From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mx0b-001b2d01.pphosted.com (mx0b-001b2d01.pphosted.com [148.163.158.5]) by sourceware.org (Postfix) with ESMTPS id CAE023857BB2 for ; Mon, 15 May 2023 07:18:09 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.2 sourceware.org CAE023857BB2 Authentication-Results: sourceware.org; dmarc=none (p=none dis=none) header.from=linux.ibm.com Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=linux.ibm.com Received: from pps.filterd (m0353723.ppops.net [127.0.0.1]) by mx0a-001b2d01.pphosted.com (8.17.1.19/8.17.1.19) with ESMTP id 34F77pLx019560 for ; Mon, 15 May 2023 07:18:09 GMT DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=ibm.com; h=from : to : cc : subject : date : message-id : in-reply-to : references : mime-version : content-transfer-encoding; s=pp1; bh=xXO5ePXZLhcfcHK1sg2Rk8O7FDEKgib3hiUzFx8LwWg=; b=h+EIw8LRU99Yj07b3TxCAR1bCJhvthgZCnrCPo46fVDmqxpwFJ8QSHeH2WoQlK91KUT3 B+wuxkZNwG6rjvrnp4XB/pbjPLJ62nvQN7/e3MivIAzdyH76eM/cn02+dPkNmXLegRh9 Hb3z4anOABmF4feB+RF1C8/iMLdlhNBxINn4IZ0EnUmyiLLQ7NHAO2qMHLRYv1v+QfTe lJBFtv4f3xxpDf5gcE+dwFhFCMUv53JhwUFHv6if2yJccVCIltMKzk4UpELrO4VlVa3t TIjiN6waReeCz1lj9YbkXqciqSxpubTYe8IA3F2g7FRoZfR3w5WWXxmkZGarbogHmrga ZQ== Received: from ppma04ams.nl.ibm.com (63.31.33a9.ip4.static.sl-reverse.com [169.51.49.99]) by mx0a-001b2d01.pphosted.com (PPS) with ESMTPS id 3qkdy9kn8w-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT) for ; Mon, 15 May 2023 07:18:09 +0000 Received: from pps.filterd (ppma04ams.nl.ibm.com [127.0.0.1]) by ppma04ams.nl.ibm.com (8.17.1.19/8.17.1.19) with ESMTP id 34F5Ydi1007919 for ; Mon, 15 May 2023 07:18:06 GMT Received: from smtprelay06.fra02v.mail.ibm.com ([9.218.2.230]) by ppma04ams.nl.ibm.com (PPS) with ESMTPS id 3qj264rwhn-1 (version=TLSv1.2 cipher=ECDHE-RSA-AES256-GCM-SHA384 bits=256 verify=NOT) for ; Mon, 15 May 2023 07:18:06 +0000 Received: from smtpav05.fra02v.mail.ibm.com (smtpav05.fra02v.mail.ibm.com [10.20.54.104]) by smtprelay06.fra02v.mail.ibm.com (8.14.9/8.14.9/NCO v10.0) with ESMTP id 34F7HwTZ38142586 (version=TLSv1/SSLv3 cipher=DHE-RSA-AES256-GCM-SHA384 bits=256 verify=OK); Mon, 15 May 2023 07:17:58 GMT Received: from smtpav05.fra02v.mail.ibm.com (unknown [127.0.0.1]) by IMSVA (Postfix) with ESMTP id 7C28220049; Mon, 15 May 2023 07:17:58 +0000 (GMT) Received: from smtpav05.fra02v.mail.ibm.com (unknown [127.0.0.1]) by IMSVA (Postfix) with ESMTP id 5628A20040; Mon, 15 May 2023 07:17:58 +0000 (GMT) Received: from a8345010.lnxne.boe (unknown [9.152.108.100]) by smtpav05.fra02v.mail.ibm.com (Postfix) with ESMTPS; Mon, 15 May 2023 07:17:58 +0000 (GMT) From: Stefan Schulze Frielinghaus To: krebbel@linux.ibm.com, gcc-patches@gcc.gnu.org Cc: Stefan Schulze Frielinghaus Subject: [PATCH 1/3] s390: Refactor block operation cpymem Date: Mon, 15 May 2023 09:17:36 +0200 Message-Id: <20230515071738.563660-2-stefansf@linux.ibm.com> X-Mailer: git-send-email 2.39.2 In-Reply-To: <20230515071738.563660-1-stefansf@linux.ibm.com> References: <20230515071738.563660-1-stefansf@linux.ibm.com> MIME-Version: 1.0 Content-Transfer-Encoding: 8bit X-TM-AS-GCONF: 00 X-Proofpoint-ORIG-GUID: zVlsZ6kO71pSlnFsXA3tvwjgKnH5ODMg X-Proofpoint-GUID: zVlsZ6kO71pSlnFsXA3tvwjgKnH5ODMg X-Proofpoint-Virus-Version: vendor=baseguard engine=ICAP:2.0.254,Aquarius:18.0.942,Hydra:6.0.573,FMLib:17.11.170.22 definitions=2023-05-15_04,2023-05-05_01,2023-02-09_01 X-Proofpoint-Spam-Details: rule=outbound_notspam policy=outbound score=0 phishscore=0 adultscore=0 mlxscore=0 suspectscore=0 mlxlogscore=999 malwarescore=0 priorityscore=1501 impostorscore=0 bulkscore=0 lowpriorityscore=0 clxscore=1015 spamscore=0 classifier=spam adjust=0 reason=mlx scancount=1 engine=8.12.0-2304280000 definitions=main-2305150062 X-Spam-Status: No, score=-8.3 required=5.0 tests=BAYES_00,DKIM_SIGNED,DKIM_VALID,DKIM_VALID_EF,GIT_PATCH_0,RCVD_IN_MSPIKE_H2,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: Do not perform a libc function call into memcpy in case the size is not a compile-time constant but bounded and the upper bound is less than or equal to 256 bytes. gcc/ChangeLog: * config/s390/s390-protos.h (s390_expand_cpymem): Change function signature. * config/s390/s390.cc (s390_expand_cpymem): For memcpy's less than or equal to 256 byte do not perform a libc call. (s390_expand_insv): Adapt new function signature of s390_expand_cpymem. * config/s390/s390.md: Change expander into a version which takes 8 operands. --- gcc/config/s390/s390-protos.h | 2 +- gcc/config/s390/s390.cc | 84 +++++++++++++++++++++++++++-------- gcc/config/s390/s390.md | 10 +++-- 3 files changed, 74 insertions(+), 22 deletions(-) diff --git a/gcc/config/s390/s390-protos.h b/gcc/config/s390/s390-protos.h index 67fe09e732d..2c7495ca247 100644 --- a/gcc/config/s390/s390-protos.h +++ b/gcc/config/s390/s390-protos.h @@ -107,7 +107,7 @@ extern void s390_reload_symref_address (rtx , rtx , rtx , bool); extern void s390_expand_plus_operand (rtx, rtx, rtx); extern void emit_symbolic_move (rtx *); extern void s390_load_address (rtx, rtx); -extern bool s390_expand_cpymem (rtx, rtx, rtx); +extern bool s390_expand_cpymem (rtx, rtx, rtx, rtx, rtx); extern void s390_expand_setmem (rtx, rtx, rtx); extern bool s390_expand_cmpmem (rtx, rtx, rtx, rtx); extern void s390_expand_vec_strlen (rtx, rtx, rtx); diff --git a/gcc/config/s390/s390.cc b/gcc/config/s390/s390.cc index 505de995da8..95ea5e8d009 100644 --- a/gcc/config/s390/s390.cc +++ b/gcc/config/s390/s390.cc @@ -5650,27 +5650,27 @@ legitimize_reload_address (rtx ad, machine_mode mode ATTRIBUTE_UNUSED, return NULL_RTX; } -/* Emit code to move LEN bytes from DST to SRC. */ +/* Emit code to move LEN bytes from SRC to DST. */ bool -s390_expand_cpymem (rtx dst, rtx src, rtx len) +s390_expand_cpymem (rtx dst, rtx src, rtx len, rtx min_len_rtx, rtx max_len_rtx) { - /* When tuning for z10 or higher we rely on the Glibc functions to - do the right thing. Only for constant lengths below 64k we will - generate inline code. */ - if (s390_tune >= PROCESSOR_2097_Z10 - && (GET_CODE (len) != CONST_INT || INTVAL (len) > (1<<16))) - return false; + /* Exit early in case nothing has to be done. */ + if (CONST_INT_P (len) && UINTVAL (len) == 0) + return true; + + unsigned HOST_WIDE_INT min_len = UINTVAL (min_len_rtx); + unsigned HOST_WIDE_INT max_len + = max_len_rtx ? UINTVAL (max_len_rtx) : HOST_WIDE_INT_M1U; /* Expand memcpy for constant length operands without a loop if it is shorter that way. With a constant length argument a memcpy loop (without pfd) is 36 bytes -> 6 * mvc */ - if (GET_CODE (len) == CONST_INT - && INTVAL (len) >= 0 - && INTVAL (len) <= 256 * 6 - && (!TARGET_MVCLE || INTVAL (len) <= 256)) + if (CONST_INT_P (len) + && UINTVAL (len) <= 6 * 256 + && (!TARGET_MVCLE || UINTVAL (len) <= 256)) { HOST_WIDE_INT o, l; @@ -5681,14 +5681,57 @@ s390_expand_cpymem (rtx dst, rtx src, rtx len) emit_insn (gen_cpymem_short (newdst, newsrc, GEN_INT (l > 256 ? 255 : l - 1))); } + + return true; } - else if (TARGET_MVCLE) + else if (TARGET_MVCLE + && (s390_tune < PROCESSOR_2097_Z10 + || (CONST_INT_P (len) && UINTVAL (len) <= (1 << 16)))) { emit_insn (gen_cpymem_long (dst, src, convert_to_mode (Pmode, len, 1))); + return true; } - else + /* Non-constant length and no loop required. */ + else if (!CONST_INT_P (len) && max_len <= 256) + { + rtx_code_label *end_label; + + if (min_len == 0) + { + end_label = gen_label_rtx (); + emit_cmp_and_jump_insns (len, const0_rtx, EQ, NULL_RTX, + GET_MODE (len), 1, end_label, + profile_probability::very_unlikely ()); + } + + rtx lenm1 = expand_binop (GET_MODE (len), add_optab, len, constm1_rtx, + NULL_RTX, 1, OPTAB_DIRECT); + + /* Prefer a vectorized implementation over one which makes use of an + execute instruction since it is faster (although it increases register + pressure). */ + if (max_len <= 16 && TARGET_VX) + { + rtx tmp = gen_reg_rtx (V16QImode); + lenm1 = convert_to_mode (SImode, lenm1, 1); + emit_insn (gen_vllv16qi (tmp, lenm1, src)); + emit_insn (gen_vstlv16qi (tmp, lenm1, dst)); + } + else if (TARGET_Z15) + emit_insn (gen_mvcrl (dst, src, convert_to_mode (SImode, lenm1, 1))); + else + emit_insn ( + gen_cpymem_short (dst, src, convert_to_mode (Pmode, lenm1, 1))); + + if (min_len == 0) + emit_label (end_label); + + return true; + } + + else if (s390_tune < PROCESSOR_2097_Z10 || (CONST_INT_P (len) && UINTVAL (len) <= (1 << 16))) { rtx dst_addr, src_addr, count, blocks, temp; rtx_code_label *loop_start_label = gen_label_rtx (); @@ -5706,8 +5749,9 @@ s390_expand_cpymem (rtx dst, rtx src, rtx len) blocks = gen_reg_rtx (mode); convert_move (count, len, 1); - emit_cmp_and_jump_insns (count, const0_rtx, - EQ, NULL_RTX, mode, 1, end_label); + if (min_len == 0) + emit_cmp_and_jump_insns (count, const0_rtx, EQ, NULL_RTX, mode, 1, + end_label); emit_move_insn (dst_addr, force_operand (XEXP (dst, 0), NULL_RTX)); emit_move_insn (src_addr, force_operand (XEXP (src, 0), NULL_RTX)); @@ -5767,8 +5811,11 @@ s390_expand_cpymem (rtx dst, rtx src, rtx len) emit_insn (gen_cpymem_short (dst, src, convert_to_mode (Pmode, count, 1))); emit_label (end_label); + + return true; } - return true; + + return false; } /* Emit code to set LEN bytes at DST to VAL. @@ -6599,7 +6646,8 @@ s390_expand_insv (rtx dest, rtx op1, rtx op2, rtx src) dest = adjust_address (dest, BLKmode, 0); set_mem_size (dest, size); - s390_expand_cpymem (dest, src_mem, GEN_INT (size)); + rtx size_rtx = GEN_INT (size); + s390_expand_cpymem (dest, src_mem, size_rtx, size_rtx, size_rtx); return true; } diff --git a/gcc/config/s390/s390.md b/gcc/config/s390/s390.md index 00d39608e1d..d9ce287ab85 100644 --- a/gcc/config/s390/s390.md +++ b/gcc/config/s390/s390.md @@ -3341,11 +3341,15 @@ (define_expand "cpymem" [(set (match_operand:BLK 0 "memory_operand" "") ; destination (match_operand:BLK 1 "memory_operand" "")) ; source - (use (match_operand:GPR 2 "general_operand" "")) ; count - (match_operand 3 "" "")] + (use (match_operand:GPR 2 "general_operand" "")) ; size + (match_operand 3 "") ; align + (match_operand 4 "") ; expected align + (match_operand 5 "") ; expected size + (match_operand 6 "") ; minimal size + (match_operand 7 "")] ; maximal size "" { - if (s390_expand_cpymem (operands[0], operands[1], operands[2])) + if (s390_expand_cpymem (operands[0], operands[1], operands[2], operands[6], operands[7])) DONE; else FAIL; -- 2.39.2