> -Original Message-
> From: Richard Sandiford
> Sent: Thursday, January 9, 2025 3:09 PM
> To: Tamar Christina
> Cc: gcc-patches@gcc.gnu.org; nd ; Richard Earnshaw
> ; ktkac...@gcc.gnu.org
> Subject: Re: [PATCH]AArch64: Fix costing of emulated gathers/scatters
&g
Tamar Christina writes:
>> > + After the final loads are done it issues a
>> > + vec_construct to recreate the vector from the scalar. For costing
>> > when
>> > + we see a vec_to_scalar on a stmt with VMAT_GATHER_SCATTER we are
>> dealing
>> > + with an emulated instruction and
> -Original Message-
> From: Richard Sandiford
> Sent: Wednesday, January 8, 2025 10:30 AM
> To: Tamar Christina
> Cc: gcc-patches@gcc.gnu.org; nd ; Richard Earnshaw
> ; ktkac...@gcc.gnu.org
> Subject: Re: [PATCH]AArch64: Fix costing of emulated gathers/scatters
&g
Tamar Christina writes:
>> >> i.e. we use separate address arithmetic and avoid UMOVs. Counting
>> >> two loads and one store for each element of the scatter store seems
>> >> like overkill for that.
>> >
>> > Hmm agreed..
>> >
>> > How about for stores we increase the load counts by count / 2?
>
> >> i.e. we use separate address arithmetic and avoid UMOVs. Counting
> >> two loads and one store for each element of the scatter store seems
> >> like overkill for that.
> >
> > Hmm agreed..
> >
> > How about for stores we increase the load counts by count / 2?
> >
> > This would account for th
Tamar Christina writes:
>> >> So I think ideally, we should try to detect whether the indices come
>> >> directly from memory or are the result of arithmetic. In the former case,
>> >> we should do the loads adjustment above. In the latter case, we should
>> >> keep the vec_to_scalar accounting
> -Original Message-
> From: Richard Sandiford
> Sent: Thursday, January 2, 2025 5:54 PM
> To: Tamar Christina
> Cc: gcc-patches@gcc.gnu.org; nd ; Richard Earnshaw
> ; ktkac...@gcc.gnu.org
> Subject: Re: [PATCH]AArch64: Fix costing of emulated gathers/scatters
&g
Tamar Christina writes:
>> > [...]
>> > #define iterations 10
>> > #define LEN_1D 32000
>> >
>> > float a[LEN_1D], b[LEN_1D];
>> >
>> > float
>> > s4115 (int *ip)
>> > {
>> > float sum = 0.;
>> > for (int i = 0; i < LEN_1D; i++)
>> > {
>> > sum += a[i] * b[ip[i]];
>
> -Original Message-
> From: Richard Sandiford
> Sent: Thursday, January 2, 2025 4:52 PM
> To: Tamar Christina
> Cc: gcc-patches@gcc.gnu.org; nd ; Richard Earnshaw
> ; ktkac...@gcc.gnu.org
> Subject: Re: [PATCH]AArch64: Fix costing of emulated gathers/scatters
&g
Tamar Christina writes:
> Hi All,
>
> When a target does not support gathers and scatters the vectorizer tries to
> emulate these using scalar loads/stores and a reconstruction of vectors from
> scalar.
>
> The loads are still marked with VMAT_GATHER_SCATTER to indicate that they are
> gather/scat
Hi All,
When a target does not support gathers and scatters the vectorizer tries to
emulate these using scalar loads/stores and a reconstruction of vectors from
scalar.
The loads are still marked with VMAT_GATHER_SCATTER to indicate that they are
gather/scatters, however the vectorizer also asks
11 matches
Mail list logo
