From: Kyrylo Tkachov <Kyrylo.Tkachov@arm.com>
To: Richard Sandiford <Richard.Sandiford@arm.com>,
Philipp Tomsich <philipp.tomsich@vrull.eu>
Cc: "gcc-patches@gcc.gnu.org" <gcc-patches@gcc.gnu.org>,
JiangNing Liu <jiangning.liu@amperecomputing.com>,
Christoph Muellner <christoph.muellner@vrull.eu>
Subject: RE: [PATCH] aarch64: Add support for Ampere-1A (-mcpu=ampere1a) CPU
Date: Mon, 14 Nov 2022 11:26:22 +0000 [thread overview]
Message-ID: <PAXPR08MB692647F65D05EE2A37EFE5BF93059@PAXPR08MB6926.eurprd08.prod.outlook.com> (raw)
In-Reply-To: <mptleodn63y.fsf@arm.com>
> -----Original Message-----
> From: Gcc-patches <gcc-patches-
> bounces+kyrylo.tkachov=arm.com@gcc.gnu.org> On Behalf Of Richard
> Sandiford via Gcc-patches
> Sent: Monday, November 14, 2022 11:11 AM
> To: Philipp Tomsich <philipp.tomsich@vrull.eu>
> Cc: gcc-patches@gcc.gnu.org; JiangNing Liu
> <jiangning.liu@amperecomputing.com>; Christoph Muellner
> <christoph.muellner@vrull.eu>
> Subject: Re: [PATCH] aarch64: Add support for Ampere-1A (-
> mcpu=ampere1a) CPU
>
> Philipp Tomsich <philipp.tomsich@vrull.eu> writes:
> > This patch adds support for Ampere-1A CPU:
> > - recognize the name of the core and provide detection for -mcpu=native,
> > - updated extra_costs,
> > - adds a new fusion pair for (A+B+1 and A-B-1).
> >
> > Ampere-1A and Ampere-1 have more timing difference than the extra
> > costs indicate, but these don't propagate through to the headline
> > items in our extra costs (e.g. the change in latency for scalar sqrt
> > doesn't have a corresponding table entry).
> >
> > gcc/ChangeLog:
> >
> > * config/aarch64/aarch64-cores.def (AARCH64_CORE): Add
> ampere1a.
> > * config/aarch64/aarch64-cost-tables.h: Add ampere1a_extra_costs.
> > * config/aarch64/aarch64-fusion-pairs.def (AARCH64_FUSION_PAIR):
> > Define a new fusion pair for A+B+1/A-B-1 (i.e., add/subtract two
> > registers and then +1/-1).
> > * config/aarch64/aarch64-tune.md: Regenerate.
> > * config/aarch64/aarch64.cc (aarch_macro_fusion_pair_p):
> Implement
> > idiom-matcher for the new fusion pair.
>
> OK except for a minor formatting nit:
>
> >
> > Signed-off-by: Philipp Tomsich <philipp.tomsich@vrull.eu>
> > ---
> >
> > gcc/config/aarch64/aarch64-cores.def | 1 +
> > gcc/config/aarch64/aarch64-cost-tables.h | 107 ++++++++++++++++++++
> > gcc/config/aarch64/aarch64-fusion-pairs.def | 1 +
> > gcc/config/aarch64/aarch64-tune.md | 2 +-
> > gcc/config/aarch64/aarch64.cc | 63 ++++++++++++
> > 5 files changed, 173 insertions(+), 1 deletion(-)
> >
> > diff --git a/gcc/config/aarch64/aarch64-cores.def
> b/gcc/config/aarch64/aarch64-cores.def
> > index d2671778928..aead587cec1 100644
> > --- a/gcc/config/aarch64/aarch64-cores.def
> > +++ b/gcc/config/aarch64/aarch64-cores.def
> > @@ -70,6 +70,7 @@ AARCH64_CORE("thunderxt83", thunderxt83,
> thunderx, V8A, (CRC, CRYPTO), thu
> >
> > /* Ampere Computing ('\xC0') cores. */
> > AARCH64_CORE("ampere1", ampere1, cortexa57, V8_6A, (F16, RNG, AES,
> SHA3), ampere1, 0xC0, 0xac3, -1)
> > +AARCH64_CORE("ampere1a", ampere1a, cortexa57, V8_6A, (F16, RNG,
> AES, SHA3, MEMTAG), ampere1a, 0xC0, 0xac4, -1)
> > /* Do not swap around "emag" and "xgene1",
> > this order is required to handle variant correctly. */
> > AARCH64_CORE("emag", emag, xgene1, V8A, (CRC, CRYPTO),
> emag, 0x50, 0x000, 3)
> > diff --git a/gcc/config/aarch64/aarch64-cost-tables.h
> b/gcc/config/aarch64/aarch64-cost-tables.h
> > index 760d7b30368..48522606fbe 100644
> > --- a/gcc/config/aarch64/aarch64-cost-tables.h
> > +++ b/gcc/config/aarch64/aarch64-cost-tables.h
> > @@ -775,4 +775,111 @@ const struct cpu_cost_table
> ampere1_extra_costs =
> > }
> > };
> >
> > +const struct cpu_cost_table ampere1a_extra_costs =
> > +{
> > + /* ALU */
> > + {
> > + 0, /* arith. */
> > + 0, /* logical. */
> > + 0, /* shift. */
> > + COSTS_N_INSNS (1), /* shift_reg. */
> > + 0, /* arith_shift. */
> > + COSTS_N_INSNS (1), /* arith_shift_reg. */
> > + 0, /* log_shift. */
> > + COSTS_N_INSNS (1), /* log_shift_reg. */
> > + 0, /* extend. */
> > + COSTS_N_INSNS (1), /* extend_arith. */
> > + 0, /* bfi. */
> > + 0, /* bfx. */
> > + 0, /* clz. */
> > + 0, /* rev. */
> > + 0, /* non_exec. */
> > + true /* non_exec_costs_exec. */
> > + },
> > + {
> > + /* MULT SImode */
> > + {
> > + COSTS_N_INSNS (3), /* simple. */
> > + COSTS_N_INSNS (3), /* flag_setting. */
> > + COSTS_N_INSNS (3), /* extend. */
> > + COSTS_N_INSNS (4), /* add. */
> > + COSTS_N_INSNS (4), /* extend_add. */
> > + COSTS_N_INSNS (19) /* idiv. */
> > + },
> > + /* MULT DImode */
> > + {
> > + COSTS_N_INSNS (3), /* simple. */
> > + 0, /* flag_setting (N/A). */
> > + COSTS_N_INSNS (3), /* extend. */
> > + COSTS_N_INSNS (4), /* add. */
> > + COSTS_N_INSNS (4), /* extend_add. */
> > + COSTS_N_INSNS (35) /* idiv. */
> > + }
> > + },
> > + /* LD/ST */
> > + {
> > + COSTS_N_INSNS (4), /* load. */
> > + COSTS_N_INSNS (4), /* load_sign_extend. */
> > + 0, /* ldrd (n/a). */
> > + 0, /* ldm_1st. */
> > + 0, /* ldm_regs_per_insn_1st. */
> > + 0, /* ldm_regs_per_insn_subsequent. */
> > + COSTS_N_INSNS (5), /* loadf. */
> > + COSTS_N_INSNS (5), /* loadd. */
> > + COSTS_N_INSNS (5), /* load_unaligned. */
> > + 0, /* store. */
> > + 0, /* strd. */
> > + 0, /* stm_1st. */
> > + 0, /* stm_regs_per_insn_1st. */
> > + 0, /* stm_regs_per_insn_subsequent. */
> > + COSTS_N_INSNS (2), /* storef. */
> > + COSTS_N_INSNS (2), /* stored. */
> > + COSTS_N_INSNS (2), /* store_unaligned. */
> > + COSTS_N_INSNS (3), /* loadv. */
> > + COSTS_N_INSNS (3) /* storev. */
> > + },
> > + {
> > + /* FP SFmode */
> > + {
> > + COSTS_N_INSNS (25), /* div. */
> > + COSTS_N_INSNS (4), /* mult. */
> > + COSTS_N_INSNS (4), /* mult_addsub. */
> > + COSTS_N_INSNS (4), /* fma. */
> > + COSTS_N_INSNS (4), /* addsub. */
> > + COSTS_N_INSNS (2), /* fpconst. */
> > + COSTS_N_INSNS (4), /* neg. */
> > + COSTS_N_INSNS (4), /* compare. */
> > + COSTS_N_INSNS (4), /* widen. */
> > + COSTS_N_INSNS (4), /* narrow. */
> > + COSTS_N_INSNS (4), /* toint. */
> > + COSTS_N_INSNS (4), /* fromint. */
> > + COSTS_N_INSNS (4) /* roundint. */
> > + },
> > + /* FP DFmode */
> > + {
> > + COSTS_N_INSNS (34), /* div. */
> > + COSTS_N_INSNS (5), /* mult. */
> > + COSTS_N_INSNS (5), /* mult_addsub. */
> > + COSTS_N_INSNS (5), /* fma. */
> > + COSTS_N_INSNS (5), /* addsub. */
> > + COSTS_N_INSNS (2), /* fpconst. */
> > + COSTS_N_INSNS (5), /* neg. */
> > + COSTS_N_INSNS (5), /* compare. */
> > + COSTS_N_INSNS (5), /* widen. */
> > + COSTS_N_INSNS (5), /* narrow. */
> > + COSTS_N_INSNS (6), /* toint. */
> > + COSTS_N_INSNS (6), /* fromint. */
> > + COSTS_N_INSNS (5) /* roundint. */
> > + }
> > + },
> > + /* Vector */
> > + {
> > + COSTS_N_INSNS (3), /* alu. */
> > + COSTS_N_INSNS (3), /* mult. */
> > + COSTS_N_INSNS (2), /* movi. */
> > + COSTS_N_INSNS (2), /* dup. */
> > + COSTS_N_INSNS (2) /* extract. */
> > + }
> > +};
> > +
> > #endif
> > diff --git a/gcc/config/aarch64/aarch64-fusion-pairs.def
> b/gcc/config/aarch64/aarch64-fusion-pairs.def
> > index c064fb9b85d..3990f77c127 100644
> > --- a/gcc/config/aarch64/aarch64-fusion-pairs.def
> > +++ b/gcc/config/aarch64/aarch64-fusion-pairs.def
> > @@ -36,5 +36,6 @@ AARCH64_FUSION_PAIR ("cmp+branch",
> CMP_BRANCH)
> > AARCH64_FUSION_PAIR ("aes+aesmc", AES_AESMC)
> > AARCH64_FUSION_PAIR ("alu+branch", ALU_BRANCH)
> > AARCH64_FUSION_PAIR ("alu+cbz", ALU_CBZ)
> > +AARCH64_FUSION_PAIR ("A+B+1/A-B-1", ADDSUB_2REG_CONST1)
These names can be used from the command line through the developer-only -moverride option.
I think maybe "addsub_2reg_const1" would be a better name here? (I wonder if '/' can cause trouble in some shells)
Thanks,
Kyrilll
> >
> > #undef AARCH64_FUSION_PAIR
> > diff --git a/gcc/config/aarch64/aarch64-tune.md
> b/gcc/config/aarch64/aarch64-tune.md
> > index 22ec1be5a4c..b7d6fc8cc88 100644
> > --- a/gcc/config/aarch64/aarch64-tune.md
> > +++ b/gcc/config/aarch64/aarch64-tune.md
> > @@ -1,5 +1,5 @@
> > ;; -*- buffer-read-only: t -*-
> > ;; Generated automatically by gentune.sh from aarch64-cores.def
> > (define_attr "tune"
> > -
> "cortexa34,cortexa35,cortexa53,cortexa57,cortexa72,cortexa73,thun
> derx,thunderxt88p1,thunderxt88,octeontx,octeontxt81,octeontxt83,thunder
> xt81,thunderxt83,ampere1,emag,xgene1,falkor,qdf24xx,exynosm1,phecda,t
> hunderx2t99p1,vulcan,thunderx2t99,cortexa55,cortexa75,cortexa76,cortexa
> 76ae,cortexa77,cortexa78,cortexa78ae,cortexa78c,cortexa65,cortexa65ae,co
> rtexx1,cortexx1c,ares,neoversen1,neoversee1,octeontx2,octeontx2t98,octeo
> ntx2t96,octeontx2t93,octeontx2f95,octeontx2f95n,octeontx2f95mm,a64fx,ts
> v110,thunderx3t110,zeus,neoversev1,neoverse512tvb,saphira,cortexa57cort
> exa53,cortexa72cortexa53,cortexa73cortexa35,cortexa73cortexa53,cortexa7
> 5cortexa55,cortexa76cortexa55,cortexr82,cortexa510,cortexa710,cortexa715
> ,cortexx2,neoversen2,demeter,neoversev2"
> > +
> "cortexa34,cortexa35,cortexa53,cortexa57,cortexa72,cortexa73,thun
> derx,thunderxt88p1,thunderxt88,octeontx,octeontxt81,octeontxt83,thunder
> xt81,thunderxt83,ampere1,ampere1a,emag,xgene1,falkor,qdf24xx,exynosm
> 1,phecda,thunderx2t99p1,vulcan,thunderx2t99,cortexa55,cortexa75,cortexa
> 76,cortexa76ae,cortexa77,cortexa78,cortexa78ae,cortexa78c,cortexa65,corte
> xa65ae,cortexx1,cortexx1c,ares,neoversen1,neoversee1,octeontx2,octeontx
> 2t98,octeontx2t96,octeontx2t93,octeontx2f95,octeontx2f95n,octeontx2f95m
> m,a64fx,tsv110,thunderx3t110,zeus,neoversev1,neoverse512tvb,saphira,cor
> texa57cortexa53,cortexa72cortexa53,cortexa73cortexa35,cortexa73cortexa5
> 3,cortexa75cortexa55,cortexa76cortexa55,cortexr82,cortexa510,cortexa710,c
> ortexa715,cortexx2,neoversen2,demeter,neoversev2"
> > (const (symbol_ref "((enum attr_tune) aarch64_tune)")))
> > diff --git a/gcc/config/aarch64/aarch64.cc b/gcc/config/aarch64/aarch64.cc
> > index d1f979ebcf8..1f7777b5214 100644
> > --- a/gcc/config/aarch64/aarch64.cc
> > +++ b/gcc/config/aarch64/aarch64.cc
> > @@ -1921,6 +1921,43 @@ static const struct tune_params
> ampere1_tunings =
> > &ere1_prefetch_tune
> > };
> >
> > +static const struct tune_params ampere1a_tunings =
> > +{
> > + &ere1a_extra_costs,
> > + &generic_addrcost_table,
> > + &generic_regmove_cost,
> > + &ere1_vector_cost,
> > + &generic_branch_cost,
> > + &generic_approx_modes,
> > + SVE_NOT_IMPLEMENTED, /* sve_width */
> > + { 4, /* load_int. */
> > + 4, /* store_int. */
> > + 4, /* load_fp. */
> > + 4, /* store_fp. */
> > + 4, /* load_pred. */
> > + 4 /* store_pred. */
> > + }, /* memmov_cost. */
> > + 4, /* issue_rate */
> > + (AARCH64_FUSE_ADRP_ADD | AARCH64_FUSE_AES_AESMC |
> > + AARCH64_FUSE_MOV_MOVK | AARCH64_FUSE_MOVK_MOVK |
> > + AARCH64_FUSE_ALU_BRANCH /* adds, ands, bics, ccmp, ccmn */ |
> > + AARCH64_FUSE_CMP_BRANCH | AARCH64_FUSE_ALU_CBZ |
> > + AARCH64_FUSE_ADDSUB_2REG_CONST1),
> > + /* fusible_ops */
> > + "32", /* function_align. */
> > + "4", /* jump_align. */
> > + "32:16", /* loop_align. */
> > + 2, /* int_reassoc_width. */
> > + 4, /* fp_reassoc_width. */
> > + 2, /* vec_reassoc_width. */
> > + 2, /* min_div_recip_mul_sf. */
> > + 2, /* min_div_recip_mul_df. */
> > + 0, /* max_case_values. */
> > + tune_params::AUTOPREFETCHER_WEAK, /* autoprefetcher_model. */
> > + (AARCH64_EXTRA_TUNE_NONE), /* tune_flags. */
> > + &ere1_prefetch_tune
> > +};
> > +
> > static const advsimd_vec_cost neoversev1_advsimd_vector_cost =
> > {
> > 2, /* int_stmt_cost */
> > @@ -25539,6 +25576,32 @@ aarch_macro_fusion_pair_p (rtx_insn *prev,
> rtx_insn *curr)
> > }
> > }
> >
> > + /* Fuse A+B+1 and A-B-1 */
> > + if (simple_sets_p && aarch64_fusion_enabled_p
> (AARCH64_FUSE_ADDSUB_2REG_CONST1))
>
> Line exceeds 80 chars.
>
> Thanks,
> Richard
>
> > + {
> > + /* We're trying to match:
> > + prev == (set (r0) (plus (r0) (r1)))
> > + curr == (set (r0) (plus (r0) (const_int 1)))
> > + or:
> > + prev == (set (r0) (minus (r0) (r1)))
> > + curr == (set (r0) (plus (r0) (const_int -1))) */
> > +
> > + rtx prev_src = SET_SRC (prev_set);
> > + rtx curr_src = SET_SRC (curr_set);
> > +
> > + int polarity = 1;
> > + if (GET_CODE (prev_src) == MINUS)
> > + polarity = -1;
> > +
> > + if (GET_CODE (curr_src) == PLUS
> > + && (GET_CODE (prev_src) == PLUS || GET_CODE (prev_src) ==
> MINUS)
> > + && CONST_INT_P (XEXP (curr_src, 1))
> > + && INTVAL (XEXP (curr_src, 1)) == polarity
> > + && REG_P (XEXP (curr_src, 0))
> > + && REGNO (SET_DEST (prev_set)) == REGNO (XEXP (curr_src, 0)))
> > + return true;
> > + }
> > +
> > return false;
> > }
prev parent reply other threads:[~2022-11-14 11:26 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-11-13 16:02 Philipp Tomsich
2022-11-14 11:11 ` Richard Sandiford
2022-11-14 11:26 ` Kyrylo Tkachov [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=PAXPR08MB692647F65D05EE2A37EFE5BF93059@PAXPR08MB6926.eurprd08.prod.outlook.com \
--to=kyrylo.tkachov@arm.com \
--cc=Richard.Sandiford@arm.com \
--cc=christoph.muellner@vrull.eu \
--cc=gcc-patches@gcc.gnu.org \
--cc=jiangning.liu@amperecomputing.com \
--cc=philipp.tomsich@vrull.eu \
/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).