| Technote PT 545 | October 1990 |
Last reviewed: 8/1/92
Now that MPW includes the "far" addressing model, how does one use the assembler with the far model? For example, how does one generate a "jsr.l symbol" that the linker will recognize as a "far" reference?
___
The MPW Assembler does support the new "far" runtime model; it's documented in the "MPW 3.2 Run-Time Architecture" release note, included with recent versions of MPW that support the new model. (The path to this note on E.T.O. #4 is E.T.O. #4:Tools - Objects:Macintosh Programmer's Workshop:MPW QR4:QR4 Notes:MPW QR4 Run-Time Architecture.)
The specific example to your parenthetical question is "JSR (symbol).L", but you should see the release note for background information and important caveats.
Last reviewed: 6/14/93
Where are the MPW 3.2 Assembler "generic instruction conversions"? After upgrading from MPW Assembler 3.1 to 3.2, I noticed while debugging some of my code that an Adda instruction was converted to Lea for me. While I appreciate this (it saves cycles) I didn't expect it nor can I find where it's documented. The cover letter that came with 3.2 says that Asm has "new optimizations" but doesn't elaborate. I have found Appendix A in the 3.0 documentation that describes some obvious optimizations. Where is the corresponding list of optimizations for 3.2?
___
You correctly noted that the 3.2 Assembler supports optimization of instructions which were previously not optimized. The following additional optimizations are available:
ADD #<Data>,An [-8 <= data <= -1] =====> SUBQ #-<Data>,An ADDA #<Data>,An [-8 <= data <= -1] =====> SUBQ #-<Data>,An ADDI #<Data>,An [-8 <= data <= -1] =====> SUBQ #-<Data>,An SUB #<Data>,An [-8 <= data <= -1] =====> ADDQ #-<Data>,An SUBA #<Data>,An [-8 <= data <= -1] =====> ADDQ #-<Data>,An SUBI #<Data>,An [-8 <= data <= -1] =====> ADDQ #-<Data>,An
These optimizations only occur when the Assembler understands the operand to be a negative number. Since the Assembler operates on 32-bit arithmetic, the operand value must be written as a negative number to activate the optimization, regardless of the size (B, W, L) of the instruction. Thus, in a byte operation, the operand must specify "-1" instead of "$FF," even though, unoptimized, they would both compile to the same instruction.
ADDI #<Data>,<EA3> [+1 <= data <= +8] =====> ADDQ #<Data>,<EA3> SUBI #<Data>,<EA3> [+1 <= data <= +8] =====> SUBQ #<Data>,<EA3> ADD.I #<data>,An [-32767 <= data <= +32767] =====> LEA <Data>(An),An
Last reviewed: 6/14/93
When I try to turn on Macintosh interrupts after simple diagnostics, the MPW Assembler reports an error, as follows:
# 303: ANDI.W #~$0700,SR ### Error 54 ### Illegal/missing operands or opcode not allowed for target machine
Is there a way to get this to work?
___
Using the tilde character on a one's complement operation in MPW Assembler can give confusing results. The assembler, in its infinite wisdom, treats all constants as signed 32-bit values. Therefore, $0700 is really treated as $00000700. The one's complement of this is $FFFFF8FF, which is too large a value for an ANDI.W instruction. The syntax that will work is:
ANDI.W #~$FFFF0700,SR