Check mask type when doing cond_op related gimple simplification.

  Ping.

  Bootstrapped and regtest on x86_64-linux-gnu{-m32,}, aarch64-unknown-linux-gnu{-m32,}
  Ok for trunk?

gcc/ChangeLog:

	PR middle-end/102080
	* match.pd: Check mask type when doing cond_op related gimple
	simplification.
	* tree.c (is_truth_type_for): New function.
	* tree.h (is_truth_type_for): New declaration.

gcc/testsuite/ChangeLog:

	PR middle-end/102080
	* gcc.target/i386/pr102080.c: New test.
---
 gcc/match.pd                             |  8 +++----
 gcc/testsuite/gcc.target/i386/pr102080.c | 19 ++++++++++++++++
 gcc/tree.c                               | 29 ++++++++++++++++++++++++
 gcc/tree.h                               |  1 +
 4 files changed, 53 insertions(+), 4 deletions(-)
 create mode 100644 gcc/testsuite/gcc.target/i386/pr102080.c
  

On Thu, Sep 16, 2021 at 8:27 AM liuhongt <hongtao.liu@intel.com> wrote:
>
> Ping.
>
>   Bootstrapped and regtest on x86_64-linux-gnu{-m32,}, aarch64-unknown-linux-gnu{-m32,}
>   Ok for trunk?
>
> gcc/ChangeLog:
>
>         PR middle-end/102080
>         * match.pd: Check mask type when doing cond_op related gimple
>         simplification.
>         * tree.c (is_truth_type_for): New function.
>         * tree.h (is_truth_type_for): New declaration.
>
> gcc/testsuite/ChangeLog:
>
>         PR middle-end/102080
>         * gcc.target/i386/pr102080.c: New test.
> ---
>  gcc/match.pd                             |  8 +++----
>  gcc/testsuite/gcc.target/i386/pr102080.c | 19 ++++++++++++++++
>  gcc/tree.c                               | 29 ++++++++++++++++++++++++
>  gcc/tree.h                               |  1 +
>  4 files changed, 53 insertions(+), 4 deletions(-)
>  create mode 100644 gcc/testsuite/gcc.target/i386/pr102080.c
>
> diff --git a/gcc/match.pd b/gcc/match.pd
> index 008f7758c96..41f9e6d97f0 100644
> --- a/gcc/match.pd
> +++ b/gcc/match.pd
> @@ -7020,13 +7020,13 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
>    (vec_cond @0 (view_convert? (uncond_op@4 @1 @2)) @3)
>    (with { tree op_type = TREE_TYPE (@4); }
>     (if (vectorized_internal_fn_supported_p (as_internal_fn (cond_op), op_type)
> -       && element_precision (type) == element_precision (op_type))
> +       && is_truth_type_for (op_type, TREE_TYPE (@0)))
>      (view_convert (cond_op @0 @1 @2 (view_convert:op_type @3))))))
>   (simplify
>    (vec_cond @0 @1 (view_convert? (uncond_op@4 @2 @3)))
>    (with { tree op_type = TREE_TYPE (@4); }
>     (if (vectorized_internal_fn_supported_p (as_internal_fn (cond_op), op_type)
> -       && element_precision (type) == element_precision (op_type))
> +       && is_truth_type_for (op_type, TREE_TYPE (@0)))
>      (view_convert (cond_op (bit_not @0) @2 @3 (view_convert:op_type @1)))))))
>
>  /* Same for ternary operations.  */
> @@ -7036,13 +7036,13 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
>    (vec_cond @0 (view_convert? (uncond_op@5 @1 @2 @3)) @4)
>    (with { tree op_type = TREE_TYPE (@5); }
>     (if (vectorized_internal_fn_supported_p (as_internal_fn (cond_op), op_type)
> -       && element_precision (type) == element_precision (op_type))
> +       && is_truth_type_for (op_type, TREE_TYPE (@0)))
>      (view_convert (cond_op @0 @1 @2 @3 (view_convert:op_type @4))))))
>   (simplify
>    (vec_cond @0 @1 (view_convert? (uncond_op@5 @2 @3 @4)))
>    (with { tree op_type = TREE_TYPE (@5); }
>     (if (vectorized_internal_fn_supported_p (as_internal_fn (cond_op), op_type)
> -       && element_precision (type) == element_precision (op_type))
> +       && is_truth_type_for (op_type, TREE_TYPE (@0)))
>      (view_convert (cond_op (bit_not @0) @2 @3 @4
>                   (view_convert:op_type @1)))))))
>  #endif
> diff --git a/gcc/testsuite/gcc.target/i386/pr102080.c b/gcc/testsuite/gcc.target/i386/pr102080.c
> new file mode 100644
> index 00000000000..4c5ee32ee63
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/i386/pr102080.c
> @@ -0,0 +1,19 @@
> +/* { dg-do compile } */
> +/* { dg-options "-O2" } */
> +
> +#include<immintrin.h>
> +typedef float __m256 __attribute__((__vector_size__(32)));
> +__m256 _mm256_blendv_ps___Y, _mm256_blendv_ps___M, _mm256_mul_ps___A,
> +  _mm256_mul_ps___B, IfThenElse___trans_tmp_9;
> +
> +void
> +__attribute__ ((target("avx")))
> +IfThenElse (__m256 no) {
> +  IfThenElse___trans_tmp_9 = _mm256_blendv_ps (no, _mm256_blendv_ps___Y, _mm256_blendv_ps___M);
> +}
> +void
> +__attribute__ ((target("avx512vl")))
> +EncodedFromDisplay() {
> +  __m256 __trans_tmp_11 = _mm256_mul_ps___A * _mm256_mul_ps___B;
> +  IfThenElse(__trans_tmp_11);
> +}
> diff --git a/gcc/tree.c b/gcc/tree.c
> index 3d15948fd1a..994775d7314 100644
> --- a/gcc/tree.c
> +++ b/gcc/tree.c
> @@ -10737,6 +10737,35 @@ signed_type_for (tree type)
>    return signed_or_unsigned_type_for (0, type);
>  }
>
> +/* - For VECTOR_TYPEs:
> +    - The truth type must be a VECTOR_BOOLEAN_TYPE.
> +    - The number of elements must match (known_eq).
> +    - targetm.vectorize.get_mask_mode exists, and exactly
> +      the same mode as the truth type.
> +   - Otherwise, the truth type must be a BOOLEAN_TYPE
> +     or useless_type_conversion_p to BOOLEAN_TYPE.  */
> +bool
> +is_truth_type_for (tree type, tree truth_type)
> +{
> +  machine_mode mask_mode = TYPE_MODE (truth_type);
> +  machine_mode vmode = TYPE_MODE (type);
> +  machine_mode tmask_mode;
> +
> +  if (TREE_CODE (type) == VECTOR_TYPE)
> +    {
> +      if (VECTOR_BOOLEAN_TYPE_P (truth_type)
> +         && known_eq (TYPE_VECTOR_SUBPARTS (type),
> +                      TYPE_VECTOR_SUBPARTS (truth_type))
> +         && targetm.vectorize.get_mask_mode (vmode).exists (&tmask_mode)
> +         && tmask_mode == mask_mode)
> +       return true;
> +
> +      return false;
> +    }
> +
> +  return useless_type_conversion_p (boolean_type_node, truth_type);
> +}
> +
>  /* If TYPE is a vector type, return a signed integer vector type with the
>     same width and number of subparts. Otherwise return boolean_type_node.  */
>
> diff --git a/gcc/tree.h b/gcc/tree.h
> index 7274ba75f59..06c027da450 100644
> --- a/gcc/tree.h
> +++ b/gcc/tree.h
> @@ -4591,6 +4591,7 @@ extern tree build_string_literal (unsigned, const char * = NULL,
>  extern tree signed_or_unsigned_type_for (int, tree);
>  extern tree signed_type_for (tree);
>  extern tree unsigned_type_for (tree);
> +bool is_truth_type_for (tree, tree);

just for style add 'extern' to follow surrounding code.

The patch looks OK to me with that change.

Thanks,
Richard.

>  extern tree truth_type_for (tree);
>  extern tree build_pointer_type_for_mode (tree, machine_mode, bool);
>  extern tree build_pointer_type (tree);
> --
> 2.27.0
>