"Stubbs, Andrew" writes:
> Is GET_INNER_MODE valid for scalers though?
Yeah, GET_MODE_INNER (x) == x for scalars. That makes GET_MODE_INNER
useful for stripping vectorness or complexness without having to check
for them first. It's also useful when working out what the valid shift
ranges are fo
On 10 Aug 2020 17:23, Richard Sandiford wrote:
Andrew Stubbs writes:
> On 06/08/2020 04:54, Richard Sandiford wrote:
>>> diff --git a/gcc/emit-rtl.c b/gcc/emit-rtl.c
>>> index f9b0e9714d9..d7067989ad7 100644
>>> --- a/gcc/emit-rtl.c
>>> +++ b/gcc/emit-rtl.c
>>> @@ -947,6 +947,11 @@ validate_sub
Andrew Stubbs writes:
> On 06/08/2020 04:54, Richard Sandiford wrote:
>>> diff --git a/gcc/emit-rtl.c b/gcc/emit-rtl.c
>>> index f9b0e9714d9..d7067989ad7 100644
>>> --- a/gcc/emit-rtl.c
>>> +++ b/gcc/emit-rtl.c
>>> @@ -947,6 +947,11 @@ validate_subreg (machine_mode omode, machine_mode
>>> imode,
On 06/08/2020 04:54, Richard Sandiford wrote:
diff --git a/gcc/emit-rtl.c b/gcc/emit-rtl.c
index f9b0e9714d9..d7067989ad7 100644
--- a/gcc/emit-rtl.c
+++ b/gcc/emit-rtl.c
@@ -947,6 +947,11 @@ validate_subreg (machine_mode omode, machine_mode imode,
else if (VECTOR_MODE_P (omode)
&&
Andrew Stubbs writes:
> This patch is a prerequisite for some patches I have to add multiple
> vector sizes on amdgcn.
>
> The problem is that validate_subreg rejects things like this:
>
>(subreg:V32SF (reg:V64SF) 0)
>
> These are commonly generated by the compiler. The integer equivalents
>
This patch is a prerequisite for some patches I have to add multiple
vector sizes on amdgcn.
The problem is that validate_subreg rejects things like this:
(subreg:V32SF (reg:V64SF) 0)
These are commonly generated by the compiler. The integer equivalents
work fine.
To be honest, I don't kn
