Loading...
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 | /* * arch/xtensa/lib/memset.S * * ANSI C standard library function memset * (Well, almost. .fixup code might return zero.) * * This file is subject to the terms and conditions of the GNU General * Public License. See the file "COPYING" in the main directory of * this archive for more details. * * Copyright (C) 2002 Tensilica Inc. */ #include <linux/linkage.h> #include <variant/core.h> #include <asm/asmmacro.h> /* * void *memset(void *dst, int c, size_t length) * * The algorithm is as follows: * Create a word with c in all byte positions * If the destination is aligned, * do 16B chucks with a loop, and then finish up with * 8B, 4B, 2B, and 1B stores conditional on the length. * If destination is unaligned, align it by conditionally * setting 1B and 2B and then go to aligned case. * This code tries to use fall-through branches for the common * case of an aligned destination (except for the branches to * the alignment labels). */ .text ENTRY(__memset) WEAK(memset) entry sp, 16 # minimal stack frame # a2/ dst, a3/ c, a4/ length extui a3, a3, 0, 8 # mask to just 8 bits slli a7, a3, 8 # duplicate character in all bytes of word or a3, a3, a7 # ... slli a7, a3, 16 # ... or a3, a3, a7 # ... mov a5, a2 # copy dst so that a2 is return value movi a6, 3 # for alignment tests bany a2, a6, .Ldstunaligned # if dst is unaligned .L0: # return here from .Ldstunaligned when dst is aligned srli a7, a4, 4 # number of loop iterations with 16B # per iteration bnez a4, .Laligned retw /* * Destination is word-aligned. */ # set 16 bytes per iteration for word-aligned dst .align 4 # 1 mod 4 alignment for LOOPNEZ .byte 0 # (0 mod 4 alignment for LBEG) .Laligned: #if XCHAL_HAVE_LOOPS loopnez a7, .Loop1done #else /* !XCHAL_HAVE_LOOPS */ beqz a7, .Loop1done slli a6, a7, 4 add a6, a6, a5 # a6 = end of last 16B chunk #endif /* !XCHAL_HAVE_LOOPS */ .Loop1: EX(10f) s32i a3, a5, 0 EX(10f) s32i a3, a5, 4 EX(10f) s32i a3, a5, 8 EX(10f) s32i a3, a5, 12 addi a5, a5, 16 #if !XCHAL_HAVE_LOOPS blt a5, a6, .Loop1 #endif /* !XCHAL_HAVE_LOOPS */ .Loop1done: bbci.l a4, 3, .L2 # set 8 bytes EX(10f) s32i a3, a5, 0 EX(10f) s32i a3, a5, 4 addi a5, a5, 8 .L2: bbci.l a4, 2, .L3 # set 4 bytes EX(10f) s32i a3, a5, 0 addi a5, a5, 4 .L3: bbci.l a4, 1, .L4 # set 2 bytes EX(10f) s16i a3, a5, 0 addi a5, a5, 2 .L4: bbci.l a4, 0, .L5 # set 1 byte EX(10f) s8i a3, a5, 0 .L5: .Lret1: retw /* * Destination is unaligned */ .Ldstunaligned: bltui a4, 8, .Lbyteset # do short copies byte by byte bbci.l a5, 0, .L20 # branch if dst alignment half-aligned # dst is only byte aligned # set 1 byte EX(10f) s8i a3, a5, 0 addi a5, a5, 1 addi a4, a4, -1 # now retest if dst aligned bbci.l a5, 1, .L0 # if now aligned, return to main algorithm .L20: # dst half-aligned # set 2 bytes EX(10f) s16i a3, a5, 0 addi a5, a5, 2 addi a4, a4, -2 j .L0 # dst is now aligned, return to main algorithm /* * Byte by byte set */ .align 4 .byte 0 # 1 mod 4 alignment for LOOPNEZ # (0 mod 4 alignment for LBEG) .Lbyteset: #if XCHAL_HAVE_LOOPS loopnez a4, .Lbytesetdone #else /* !XCHAL_HAVE_LOOPS */ beqz a4, .Lbytesetdone add a6, a5, a4 # a6 = ending address #endif /* !XCHAL_HAVE_LOOPS */ .Lbyteloop: EX(10f) s8i a3, a5, 0 addi a5, a5, 1 #if !XCHAL_HAVE_LOOPS blt a5, a6, .Lbyteloop #endif /* !XCHAL_HAVE_LOOPS */ .Lbytesetdone: retw ENDPROC(__memset) .section .fixup, "ax" .align 4 /* We return zero if a failure occurred. */ 10: movi a2, 0 retw |