From: Daniel Henrique Barboza <danielhb413@gmail.com>
To: Richard Henderson <richard.henderson@linaro.org>, qemu-devel@nongnu.org
Cc: "Cédric Le Goater" <clg@kaod.org>,
"David Gibson" <david@gibson.dropbear.id.au>,
"Greg Kurz" <groug@kaod.org>,
qemu-ppc@nongnu.org
Subject: Re: [PATCH v2 20/25] target/ppc: Rewrite trans_ADDG6S
Date: Tue, 7 Mar 2023 18:51:56 -0300 [thread overview]
Message-ID: <4951ac67-2bfd-b6db-4cb6-94d7ead96325@gmail.com> (raw)
In-Reply-To: <20230307183503.2512684-21-richard.henderson@linaro.org>
On 3/7/23 15:34, Richard Henderson wrote:
> Compute all carry bits in parallel instead of a loop.
>
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
Hmmmm the function was added by 6addef4d27268 9 months ago. All tcg ops you used
here were available back then.
I guess this existing implementation was an oversight on our end.
Reviewed-by: Daniel Henrique Barboza <danielhb413@gmail.com>
> Cc: Daniel Henrique Barboza <danielhb413@gmail.com>
> Cc: Cédric Le Goater <clg@kaod.org>
> Cc: David Gibson <david@gibson.dropbear.id.au>
> Cc: Greg Kurz <groug@kaod.org>
> Cc: qemu-ppc@nongnu.org
> ---
> target/ppc/translate/fixedpoint-impl.c.inc | 44 +++++++++++-----------
> 1 file changed, 23 insertions(+), 21 deletions(-)
>
> diff --git a/target/ppc/translate/fixedpoint-impl.c.inc b/target/ppc/translate/fixedpoint-impl.c.inc
> index 20ea484c3d..02d86b77a8 100644
> --- a/target/ppc/translate/fixedpoint-impl.c.inc
> +++ b/target/ppc/translate/fixedpoint-impl.c.inc
> @@ -484,33 +484,35 @@ static bool trans_PEXTD(DisasContext *ctx, arg_X *a)
>
> static bool trans_ADDG6S(DisasContext *ctx, arg_X *a)
> {
> - const uint64_t carry_bits = 0x1111111111111111ULL;
> - TCGv t0, t1, carry, zero = tcg_constant_tl(0);
> + const target_ulong carry_bits = (target_ulong)-1 / 0xf;
> + TCGv in1, in2, carryl, carryh, tmp;
> + TCGv zero = tcg_constant_tl(0);
>
> REQUIRE_INSNS_FLAGS2(ctx, BCDA_ISA206);
>
> - t0 = tcg_temp_new();
> - t1 = tcg_const_tl(0);
> - carry = tcg_const_tl(0);
> + in1 = cpu_gpr[a->ra];
> + in2 = cpu_gpr[a->rb];
> + tmp = tcg_temp_new();
> + carryl = tcg_temp_new();
> + carryh = tcg_temp_new();
>
> - for (int i = 0; i < 16; i++) {
> - tcg_gen_shri_tl(t0, cpu_gpr[a->ra], i * 4);
> - tcg_gen_andi_tl(t0, t0, 0xf);
> - tcg_gen_add_tl(t1, t1, t0);
> + /* Addition with carry. */
> + tcg_gen_add2_tl(carryl, carryh, in1, zero, in2, zero);
> + /* Addition without carry. */
> + tcg_gen_xor_tl(tmp, in1, in2);
> + /* Difference between the two is carry in to each bit. */
> + tcg_gen_xor_tl(carryl, carryl, tmp);
>
> - tcg_gen_shri_tl(t0, cpu_gpr[a->rb], i * 4);
> - tcg_gen_andi_tl(t0, t0, 0xf);
> - tcg_gen_add_tl(t1, t1, t0);
> + /*
> + * The carry-out that we're looking for is the carry-in to
> + * the next nibble. Shift the double-word down one nibble,
> + * which puts all of the bits back into one word.
> + */
> + tcg_gen_extract2_tl(carryl, carryl, carryh, 4);
>
> - tcg_gen_andi_tl(t1, t1, 0x10);
> - tcg_gen_setcond_tl(TCG_COND_NE, t1, t1, zero);
> -
> - tcg_gen_shli_tl(t0, t1, i * 4);
> - tcg_gen_or_tl(carry, carry, t0);
> - }
> -
> - tcg_gen_xori_tl(carry, carry, (target_long)carry_bits);
> - tcg_gen_muli_tl(cpu_gpr[a->rt], carry, 6);
> + /* Invert, isolate the carry bits, and produce 6's. */
> + tcg_gen_andc_tl(carryl, tcg_constant_tl(carry_bits), carryl);
> + tcg_gen_muli_tl(cpu_gpr[a->rt], carryl, 6);
> return true;
> }
>
next prev parent reply other threads:[~2023-03-07 21:52 UTC|newest]
Thread overview: 51+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-03-07 18:34 [PATCH v2 00/25] tcg: Remove tcg_const_* Richard Henderson
2023-03-07 18:34 ` [PATCH v2 01/25] target/arm: Use rmode >= 0 for need_rmode Richard Henderson
2023-03-07 18:34 ` [PATCH v2 02/25] target/arm: Handle FPROUNDING_ODD in arm_rmode_to_sf Richard Henderson
2023-03-08 17:25 ` Philippe Mathieu-Daudé
2023-03-07 18:34 ` [PATCH v2 03/25] target/arm: Improve arm_rmode_to_sf Richard Henderson
2023-03-07 18:34 ` [PATCH v2 04/25] target/arm: Consistently use ARMFPRounding during translation Richard Henderson
2023-03-07 18:34 ` [PATCH v2 05/25] target/arm: Create gen_set_rmode, gen_restore_rmode Richard Henderson
2023-03-09 9:48 ` Philippe Mathieu-Daudé
2023-03-07 18:34 ` [PATCH v2 06/25] target/arm: Improve trans_BFCI Richard Henderson
2023-03-07 23:05 ` Philippe Mathieu-Daudé
2023-03-07 18:34 ` [PATCH v2 07/25] target/arm: Avoid tcg_const_ptr in gen_sve_{ldr, str} Richard Henderson
2023-03-07 18:34 ` [PATCH v2 08/25] target/arm: Avoid tcg_const_* in translate-mve.c Richard Henderson
2023-03-09 13:27 ` Philippe Mathieu-Daudé
2023-03-07 18:34 ` [PATCH v2 09/25] target/arm: Avoid tcg_const_ptr in disas_simd_zip_trn Richard Henderson
2023-03-07 23:02 ` Philippe Mathieu-Daudé
2023-03-07 18:34 ` [PATCH v2 10/25] target/arm: Avoid tcg_const_ptr in handle_vec_simd_sqshrn Richard Henderson
2023-03-07 18:34 ` [PATCH v2 11/25] target/arm: Avoid tcg_const_ptr in handle_rev Richard Henderson
2023-03-07 18:34 ` [PATCH v2 12/25] target/m68k: Reject immediate as destination in gen_ea_mode Richard Henderson
2023-03-09 12:32 ` Laurent Vivier
2023-03-09 16:27 ` Richard Henderson
2023-03-07 18:34 ` [PATCH v2 13/25] target/m68k: Use tcg_constant_i32 " Richard Henderson
2023-03-07 18:34 ` [PATCH v2 14/25] target/ppc: Avoid tcg_const_i64 in do_vcntmb Richard Henderson
2023-03-07 21:42 ` Daniel Henrique Barboza
2023-03-09 10:18 ` Philippe Mathieu-Daudé
2023-03-07 18:34 ` [PATCH v2 15/25] target/ppc: Avoid tcg_const_* in vmx-impl.c.inc Richard Henderson
2023-03-07 21:42 ` Daniel Henrique Barboza
2023-03-09 9:49 ` Philippe Mathieu-Daudé
2023-03-07 18:34 ` [PATCH v2 16/25] target/ppc: Avoid tcg_const_* in xxeval Richard Henderson
2023-03-07 21:42 ` Daniel Henrique Barboza
2023-03-09 9:51 ` Philippe Mathieu-Daudé
2023-03-07 18:34 ` [PATCH v2 17/25] target/ppc: Avoid tcg_const_* in vsx-impl.c.inc Richard Henderson
2023-03-07 21:42 ` Daniel Henrique Barboza
2023-03-09 9:52 ` Philippe Mathieu-Daudé
2023-03-07 18:34 ` [PATCH v2 18/25] target/ppc: Avoid tcg_const_* in fp-impl.c.inc Richard Henderson
2023-03-07 21:43 ` Daniel Henrique Barboza
2023-03-09 9:54 ` Philippe Mathieu-Daudé
2023-03-07 18:34 ` [PATCH v2 19/25] target/ppc: Avoid tcg_const_* in power8-pmu-regs.c.inc Richard Henderson
2023-03-07 21:43 ` Daniel Henrique Barboza
2023-03-09 9:54 ` Philippe Mathieu-Daudé
2023-03-07 18:34 ` [PATCH v2 20/25] target/ppc: Rewrite trans_ADDG6S Richard Henderson
2023-03-07 21:51 ` Daniel Henrique Barboza [this message]
2023-03-07 22:34 ` Richard Henderson
2023-03-07 18:34 ` [PATCH v2 21/25] target/ppc: Fix gen_tlbsx_booke206 Richard Henderson
2023-03-07 21:44 ` Daniel Henrique Barboza
2023-03-07 18:35 ` [PATCH v2 22/25] target/ppc: Avoid tcg_const_* in translate.c Richard Henderson
2023-03-07 21:44 ` Daniel Henrique Barboza
2023-03-09 10:02 ` Philippe Mathieu-Daudé
2023-03-07 18:35 ` [PATCH v2 23/25] target/tricore: Use min/max for saturate Richard Henderson
2023-03-09 10:09 ` Philippe Mathieu-Daudé
2023-03-07 18:35 ` [PATCH v2 24/25] tcg: Drop tcg_const_*_vec Richard Henderson
2023-03-07 18:35 ` [PATCH v2 25/25] tcg: Drop tcg_const_* 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=4951ac67-2bfd-b6db-4cb6-94d7ead96325@gmail.com \
--to=danielhb413@gmail.com \
--cc=clg@kaod.org \
--cc=david@gibson.dropbear.id.au \
--cc=groug@kaod.org \
--cc=qemu-devel@nongnu.org \
--cc=qemu-ppc@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).