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)
Richard Henderson writes:
> On 12/22/2011 11:33 AM, Richard Sandiford wrote:
>> was it simply the mode punning from HI to V4HI that stops us from
>> using the loongson_pinsr_N insns? E.g. something like:
>
> Mostly...
>
>> I realise this isn't exactly simpler in code terms, but it avoids the
>> u
On 12/22/2011 11:33 AM, Richard Sandiford wrote:
> was it simply the mode punning from HI to V4HI that stops us from
> using the loongson_pinsr_N insns? E.g. something like:
Mostly...
> I realise this isn't exactly simpler in code terms, but it avoids the
> unspec and avoids the dual mode on mat
Looks good, but with this:
Richard Henderson writes:
> +(define_insn "*vec_setv4hi"
> + [(set (match_operand:V4HI 0 "register_operand" "=f")
> + (unspec:V4HI [(match_operand:V4HI 1 "register_operand" "f")
> + (match_operand:SI 2 "register_operand" "f")
> +
---
gcc/config/mips/loongson.h |4 +-
gcc/config/mips/loongson.md| 200 ++-
gcc/config/mips/mips-modes.def | 12 ++-
gcc/config/mips/mips-protos.h |1 +
gcc/config/mips/mips-ps-3d.md | 225 +--
gcc/config/mip
