On Wed, Nov 27, 2024 at 1:20 AM Andrew Pinski <quic_apin...@quicinc.com> wrote:
>
> For a double conversion, we will simplify it into a conversion
> with an and if the outer type and inside precision matches and
> the intra precision is smaller and unsigned. We should be able
> to extend this to where the outer precision is larger too.
> This is a good canonicalization too.
OK.
Thanks,
Richard.
> Bootstrapped and tested on x86_64-linux-gnu.
>
> PR tree-optimization/117776
> gcc/ChangeLog:
>
> * match.pd (nested int casts): Allow for the case
> where the final prec is greater than the original
> prec.
>
> gcc/testsuite/ChangeLog:
>
> * g++.dg/vect/pr117776.cc: New test.
> * gcc.dg/tree-ssa/cast-3.c: New test.
>
> Signed-off-by: Andrew Pinski <quic_apin...@quicinc.com>
> ---
> gcc/match.pd | 4 +-
> gcc/testsuite/g++.dg/vect/pr117776.cc | 62 ++++++++++++++++++++++++++
> gcc/testsuite/gcc.dg/tree-ssa/cast-3.c | 25 +++++++++++
> 3 files changed, 89 insertions(+), 2 deletions(-)
> create mode 100644 gcc/testsuite/g++.dg/vect/pr117776.cc
> create mode 100644 gcc/testsuite/gcc.dg/tree-ssa/cast-3.c
>
> diff --git a/gcc/match.pd b/gcc/match.pd
> index ad354274b2a..5576867e9fc 100644
> --- a/gcc/match.pd
> +++ b/gcc/match.pd
> @@ -5473,8 +5473,8 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
> canonicalized as bitwise and of a mask. */
> (if (GIMPLE /* PR70366: doing this in GENERIC breaks -Wconversion. */
> && final_int && inter_int && inside_int
> - && final_prec == inside_prec
> - && final_prec > inter_prec
> + && final_prec >= inside_prec
> + && inside_prec > inter_prec
> && inter_unsignedp)
> (convert (bit_and @0 { wide_int_to_tree
> (inside_type,
> diff --git a/gcc/testsuite/g++.dg/vect/pr117776.cc
> b/gcc/testsuite/g++.dg/vect/pr117776.cc
> new file mode 100644
> index 00000000000..cbb8079bd91
> --- /dev/null
> +++ b/gcc/testsuite/g++.dg/vect/pr117776.cc
> @@ -0,0 +1,62 @@
> +// { dg-do compile }
> +// { dg-require-effective-target vect_int }
> +
> +// PR tree-optimization/117776
> +
> +typedef __SIZE_TYPE__ size_t;
> +typedef unsigned int u32;
> +typedef unsigned char u8;
> +
> +static inline const bool
> +is_even_bool(u8 n)
> +{
> + return n % 2 == 0;
> +}
> +
> +static inline const u32
> +is_even_u32(u8 n)
> +{
> + return n % 2 == 0;
> +}
> +
> +static inline
> +const u32 is_even_convert(u8 n)
> +{
> + return is_even_bool(n);
> +}
> +
> +u32 count_even_v1(u8 *data, size_t len)
> +{
> + u32 ret = 0;
> + for (size_t i = 0; i < len; i++)
> + ret += is_even_bool(data[i]);
> + return ret;
> +}
> +
> +u32 count_even_v2(u8 *data, size_t len)
> +{
> + u32 ret = 0;
> + for (size_t i = 0; i < len; i++)
> + ret += is_even_u32(data[i]);
> + return ret;
> +}
> +
> +u32 count_even_v3(u8 *data, size_t len)
> +{
> + u32 ret = 0;
> + for (size_t i = 0; i < len; i++)
> + ret += is_even_convert(data[i]);
> + return ret;
> +}
> +
> +u32 count_even_v4(u8 *data, size_t len)
> +{
> + u32 ret = 0;
> + for (size_t i = 0; i < len; i++)
> + ret += static_cast<u32>(is_even_bool(data[i]));
> + return ret;
> +}
> +
> +// All 4 count_even_v functions should be vectorized
> +
> +// { dg-final { scan-tree-dump-times "LOOP VECTORIZED" 4 "vect" } }
> diff --git a/gcc/testsuite/gcc.dg/tree-ssa/cast-3.c
> b/gcc/testsuite/gcc.dg/tree-ssa/cast-3.c
> new file mode 100644
> index 00000000000..37214ecb915
> --- /dev/null
> +++ b/gcc/testsuite/gcc.dg/tree-ssa/cast-3.c
> @@ -0,0 +1,25 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O2 -fdump-tree-optimized" } */
> +
> +/* PR tree-optimization/117776 */
> +
> +/* These 2 functions should be optimized to just `return 1;`
> + as `(int)(unsigned smaller)medium` is the same as
> + `(int)medium&smaller_mask` where smaller_mask in this case is
> + either `1` (_Bool) or `0xff` (char). */
> +int f(unsigned char a)
> +{
> + _Bool odd = (a & 1) != 0;
> + int odd1 = (a & 1) != 0;
> + return odd == odd1;
> +}
> +int g(unsigned short a)
> +{
> + unsigned char b = a;
> + int d = b;
> + unsigned short c = a&0xff;
> + int e = c;
> + return d == e;
> +}
> +
> +/* { dg-final { scan-tree-dump-times "return 1" 2 "optimized" } } */
> --
> 2.43.0
>
