/* * Copyright (C) 2008 The Android Open Source Project * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * * Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * * Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in * the documentation and/or other materials provided with the * distribution. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE * COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS * OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF * SUCH DAMAGE. */ #include .text .global memcmp .type memcmp, %function .align 4 /* * Optimized memcmp() for ARM9. * This would not be optimal on XScale or ARM11, where more prefetching * and use of PLD will be needed. * The 2 major optimzations here are * (1) The main loop compares 16 bytes at a time * (2) The loads are scheduled in a way they won't stall */ memcmp: PLD (r0, #0) PLD (r1, #0) /* take of the case where length is 0 or the buffers are the same */ cmp r0, r1 cmpne r2, #0 moveq r0, #0 bxeq lr /* save registers */ stmfd sp!, {r4, lr} PLD (r0, #32) PLD (r1, #32) /* since r0 hold the result, move the first source * pointer somewhere else */ mov r4, r0 /* make sure we have at least 8+4 bytes, this simplify things below * and avoid some overhead for small blocks */ cmp r2, #(8+4) bmi 8f /* align first pointer to word boundary * offset = -src & 3 */ rsb r3, r4, #0 ands r3, r3, #3 beq 0f /* align first pointer */ sub r2, r2, r3 1: ldrb r0, [r4], #1 ldrb ip, [r1], #1 subs r0, r0, ip bne 9f subs r3, r3, #1 bne 1b 0: /* here the first pointer is aligned, and we have at least 4 bytes * to process. */ /* see if the pointers are congruent */ eor r0, r4, r1 ands r0, r0, #3 bne 5f /* congruent case, 32 bytes per iteration * We need to make sure there are at least 32+4 bytes left * because we effectively read ahead one word, and we could * read past the buffer (and segfault) if we're not careful. */ ldr ip, [r1] subs r2, r2, #(32 + 4) bmi 1f 0: PLD (r4, #64) PLD (r1, #64) ldr r0, [r4], #4 ldr lr, [r1, #4]! eors r0, r0, ip ldreq r0, [r4], #4 ldreq ip, [r1, #4]! eoreqs r0, r0, lr ldreq r0, [r4], #4 ldreq lr, [r1, #4]! eoreqs r0, r0, ip ldreq r0, [r4], #4 ldreq ip, [r1, #4]! eoreqs r0, r0, lr ldreq r0, [r4], #4 ldreq lr, [r1, #4]! eoreqs r0, r0, ip ldreq r0, [r4], #4 ldreq ip, [r1, #4]! eoreqs r0, r0, lr ldreq r0, [r4], #4 ldreq lr, [r1, #4]! eoreqs r0, r0, ip ldreq r0, [r4], #4 ldreq ip, [r1, #4]! eoreqs r0, r0, lr bne 2f subs r2, r2, #32 bhs 0b /* do we have at least 4 bytes left? */ 1: adds r2, r2, #(32 - 4 + 4) bmi 4f /* finish off 4 bytes at a time */ 3: ldr r0, [r4], #4 ldr ip, [r1], #4 eors r0, r0, ip bne 2f subs r2, r2, #4 bhs 3b /* are we done? */ 4: adds r2, r2, #4 moveq r0, #0 beq 9f /* finish off the remaining bytes */ b 8f 2: /* the last 4 bytes are different, restart them */ sub r4, r4, #4 sub r1, r1, #4 mov r2, #4 /* process the last few bytes */ 8: ldrb r0, [r4], #1 ldrb ip, [r1], #1 // stall subs r0, r0, ip bne 9f subs r2, r2, #1 bne 8b 9: /* restore registers and return */ ldmfd sp!, {r4, lr} bx lr 5: /*************** non-congruent case ***************/ and r0, r1, #3 cmp r0, #2 bne 4f /* here, offset is 2 (16-bits aligned, special cased) */ /* make sure we have at least 16 bytes to process */ subs r2, r2, #16 addmi r2, r2, #16 bmi 8b /* align the unaligned pointer */ bic r1, r1, #3 ldr lr, [r1], #4 6: PLD (r1, #64) PLD (r4, #64) mov ip, lr, lsr #16 ldr lr, [r1], #4 ldr r0, [r4], #4 orr ip, ip, lr, lsl #16 eors r0, r0, ip moveq ip, lr, lsr #16 ldreq lr, [r1], #4 ldreq r0, [r4], #4 orreq ip, ip, lr, lsl #16 eoreqs r0, r0, ip moveq ip, lr, lsr #16 ldreq lr, [r1], #4 ldreq r0, [r4], #4 orreq ip, ip, lr, lsl #16 eoreqs r0, r0, ip moveq ip, lr, lsr #16 ldreq lr, [r1], #4 ldreq r0, [r4], #4 orreq ip, ip, lr, lsl #16 eoreqs r0, r0, ip bne 7f subs r2, r2, #16 bhs 6b sub r1, r1, #2 /* are we done? */ adds r2, r2, #16 moveq r0, #0 beq 9b /* finish off the remaining bytes */ b 8b 7: /* fix up the 2 pointers and fallthrough... */ sub r1, r1, #(4+2) sub r4, r4, #4 mov r2, #4 b 8b 4: /*************** offset is 1 or 3 (less optimized) ***************/ stmfd sp!, {r5, r6, r7} // r5 = rhs // r6 = lhs // r7 = scratch mov r5, r0, lsl #3 /* r5 = right shift */ rsb r6, r5, #32 /* r6 = left shift */ /* align the unaligned pointer */ bic r1, r1, #3 ldr r7, [r1], #4 sub r2, r2, #8 6: mov ip, r7, lsr r5 ldr r7, [r1], #4 ldr r0, [r4], #4 orr ip, ip, r7, lsl r6 eors r0, r0, ip moveq ip, r7, lsr r5 ldreq r7, [r1], #4 ldreq r0, [r4], #4 orreq ip, ip, r7, lsl r6 eoreqs r0, r0, ip bne 7f subs r2, r2, #8 bhs 6b sub r1, r1, r6, lsr #3 ldmfd sp!, {r5, r6, r7} /* are we done? */ adds r2, r2, #8 moveq r0, #0 beq 9b /* finish off the remaining bytes */ b 8b 7: /* fix up the 2 pointers and fallthrough... */ sub r1, r1, #4 sub r1, r1, r6, lsr #3 sub r4, r4, #4 mov r2, #4 ldmfd sp!, {r5, r6, r7} b 8b