Re: [PATCH] MATCH: Optimize COND_ADD_LEN reduction pattern

[Richard Biener <richard.guenther@gmail.com>]

On Tue, Sep 26, 2023 at 9:13 AM Juzhe-Zhong <juzhe.zhong@rivai.ai> wrote:
>
>
> This patch leverage this commit: https://gcc.gnu.org/git/gitweb.cgi?p=gcc.git;h=62b505a4d5fc89
> to optimize COND_LEN_ADD reduction pattern.
>
> We are doing optimization of VEC_COND_EXPR + COND_LEN_ADD -> COND_LEN_ADD.
>
> Consider thsi following case:
>
> #include <stdint.h>
>
> void
> pr11594 (uint64_t *restrict a, uint64_t *restrict b, int loop_size)
> {
>   uint64_t result = 0;
>
>   for (int i = 0; i < loop_size; i++)
>     {
>       if (b[i] <= a[i])
>         {
>           result += a[i];
>         }
>     }
>
>   a[0] = result;
> }
>
> Before this patch:
>         vsetvli a7,zero,e64,m1,ta,ma
>         vmv.v.i v2,0
>         vmv1r.v v3,v2                    --- redundant
> .L3:
>         vsetvli a5,a2,e64,m1,ta,ma
>         vle64.v v1,0(a3)
>         vle64.v v0,0(a1)
>         slli    a6,a5,3
>         vsetvli a7,zero,e64,m1,ta,ma
>         sub     a2,a2,a5
>         vmsleu.vv       v0,v0,v1
>         add     a1,a1,a6
>         vmerge.vvm      v1,v3,v1,v0     ---- redundant.
>         add     a3,a3,a6
>         vsetvli zero,a5,e64,m1,tu,ma
>         vadd.vv v2,v2,v1
>         bne     a2,zero,.L3
>         li      a5,0
>         vsetvli a4,zero,e64,m1,ta,ma
>         vmv.s.x v1,a5
>         vredsum.vs      v2,v2,v1
>         vmv.x.s a5,v2
>         sd      a5,0(a0)
>         ret
>
> After this patch:
>
>         vsetvli a6,zero,e64,m1,ta,ma
>         vmv.v.i v1,0
> .L3:
>         vsetvli a5,a2,e64,m1,ta,ma
>         vle64.v v2,0(a4)
>         vle64.v v0,0(a1)
>         slli    a3,a5,3
>         vsetvli a6,zero,e64,m1,ta,ma
>         sub     a2,a2,a5
>         vmsleu.vv       v0,v0,v2
>         add     a1,a1,a3
>         vsetvli zero,a5,e64,m1,tu,mu
>         add     a4,a4,a3
>         vadd.vv v1,v1,v2,v0.t
>         bne     a2,zero,.L3
>         li      a5,0
>         vsetivli        zero,1,e64,m1,ta,ma
>         vmv.s.x v2,a5
>         vsetvli a5,zero,e64,m1,ta,ma
>         vredsum.vs      v1,v1,v2
>         vmv.x.s a5,v1
>         sd      a5,0(a0)
>         ret
>
> Bootstrap && Regression is running.
>
> Ok for trunk when testing passes ?
>
>         PR tree-optimization/111594
>         PR tree-optimization/110660
>
> gcc/ChangeLog:
>
>         * match.pd: Optimize COND_LEN_ADD reduction.
>
> gcc/testsuite/ChangeLog:
>
>         * gcc.target/riscv/rvv/autovec/cond/cond_reduc-1.c: New test.
>         * gcc.target/riscv/rvv/autovec/cond/pr111594.c: New test.
>
> ---
>  gcc/match.pd                                  | 13 +++++++++
>  .../riscv/rvv/autovec/cond/cond_reduc-1.c     | 29 +++++++++++++++++++
>  .../riscv/rvv/autovec/cond/pr111594.c         | 22 ++++++++++++++
>  3 files changed, 64 insertions(+)
>  create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/cond/cond_reduc-1.c
>  create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/autovec/cond/pr111594.c
>
> diff --git a/gcc/match.pd b/gcc/match.pd
> index a17778fbaa6..af8d12c138e 100644
> --- a/gcc/match.pd
> +++ b/gcc/match.pd
> @@ -8866,6 +8866,19 @@ and,
>    (IFN_COND_ADD @0 @1 (vec_cond @2 @3 integer_zerop) @1)
>     (IFN_COND_ADD (bit_and @0 @2) @1 @3 @1))
>
> +/* Detect simplication for a conditional length reduction where
> +
> +   a = mask ? b : 0
> +   c = i < len + bias ? d + a : d
> +
> +   is turned into
> +
> +   c = mask && i < len ? d + b : d.  */
> +(simplify
> +  (IFN_COND_LEN_ADD integer_minus_onep @0 (vec_cond @1 @2 zerop) @0 @3 @4)

I think you want intger_truep instead of integer_minus_onep for
readability.  Since you
use zerop here can you also adjust the preceeding pattern?

> +   (if (!HONOR_NANS (type) && !HONOR_SIGNED_ZEROS (type))

it might be better to check ANY_INTEGRAL_TYPE_P (type) ||
fold_real_zero_addition_p (type, NULL_TREE, @5, 0)
your change misses HONOR_SIGN_DEPENDENT_ROUNDING I think.

> +    (IFN_COND_LEN_ADD @1 @0 @2 @0 @3 @4)))
> +



>  /* For pointers @0 and @2 and nonnegative constant offset @1, look for
>     expressions like:
>
> diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/cond/cond_reduc-1.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/cond/cond_reduc-1.c
> new file mode 100644
> index 00000000000..db6f9d1ec6c
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/cond/cond_reduc-1.c
> @@ -0,0 +1,29 @@
> +/* { dg-do compile } */
> +/* { dg-additional-options "-march=rv64gcv_zvfh -mabi=lp64d -fno-vect-cost-model -ffast-math -fdump-tree-optimized" } */
> +
> +#include <stdint-gcc.h>
> +
> +#define COND_REDUCTION(TYPE)                                                   \
> +  TYPE foo##TYPE (TYPE *restrict a, TYPE *restrict b, int loop_size)           \
> +  {                                                                            \
> +    TYPE result = 0;                                                           \
> +    for (int i = 0; i < loop_size; i++)                                        \
> +      if (b[i] <= a[i])                                                        \
> +       result += a[i];                                                        \
> +    return result;                                                             \
> +  }
> +
> +COND_REDUCTION (int8_t)
> +COND_REDUCTION (int16_t)
> +COND_REDUCTION (int32_t)
> +COND_REDUCTION (int64_t)
> +COND_REDUCTION (uint8_t)
> +COND_REDUCTION (uint16_t)
> +COND_REDUCTION (uint32_t)
> +COND_REDUCTION (uint64_t)
> +COND_REDUCTION (_Float16)
> +COND_REDUCTION (float)
> +COND_REDUCTION (double)
> +
> +/* { dg-final { scan-tree-dump-not "VCOND_MASK" "optimized" } } */
> +/* { dg-final { scan-tree-dump-times "COND_LEN_ADD" 11 "optimized" } } */
> diff --git a/gcc/testsuite/gcc.target/riscv/rvv/autovec/cond/pr111594.c b/gcc/testsuite/gcc.target/riscv/rvv/autovec/cond/pr111594.c
> new file mode 100644
> index 00000000000..6d81b26fbd0
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/riscv/rvv/autovec/cond/pr111594.c
> @@ -0,0 +1,22 @@
> +/* { dg-do compile } */
> +/* { dg-additional-options "-march=rv64gcv -mabi=lp64d -fno-vect-cost-model -ffast-math" } */
> +
> +#include <stdint-gcc.h>
> +
> +void
> +pr11594 (uint64_t *restrict a, uint64_t *restrict b, int loop_size)
> +{
> +  uint64_t result = 0;
> +
> +  for (int i = 0; i < loop_size; i++)
> +    {
> +      if (b[i] <= a[i])
> +       {
> +         result += a[i];
> +       }
> +    }
> +
> +  a[0] = result;
> +}
> +
> +/* { dg-final { scan-assembler-not {vmerge} } } */
> --
> 2.36.3
>