| lkcl | cesar[m]1: ha! cool! you spotted the zero-reg thing. i will comment it in get_predint, i forgot to | 10:55 |
|---|---|---|
| cesar[m]1 | We could also make get_predint output an "is always" signal, instead of relying on zero-reg. More elegant. | 10:59 |
| lkcl | true. | 12:07 |
| lkcl | gimme 1 sec... | 12:08 |
| lkcl | done | 12:12 |
| lkcl | segher: moornin. can i ask: are there any circumstances where scalar pattern-matches in rs6000.md returns arithmetic expressions for register numbers? "addi 5+3, r0, 5" rather than "addi r8, r0, 5" or just "addi 8, 0, 5"? | 12:13 |
| segher | lkcl: internal to GCC, there is only RTL here. when output as assembler code, we do not do calculations with register numbers in assembler code: that would not work for rN and %rN. there are %Ln, %Yn, %Zn for this | 16:17 |
| lkcl | segher: ahh that's excellent | 16:42 |
| segher | in many cases we do not use the output modifiers; instead, some other code calculate what register to use | 16:45 |
| segher | operands[4] = operands[1] + 1; etc. | 16:45 |
| segher | internal to GCC, register names are just numbers | 16:46 |
| lkcl | that would make sense | 17:07 |
| lkcl | faster than communicating the same thing in ASCII | 17:07 |
Generated by irclog2html.py 2.17.1 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!