From: Richard Henderson <richard.henderson@linaro.org>
To: Daniel Henrique Barboza <dbarboza@ventanamicro.com>,
qemu-devel@nongnu.org
Cc: qemu-riscv@nongnu.org, alistair.francis@wdc.com,
bmeng@tinylab.org, liwei1518@gmail.com,
zhiwei_liu@linux.alibaba.com, palmer@rivosinc.com,
max.chou@sifive.com
Subject: Re: [PATCH v3 2/5] target/riscv: remove 'over' brconds from vector trans
Date: Tue, 20 Feb 2024 10:19:20 -1000 [thread overview]
Message-ID: <e9cdebea-555f-4eb4-9a71-305f3ef92c42@linaro.org> (raw)
In-Reply-To: <20240220192607.141880-3-dbarboza@ventanamicro.com>
On 2/20/24 09:26, Daniel Henrique Barboza wrote:
> Most of the vector translations has this following pattern at the start:
>
> TCGLabel *over = gen_new_label();
> tcg_gen_brcond_tl(TCG_COND_GEU, cpu_vstart, cpu_vl, over);
>
> And then right at the end:
>
> gen_set_label(over);
> return true;
>
> This means that if vstart >= vl we'll not set vstart = 0 at the end of
> the insns - this is done inside the helper that is being skipped. The
> reason why this pattern hasn't been a bigger problem is because the
> conditional vstart >= vl is very rare.
>
> Checking all the helpers in vector_helper.c we see all of them with a
> pattern like this:
>
> for (i = env->vstart; i < vl; i++) {
> (...)
> }
> env->vstart = 0;
>
> Thus they can handle vstart >= vl case gracefully, with the benefit of
> setting env->vstart = 0 during the process.
>
> Remove all 'over' conditionals and let the helper set env->vstart = 0
> every time.
>
> Suggested-by: Richard Henderson<richard.henderson@linaro.org>
> Signed-off-by: Daniel Henrique Barboza<dbarboza@ventanamicro.com>
> ---
> target/riscv/insn_trans/trans_rvv.c.inc | 116 -----------------------
> target/riscv/insn_trans/trans_rvvk.c.inc | 18 ----
> 2 files changed, 134 deletions(-)
Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
r~
next prev parent reply other threads:[~2024-02-20 20:20 UTC|newest]
Thread overview: 12+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-02-20 19:26 [PATCH v3 0/5] riscv: set vstart_eq_zero on mark_vs_dirty Daniel Henrique Barboza
2024-02-20 19:26 ` [PATCH v3 1/5] trans_rvv.c.inc: mark_vs_dirty() before stores Daniel Henrique Barboza
2024-02-20 20:17 ` Richard Henderson
2024-02-20 21:10 ` Daniel Henrique Barboza
2024-02-20 19:26 ` [PATCH v3 2/5] target/riscv: remove 'over' brconds from vector trans Daniel Henrique Barboza
2024-02-20 20:19 ` Richard Henderson [this message]
2024-02-20 19:26 ` [PATCH v3 3/5] target/riscv/vector_helper.c: set vstart = 0 in GEN_VEXT_VSLIDEUP_VX() Daniel Henrique Barboza
2024-02-20 20:20 ` Richard Henderson
2024-02-20 19:26 ` [PATCH v3 4/5] trans_rvv.c.inc: remove redundant mark_vs_dirty() calls Daniel Henrique Barboza
2024-02-20 20:20 ` Richard Henderson
2024-02-20 19:26 ` [PATCH v3 5/5] target/riscv: Clear vstart_qe_zero flag Daniel Henrique Barboza
2024-02-20 20:22 ` Richard Henderson
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=e9cdebea-555f-4eb4-9a71-305f3ef92c42@linaro.org \
--to=richard.henderson@linaro.org \
--cc=alistair.francis@wdc.com \
--cc=bmeng@tinylab.org \
--cc=dbarboza@ventanamicro.com \
--cc=liwei1518@gmail.com \
--cc=max.chou@sifive.com \
--cc=palmer@rivosinc.com \
--cc=qemu-devel@nongnu.org \
--cc=qemu-riscv@nongnu.org \
--cc=zhiwei_liu@linux.alibaba.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 NNTP newsgroup(s).