Nit: for title, should be SAT_SUB(x, 1) instead of SAT_SUB(x - 1)?
Pan
-----Original Message-----
From: Andrew Pinski <pins...@gmail.com>
Sent: Wednesday, October 23, 2024 1:46 PM
To: Li Xu <xu...@eswincomputing.com>
Cc: gcc-patches@gcc.gnu.org; kito.ch...@gmail.com; richard.guent...@gmail.com;
tamar.christ...@arm.com; juzhe.zh...@rivai.ai; Li, Pan2 <pan2...@intel.com>;
jeffreya...@gmail.com; rdapp....@gmail.com
Subject: Re: [PATCH 1/2] Match: Simplify unsigned scalar sat_sub(x -1) to (x -
x != 0)
On Tue, Oct 22, 2024 at 10:39 PM Li Xu <xu...@eswincomputing.com> wrote:
>
> From: xuli <xu...@eswincomputing.com>
>
> When the imm operand op1=1 in the unsigned scalar sat_sub form2 below,
> we can simplify (x != 0 ? x + max : 0) to (x - x != 0), thereby eliminating
> a branch instruction.
>
> Form2:
> T __attribute__((noinline)) \
> sat_u_sub_imm##IMM##_##T##_fmt_2 (T x) \
> { \
> return x >= (T)IMM ? x - (T)IMM : 0; \
> }
>
> Take below form 2 as example:
> DEF_SAT_U_SUB_IMM_FMT_2(uint8_t, 1)
>
> Before this patch:
> __attribute__((noinline))
> uint8_t sat_u_sub_imm1_uint8_t_fmt_2 (uint8_t x)
> {
> uint8_t _1;
> uint8_t _3;
>
> <bb 2> [local count: 1073741824]:
> if (x_2(D) != 0)
> goto <bb 3>; [50.00%]
> else
> goto <bb 4>; [50.00%]
>
> <bb 3> [local count: 536870912]:
> _3 = x_2(D) + 255;
>
> <bb 4> [local count: 1073741824]:
> # _1 = PHI <x_2(D)(2), _3(3)>
> return _1;
>
> }
>
> Assembly code:
> sat_u_sub_imm1_uint8_t_fmt_2:
> beq a0,zero,.L2
> addiw a0,a0,-1
> andi a0,a0,0xff
> .L2:
> ret
>
> After this patch:
> __attribute__((noinline))
> uint8_t sat_u_sub_imm1_uint8_t_fmt_2 (uint8_t x)
> {
> _Bool _1;
> unsigned char _2;
> uint8_t _4;
>
> <bb 2> [local count: 1073741824]:
> _1 = x_3(D) != 0;
> _2 = (unsigned char) _1;
> _4 = x_3(D) - _2;
> return _4;
>
> }
>
> Assembly code:
> sat_u_sub_imm1_uint8_t_fmt_2:
> snez a5,a0
> subw a0,a0,a5
> andi a0,a0,0xff
> ret
>
> The below test suites are passed for this patch:
> 1. The rv64gcv fully regression tests.
> 2. The x86 bootstrap tests.
> 3. The x86 fully regression tests.
>
> Signed-off-by: Li Xu <xu...@eswincomputing.com>
> gcc/ChangeLog:
>
> * match.pd: Simplify (x != 0 ? x + max : 0) to (x - x != 0).
> ---
> gcc/match.pd | 16 ++++++++++++++++
> 1 file changed, 16 insertions(+)
>
> diff --git a/gcc/match.pd b/gcc/match.pd
> index 0455dfa6993..9dcb9555cc0 100644
> --- a/gcc/match.pd
> +++ b/gcc/match.pd
> @@ -3383,6 +3383,22 @@ DEFINE_INT_AND_FLOAT_ROUND_FN (RINT)
> }
> (if (wi::eq_p (sum, wi::uhwi (0, precision)))))))
>
> +/* The boundary condition for case 10: IMM = 1:
> + SAT_U_SUB = X >= IMM ? (X - IMM) : 0.
> + simplify (X != 0 ? X + max : 0) to (X - X != 0). */
> +(simplify
> + (cond (ne @0 integer_zerop) (plus @0 INTEGER_CST@1) integer_zerop)
I think you could use integer_all_onesp@1 here rather than the check
you are doing on the INTEGER_CST.
Then you only need the INTEGRAL_TYPE_P check.
You should also add a few testcases for this. See
`gcc.dg/tree-ssa/phiopt*.c` for similar testcases.
Thanks,
Andrew Pinski
> + (if (INTEGRAL_TYPE_P (type) && TYPE_UNSIGNED (type)
> + && types_match (type, @0))
> + (with
> + {
> + unsigned precision = TYPE_PRECISION (type);
> + wide_int max = wi::mask (precision, false, precision);
> + wide_int c1 = wi::to_wide (@1);
> + }
> + (if (wi::eq_p (c1, max))
> + (minus @0 (convert (ne @0 { build_zero_cst (type); })))))))
> +
> /* Signed saturation sub, case 1:
> T minus = (T)((UT)X - (UT)Y);
> SAT_S_SUB = (X ^ Y) & (X ^ minus) < 0 ? (-(T)(X < 0) ^ MAX) : minus;
> --
> 2.17.1
>
