On 12/22/2011 01:16 PM, Richard Sandiford wrote: >> That's just re-aligning the comment columns so that V16QI fits. > > OK, but there wasn't a V16QI comment that I could see.
Heh. It was supposed to be on ... >>>> +/* Double-sized vector modes for vec_concat. */ >>>> +VECTOR_MODE (INT, QI, 16); >>>> +VECTOR_MODE (INT, HI, 8); >>>> +VECTOR_MODE (INT, SI, 4); >>>> +VECTOR_MODE (FLOAT, SF, 4); ... these. I wonder where those got lost. > I'd still rather be safe than sorry when it comes to ABI stuff. Ok. > OK, but just so that I understand: does that mean we're missing some > single-operand cases that ought to be there, and this code is just > stopping us from ICEing when we hit them? Would it be worth having > a gcc_checking_assert there? I don't know if we're missing any cases. We may not be; the patterns we're using here are certainly simpler than those on x86. I can certainly add an assert... r~
