From patchwork Thu Dec 8 10:02:58 2011 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Dr. David Alan Gilbert" X-Patchwork-Id: 5537 Return-Path: X-Original-To: patchwork@peony.canonical.com Delivered-To: patchwork@peony.canonical.com Received: from fiordland.canonical.com (fiordland.canonical.com [91.189.94.145]) by peony.canonical.com (Postfix) with ESMTP id A9FDC23E1A for ; Thu, 8 Dec 2011 10:03:09 +0000 (UTC) Received: from mail-bw0-f52.google.com (mail-bw0-f52.google.com [209.85.214.52]) by fiordland.canonical.com (Postfix) with ESMTP id 87967A18111 for ; Thu, 8 Dec 2011 10:03:09 +0000 (UTC) Received: by bke17 with SMTP id 17so1941823bke.11 for ; Thu, 08 Dec 2011 02:03:09 -0800 (PST) Received: by 10.205.127.12 with SMTP id gy12mr1108255bkc.108.1323338587979; Thu, 08 Dec 2011 02:03:07 -0800 (PST) X-Forwarded-To: linaro-patchwork@canonical.com X-Forwarded-For: patch@linaro.org linaro-patchwork@canonical.com Delivered-To: patches@linaro.org Received: by 10.205.129.2 with SMTP id hg2cs90107bkc; Thu, 8 Dec 2011 02:03:07 -0800 (PST) Received: by 10.180.74.211 with SMTP id w19mr4065021wiv.7.1323338586075; Thu, 08 Dec 2011 02:03:06 -0800 (PST) Received: from mail-ww0-f50.google.com (mail-ww0-f50.google.com [74.125.82.50]) by mx.google.com with ESMTPS id fe5si3245804wbb.30.2011.12.08.02.03.05 (version=TLSv1/SSLv3 cipher=OTHER); Thu, 08 Dec 2011 02:03:06 -0800 (PST) Received-SPF: neutral (google.com: 74.125.82.50 is neither permitted nor denied by best guess record for domain of david.gilbert@linaro.org) client-ip=74.125.82.50; Authentication-Results: mx.google.com; spf=neutral (google.com: 74.125.82.50 is neither permitted nor denied by best guess record for domain of david.gilbert@linaro.org) smtp.mail=david.gilbert@linaro.org Received: by wgbdr11 with SMTP id dr11so3055910wgb.31 for ; Thu, 08 Dec 2011 02:03:05 -0800 (PST) Received: by 10.181.11.226 with SMTP id el2mr3795589wid.64.1323338585771; Thu, 08 Dec 2011 02:03:05 -0800 (PST) Received: from davesworkthinkpad (tu006.demon.co.uk. [83.105.84.125]) by mx.google.com with ESMTPS id x8sm7779182wix.17.2011.12.08.02.03.04 (version=TLSv1/SSLv3 cipher=OTHER); Thu, 08 Dec 2011 02:03:05 -0800 (PST) Date: Thu, 8 Dec 2011 10:02:58 +0000 From: "Dr. David Alan Gilbert" To: "Joseph S. Myers" Cc: libc-ports@sourceware.org, patches@linaro.org Subject: [ARM] architecture specific subdirectories & optimised memchr [V6] Message-ID: <20111208100257.GA6917@davesworkthinkpad> References: <20111202173217.GA27139@davesworkthinkpad> MIME-Version: 1.0 Content-Disposition: inline In-Reply-To: User-Agent: Mutt/1.5.20 (2009-06-14) This is v6 of my ARM memchr + version specific subdirectory patch. It's against svn rev r15869. Relative to v5 I've: Removed the cbz in the align loop (thanks Joseph for spotting that!) Relative to v4 I've: 1) Taken on board some of the comments from my submission to newlib. Including the use or ldrd rather than ldm. 2) Added a pattern so that ARM_ARCH_8 or 9 fall back to v7 so that it should work for 32bit ARMv8. Note that ARMv8 is deprecating some features in 32bit according to the docs on ARMs website, although they should still work in v8. Dave Index: ports/sysdeps/arm/preconfigure =================================================================== --- ports/sysdeps/arm/preconfigure (revision 15384) +++ ports/sysdeps/arm/preconfigure (working copy) @@ -3,6 +3,38 @@ base_machine=arm case $config_os in linux-gnueabi*) + # Lets ask the compiler which ARM family we've got + # Unfortunately it doesn't define any flags for implementations + # that you might pass to -mcpu or -mtune + # Note if you add patterns here you must ensure that + # an appropriate directory exists in sysdeps/arm/eabi + archcppflag=`echo "" | + $CC $CFLAGS $CPPFLAGS -E -dM - | + grep __ARM_ARCH | + sed -e 's/^#define //' -e 's/ .*//'` + + case x$archcppflag in + x__ARM_ARCH_[89]*__) + machine=armv7 + echo "Found compiler is configured for something newer than v7 - using v7" + ;; + + x__ARM_ARCH_7A__) + machine=armv7 + echo "Found compiler is configured for $machine" + ;; + + x__ARM_ARCH_6T2__) + machine=armv6t2 + echo "Found compiler is configured for $machine" + ;; + + *) + machine=arm + echo 2>&1 "arm/preconfigure: Did not find ARM architecture type; using default" + ;; + esac + machine=arm/eabi/$machine if [ "${CFLAGS+set}" != "set" ]; then CFLAGS="-g -O2" Index: ports/sysdeps/arm/eabi/armv6t2/memchr.S =================================================================== --- ports/sysdeps/arm/eabi/armv6t2/memchr.S (revision 0) +++ ports/sysdeps/arm/eabi/armv6t2/memchr.S (revision 0) @@ -0,0 +1,164 @@ +/* Copyright (C) 2011 Free Software Foundation, Inc. + This file is part of the GNU C Library. + Code contributed by Dave Gilbert + + 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, write to the Free + Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA + 02111-1307 USA. */ + +#include + +@ This memchr routine is optimised on a Cortex-A9 and should work on all ARMv7 +@ and ARMv6T2 processors. It has a fast path for short sizes, and has an +@ optimised path for large data sets; the worst case is finding the match early +@ in a large data set. +@ Note: The use of cbz/cbnz means it's Thumb only + +@ 2011-07-15 david.gilbert@linaro.org +@ Copy from Cortex strings release 21 and change license +@ http://bazaar.launchpad.net/~linaro-toolchain-dev/cortex-strings/trunk/view/head:/src/linaro-a9/memchr.S +@ Change function declarations/entry/exit +@ 2011-12-01 david.gilbert@linaro.org +@ Add some fixes from comments received (including use of ldrd instead ldm) +@ 2011-12-07 david.gilbert@linaro.org +@ Removed cbz from align loop - can't be taken + +@ this lets us check a flag in a 00/ff byte easily in either endianness +#ifdef __ARMEB__ +#define CHARTSTMASK(c) 1<<(31-(c*8)) +#else +#define CHARTSTMASK(c) 1<<(c*8) +#endif + .syntax unified + + .text + .thumb + +@ --------------------------------------------------------------------------- + .thumb_func + .global memchr + .type memchr,%function +ENTRY(memchr) + @ r0 = start of memory to scan + @ r1 = character to look for + @ r2 = length + @ returns r0 = pointer to character or NULL if not found + and r1,r1,#0xff @ Don't think we can trust the caller to actually pass a char + + cmp r2,#16 @ If it's short don't bother with anything clever + blt 20f + + tst r0, #7 @ If it's already aligned skip the next bit + beq 10f + + @ Work up to an aligned point +5: + ldrb r3, [r0],#1 + subs r2, r2, #1 + cmp r3, r1 + beq 50f @ If it matches exit found + tst r0, #7 + bne 5b @ If not aligned yet then do next byte + +10: + @ At this point, we are aligned, we know we have at least 8 bytes to work with + push {r4,r5,r6,r7} + cfi_adjust_cfa_offset (16) + cfi_rel_offset (r4, 0) + cfi_rel_offset (r5, 4) + cfi_rel_offset (r6, 8) + cfi_rel_offset (r7, 12) + + cfi_remember_state + + orr r1, r1, r1, lsl #8 @ expand the match word across to all bytes + orr r1, r1, r1, lsl #16 + bic r4, r2, #7 @ Number of double words to work with * 8 + mvns r7, #0 @ all F's + movs r3, #0 + +15: + ldrd r5,r6, [r0],#8 + subs r4, r4, #8 + eor r5,r5, r1 @ Get it so that r5,r6 have 00's where the bytes match the target + eor r6,r6, r1 + uadd8 r5, r5, r7 @ Parallel add 0xff - sets the GE bits for anything that wasn't 0 + sel r5, r3, r7 @ bytes are 00 for none-00 bytes, or ff for 00 bytes - NOTE INVERSION + uadd8 r6, r6, r7 @ Parallel add 0xff - sets the GE bits for anything that wasn't 0 + sel r6, r5, r7 @ chained....bytes are 00 for none-00 bytes, or ff for 00 bytes - NOTE INVERSION + cbnz r6, 60f + bne 15b @ (Flags from the subs above) If not run out of bytes then go around again + + pop {r4,r5,r6,r7} + cfi_adjust_cfa_offset (-16) + cfi_restore (r4) + cfi_restore (r5) + cfi_restore (r6) + cfi_restore (r7) + + and r1,r1,#0xff @ Get r1 back to a single character from the expansion above + and r2,r2,#7 @ Leave the count remaining as the number after the double words have been done + +20: + cbz r2, 40f @ 0 length or hit the end already then not found + +21: @ Post aligned section, or just a short call + ldrb r3,[r0],#1 + subs r2,r2,#1 + eor r3,r3,r1 @ r3 = 0 if match - doesn't break flags from sub + cbz r3, 50f + bne 21b @ on r2 flags + +40: + movs r0,#0 @ not found + DO_RET(lr) + +50: + subs r0,r0,#1 @ found + DO_RET(lr) + +60: @ We're here because the fast path found a hit - now we have to track down exactly which word it was + @ r0 points to the start of the double word after the one that was tested + @ r5 has the 00/ff pattern for the first word, r6 has the chained value + cfi_restore_state + cmp r5, #0 + itte eq + moveq r5, r6 @ the end is in the 2nd word + subeq r0,r0,#3 @ Points to 2nd byte of 2nd word + subne r0,r0,#7 @ or 2nd byte of 1st word + + @ r0 currently points to the 2nd byte of the word containing the hit + tst r5, # CHARTSTMASK(0) @ 1st character + bne 61f + adds r0,r0,#1 + tst r5, # CHARTSTMASK(1) @ 2nd character + ittt eq + addeq r0,r0,#1 + tsteq r5, # (3<<15) @ 2nd & 3rd character + @ If not the 3rd must be the last one + addeq r0,r0,#1 + +61: + pop {r4,r5,r6,r7} + cfi_adjust_cfa_offset (-16) + cfi_restore (r4) + cfi_restore (r5) + cfi_restore (r6) + cfi_restore (r7) + + subs r0,r0,#1 + DO_RET(lr) + +END(memchr) +libc_hidden_builtin_def (memchr) Index: ports/sysdeps/arm/eabi/armv7/Implies =================================================================== --- ports/sysdeps/arm/eabi/armv7/Implies (revision 0) +++ ports/sysdeps/arm/eabi/armv7/Implies (revision 0) @@ -0,0 +1,2 @@ +# We can do everything that 6T2 can +arm/eabi/armv6t2