Functioning of TM in assembler

himanshupant · New User Joined: 21 Mar 2007 Posts: 46 Location: India

I have the below statement

TM SWITCH,X'00'

In the above case what would be the condition code set to after the TM for assuming

1) SWITCH is all zeros
2) SWITCH is all Ones
3) SWITCH is mixed zeros and ones.

For any other mask i.e. one not having zeroes , the IBM manual says the condition code is

0 all selected bits are zero
1 selected bits are mixed
2 not used!
3 all selected bits are one

but for the above case when the mask itself is zeroes , no bit would be selected .So how will the instruction behave ?

William Thompson · Posted: Thu Mar 29, 2007 6:53 pm

I guess that the cc wouldn't change....
Checking my POP shows "0 Selected bits all-zero; mask is all-zero".
Good guess Bill...

dick scherrer · Posted: Fri Mar 30, 2007 12:45 am

Hello,

Test under Mask (TM) assumes that you are looking for an "on" condition in one or more of the tested bits.

If you don't want to test for at least one bit on, there is no need to issue the TM.

William Thompson · Posted: Fri Mar 30, 2007 1:21 am

Is TM an EXecutable instruction?

dick scherrer · Posted: Fri Mar 30, 2007 1:43 am

Hi Bill,

Yes, it is a way to test bit patterns. The machine code is x'91' (no i've not memorized the op-codes

).

William Thompson · Posted: Fri Mar 30, 2007 1:59 am

No, I mean is it and instruction that you can EX, overlaying the bit pattern part of the instruction like you can when you EX a MVC and overlay the length?

himanshupant · New User Joined: 21 Mar 2007 Posts: 46 Location: India

I agree Dick that with the test mask x'00' no bit would be tested .But the case I am pointing out has occured in a production code.So I was wondering that it might be changing the condition code somehow , because after this TM there is a BNO condition.

dick scherrer · Posted: Sat Mar 31, 2007 12:35 am

Hello,

Without having the code to experiment with, my guess is that you have a rather elaborate unconditional branch. . .

Bill, i'm not sure if it can be EXed or not.

enrico-sorichetti · Posted: Sat Mar 31, 2007 2:33 am

Hi !

quoting from the esa/390 principles of operations

William Thompson · Posted: Sat Mar 31, 2007 2:38 am

Hi back,

bengtpelle · Posted: Sat Mar 31, 2007 11:50 pm

I do not think that you got the full picture. From reading above, I noticed that you had a BNO after the TM for X'00".
This BNO is useless and it will never be executed.
It is true that the TM can be executed, however if it is, the BNO should be placed after the EX instruction, not after the TM.

dick scherrer · Posted: Sun Apr 01, 2007 12:32 am

Hello,

In the initial post and the follow up from himanshupant, there was no EX mentioned.

The EX discussion was (thinking out loud here) wondering if the TM might be configured on the fly rather than always use the x'00' mask.

Why would the BNO "never be executed"? I believe it will be executed every time thru the code and will always do the same thing - as mentioned earlier, "a rather elaborate unconditional branch". . . .

William Thompson · Posted: Sun Apr 01, 2007 12:35 am

bengtpelle · Posted: Sun Apr 01, 2007 12:59 am

XI, NI would work.
To figure that out, there should be a label on the TM instruction.
A dump could tell us if the TM was modified.

Here is a totaly unrelated idea....
Have you ever tried to Execute a BALR instruction????
Quite facinating thought on program flow.
I have used it to protect program logic from debugging.

William Thompson · Posted: Sun Apr 01, 2007 1:19 am

Nasty.....But we don't code that way anymore, do we....

dick scherrer · Posted: Sun Apr 01, 2007 1:24 am

Hello,

Whatever does

bengtpelle · Posted: Sun Apr 01, 2007 1:33 am

No we don't, but I wish we could.