Assembler programe is abending in divide instruction

thanoozm · New User Joined: 23 May 2007 Posts: 26 Location: hyderabad

Hi ,

Divide instruction is failing in assembler program.

I have moved X`0000D221` into register R5.

SLA R5,5
LA R4,0
D R4,F`12` – this instruction is giving abend.

Thanks
Thanooz.

Robert Sample · Posted: Fri Dec 04, 2009 6:53 pm

Okay, you didn't tell us which abend you're getting. You didn't give us any of the diagnostic data put out by the system. You didn't give us anything but three lines of assembler.

So just what do you expect us to do with so little? This isn't psychic day. If it was, I'd suggest line 379 is where your problem is.

PeterHolland · Posted: Fri Dec 04, 2009 7:09 pm

The SLA instruction i dont understand.

You load the uneven register with a value,
you zero the even register
and then divide the even register with a register or a storage value.

thanoozm · New User Joined: 23 May 2007 Posts: 26 Location: hyderabad

Hi Robert,

System is giving S0c9 error. SYSabend dump is 235,00 pages. I will attach that.

Thanks,
thanooz.

Garry Carroll · Posted: Fri Dec 04, 2009 7:47 pm

First thing that leaps to mind is the absence of an equals sign on the divide....

Robert Sample · Posted: Fri Dec 04, 2009 7:57 pm

You certainly don't need to attach the whole dump -- there is more diagnostic data produced than just a dump! Giving us the abending instruction, registers and PSW for example would be a good start. From the POP:

thanoozm · New User Joined: 23 May 2007 Posts: 26 Location: hyderabad

Hi Robert,

I have tried to display the value before divide instruction. It has displayed correct value '00000691'.

Robert Sample · Posted: Fri Dec 04, 2009 8:44 pm

The D instruction -- what is the generated machine code for 5 instructions before and after it? This is listed on the assembler output to the left of the mnemonic and operands.

Also, note that there is absolutely no question about the register containing a valid value -- this abend occurs because you are dividing the number in the register by zero (for example). So telling us the register has x'00000691' is not providing any useful data to us.

thanoozm · New User Joined: 23 May 2007 Posts: 26 Location: hyderabad

Hi Robert,

Listing before 5 instructions.

Bill O'Boyle · Posted: Fri Dec 04, 2009 9:03 pm

The proper way to load the odd-register prior to issuing a D/DR instruction and preserve the sign (positive or negative, regardless) is to load the even register first and then shift right 32-Bits into the adjacent odd-register. Example -

Garry Carroll · Posted: Fri Dec 04, 2009 9:22 pm

Dear God!

What part of

Robert Sample · Posted: Fri Dec 04, 2009 9:44 pm

I suspect your registers are getting out of whack somewhere. Your code has the divide instruction generating as 5D40 3008, so register 3 offset 008 is being used to pull in the F'12' Yet your dump shows

Bill O'Boyle · Posted: Fri Dec 04, 2009 9:50 pm

Once the ICM has been issued and R5 is greater than F'0', but it is less than F'16' (X'00000010'), when you issue an SRL R5,5 (same as D R5,=F'16'), the result in R5 will be F'0'. This is how the S0C9 is being raised.

Let's say R5 equals F'15' (B'00001111'). When you shift-right 5-Bits, the B'00001111' becomes B'00000000'.

Bill

thanoozm · New User Joined: 23 May 2007 Posts: 26 Location: hyderabad

Hi Bill,

That is correct, my value in R5 is `0000D221` after right shift it will become `00000691`.

I am geting correct value upto that.

thanoozm · New User Joined: 23 May 2007 Posts: 26 Location: hyderabad

Hi Robert,

I am not able to see the adress off R3 in dump, Please let me know where you seen.

thanks,
Thanooz.

Bill O'Boyle · Posted: Fri Dec 04, 2009 10:24 pm

Bill O'Boyle · Posted: Fri Dec 04, 2009 10:36 pm

The sub-program's base register (R3) may be corrupted.

Are you loading R3 from R15 when the sub-program gains control?

As an alternative, instead of using a literal F'12' with the D instruction, load a spare register (R10 for example), with an F'12' (LA R10,12). Then change your D to a Divide Register (DR R4,R10) and see if that makes a difference.

Is this D always causing an S0C9 or does it work sometimes?

Bill

thanoozm · New User Joined: 23 May 2007 Posts: 26 Location: hyderabad

Hi Bill,

It worked some times correctly. Today it is giving problem.

thanks,
Thanooz.

PeterHolland · Posted: Fri Dec 04, 2009 11:32 pm

This is what i should do :

Robert Sample · Posted: Fri Dec 04, 2009 11:40 pm

Thanooz: in your dump1.txt file, page 27, line 1472 has the first TCB (Task Control Block); the register values are immediately after the TCB offsets and data values.

thanoozm · New User Joined: 23 May 2007 Posts: 26 Location: hyderabad

Hi All,

Thanks for your help. I have changed register loading, now it is working fine.

Thanks to robert and bill.

thanoozm · New User Joined: 23 May 2007 Posts: 26 Location: hyderabad

Hi Robert,

Where can i see the memory location 00000072 and it's values. I am searching for that and i didn't find. I find the R3 value and i couldn't find the memory value, This for know the things.

problem is resolved, thanks for your help.

Thanks,
Thanooz.

Robert Sample · Posted: Sun Dec 06, 2009 6:30 pm

Low memory locations are reserved for use by the operating system -- using location 00000072 should only be done by the system; if your registers are pointing there then it is a problem. You usually don't see them in a dump since they are rarely relevant to an application abend.

PeterHolland · Posted: Sun Dec 06, 2009 7:31 pm

Well you can access locations like that, e.g. the CVT.
But its readable not writeable.

dick scherrer · Posted: Sun Dec 06, 2009 11:54 pm

Hello,

So, unless the code references system data areas (i.e. control blocks), 72 is not going to be a proper address for the code to reference.

However, registers contain "things" other than addresses. . . Length, count, displacement, etc. . .