From: Daniel Henrique Barboza <danielhb413@gmail.com>
To: Nicholas Piggin <npiggin@gmail.com>
Cc: Richard Henderson <richard.henderson@linaro.org>,
qemu-ppc@nongnu.org, qemu-devel@nongnu.org,
qemu-stable@nongnu.org
Subject: Re: [PATCH v2 4/4] target/ppc: Rework store conditional to avoid branch
Date: Mon, 5 Jun 2023 10:42:22 -0300 [thread overview]
Message-ID: <6a1f0c3f-2063-be85-6d58-c672472d019a@gmail.com> (raw)
In-Reply-To: <20230605025445.161932-4-npiggin@gmail.com>
On 6/4/23 23:54, Nicholas Piggin wrote:
> Rework store conditional to avoid a branch in the success case.
> Change some of the variable names and layout while here so
> gen_conditional_store more closely matches gen_stqcx_.
>
> Reviewed-by: Richard Henderson <richard.henderson@linaro.org>
> Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
> ---
Queued. Thanks,
Daniel
> v2:
> - Reinstate lost DEF_MEMOP [Richard]
>
> I think the DEF_MEMOP is redundant here, but admit that's not something
> that should be changed with this patch. I will look at cleaning those up
> later.
>
> Thanks,
> Nick
>
> target/ppc/translate.c | 63 ++++++++++++++++++++----------------------
> 1 file changed, 30 insertions(+), 33 deletions(-)
>
> diff --git a/target/ppc/translate.c b/target/ppc/translate.c
> index acb99d8691..434caad258 100644
> --- a/target/ppc/translate.c
> +++ b/target/ppc/translate.c
> @@ -3813,31 +3813,32 @@ static void gen_stdat(DisasContext *ctx)
>
> static void gen_conditional_store(DisasContext *ctx, MemOp memop)
> {
> - TCGLabel *l1 = gen_new_label();
> - TCGLabel *l2 = gen_new_label();
> - TCGv t0 = tcg_temp_new();
> - int reg = rS(ctx->opcode);
> + TCGLabel *lfail;
> + TCGv EA;
> + TCGv cr0;
> + TCGv t0;
> + int rs = rS(ctx->opcode);
>
> + lfail = gen_new_label();
> + EA = tcg_temp_new();
> + cr0 = tcg_temp_new();
> + t0 = tcg_temp_new();
> +
> + tcg_gen_mov_tl(cr0, cpu_so);
> gen_set_access_type(ctx, ACCESS_RES);
> - gen_addr_reg_index(ctx, t0);
> - tcg_gen_brcond_tl(TCG_COND_NE, t0, cpu_reserve, l1);
> - tcg_gen_brcondi_tl(TCG_COND_NE, cpu_reserve_length, memop_size(memop), l1);
> + gen_addr_reg_index(ctx, EA);
> + tcg_gen_brcond_tl(TCG_COND_NE, EA, cpu_reserve, lfail);
> + tcg_gen_brcondi_tl(TCG_COND_NE, cpu_reserve_length, memop_size(memop), lfail);
>
> - t0 = tcg_temp_new();
> tcg_gen_atomic_cmpxchg_tl(t0, cpu_reserve, cpu_reserve_val,
> - cpu_gpr[reg], ctx->mem_idx,
> + cpu_gpr[rs], ctx->mem_idx,
> DEF_MEMOP(memop) | MO_ALIGN);
> tcg_gen_setcond_tl(TCG_COND_EQ, t0, t0, cpu_reserve_val);
> tcg_gen_shli_tl(t0, t0, CRF_EQ_BIT);
> - tcg_gen_or_tl(t0, t0, cpu_so);
> - tcg_gen_trunc_tl_i32(cpu_crf[0], t0);
> - tcg_gen_br(l2);
> + tcg_gen_or_tl(cr0, cr0, t0);
>
> - gen_set_label(l1);
> -
> - tcg_gen_trunc_tl_i32(cpu_crf[0], cpu_so);
> -
> - gen_set_label(l2);
> + gen_set_label(lfail);
> + tcg_gen_trunc_tl_i32(cpu_crf[0], cr0);
> tcg_gen_movi_tl(cpu_reserve, -1);
> }
>
> @@ -3891,25 +3892,26 @@ static void gen_lqarx(DisasContext *ctx)
> /* stqcx. */
> static void gen_stqcx_(DisasContext *ctx)
> {
> - TCGLabel *lab_fail, *lab_over;
> - int rs = rS(ctx->opcode);
> + TCGLabel *lfail;
> TCGv EA, t0, t1;
> + TCGv cr0;
> TCGv_i128 cmp, val;
> + int rs = rS(ctx->opcode);
>
> if (unlikely(rs & 1)) {
> gen_inval_exception(ctx, POWERPC_EXCP_INVAL_INVAL);
> return;
> }
>
> - lab_fail = gen_new_label();
> - lab_over = gen_new_label();
> + lfail = gen_new_label();
> + EA = tcg_temp_new();
> + cr0 = tcg_temp_new();
>
> + tcg_gen_mov_tl(cr0, cpu_so);
> gen_set_access_type(ctx, ACCESS_RES);
> - EA = tcg_temp_new();
> gen_addr_reg_index(ctx, EA);
> -
> - tcg_gen_brcond_tl(TCG_COND_NE, EA, cpu_reserve, lab_fail);
> - tcg_gen_brcondi_tl(TCG_COND_NE, cpu_reserve_length, 16, lab_fail);
> + tcg_gen_brcond_tl(TCG_COND_NE, EA, cpu_reserve, lfail);
> + tcg_gen_brcondi_tl(TCG_COND_NE, cpu_reserve_length, 16, lfail);
>
> cmp = tcg_temp_new_i128();
> val = tcg_temp_new_i128();
> @@ -3932,15 +3934,10 @@ static void gen_stqcx_(DisasContext *ctx)
>
> tcg_gen_setcondi_tl(TCG_COND_EQ, t0, t0, 0);
> tcg_gen_shli_tl(t0, t0, CRF_EQ_BIT);
> - tcg_gen_or_tl(t0, t0, cpu_so);
> - tcg_gen_trunc_tl_i32(cpu_crf[0], t0);
> -
> - tcg_gen_br(lab_over);
> - gen_set_label(lab_fail);
> -
> - tcg_gen_trunc_tl_i32(cpu_crf[0], cpu_so);
> + tcg_gen_or_tl(cr0, cr0, t0);
>
> - gen_set_label(lab_over);
> + gen_set_label(lfail);
> + tcg_gen_trunc_tl_i32(cpu_crf[0], cr0);
> tcg_gen_movi_tl(cpu_reserve, -1);
> }
> #endif /* defined(TARGET_PPC64) */
next prev parent reply other threads:[~2023-06-05 13:43 UTC|newest]
Thread overview: 9+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-06-05 2:54 [PATCH v2 1/4] target/ppc: Fix lqarx to set cpu_reserve Nicholas Piggin
2023-06-05 2:54 ` [PATCH v2 2/4] target/ppc: Ensure stcx size matches larx Nicholas Piggin
2023-06-05 13:42 ` Daniel Henrique Barboza
2023-06-05 2:54 ` [PATCH v2 3/4] target/ppc: Remove larx/stcx. memory barrier semantics Nicholas Piggin
2023-06-05 13:42 ` Daniel Henrique Barboza
2023-06-05 2:54 ` [PATCH v2 4/4] target/ppc: Rework store conditional to avoid branch Nicholas Piggin
2023-06-05 13:42 ` Daniel Henrique Barboza [this message]
2023-06-05 3:09 ` [PATCH v2 1/4] target/ppc: Fix lqarx to set cpu_reserve Richard Henderson
2023-06-05 13:42 ` Daniel Henrique Barboza
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=6a1f0c3f-2063-be85-6d58-c672472d019a@gmail.com \
--to=danielhb413@gmail.com \
--cc=npiggin@gmail.com \
--cc=qemu-devel@nongnu.org \
--cc=qemu-ppc@nongnu.org \
--cc=qemu-stable@nongnu.org \
--cc=richard.henderson@linaro.org \
/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).