From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: by sourceware.org (Postfix, from userid 1791) id 69A6838493ED; Fri, 13 Jan 2023 20:03:30 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 69A6838493ED DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=sourceware.org; s=default; t=1673640210; bh=XbJ+GIM0ayTAH9pmYfkaI6m+2Pba+sVTdQZT6BI/dtQ=; h=From:To:Subject:Date:From; b=kUSerKh1A7UF6w9JQgcTASjvL01sdWbxHyrdmg2qY+u8wMakdPOyB6B25yRKW9zfq 8snjpCaDh5DrSo72MOeXI0BKc0uTgiEw9LgPMaKkuxT2BO8QEMEbDEJa7RwDg6EOxf NhKLQSOPjhHX55iRgGbxnjclfi25ISldThW90uvw= Content-Type: text/plain; charset="us-ascii" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit From: Adhemerval Zanella To: glibc-cvs@sourceware.org Subject: [glibc/azanella/generic-strings] hppa: Add memcopy.h X-Act-Checkin: glibc X-Git-Author: Richard Henderson X-Git-Refname: refs/heads/azanella/generic-strings X-Git-Oldrev: bd59b6081e03f98f2779b9fce07e3fa84350961f X-Git-Newrev: 363e8fb14cf06c35706bf03945ffcdcbdf8e84e6 Message-Id: <20230113200330.69A6838493ED@sourceware.org> Date: Fri, 13 Jan 2023 20:03:30 +0000 (GMT) List-Id: https://sourceware.org/git/gitweb.cgi?p=glibc.git;h=363e8fb14cf06c35706bf03945ffcdcbdf8e84e6 commit 363e8fb14cf06c35706bf03945ffcdcbdf8e84e6 Author: Richard Henderson Date: Tue Jan 10 18:01:01 2023 -0300 hppa: Add memcopy.h GCC's combine pass cannot merge (x >> c | y << (32 - c)) into a double-word shift unless (1) the subtract is in the same basic block and (2) the result of the subtract is used exactly once. Neither condition is true for any use of MERGE. By forcing the use of a double-word shift, we not only reduce contention on SAR, but also allow the setting of SAR to be hoisted outside of a loop. Checked on hppa-linux-gnu. Diff: --- sysdeps/hppa/memcopy.h | 42 ++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 42 insertions(+) diff --git a/sysdeps/hppa/memcopy.h b/sysdeps/hppa/memcopy.h new file mode 100644 index 0000000000..0d4b4ac435 --- /dev/null +++ b/sysdeps/hppa/memcopy.h @@ -0,0 +1,42 @@ +/* Definitions for memory copy functions, PA-RISC version. + Copyright (C) 2023 Free Software Foundation, Inc. + This file is part of the GNU C Library. + + The GNU C Library is free software; you can redistribute it and/or + modify it under the terms of the GNU Lesser General Public + License as published by the Free Software Foundation; either + version 2.1 of the License, or (at your option) any later version. + + The GNU C Library is distributed in the hope that it will be useful, + but WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + Lesser General Public License for more details. + + You should have received a copy of the GNU Lesser General Public + License along with the GNU C Library. If not, see + . */ + +#include + +/* Use a single double-word shift instead of two shifts and an ior. + If the uses of MERGE were close to the computation of shl/shr, + the compiler might have been able to create this itself. + But instead that computation is well separated. + + Using an inline function instead of a macro is the easiest way + to ensure that the types are correct. */ + +#undef MERGE + +static __always_inline op_t +MERGE (op_t w0, int shl, op_t w1, int shr) +{ + _Static_assert (OPSIZ == 4 || OPSIZ == 8, "Invalid OPSIZE"); + + op_t res; + if (OPSIZ == 4) + asm ("shrpw %1,%2,%%sar,%0" : "=r"(res) : "r"(w0), "r"(w1), "q"(shr)); + else if (OPSIZ == 8) + asm ("shrpd %1,%2,%%sar,%0" : "=r"(res) : "r"(w0), "r"(w1), "q"(shr)); + return res; +}