Compiler option to eliminate ZAP instructions that fix the sign on zero results

While reviewing our RFE backlog, we have come to the conclusion that there aren't the same number of ZAPs on z14/z15 because of the vector packed instructions. As a result, we will be rejecting this RFE.

Due to processing by IBM, this request was reassigned to have the following updated attributes:
Brand - Servers and Systems Software
Product family - Programming Languages
Product - COBOL Compilers

For recording keeping, the previous attributes were:
Brand - Rational
Product family - Design & development
Product - COBOL Compilers

Thanks. The feature, for a new comp type that would have these relaxed sign code restrictions so the scope of the change could be limited to known safe cases in the code, has been prioritized, but not put into plan yet. Once it has been put into plan, this RFE will be updated.

Yes, we would be interested in using a new Comp type. We understand that this would indeed limit the scope of the modifications to the generated code.

Hi,

We've looked into this request a bit more and consulted a wider audience. The general response was that although this sounded like an interesting option there wouldn't be many users of it as they would have a very hard time guaranteeing or convincing themselves that it was safe to use. This is because they would not know in general what assumptions external consumers of the COBOL data had regarding sign codes and what might/could break if the option was used.

A suggestion was that instead of making it a compile option it might be more useful as a new comp type that would have these relaxed sign code restrictions so the scope of the change could be limited to known safe cases in the code.

May we know if you are interested in such a new comp type?

Thanks.

The emphasis in the PMR originating this request was on ZAPs following packed decimal arithmetic. That arithmetic would understand produce preferred signs, leaving only the potential negative zeros.

Straight moves would continue to use ZAP if preferred signs are required.

Hi John, just a clarification on your previous reply, may we know if you could tolerate non-preferred signs as well as tolerating negative zero.

The reason being, ZAP fixes up both (ie negative to positive zero and also ensures the preferred positive and negative signs are used for all values).

We would need this info to determine the scope of this feature item.

Thanks!

Yes we are interested in the preferred signs case. I initially thought that this option would eliminate the ZAP referred to in the initial request but it did not

We would like to know if you are also interested in the preferred signs case?
We look forward to hearing back from you.