETE) using 64 bit register. I don't see any error in my program (I compiled and ran successfully) but I am not getting result I expected when looking at storage dump. By expected result, I mean to say when I Defined Double Word of value "2", in memory I am seeing 412 as its hex representation. May be this is how it should display I don't know. Please help me understand if I am doing something run in coding or reading.

Program

Select all

***************************************************************
DS 0H DEFINE HALWORD FOR CSECT
STMH R14,R12,128(R13) Save upper register halves
STM R14,R12,12(R13) Saving Registers
BALR R12,0 ESTABLISHING BASE REGISTER
USING *,R12 ESTABLISHING ADDREASIBILITY
LA R2,SAVELOW LOAD MEMORY ADD TO REG
ST R13,4(,R2) STORE CALLER REG IN 2ND WORD
ST R2,8(,R13) STORING CALLEE REG IN 3rd WORD
LA R3,SAVEHIGH
ST R3,128(,R13) 64 BIT STORE
LR R13,R2
ADD EQU *
LG R2,VALUE1 LOAD Double Word to R2
AG R2,VALUE2 ADD AND STORE RESULT TO R2
STG R2,ADDR
WTO 'ADD'
MULTI EQU *
LG R3,VALUE3 LOAD DOUBLE WORD TO R3
MSGR R2,R3 MULTIPLY AND STORE RESULT TO R2
STG R2,MULTIR
WTO 'MULTIPLY'
DIVIDE EQU *
LG R4,VALUE4 LOAD DOUBLE WORD
DSGR R2,R4 DIVIDE AND STORE RESULT TO R2
STG R2,DIVIDER
WTO 'DIVIDE'
DC H'0'
RETURN EQU *
L R13,SAVELOW+4
* LM R14,R12,12(13) RE
LMD R14,R12,SAVEHIGH,12(R13)
BR R14
********************* DATA AREAS *******
SAVEHIGH DC 18F'0'
SAVELOW DC 18F'0'
LOOHERE DC CL8'STARTING'
VALUE1 DC X'000000007FFFFFFF'
VALUE2 DC 1D'2'
VALUE3 DC 1D'-1'
VALUE4 DC 1D'255'
ADDR DS 1D
MULTIR DS 1D
DIVIDER DS 1D

Storage Dump

Select all

00007FB8 +001A0 00000000 00000000 E2E3C1D9 E3C9D5C7 ........STARTING
00007FC8 +001B0 00000000 7FFFFFFF 41200000 00000000 ...."...........
00007FD8 +001C0 C1100000 00000000 42FF0000 00000000 A...............
00007FE8 +001D0 41200000 7FFFFFFF 3EF00000 00000000 ...."....0......
00007FF8 +001E0 C1100000 00000000 A.......

by **enrico-sorichetti** » Sat Feb 04, 2012 2:16 am

You did nothing wrong... You just did not do it :geek:

reading the manuals in HLASM bookshelf ???
http://publibz.boulder.ibm.com/cgi-bin/ ... s/ASMSH030

and specifically the
V1Rx Language Ref
http://publibz.boulder.ibm.com/cgi-bin/ ... 0711003624
at the paragraph
http://publibz.boulder.ibm.com/cgi-bin/ ... 3624&CASE=
were it is spelt in clear letters that ...

Select all

Code Constant Type Machine Format
---- ---------------- -------------------------------------------------
D Floating-point Long floating-point format; normally a doubleword

AND

Type Type Extension Description
---- ---------------- -------------------------------------------------
F D Doubleword fixed-point constant

so

Select all

000000 00000001 3 A DC F'1'
000004 00000000
000008 0000000000000001 4 A1 DC FD'1'
000010 04 5 LA DC AL1(L'A)
000011 08 6 LA1 DC AL1(L'A1)
00004 7 ELA EQU L'A
00008 8 ELA1 EQU L'A1

by **steve-myers** » Sat Feb 04, 2012 2:23 am

Well, to start: you need to do more research on how to define 64-bit integers. D'2' defines a 64-bit hexadecimal floating point 2, not a 64-bit integer 2.

by **xboss** » Sat Feb 04, 2012 4:47 am

Thank you, both for you time and links. Sometime to get up to correct resources is itself a research if you don't know the correct keyword.

by **steve-myers** » Sat Feb 04, 2012 8:12 am

Enrico - you should not have given xboss the solution; better he finds it himself! Not only is there a better chance it will stick, but learning how to do the research is probably more valuable. I recall an incident back in the 1970s when IBM would provide on site a software support. Anyway, I was working with our support on some issue - can't remember what it was, when he volunteered, "There's a couple of things I like about you; you usually admit it when you don't know something rather than try to convince me that you do know it, and you usually know where to look to try to resolve something." Now xboss at least has the former quality, but he really needs to acquire the latter.

There is another possible way to specify a 64-bit integer: FL8'value'. I've been using FL8'-1' for nearly as long as I've been writing System/360 Assembler! Granted, it doesn't align properly, but that's something that's not hard to work around if you expect it.

One other thing I noticed in the code xboss showed. I'm not sure his method to save the 64-bit registers is safe. There are two issues

He assumes his caller is providing an extension to the standard save area.
His general methodolgy is incorrect. I'm giving a freebie link here, something I don't do very often any more to try make our lurkers, and possibly our topic starters training in doing their own research. Technically the link is a little out of date, but I doubt it's changed.

Actually, the few times I've done true 64-bit arithmetic I haven't bothered to save and restore the high order bits of the registers and it hasn't seemed to do any harm.

Much safer, but almost certainly much slower is to use BAKR on entry to save the registers in the linkage stack and PR to restore the registers and return. I did a performance study on this long before 64-bit GPRs, but BAKR did store the access registers. I don't recall the numbers, but it was much slower than regular linkage.

by **xboss** » Mon Feb 06, 2012 11:30 pm

Steve,
Thanks for both compliments and critics, I am working hard to acquire latter quality as well.

Yes, with performance trade off, BAKR/PR can be used to save GR as well as Access register. But being in a learning phase I want to try out all possible ways.

Thanks,

by **steve-myers** » Tue Feb 07, 2012 2:40 am

xboss wrote:... But being in a learning phase I want to try out all possible ways. ...

Unfortunately, most supervisor types don't understand how important it is for people to experiment. For that matter, most people don't appear to understand this. Far better to screw up something in an experiment than to apply what amounts to an experiment to a production something and break it because the experiment failed.

IBM Mainframe Forum

Grande 64 bit Operation

Grande 64 bit Operation

Re: Grande 64 bit Operation

Re: Grande 64 bit Operation

Re: Grande 64 bit Operation

Re: Grande 64 bit Operation

Re: Grande 64 bit Operation

Re: Grande 64 bit Operation