I know using more precise constraints might result in better code gen
in some situations, but I am Curious what's the difference between the
using pattern condition and constraints/predicates in this case? Is
there any performance or code gen difference?
On Fri, Aug 19, 2022 at 6:07 AM apinski--- via Gcc-patches
wrote:
>
> From: Andrew Pinski
>
> This simplifies the code by adding a predicate and a constraint for 1/2/3.
> The aarch64 backend has a similar predicate called aarch64_shift_imm_
> which they use there.
>
> OK? Built and tested on riscv32-linux-gnu and riscv64-linux-gnu with no
> regressions.
>
> Thanks,
> Andrew Pinski
>
> gcc/ChangeLog:
>
> * config/riscv/constraints.md (Ds3): New constraint.
> * config/riscv/predicates.md (imm123_operand): New predicate.
> * config/riscv/bitmanip.md (*shNadd): Use Ds3 and imm123_operand.
> (*shNadduw): Likewise.
> ---
> gcc/config/riscv/bitmanip.md| 8 +++-
> gcc/config/riscv/constraints.md | 6 ++
> gcc/config/riscv/predicates.md | 5 +
> 3 files changed, 14 insertions(+), 5 deletions(-)
>
> diff --git a/gcc/config/riscv/bitmanip.md b/gcc/config/riscv/bitmanip.md
> index ebd6eee1a22..73a36f7751b 100644
> --- a/gcc/config/riscv/bitmanip.md
> +++ b/gcc/config/riscv/bitmanip.md
> @@ -32,10 +32,9 @@ (define_insn "*zero_extendsidi2_bitmanip"
> (define_insn "*shNadd"
>[(set (match_operand:X 0 "register_operand" "=r")
> (plus:X (ashift:X (match_operand:X 1 "register_operand" "r")
> - (match_operand:QI 2 "immediate_operand" "I"))
> + (match_operand:QI 2 "imm123_operand" "Ds3"))
> (match_operand:X 3 "register_operand" "r")))]
> - "TARGET_ZBA
> - && (INTVAL (operands[2]) >= 1) && (INTVAL (operands[2]) <= 3)"
> + "TARGET_ZBA"
>"sh%2add\t%0,%1,%3"
>[(set_attr "type" "bitmanip")
> (set_attr "mode" "")])
> @@ -44,11 +43,10 @@ (define_insn "*shNadduw"
>[(set (match_operand:DI 0 "register_operand" "=r")
> (plus:DI
> (and:DI (ashift:DI (match_operand:DI 1 "register_operand" "r")
> -(match_operand:QI 2 "immediate_operand" "I"))
> +(match_operand:QI 2 "imm123_operand" "Ds3"))
> (match_operand 3 "immediate_operand" ""))
> (match_operand:DI 4 "register_operand" "r")))]
>"TARGET_64BIT && TARGET_ZBA
> - && (INTVAL (operands[2]) >= 1) && (INTVAL (operands[2]) <= 3)
> && (INTVAL (operands[3]) >> INTVAL (operands[2])) == 0x"
>"sh%2add.uw\t%0,%1,%4"
>[(set_attr "type" "bitmanip")
> diff --git a/gcc/config/riscv/constraints.md b/gcc/config/riscv/constraints.md
> index bafa4188ccb..61b84875fd9 100644
> --- a/gcc/config/riscv/constraints.md
> +++ b/gcc/config/riscv/constraints.md
> @@ -54,6 +54,12 @@ (define_constraint "L"
>(and (match_code "const_int")
> (match_test "LUI_OPERAND (ival)")))
>
> +(define_constraint "Ds3"
> + "@internal
> + 1, 2 or 3 immediate"
> + (and (match_code "const_int")
> + (match_test "IN_RANGE (ival, 1, 3)")))
> +
> ;; Floating-point constant +0.0, used for FCVT-based moves when FMV is
> ;; not available in RV32.
> (define_constraint "G"
> diff --git a/gcc/config/riscv/predicates.md b/gcc/config/riscv/predicates.md
> index 79e0c1d5589..2af7f661d6f 100644
> --- a/gcc/config/riscv/predicates.md
> +++ b/gcc/config/riscv/predicates.md
> @@ -244,6 +244,11 @@ (define_predicate "imm5_operand"
>(and (match_code "const_int")
> (match_test "INTVAL (op) < 5")))
>
> +;; A const_int for sh1add/sh2add/sh3add
> +(define_predicate "imm123_operand"
> + (and (match_code "const_int")
> + (match_test "IN_RANGE (INTVAL (op), 1, 3)")))
> +
> ;; A CONST_INT operand that consists of a single run of consecutive set bits.
> (define_predicate "consecutive_bits_operand"
>(match_code "const_int")
> --
> 2.27.0
>