From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-lf1-x129.google.com (mail-lf1-x129.google.com [IPv6:2a00:1450:4864:20::129]) by sourceware.org (Postfix) with ESMTPS id 9A7FA3858D20 for ; Mon, 22 May 2023 10:03:00 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org 9A7FA3858D20 Authentication-Results: sourceware.org; dmarc=pass (p=none dis=none) header.from=gmail.com Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=gmail.com Received: by mail-lf1-x129.google.com with SMTP id 2adb3069b0e04-4f004cc54f4so6854522e87.3 for ; Mon, 22 May 2023 03:03:00 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gmail.com; s=20221208; t=1684749779; x=1687341779; h=content-transfer-encoding:cc:to:subject:message-id:date:from :in-reply-to:references:mime-version:from:to:cc:subject:date :message-id:reply-to; bh=yMpuQk5zjSrrvZ9Ra/W/FnwEJcduaVat41mRYo+XdSA=; b=l2QRZwOriAUUcSLwkAn5yNysxQThq8lMGhRRsXa3Z4J4gWYZNkv9TcUEki6NdM16nW ND92rU3NJsjGfzVupOiRpptXgCUJUzjhe4iZQvdHkTwqc9UFBSipkEj1wwrkQoDmjefW P0jyDHtlaPXEjB/aSxglMs1ydE/+DVL1SQapaP+/1rLki78ze75L2mxxK2Gu9FsVTk+V mhqC4FDtO1enQdDwqI0xvp5IOQeQhYqAxEzyEgAZTtfkHFmC5jS/iYdSyszj+SPVMK2e +H2Spu7dKLtYMtibNtEXsRcV+khciHh2Nz20TXPGTE6cqId9aa3uYVhGKWkfRlU+iPlN AcnQ== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20221208; t=1684749779; x=1687341779; h=content-transfer-encoding:cc:to:subject:message-id:date:from :in-reply-to:references:mime-version:x-gm-message-state:from:to:cc :subject:date:message-id:reply-to; bh=yMpuQk5zjSrrvZ9Ra/W/FnwEJcduaVat41mRYo+XdSA=; b=h1uEPUiNB8NhD5hPnu9g5fOgmtn+Ln97qERTVB/C9+lSHc8YCTtV0ioCTUTsQHx8uV v98UmG3oZsWss4jPidsJAxiOkv/6B2Ls6Nqx8xR964z68s3oQhrNUHcoddc+V4wcisgH Peyj9Kzpi0CeOt3rrtKYqcEedqPGIzQ09J8Kvz0AA+O5zM726GIbfn1XqbkPNX/FHnNn gZQIq0FS97jh1sUrxz7WeaY+TGgCH1gIG9N89OEwDZHEs02lCvDmHeDWHK70hJyFBNQk h99z43xancCWhy+YMZqUYnmUYxOewHTwDNQu5sUcjDLJhDN4al0bmb5Supg3PavVf5Jp StFg== X-Gm-Message-State: AC+VfDxcZAwkpLwN36RPmEhnmPRos+hsq+9EztmUnlRnaiIvZWGjHFsz nfR0JYpp565lOt0nKfI9+ZbzNBdVk9iFDzihQwo= X-Google-Smtp-Source: ACHHUZ7JPjiToTkbyvA6k87+no5pw7Jnnf9LvXT9oVSXdu50DjYu8Akp4hwmKE/ZUFrrsfjTfoxOES1Xs7UlJA5soAE= X-Received: by 2002:a19:f509:0:b0:4f3:b0e1:feb7 with SMTP id j9-20020a19f509000000b004f3b0e1feb7mr3202566lfb.37.1684749778690; Mon, 22 May 2023 03:02:58 -0700 (PDT) MIME-Version: 1.0 References: <20230518210331.11564-1-amonakov@ispras.ru> In-Reply-To: <20230518210331.11564-1-amonakov@ispras.ru> From: Richard Biener Date: Mon, 22 May 2023 12:02:46 +0200 Message-ID: Subject: Re: [PATCH] c-family: implement -ffp-contract=on To: Alexander Monakov , Marek Polacek , Jason Merrill Cc: gcc-patches@gcc.gnu.org Content-Type: text/plain; charset="UTF-8" Content-Transfer-Encoding: quoted-printable X-Spam-Status: No, score=-3.9 required=5.0 tests=BAYES_00,DKIM_SIGNED,DKIM_VALID,DKIM_VALID_AU,DKIM_VALID_EF,FREEMAIL_FROM,GIT_PATCH_0,RCVD_IN_DNSWL_NONE,SPAM_BODY,SPF_HELO_NONE,SPF_PASS,TXREP,T_SCC_BODY_TEXT_LINE autolearn=ham autolearn_force=no version=3.4.6 X-Spam-Checker-Version: SpamAssassin 3.4.6 (2021-04-09) on server2.sourceware.org List-Id: On Thu, May 18, 2023 at 11:04=E2=80=AFPM Alexander Monakov via Gcc-patches wrote: > > Implement -ffp-contract=3Don for C and C++ without changing default > behavior (=3Doff for -std=3DcNN, =3Dfast for C++ and -std=3DgnuNN). The documentation changes mention the defaults are changed for standard modes, I suppose you want to remove that hunk. > gcc/c-family/ChangeLog: > > * c-gimplify.cc (fma_supported_p): New helper. > (c_gimplify_expr) [PLUS_EXPR, MINUS_EXPR]: Implement FMA > contraction. > > gcc/ChangeLog: > > * common.opt (fp_contract_mode) [on]: Remove fallback. > * config/sh/sh.md (*fmasf4): Correct flag_fp_contract_mode test. > * doc/invoke.texi (-ffp-contract): Update. > * trans-mem.cc (diagnose_tm_1): Skip internal function calls. > --- > gcc/c-family/c-gimplify.cc | 78 ++++++++++++++++++++++++++++++++++++++ > gcc/common.opt | 3 +- > gcc/config/sh/sh.md | 2 +- > gcc/doc/invoke.texi | 8 ++-- > gcc/trans-mem.cc | 3 ++ > 5 files changed, 88 insertions(+), 6 deletions(-) > > diff --git a/gcc/c-family/c-gimplify.cc b/gcc/c-family/c-gimplify.cc > index ef5c7d919f..f7635d3b0c 100644 > --- a/gcc/c-family/c-gimplify.cc > +++ b/gcc/c-family/c-gimplify.cc > @@ -41,6 +41,8 @@ along with GCC; see the file COPYING3. If not see > #include "c-ubsan.h" > #include "tree-nested.h" > #include "context.h" > +#include "tree-pass.h" > +#include "internal-fn.h" > > /* The gimplification pass converts the language-dependent trees > (ld-trees) emitted by the parser into language-independent trees > @@ -686,6 +688,14 @@ c_build_bind_expr (location_t loc, tree block, tree = body) > return bind; > } > > +/* Helper for c_gimplify_expr: test if target supports fma-like FN. */ > + > +static bool > +fma_supported_p (enum internal_fn fn, tree type) > +{ > + return direct_internal_fn_supported_p (fn, type, OPTIMIZE_FOR_BOTH); > +} > + > /* Gimplification of expression trees. */ > > /* Do C-specific gimplification on *EXPR_P. PRE_P and POST_P are as in > @@ -739,6 +749,74 @@ c_gimplify_expr (tree *expr_p, gimple_seq *pre_p ATT= RIBUTE_UNUSED, > break; > } > > + case PLUS_EXPR: > + case MINUS_EXPR: > + { > + tree type =3D TREE_TYPE (*expr_p); > + /* For -ffp-contract=3Don we need to attempt FMA contraction only > + during initial gimplification. Late contraction across statem= ent > + boundaries would violate language semantics. */ > + if (SCALAR_FLOAT_TYPE_P (type) > + && flag_fp_contract_mode =3D=3D FP_CONTRACT_ON > + && cfun && !(cfun->curr_properties & PROP_gimple_any) > + && fma_supported_p (IFN_FMA, type)) > + { > + bool neg_mul =3D false, neg_add =3D code =3D=3D MINUS_EXPR; > + > + tree *op0_p =3D &TREE_OPERAND (*expr_p, 0); > + tree *op1_p =3D &TREE_OPERAND (*expr_p, 1); > + > + /* Look for =C2=B1(x * y) =C2=B1 z, swapping operands if nece= ssary. */ > + if (TREE_CODE (*op0_p) =3D=3D NEGATE_EXPR > + && TREE_CODE (TREE_OPERAND (*op0_p, 0)) =3D=3D MULT_EXPR) > + /* '*EXPR_P' is '-(x * y) =C2=B1 z'. This is fine. */; > + else if (TREE_CODE (*op0_p) !=3D MULT_EXPR) > + { > + std::swap (op0_p, op1_p); > + std::swap (neg_mul, neg_add); > + } > + if (TREE_CODE (*op0_p) =3D=3D NEGATE_EXPR) > + { > + op0_p =3D &TREE_OPERAND (*op0_p, 0); > + neg_mul =3D !neg_mul; > + } > + if (TREE_CODE (*op0_p) !=3D MULT_EXPR) > + break; > + auto_vec ops (3); > + ops.quick_push (TREE_OPERAND (*op0_p, 0)); > + ops.quick_push (TREE_OPERAND (*op0_p, 1)); > + ops.quick_push (*op1_p); > + > + enum internal_fn ifn =3D IFN_FMA; > + if (neg_mul) > + { > + if (fma_supported_p (IFN_FNMA, type)) > + ifn =3D IFN_FNMA; > + else > + ops[0] =3D build1 (NEGATE_EXPR, type, ops[0]); > + } > + if (neg_add) > + { > + enum internal_fn ifn2 =3D ifn =3D=3D IFN_FMA ? IFN_FMS : = IFN_FNMS; > + if (fma_supported_p (ifn2, type)) > + ifn =3D ifn2; > + else > + ops[2] =3D build1 (NEGATE_EXPR, type, ops[2]); > + } > + for (auto &&op : ops) > + if (gimplify_expr (&op, pre_p, post_p, is_gimple_val, fb_rv= alue) > + =3D=3D GS_ERROR) > + return GS_ERROR; > + > + gcall *call =3D gimple_build_call_internal_vec (ifn, ops); > + gimple_seq_add_stmt_without_update (pre_p, call); > + *expr_p =3D create_tmp_var (type); > + gimple_call_set_lhs (call, *expr_p); it would be possible to do *expr_p =3D build_call_expr_internal (ifn, type, ops[0], ops[1]. ops[2]); return GS_OK; and not worry about temporary creation and gimplifying of the operands. That would in theory also leave the possibility to do this during genericization instead (and avoid the guard against late invocation of the hook). Otherwise it looks OK, but I'll let frontend maintainers have a chance to l= ook as well. Thanks for tackling this long-standing issue. Richard. > + return GS_ALL_DONE; > + } > + break; > + } > + > default:; > } > > diff --git a/gcc/common.opt b/gcc/common.opt > index a28ca13385..3daec85aef 100644 > --- a/gcc/common.opt > +++ b/gcc/common.opt > @@ -1662,9 +1662,8 @@ Name(fp_contract_mode) Type(enum fp_contract_mode) = UnknownError(unknown floating > EnumValue > Enum(fp_contract_mode) String(off) Value(FP_CONTRACT_OFF) > > -; Not implemented, fall back to conservative FP_CONTRACT_OFF. > EnumValue > -Enum(fp_contract_mode) String(on) Value(FP_CONTRACT_OFF) > +Enum(fp_contract_mode) String(on) Value(FP_CONTRACT_ON) > > EnumValue > Enum(fp_contract_mode) String(fast) Value(FP_CONTRACT_FAST) > diff --git a/gcc/config/sh/sh.md b/gcc/config/sh/sh.md > index 4622dba012..5cb1795482 100644 > --- a/gcc/config/sh/sh.md > +++ b/gcc/config/sh/sh.md > @@ -9269,7 +9269,7 @@ (define_insn_and_split "*fmasf4" > (match_operand:SF 3 "arith_reg_operand" "0"))) > (clobber (reg:SI FPSCR_STAT_REG)) > (use (reg:SI FPSCR_MODES_REG))] > - "TARGET_SH2E && flag_fp_contract_mode !=3D FP_CONTRACT_OFF" > + "TARGET_SH2E && flag_fp_contract_mode =3D=3D FP_CONTRACT_FAST" > "fmac %1,%2,%0" > "&& can_create_pseudo_p ()" > [(parallel [(set (match_dup 0) > diff --git a/gcc/doc/invoke.texi b/gcc/doc/invoke.texi > index b92b857602..cb1e9a1d9f 100644 > --- a/gcc/doc/invoke.texi > +++ b/gcc/doc/invoke.texi > @@ -11983,10 +11983,12 @@ This option is enabled by default at optimizati= on levels @option{-O1}, > such as forming of fused multiply-add operations if the target has > native support for them. > @option{-ffp-contract=3Don} enables floating-point expression contractio= n > -if allowed by the language standard. This is currently not implemented > -and treated equal to @option{-ffp-contract=3Doff}. > +if allowed by the language standard. This is implemented for C and C++, > +where it enables contraction within one expression, but not across > +different statements. > > -The default is @option{-ffp-contract=3Dfast}. > +The default is @option{-ffp-contract=3Doff} for C in a standards complia= nt mode > +(@option{-std=3Dc11} or similar), @option{-ffp-contract=3Dfast} otherwis= e. > > @opindex fomit-frame-pointer > @item -fomit-frame-pointer > diff --git a/gcc/trans-mem.cc b/gcc/trans-mem.cc > index 4b129663e0..2174faef4c 100644 > --- a/gcc/trans-mem.cc > +++ b/gcc/trans-mem.cc > @@ -637,6 +637,9 @@ diagnose_tm_1 (gimple_stmt_iterator *gsi, bool *handl= ed_ops_p, > { > case GIMPLE_CALL: > { > + if (gimple_call_internal_p (stmt)) > + break; > + > tree fn =3D gimple_call_fn (stmt); > > if ((d->summary_flags & DIAG_TM_OUTER) =3D=3D 0 > -- > 2.39.2 >