From: Kito Cheng <kito.cheng@gmail.com>
To: juzhe.zhong@rivai.ai
Cc: gcc-patches@gcc.gnu.org, palmer@dabbelt.com, jeffreyalaw@gmail.com
Subject: Re: [PATCH V2] RISC-V: Eliminate redundant vsetvli for duplicate AVL def
Date: Mon, 24 Apr 2023 14:27:51 +0800 [thread overview]
Message-ID: <CA+yXCZDEpj5U4TFs_fakQdhGOKrNdmXR+Xr51E7QvZfyrUs3_g@mail.gmail.com> (raw)
In-Reply-To: <20230423121859.95799-1-juzhe.zhong@rivai.ai>
Committed, thanks :)
On Sun, Apr 23, 2023 at 8:19 PM <juzhe.zhong@rivai.ai> wrote:
>
> From: Juzhe-Zhong <juzhe.zhong@rivai.ai>
>
> This patch is the V2 patch:https://patchwork.sourceware.org/project/gcc/patch/20230328010124.235703-1-juzhe.zhong@rivai.ai/
>
> Address comments from Jeff. Add comments for all_avail_in_compatible_p and refine comments of codes.
>
> gcc/ChangeLog:
>
> * config/riscv/riscv-vsetvl.cc (vector_infos_manager::all_avail_in_compatible_p): New function.
> (pass_vsetvl::refine_vsetvls): Optimize vsetvls.
> * config/riscv/riscv-vsetvl.h: New function.
>
> gcc/testsuite/ChangeLog:
>
> * gcc.target/riscv/rvv/vsetvl/avl_single-102.c: New test.
>
> ---
> gcc/config/riscv/riscv-vsetvl.cc | 44 +++++++++++++++++--
> gcc/config/riscv/riscv-vsetvl.h | 1 +
> .../riscv/rvv/vsetvl/avl_single-102.c | 16 +++++++
> 3 files changed, 58 insertions(+), 3 deletions(-)
> create mode 100644 gcc/testsuite/gcc.target/riscv/rvv/vsetvl/avl_single-102.c
>
> diff --git a/gcc/config/riscv/riscv-vsetvl.cc b/gcc/config/riscv/riscv-vsetvl.cc
> index fa68b8a0462..89a45a428a4 100644
> --- a/gcc/config/riscv/riscv-vsetvl.cc
> +++ b/gcc/config/riscv/riscv-vsetvl.cc
> @@ -2446,6 +2446,26 @@ vector_infos_manager::all_same_ratio_p (sbitmap bitdata) const
> return true;
> }
>
> +/* Return TRUE if the incoming vector configuration state
> + to CFG_BB is compatible with the vector configuration
> + state in CFG_BB, FALSE otherwise. */
> +bool
> +vector_infos_manager::all_avail_in_compatible_p (const basic_block cfg_bb) const
> +{
> + const auto &info = vector_block_infos[cfg_bb->index].local_dem;
> + sbitmap avin = vector_avin[cfg_bb->index];
> + unsigned int bb_index;
> + sbitmap_iterator sbi;
> + EXECUTE_IF_SET_IN_BITMAP (avin, 0, bb_index, sbi)
> + {
> + const auto &avin_info
> + = static_cast<const vl_vtype_info &> (*vector_exprs[bb_index]);
> + if (!info.compatible_p (avin_info))
> + return false;
> + }
> + return true;
> +}
> +
> bool
> vector_infos_manager::all_same_avl_p (const basic_block cfg_bb,
> sbitmap bitdata) const
> @@ -3816,9 +3836,27 @@ pass_vsetvl::refine_vsetvls (void) const
> m_vector_manager->to_refine_vsetvls.add (rinsn);
> continue;
> }
> - rinsn = PREV_INSN (rinsn);
> - rtx new_pat = gen_vsetvl_pat (VSETVL_VTYPE_CHANGE_ONLY, info, NULL_RTX);
> - change_insn (rinsn, new_pat);
> +
> + /* If all incoming edges to a block have a vector state that is compatbile
> + with the block. In such a case we need not emit a vsetvl in the current
> + block. */
> +
> + gcc_assert (has_vtype_op (insn->rtl ()));
> + rinsn = PREV_INSN (insn->rtl ());
> + gcc_assert (vector_config_insn_p (PREV_INSN (insn->rtl ())));
> + if (m_vector_manager->all_avail_in_compatible_p (cfg_bb))
> + {
> + size_t id = m_vector_manager->get_expr_id (info);
> + if (bitmap_bit_p (m_vector_manager->vector_del[cfg_bb->index], id))
> + continue;
> + eliminate_insn (rinsn);
> + }
> + else
> + {
> + rtx new_pat
> + = gen_vsetvl_pat (VSETVL_VTYPE_CHANGE_ONLY, info, NULL_RTX);
> + change_insn (rinsn, new_pat);
> + }
> }
> }
>
> diff --git a/gcc/config/riscv/riscv-vsetvl.h b/gcc/config/riscv/riscv-vsetvl.h
> index 9041eee1281..d7a6c14e931 100644
> --- a/gcc/config/riscv/riscv-vsetvl.h
> +++ b/gcc/config/riscv/riscv-vsetvl.h
> @@ -452,6 +452,7 @@ public:
> bool all_same_ratio_p (sbitmap) const;
>
> bool all_empty_predecessor_p (const basic_block) const;
> + bool all_avail_in_compatible_p (const basic_block) const;
>
> void release (void);
> void create_bitmap_vectors (void);
> diff --git a/gcc/testsuite/gcc.target/riscv/rvv/vsetvl/avl_single-102.c b/gcc/testsuite/gcc.target/riscv/rvv/vsetvl/avl_single-102.c
> new file mode 100644
> index 00000000000..8236d4e7f18
> --- /dev/null
> +++ b/gcc/testsuite/gcc.target/riscv/rvv/vsetvl/avl_single-102.c
> @@ -0,0 +1,16 @@
> +/* { dg-do compile } */
> +/* { dg-options "-march=rv32gcv -mabi=ilp32 -fno-schedule-insns -fno-schedule-insns2 -fno-tree-vectorize -frename-registers" } */
> +
> +#include "riscv_vector.h"
> +
> +void f (int8_t* base1,int8_t* base2,int8_t* out,int n)
> +{
> + vint8mf4_t v = __riscv_vle8_v_i8mf4 (base1, 32);
> + for (int i = 0; i < n; i++){
> + v = __riscv_vor_vx_i8mf4 (v, 101, 32);
> + v = __riscv_vle8_v_i8mf4_tu (v, base2, 32);
> + }
> + __riscv_vse8_v_i8mf4 (out, v, 32);
> +}
> +
> +/* { dg-final { scan-assembler-times {vsetvli} 1 { target { no-opts "-O0" no-opts "-O1" no-opts "-Os" no-opts "-g" no-opts "-funroll-loops" } } } } */
> --
> 2.36.1
>
prev parent reply other threads:[~2023-04-24 6:28 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-04-23 12:18 juzhe.zhong
2023-04-24 6:27 ` Kito Cheng [this message]
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=CA+yXCZDEpj5U4TFs_fakQdhGOKrNdmXR+Xr51E7QvZfyrUs3_g@mail.gmail.com \
--to=kito.cheng@gmail.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=jeffreyalaw@gmail.com \
--cc=juzhe.zhong@rivai.ai \
--cc=palmer@dabbelt.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).