From patchwork Mon Jul 3 10:19:22 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Robin Dapp X-Patchwork-Id: 115288 Return-Path: Delivered-To: ouuuleilei@gmail.com Received: by 2002:a59:9f45:0:b0:3ea:f831:8777 with SMTP id v5csp414793vqx; Mon, 3 Jul 2023 03:20:36 -0700 (PDT) X-Google-Smtp-Source: ACHHUZ7J+gnCIYplr57L+NdI5c0yVC6sXbWNCFu8MH6f9IQW6IVFEQVGwC/7TvgoiF2o1cFTc1W0 X-Received: by 2002:aa7:cfda:0:b0:51a:4039:b9e with SMTP id r26-20020aa7cfda000000b0051a40390b9emr13231559edy.9.1688379636606; Mon, 03 Jul 2023 03:20:36 -0700 (PDT) ARC-Seal: i=1; a=rsa-sha256; t=1688379636; cv=none; d=google.com; s=arc-20160816; b=Nwc2z4w7EFSeT0nkIHbJ90rkow4gIPo6hlwTwf1d9apkByN5Ujl5X85Kxb1FNf99jr zfT4RFXKcifk+4VN+R0f/vlrA4r71RbXuCnUDwF31plZjYhhLK6sV8BOTBDPOT8QiomV rvP2uQAr23hBpT/aldcCho48p5QATraIWUH8IyZOHYS7/xiNTEYw++ONyLd62FIIiw7f xbU5ACRUM+7v2XDR39fA6boaWxHRnn9xCctlAyMwYzb6u0+vceLTM9MhV+MjQEW0jzo1 rDsrXXGifXhngigEhlxRcNNfQOZYdc6DhNXgVhoi5qhCEz6QNCgmq9seBCoc1La0zoRg GcuA== ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=google.com; s=arc-20160816; h=sender:errors-to:reply-to:from:list-subscribe:list-help:list-post :list-archive:list-unsubscribe:list-id:precedence :content-transfer-encoding:subject:to:content-language:cc:user-agent :mime-version:date:message-id:dmarc-filter:delivered-to :dkim-signature:dkim-filter; bh=kwHh6h+8TE98FpwNrcRQEIcLvEskkanEbJHHf4z3FeE=; fh=M5oJlVESyfaqkJ7FK5PFt0G/TBL6ohCXHWBs81g6fZI=; b=qAvm5Jh6mqwsmpKzWmYtJtSot4ou85l1PhmnX+if4O6ccPstd3MkxhYk4TkCMyeS/q /Ft07jNxQ0voSXTCpdyzkYhl+awK2L0lYuYLFiaxGhoQFJTBI8k8lUgcKx2WuWr/faQj xWypSCPW7rFDmXXRs+4u08LtBNeioX06dES49Yuwu3HUJ4ckGL+Tza7Zdk6/GcQojhCg wXtm55luOE3wA70LvGgTaI1SD0kbLxY0RWQFHayNOQNqlYuea4xg7Rj4iKqlBbm9ZDCK jfAByKnySHoDh9iIwRi7F6xiVV9f0IrsYvJPUk8Jd/Csz+CPdH+6MQcBRh22KzcXcuhe qGmQ== ARC-Authentication-Results: i=1; mx.google.com; dkim=pass header.i=@gcc.gnu.org header.s=default header.b=vX4bh5Hk; spf=pass (google.com: domain of gcc-patches-bounces+ouuuleilei=gmail.com@gcc.gnu.org designates 8.43.85.97 as permitted sender) smtp.mailfrom="gcc-patches-bounces+ouuuleilei=gmail.com@gcc.gnu.org"; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=gnu.org Received: from server2.sourceware.org (ip-8-43-85-97.sourceware.org. [8.43.85.97]) by mx.google.com with ESMTPS id j26-20020a50ed1a000000b0051dddbd08cdsi5327885eds.356.2023.07.03.03.20.36 for (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Mon, 03 Jul 2023 03:20:36 -0700 (PDT) Received-SPF: pass (google.com: domain of gcc-patches-bounces+ouuuleilei=gmail.com@gcc.gnu.org designates 8.43.85.97 as permitted sender) client-ip=8.43.85.97; Authentication-Results: mx.google.com; dkim=pass header.i=@gcc.gnu.org header.s=default header.b=vX4bh5Hk; spf=pass (google.com: domain of gcc-patches-bounces+ouuuleilei=gmail.com@gcc.gnu.org designates 8.43.85.97 as permitted sender) smtp.mailfrom="gcc-patches-bounces+ouuuleilei=gmail.com@gcc.gnu.org"; dmarc=pass (p=NONE sp=NONE dis=NONE) header.from=gnu.org Received: from server2.sourceware.org (localhost [IPv6:::1]) by sourceware.org (Postfix) with ESMTP id E6C65385772B for ; Mon, 3 Jul 2023 10:20:13 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org E6C65385772B DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=gcc.gnu.org; s=default; t=1688379613; bh=kwHh6h+8TE98FpwNrcRQEIcLvEskkanEbJHHf4z3FeE=; h=Date:Cc:To:Subject:List-Id:List-Unsubscribe:List-Archive: List-Post:List-Help:List-Subscribe:From:Reply-To:From; b=vX4bh5Hkk7P3vxVUgLCSLA3OZWc31GQmEXs6iF5WX5IncKn/PawiWrREPldbczmh0 KZmqCdVo8Mk37LvkJLjvgIZdkItnK7pk2B5nlkjFyZQQG1jFL27wd6XwB4AIXsbqL2 P9vq+8uhqtKg67OYjglvUHUxk3FZcsS32Ljhg0wA= X-Original-To: gcc-patches@gcc.gnu.org Delivered-To: gcc-patches@gcc.gnu.org Received: from mail-ed1-x535.google.com (mail-ed1-x535.google.com [IPv6:2a00:1450:4864:20::535]) by sourceware.org (Postfix) with ESMTPS id AAA043858D1E for ; Mon, 3 Jul 2023 10:19:25 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org AAA043858D1E Received: by mail-ed1-x535.google.com with SMTP id 4fb4d7f45d1cf-51cff235226so7691149a12.0 for ; Mon, 03 Jul 2023 03:19:25 -0700 (PDT) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20221208; t=1688379564; x=1690971564; h=content-transfer-encoding:subject:from:to:content-language:cc :user-agent:mime-version:date:message-id:x-gm-message-state:from:to :cc:subject:date:message-id:reply-to; bh=kwHh6h+8TE98FpwNrcRQEIcLvEskkanEbJHHf4z3FeE=; b=GKoCAQb9YyqfWy8jpRgDJh+KP/Xrv32EZQagXu/6evFejBBgLqhyaTi+sUejtjJt3z khAwDU6/6tj0QJhNIR+SxfuTWs8tIVZR6ljClGRbSFtGCgaGOImMx4lzrqC3XYmrXJGr XsPrY4iK6cTKDOm5g10YNosZIyFggDdvi56Ycu5W1/Ltmg8VNny63zLwtMyuAUs4gNSn 3YTzN6qsIwUJ+g7SHy4WKVLZo4pvBG8saLw25sLpWdPUFO2zMbPVK0dU3berd555ZKG/ zsaF0abK4jDZvg6bTTZbhKrt6zErdHsT3X1bw62JvUhzHyupYZiAv9qttLpeXkDku1dj SmdA== X-Gm-Message-State: ABy/qLb4Hw+hwMtFxvRW0xi4lI3hTfl4MqUyAPaq/Qu4Obog9QNmX9ET 9mpJgZrVAYOgBQA/uK1ZIdGL+r9DgWg= X-Received: by 2002:a05:6402:2055:b0:51d:9ec7:15c3 with SMTP id bc21-20020a056402205500b0051d9ec715c3mr8554989edb.20.1688379563681; Mon, 03 Jul 2023 03:19:23 -0700 (PDT) Received: from [192.168.1.23] (ip-046-005-130-086.um12.pools.vodafone-ip.de. [46.5.130.86]) by smtp.gmail.com with ESMTPSA id j18-20020a50ed12000000b0051dada80305sm6718764eds.63.2023.07.03.03.19.22 (version=TLS1_3 cipher=TLS_AES_128_GCM_SHA256 bits=128/128); Mon, 03 Jul 2023 03:19:23 -0700 (PDT) Message-ID: <5104f927-d7db-c148-911c-0ca8e783a609@gmail.com> Date: Mon, 3 Jul 2023 12:19:22 +0200 MIME-Version: 1.0 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101 Thunderbird/102.12.0 Cc: rdapp.gcc@gmail.com, "richard.sandiford" , Richard Biener , Xionghu Luo Content-Language: en-US To: gcc-patches Subject: [PATCH] gimple-isel: Recognize vec_extract pattern. X-Spam-Status: No, score=-9.3 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, 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 X-BeenThere: gcc-patches@gcc.gnu.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Gcc-patches mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-Patchwork-Original-From: Robin Dapp via Gcc-patches From: Robin Dapp Reply-To: Robin Dapp Errors-To: gcc-patches-bounces+ouuuleilei=gmail.com@gcc.gnu.org Sender: "Gcc-patches" X-getmail-retrieved-from-mailbox: =?utf-8?q?INBOX?= X-GMAIL-THRID: =?utf-8?q?1770394365442442553?= X-GMAIL-MSGID: =?utf-8?q?1770394365442442553?= Hi, In gimple-isel we already deduce a vec_set pattern from an ARRAY_REF(VIEW_CONVERT_EXPR). This patch does the same for a vec_extract. The code is largely similar to the vec_set one including the addition of a can_vec_extract_var_idx_p function in optabs.cc to check if the backend can handle a register operand as index. We already have can_vec_extract in optabs-query but that one checks whether we can extract specific modes. With the introduction of an internal function for vec_extract the expander must not FAIL. For vec_set this has already been the case so adjust the documentation accordingly. Additionally, clarify the wording of the vector-vector case for vec_extract. During testing I noticed that the aarch64 simd vec_extract expander is the only one that FAILs. Richard is currently testing a patch that tries to remove this. Bootstrap and testsuite was unchanged on x86 and power. I was a bit torn whether to add a separate function to recognize vec_extract or not and ended up doing it inline with several is_extract checks. Regards Robin gcc/ChangeLog: * doc/md.texi: Document that vec_set and vec_extract must not fail. * gimple-isel.cc (gimple_expand_vec_set_expr): Rename this... (gimple_expand_vec_set_extract_expr): ...to this. (gimple_expand_vec_exprs): Call renamed function. * internal-fn.cc (vec_extract_direct): Add. (expand_vec_extract_optab_fn): New function to expand vec_extract optab. (direct_vec_extract_optab_supported_p): Add. * internal-fn.def (VEC_EXTRACT): Add. * optabs.cc (can_vec_extract_var_idx_p): New function. * optabs.h (can_vec_extract_var_idx_p): Declare. --- gcc/doc/md.texi | 7 +++- gcc/gimple-isel.cc | 85 +++++++++++++++++++++++++++++++++++++-------- gcc/internal-fn.cc | 39 +++++++++++++++++++++ gcc/internal-fn.def | 1 + gcc/optabs.cc | 24 +++++++++++++ gcc/optabs.h | 1 + 6 files changed, 141 insertions(+), 16 deletions(-) diff --git a/gcc/doc/md.texi b/gcc/doc/md.texi index 9648fdc846a..c61602fb04d 100644 --- a/gcc/doc/md.texi +++ b/gcc/doc/md.texi @@ -5074,6 +5074,8 @@ of the result should be stored to memory. Set given field in the vector value. Operand 0 is the vector to modify, operand 1 is new value of field and operand 2 specify the field index. +This pattern is not allowed to @code{FAIL}. + @cindex @code{vec_extract@var{m}@var{n}} instruction pattern @item @samp{vec_extract@var{m}@var{n}} Extract given field from the vector value. Operand 1 is the vector, operand 2 @@ -5081,7 +5083,10 @@ specify field index and operand 0 place to store value into. The @var{n} mode is the mode of the field or vector of fields that should be extracted, should be either element mode of the vector mode @var{m}, or a vector mode with the same element mode and smaller number of elements. -If @var{n} is a vector mode, the index is counted in units of that mode. +If @var{n} is a vector mode the index is counted in multiples of +mode @var{n}. + +This pattern is not allowed to @code{FAIL}. @cindex @code{vec_init@var{m}@var{n}} instruction pattern @item @samp{vec_init@var{m}@var{n}} diff --git a/gcc/gimple-isel.cc b/gcc/gimple-isel.cc index ef688ddb57f..d6c560b63dd 100644 --- a/gcc/gimple-isel.cc +++ b/gcc/gimple-isel.cc @@ -42,15 +42,26 @@ along with GCC; see the file COPYING3. If not see /* Expand all ARRAY_REF(VIEW_CONVERT_EXPR) gimple assignments into calls to internal function based on vector type of selected expansion. - i.e.: + + For vec_set: + VIEW_CONVERT_EXPR(u)[_1] = i_4(D); => _7 = u; _8 = .VEC_SET (_7, i_4(D), _1); - u = _8; */ + u = _8; + + For vec_extract: + + _3 = VIEW_CONVERT_EXPR(vD.2208)[idx_2(D)]; + => + _4 = vD.2208; + _5 = .VEC_EXTRACT (_4, idx_2(D)); + _3 = _5; */ static bool -gimple_expand_vec_set_expr (struct function *fun, gimple_stmt_iterator *gsi) +gimple_expand_vec_set_extract_expr (struct function *fun, + gimple_stmt_iterator *gsi) { enum tree_code code; gcall *new_stmt = NULL; @@ -62,45 +73,88 @@ gimple_expand_vec_set_expr (struct function *fun, gimple_stmt_iterator *gsi) if (!stmt) return false; + bool is_extract = false; + tree lhs = gimple_assign_lhs (stmt); + tree rhs = gimple_assign_rhs1 (stmt); + tree val, op0; code = TREE_CODE (lhs); - if (code != ARRAY_REF) - return false; + if (code == ARRAY_REF) + { + /* Assume it is a vec_set. */ + val = rhs; + op0 = TREE_OPERAND (lhs, 0); + } + else + { + /* It can still be a vec_extract. */ + code = TREE_CODE (rhs); + if (code != ARRAY_REF) + return false; + + /* Sides are swapped for vec_extract. */ + is_extract = true; + val = lhs; + op0 = TREE_OPERAND (rhs, 0); + } - tree val = gimple_assign_rhs1 (stmt); - tree op0 = TREE_OPERAND (lhs, 0); if (TREE_CODE (op0) == VIEW_CONVERT_EXPR && DECL_P (TREE_OPERAND (op0, 0)) && VECTOR_TYPE_P (TREE_TYPE (TREE_OPERAND (op0, 0))) && TYPE_MODE (TREE_TYPE (lhs)) == TYPE_MODE (TREE_TYPE (TREE_TYPE (TREE_OPERAND (op0, 0))))) { - tree pos = TREE_OPERAND (lhs, 1); + tree pos; + if (!is_extract) + pos = TREE_OPERAND (lhs, 1); + else + pos = TREE_OPERAND (rhs, 1); + tree view_op0 = TREE_OPERAND (op0, 0); machine_mode outermode = TYPE_MODE (TREE_TYPE (view_op0)); + machine_mode extract_mode; + if (is_extract) + extract_mode = TYPE_MODE (TREE_TYPE (lhs)); + if (auto_var_in_fn_p (view_op0, fun->decl) - && !TREE_ADDRESSABLE (view_op0) && can_vec_set_var_idx_p (outermode)) + && !TREE_ADDRESSABLE (view_op0) + && ((!is_extract && can_vec_set_var_idx_p (outermode)) + || (is_extract && + can_vec_extract_var_idx_p (outermode, extract_mode)))) { location_t loc = gimple_location (stmt); tree var_src = make_ssa_name (TREE_TYPE (view_op0)); - tree var_dst = make_ssa_name (TREE_TYPE (view_op0)); + tree var_dst; + if (!is_extract) + var_dst = make_ssa_name (TREE_TYPE (view_op0)); + else + var_dst = make_ssa_name (TREE_TYPE (lhs)); ass_stmt = gimple_build_assign (var_src, view_op0); gimple_set_vuse (ass_stmt, gimple_vuse (stmt)); gimple_set_location (ass_stmt, loc); gsi_insert_before (gsi, ass_stmt, GSI_SAME_STMT); - new_stmt - = gimple_build_call_internal (IFN_VEC_SET, 3, var_src, val, pos); + if (!is_extract) + new_stmt + = gimple_build_call_internal (IFN_VEC_SET, 3, var_src, val, pos); + else + new_stmt + = gimple_build_call_internal (IFN_VEC_EXTRACT, 2, var_src, pos); + gimple_call_set_lhs (new_stmt, var_dst); gimple_set_location (new_stmt, loc); gsi_insert_before (gsi, new_stmt, GSI_SAME_STMT); - ass_stmt = gimple_build_assign (view_op0, var_dst); + if (!is_extract) + ass_stmt = gimple_build_assign (view_op0, var_dst); + else + ass_stmt = gimple_build_assign (lhs, var_dst); gimple_set_location (ass_stmt, loc); + if (!is_extract) + gimple_move_vops (ass_stmt, stmt); gsi_insert_before (gsi, ass_stmt, GSI_SAME_STMT); basic_block bb = gimple_bb (stmt); - gimple_move_vops (ass_stmt, stmt); if (gsi_remove (gsi, true) && gimple_purge_dead_eh_edges (bb)) cfg_changed = true; @@ -317,7 +371,8 @@ gimple_expand_vec_exprs (struct function *fun) gsi_replace (&gsi, g, false); } - cfg_changed |= gimple_expand_vec_set_expr (fun, &gsi); + cfg_changed |= gimple_expand_vec_set_extract_expr (fun, &gsi); + if (gsi_end_p (gsi)) break; } diff --git a/gcc/internal-fn.cc b/gcc/internal-fn.cc index 9017176dc7a..cbdb7e64d7c 100644 --- a/gcc/internal-fn.cc +++ b/gcc/internal-fn.cc @@ -175,6 +175,7 @@ init_internal_fns () #define len_store_direct { 3, 3, false } #define len_maskstore_direct { 4, 3, false } #define vec_set_direct { 3, 3, false } +#define vec_extract_direct { 3, 3, false } #define unary_direct { 0, 0, true } #define unary_convert_direct { -1, 0, true } #define binary_direct { 0, 0, true } @@ -3126,6 +3127,43 @@ expand_vec_set_optab_fn (internal_fn, gcall *stmt, convert_optab optab) gcc_unreachable (); } +/* Expand VEC_EXTRACT optab internal function. */ + +static void +expand_vec_extract_optab_fn (internal_fn, gcall *stmt, convert_optab optab) +{ + tree lhs = gimple_call_lhs (stmt); + tree op0 = gimple_call_arg (stmt, 0); + tree op1 = gimple_call_arg (stmt, 1); + + rtx target = expand_expr (lhs, NULL_RTX, VOIDmode, EXPAND_WRITE); + + machine_mode outermode = TYPE_MODE (TREE_TYPE (op0)); + machine_mode extract_mode = TYPE_MODE (TREE_TYPE (lhs)); + + rtx src = expand_normal (op0); + rtx pos = expand_normal (op1); + + class expand_operand ops[3]; + enum insn_code icode = convert_optab_handler (optab, outermode, + extract_mode); + + if (icode != CODE_FOR_nothing) + { + create_output_operand (&ops[0], target, extract_mode); + create_input_operand (&ops[1], src, outermode); + create_convert_operand_from (&ops[2], pos, + TYPE_MODE (TREE_TYPE (op1)), true); + if (maybe_expand_insn (icode, 3, ops)) + { + if (!rtx_equal_p (target, ops[0].value)) + emit_move_insn (target, ops[0].value); + return; + } + } + gcc_unreachable (); +} + static void expand_ABNORMAL_DISPATCHER (internal_fn, gcall *) { @@ -3976,6 +4014,7 @@ multi_vector_optab_supported_p (convert_optab optab, tree_pair types, #define direct_mask_fold_left_optab_supported_p direct_optab_supported_p #define direct_check_ptrs_optab_supported_p direct_optab_supported_p #define direct_vec_set_optab_supported_p direct_optab_supported_p +#define direct_vec_extract_optab_supported_p direct_optab_supported_p /* Return the optab used by internal function FN. */ diff --git a/gcc/internal-fn.def b/gcc/internal-fn.def index bc947c0fde7..032624b1547 100644 --- a/gcc/internal-fn.def +++ b/gcc/internal-fn.def @@ -176,6 +176,7 @@ DEF_INTERNAL_OPTAB_FN (VCONDEQ, 0, vcondeq, vec_cond) DEF_INTERNAL_OPTAB_FN (VCOND_MASK, 0, vcond_mask, vec_cond_mask) DEF_INTERNAL_OPTAB_FN (VEC_SET, 0, vec_set, vec_set) +DEF_INTERNAL_OPTAB_FN (VEC_EXTRACT, 0, vec_extract, vec_extract) DEF_INTERNAL_OPTAB_FN (LEN_STORE, 0, len_store, len_store) DEF_INTERNAL_OPTAB_FN (LEN_MASK_STORE, 0, len_maskstore, len_maskstore) diff --git a/gcc/optabs.cc b/gcc/optabs.cc index 89ab55b72ae..4440bc505b1 100644 --- a/gcc/optabs.cc +++ b/gcc/optabs.cc @@ -4417,6 +4417,30 @@ can_vec_set_var_idx_p (machine_mode vec_mode) && insn_operand_matches (icode, 2, reg3); } +/* Return whether the backend can emit a vec_extract instruction with + a non-constant index. */ +bool +can_vec_extract_var_idx_p (machine_mode vec_mode, machine_mode extr_mode) +{ + if (!VECTOR_MODE_P (vec_mode)) + return false; + + rtx reg1 = alloca_raw_REG (extr_mode, LAST_VIRTUAL_REGISTER + 1); + rtx reg2 = alloca_raw_REG (vec_mode, LAST_VIRTUAL_REGISTER + 2); + + enum insn_code icode = convert_optab_handler (vec_extract_optab, + vec_mode, extr_mode); + + const struct insn_data_d *data = &insn_data[icode]; + machine_mode idx_mode = data->operand[2].mode; + + rtx reg3 = alloca_raw_REG (idx_mode, LAST_VIRTUAL_REGISTER + 3); + + return icode != CODE_FOR_nothing && insn_operand_matches (icode, 0, reg1) + && insn_operand_matches (icode, 1, reg2) + && insn_operand_matches (icode, 2, reg3); +} + /* This function is called when we are going to emit a compare instruction that compares the values found in X and Y, using the rtl operator COMPARISON. diff --git a/gcc/optabs.h b/gcc/optabs.h index 29ccbe9235e..f14d012d92f 100644 --- a/gcc/optabs.h +++ b/gcc/optabs.h @@ -259,6 +259,7 @@ extern bool can_vcond_compare_p (enum rtx_code, machine_mode, machine_mode); /* Return whether the backend can emit vector set instructions for inserting element into vector at variable index position. */ extern bool can_vec_set_var_idx_p (machine_mode); +extern bool can_vec_extract_var_idx_p (machine_mode, machine_mode); extern rtx prepare_operand (enum insn_code, rtx, int, machine_mode, machine_mode, int);