> David Eisenberg wrote: > I want to generate baseless code using as few instructions as possible, > Can my macro modify the resulting condition code?
I'm horrified this was your solution especially considering all the other options available. 1. Regardless of programming language or whatever you are doing, you should always ask yourself if you are making a choice for the right reason. Far too often we overthink a decision or problem only to find we are solving problems that will never exist. 2. In the next 5 years, how many programs will need that extra register and you are willing to give up addressability to constants? For most people, it will only be at couple. Don't disable a useful HLAsm feature because someone convinced you that baseless is important. Your only job at this time is to make that extra register a simple choice if if's more important than constants. Needlessly removing the base reg is a terrible idea but preparation is important. The truth is that constant addressability is more important than as an 11th register. Truly baseless is often not the answer. For me, baseless was not a solution when I added entries to a large TSO parser table that caused addressability problems. Each addressability problem is unique and, in this situation, moving the table after all constants (including LTORG) was the only acceptable solution. It's understandable you want to solve a problem but consider the cost/risk/benefit of going beyond preparation. 3. Forget about converting B to J by hand because you can't modify IBM macros. The correct and very simple solution is to internally translate B to J. OPSYN B instructions to a macro. The macro only coverts B to J when it doesn't involve a register (no parenthesis). There are a couple of extremely unlikely exceptions but it's doubtful anyone will experience these exceptions. 4. Inform everyone of how to use their new choices. If they need a second base reg, simply change the using to the start of the constants because executable code does not need addressability since you B is now J (with a couple of IBM macro exceptions). If they absolutely need the extra register, then they need to decide how to best handle constants. 5. If you insist on conversion instead of prepping for baseless code, then instead of CLHHSI, use LARL R1,=X'1111' before the CLC. No sense in making this complicated. 6. Depending on CC outside the macro should be avoided if possible. 7. I suggest you look at using the structured macros available in the IBM HLASM Took Kit. Your CLHHSI problem is about esthetics. You may find coding IF (CLHHSI,AA,LE,X'1111') more palatable.
