Index: head/sys/arm/arm/bcopyinout_xscale.S =================================================================== --- head/sys/arm/arm/bcopyinout_xscale.S (revision 275417) +++ head/sys/arm/arm/bcopyinout_xscale.S (revision 275418) @@ -1,939 +1,940 @@ /* $NetBSD: bcopyinout_xscale.S,v 1.3 2003/12/15 09:27:18 scw Exp $ */ /*- * Copyright 2003 Wasabi Systems, Inc. * All rights reserved. * * Written by Steve C. Woodford for Wasabi Systems, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. 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. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed for the NetBSD Project by * Wasabi Systems, Inc. * 4. The name of Wasabi Systems, Inc. may not be used to endorse * or promote products derived from this software without specific prior * written permission. * * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``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 WASABI SYSTEMS, INC * 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 __FBSDID("$FreeBSD$"); + .syntax unified .text .align 0 #ifdef _ARM_ARCH_6 #define GET_PCB(tmp) \ mrc p15, 0, tmp, c13, c0, 4; \ add tmp, tmp, #(TD_PCB) #else .Lcurpcb: .word _C_LABEL(__pcpu) + PC_CURPCB #define GET_PCB(tmp) \ ldr tmp, .Lcurpcb #endif /* * r0 = user space address * r1 = kernel space address * r2 = length * * Copies bytes from user space to kernel space */ ENTRY(copyin) cmp r2, #0x00 movle r0, #0x00 movle pc, lr /* Bail early if length is <= 0 */ ldr r3, .L_arm_memcpy ldr r3, [r3] cmp r3, #0 beq .Lnormal ldr r3, .L_min_memcpy_size ldr r3, [r3] cmp r2, r3 blt .Lnormal stmfd sp!, {r0-r2, r4, lr} mov r3, r0 mov r0, r1 mov r1, r3 mov r3, #2 /* SRC_IS_USER */ ldr r4, .L_arm_memcpy mov lr, pc ldr pc, [r4] cmp r0, #0 ldmfd sp!, {r0-r2, r4, lr} moveq r0, #0 RETeq .Lnormal: stmfd sp!, {r10-r11, lr} GET_PCB(r10) ldr r10, [r10] mov r3, #0x00 adr ip, .Lcopyin_fault ldr r11, [r10, #PCB_ONFAULT] str ip, [r10, #PCB_ONFAULT] bl .Lcopyin_guts str r11, [r10, #PCB_ONFAULT] mov r0, #0x00 ldmfd sp!, {r10-r11, pc} .Lcopyin_fault: ldr r0, =EFAULT str r11, [r10, #PCB_ONFAULT] cmp r3, #0x00 - ldmgtfd sp!, {r4-r7} /* r3 > 0 Restore r4-r7 */ - ldmltfd sp!, {r4-r9} /* r3 < 0 Restore r4-r9 */ + ldmfdgt sp!, {r4-r7} /* r3 > 0 Restore r4-r7 */ + ldmfdlt sp!, {r4-r9} /* r3 < 0 Restore r4-r9 */ ldmfd sp!, {r10-r11, pc} .Lcopyin_guts: pld [r0] /* Word-align the destination buffer */ ands ip, r1, #0x03 /* Already word aligned? */ beq .Lcopyin_wordaligned /* Yup */ rsb ip, ip, #0x04 cmp r2, ip /* Enough bytes left to align it? */ blt .Lcopyin_l4_2 /* Nope. Just copy bytewise */ sub r2, r2, ip rsbs ip, ip, #0x03 addne pc, pc, ip, lsl #3 nop ldrbt ip, [r0], #0x01 strb ip, [r1], #0x01 ldrbt ip, [r0], #0x01 strb ip, [r1], #0x01 ldrbt ip, [r0], #0x01 strb ip, [r1], #0x01 cmp r2, #0x00 /* All done? */ RETeq /* Destination buffer is now word aligned */ .Lcopyin_wordaligned: ands ip, r0, #0x03 /* Is src also word-aligned? */ bne .Lcopyin_bad_align /* Nope. Things just got bad */ cmp r2, #0x08 /* Less than 8 bytes remaining? */ blt .Lcopyin_w_less_than8 /* Quad-align the destination buffer */ tst r1, #0x07 /* Already quad aligned? */ - ldrnet ip, [r0], #0x04 + ldrtne ip, [r0], #0x04 strne ip, [r1], #0x04 subne r2, r2, #0x04 stmfd sp!, {r4-r9} /* Free up some registers */ mov r3, #-1 /* Signal restore r4-r9 */ /* Destination buffer quad aligned, source is word aligned */ subs r2, r2, #0x80 blt .Lcopyin_w_lessthan128 /* Copy 128 bytes at a time */ .Lcopyin_w_loop128: ldrt r4, [r0], #0x04 /* LD:00-03 */ ldrt r5, [r0], #0x04 /* LD:04-07 */ pld [r0, #0x18] /* Prefetch 0x20 */ ldrt r6, [r0], #0x04 /* LD:08-0b */ ldrt r7, [r0], #0x04 /* LD:0c-0f */ ldrt r8, [r0], #0x04 /* LD:10-13 */ ldrt r9, [r0], #0x04 /* LD:14-17 */ strd r4, [r1], #0x08 /* ST:00-07 */ ldrt r4, [r0], #0x04 /* LD:18-1b */ ldrt r5, [r0], #0x04 /* LD:1c-1f */ strd r6, [r1], #0x08 /* ST:08-0f */ ldrt r6, [r0], #0x04 /* LD:20-23 */ ldrt r7, [r0], #0x04 /* LD:24-27 */ pld [r0, #0x18] /* Prefetch 0x40 */ strd r8, [r1], #0x08 /* ST:10-17 */ ldrt r8, [r0], #0x04 /* LD:28-2b */ ldrt r9, [r0], #0x04 /* LD:2c-2f */ strd r4, [r1], #0x08 /* ST:18-1f */ ldrt r4, [r0], #0x04 /* LD:30-33 */ ldrt r5, [r0], #0x04 /* LD:34-37 */ strd r6, [r1], #0x08 /* ST:20-27 */ ldrt r6, [r0], #0x04 /* LD:38-3b */ ldrt r7, [r0], #0x04 /* LD:3c-3f */ strd r8, [r1], #0x08 /* ST:28-2f */ ldrt r8, [r0], #0x04 /* LD:40-43 */ ldrt r9, [r0], #0x04 /* LD:44-47 */ pld [r0, #0x18] /* Prefetch 0x60 */ strd r4, [r1], #0x08 /* ST:30-37 */ ldrt r4, [r0], #0x04 /* LD:48-4b */ ldrt r5, [r0], #0x04 /* LD:4c-4f */ strd r6, [r1], #0x08 /* ST:38-3f */ ldrt r6, [r0], #0x04 /* LD:50-53 */ ldrt r7, [r0], #0x04 /* LD:54-57 */ strd r8, [r1], #0x08 /* ST:40-47 */ ldrt r8, [r0], #0x04 /* LD:58-5b */ ldrt r9, [r0], #0x04 /* LD:5c-5f */ strd r4, [r1], #0x08 /* ST:48-4f */ ldrt r4, [r0], #0x04 /* LD:60-63 */ ldrt r5, [r0], #0x04 /* LD:64-67 */ pld [r0, #0x18] /* Prefetch 0x80 */ strd r6, [r1], #0x08 /* ST:50-57 */ ldrt r6, [r0], #0x04 /* LD:68-6b */ ldrt r7, [r0], #0x04 /* LD:6c-6f */ strd r8, [r1], #0x08 /* ST:58-5f */ ldrt r8, [r0], #0x04 /* LD:70-73 */ ldrt r9, [r0], #0x04 /* LD:74-77 */ strd r4, [r1], #0x08 /* ST:60-67 */ ldrt r4, [r0], #0x04 /* LD:78-7b */ ldrt r5, [r0], #0x04 /* LD:7c-7f */ strd r6, [r1], #0x08 /* ST:68-6f */ strd r8, [r1], #0x08 /* ST:70-77 */ subs r2, r2, #0x80 strd r4, [r1], #0x08 /* ST:78-7f */ bge .Lcopyin_w_loop128 .Lcopyin_w_lessthan128: adds r2, r2, #0x80 /* Adjust for extra sub */ - ldmeqfd sp!, {r4-r9} + ldmfdeq sp!, {r4-r9} RETeq subs r2, r2, #0x20 blt .Lcopyin_w_lessthan32 /* Copy 32 bytes at a time */ .Lcopyin_w_loop32: ldrt r4, [r0], #0x04 ldrt r5, [r0], #0x04 pld [r0, #0x18] ldrt r6, [r0], #0x04 ldrt r7, [r0], #0x04 ldrt r8, [r0], #0x04 ldrt r9, [r0], #0x04 strd r4, [r1], #0x08 ldrt r4, [r0], #0x04 ldrt r5, [r0], #0x04 strd r6, [r1], #0x08 strd r8, [r1], #0x08 subs r2, r2, #0x20 strd r4, [r1], #0x08 bge .Lcopyin_w_loop32 .Lcopyin_w_lessthan32: adds r2, r2, #0x20 /* Adjust for extra sub */ - ldmeqfd sp!, {r4-r9} + ldmfdeq sp!, {r4-r9} RETeq /* Return now if done */ and r4, r2, #0x18 rsb r5, r4, #0x18 subs r2, r2, r4 add pc, pc, r5, lsl #1 nop /* At least 24 bytes remaining */ ldrt r4, [r0], #0x04 ldrt r5, [r0], #0x04 nop strd r4, [r1], #0x08 /* At least 16 bytes remaining */ ldrt r4, [r0], #0x04 ldrt r5, [r0], #0x04 nop strd r4, [r1], #0x08 /* At least 8 bytes remaining */ ldrt r4, [r0], #0x04 ldrt r5, [r0], #0x04 nop strd r4, [r1], #0x08 /* Less than 8 bytes remaining */ ldmfd sp!, {r4-r9} RETeq /* Return now if done */ mov r3, #0x00 .Lcopyin_w_less_than8: subs r2, r2, #0x04 - ldrget ip, [r0], #0x04 + ldrtge ip, [r0], #0x04 strge ip, [r1], #0x04 RETeq /* Return now if done */ addlt r2, r2, #0x04 ldrbt ip, [r0], #0x01 cmp r2, #0x02 - ldrgebt r2, [r0], #0x01 + ldrbtge r2, [r0], #0x01 strb ip, [r1], #0x01 - ldrgtbt ip, [r0] - strgeb r2, [r1], #0x01 - strgtb ip, [r1] + ldrbtgt ip, [r0] + strbge r2, [r1], #0x01 + strbgt ip, [r1] RET /* * At this point, it has not been possible to word align both buffers. * The destination buffer (r1) is word aligned, but the source buffer * (r0) is not. */ .Lcopyin_bad_align: stmfd sp!, {r4-r7} mov r3, #0x01 bic r0, r0, #0x03 cmp ip, #2 ldrt ip, [r0], #0x04 bgt .Lcopyin_bad3 beq .Lcopyin_bad2 b .Lcopyin_bad1 .Lcopyin_bad1_loop16: #ifdef __ARMEB__ mov r4, ip, lsl #8 #else mov r4, ip, lsr #8 #endif ldrt r5, [r0], #0x04 pld [r0, #0x018] ldrt r6, [r0], #0x04 ldrt r7, [r0], #0x04 ldrt ip, [r0], #0x04 #ifdef __ARMEB__ orr r4, r4, r5, lsr #24 mov r5, r5, lsl #8 orr r5, r5, r6, lsr #24 mov r6, r6, lsl #8 orr r6, r6, r7, lsr #24 mov r7, r7, lsl #8 orr r7, r7, ip, lsr #24 #else orr r4, r4, r5, lsl #24 mov r5, r5, lsr #8 orr r5, r5, r6, lsl #24 mov r6, r6, lsr #8 orr r6, r6, r7, lsl #24 mov r7, r7, lsr #8 orr r7, r7, ip, lsl #24 #endif str r4, [r1], #0x04 str r5, [r1], #0x04 str r6, [r1], #0x04 str r7, [r1], #0x04 .Lcopyin_bad1: subs r2, r2, #0x10 bge .Lcopyin_bad1_loop16 adds r2, r2, #0x10 - ldmeqfd sp!, {r4-r7} + ldmfdeq sp!, {r4-r7} RETeq /* Return now if done */ subs r2, r2, #0x04 sublt r0, r0, #0x03 blt .Lcopyin_l4 .Lcopyin_bad1_loop4: #ifdef __ARMEB__ mov r4, ip, lsl #8 #else mov r4, ip, lsr #8 #endif ldrt ip, [r0], #0x04 subs r2, r2, #0x04 #ifdef __ARMEB__ orr r4, r4, ip, lsr #24 #else orr r4, r4, ip, lsl #24 #endif str r4, [r1], #0x04 bge .Lcopyin_bad1_loop4 sub r0, r0, #0x03 b .Lcopyin_l4 .Lcopyin_bad2_loop16: #ifdef __ARMEB__ mov r4, ip, lsl #16 #else mov r4, ip, lsr #16 #endif ldrt r5, [r0], #0x04 pld [r0, #0x018] ldrt r6, [r0], #0x04 ldrt r7, [r0], #0x04 ldrt ip, [r0], #0x04 #ifdef __ARMEB__ orr r4, r4, r5, lsr #16 mov r5, r5, lsl #16 orr r5, r5, r6, lsr #16 mov r6, r6, lsl #16 orr r6, r6, r7, lsr #16 mov r7, r7, lsl #16 orr r7, r7, ip, lsr #16 #else orr r4, r4, r5, lsl #16 mov r5, r5, lsr #16 orr r5, r5, r6, lsl #16 mov r6, r6, lsr #16 orr r6, r6, r7, lsl #16 mov r7, r7, lsr #16 orr r7, r7, ip, lsl #16 #endif str r4, [r1], #0x04 str r5, [r1], #0x04 str r6, [r1], #0x04 str r7, [r1], #0x04 .Lcopyin_bad2: subs r2, r2, #0x10 bge .Lcopyin_bad2_loop16 adds r2, r2, #0x10 - ldmeqfd sp!, {r4-r7} + ldmfdeq sp!, {r4-r7} RETeq /* Return now if done */ subs r2, r2, #0x04 sublt r0, r0, #0x02 blt .Lcopyin_l4 .Lcopyin_bad2_loop4: #ifdef __ARMEB__ mov r4, ip, lsl #16 #else mov r4, ip, lsr #16 #endif ldrt ip, [r0], #0x04 subs r2, r2, #0x04 #ifdef __ARMEB__ orr r4, r4, ip, lsr #16 #else orr r4, r4, ip, lsl #16 #endif str r4, [r1], #0x04 bge .Lcopyin_bad2_loop4 sub r0, r0, #0x02 b .Lcopyin_l4 .Lcopyin_bad3_loop16: #ifdef __ARMEB__ mov r4, ip, lsl #24 #else mov r4, ip, lsr #24 #endif ldrt r5, [r0], #0x04 pld [r0, #0x018] ldrt r6, [r0], #0x04 ldrt r7, [r0], #0x04 ldrt ip, [r0], #0x04 #ifdef __ARMEB__ orr r4, r4, r5, lsr #8 mov r5, r5, lsl #24 orr r5, r5, r6, lsr #8 mov r6, r6, lsl #24 orr r6, r6, r7, lsr #8 mov r7, r7, lsl #24 orr r7, r7, ip, lsr #8 #else orr r4, r4, r5, lsl #8 mov r5, r5, lsr #24 orr r5, r5, r6, lsl #8 mov r6, r6, lsr #24 orr r6, r6, r7, lsl #8 mov r7, r7, lsr #24 orr r7, r7, ip, lsl #8 #endif str r4, [r1], #0x04 str r5, [r1], #0x04 str r6, [r1], #0x04 str r7, [r1], #0x04 .Lcopyin_bad3: subs r2, r2, #0x10 bge .Lcopyin_bad3_loop16 adds r2, r2, #0x10 - ldmeqfd sp!, {r4-r7} + ldmfdeq sp!, {r4-r7} RETeq /* Return now if done */ subs r2, r2, #0x04 sublt r0, r0, #0x01 blt .Lcopyin_l4 .Lcopyin_bad3_loop4: #ifdef __ARMEB__ mov r4, ip, lsl #24 #else mov r4, ip, lsr #24 #endif ldrt ip, [r0], #0x04 subs r2, r2, #0x04 #ifdef __ARMEB__ orr r4, r4, ip, lsr #8 #else orr r4, r4, ip, lsl #8 #endif str r4, [r1], #0x04 bge .Lcopyin_bad3_loop4 sub r0, r0, #0x01 .Lcopyin_l4: ldmfd sp!, {r4-r7} mov r3, #0x00 adds r2, r2, #0x04 RETeq .Lcopyin_l4_2: rsbs r2, r2, #0x03 addne pc, pc, r2, lsl #3 nop ldrbt ip, [r0], #0x01 strb ip, [r1], #0x01 ldrbt ip, [r0], #0x01 strb ip, [r1], #0x01 ldrbt ip, [r0] strb ip, [r1] RET END(copyin) /* * r0 = kernel space address * r1 = user space address * r2 = length * * Copies bytes from kernel space to user space */ ENTRY(copyout) cmp r2, #0x00 movle r0, #0x00 movle pc, lr /* Bail early if length is <= 0 */ ldr r3, .L_arm_memcpy ldr r3, [r3] cmp r3, #0 beq .Lnormale ldr r3, .L_min_memcpy_size ldr r3, [r3] cmp r2, r3 blt .Lnormale stmfd sp!, {r0-r2, r4, lr} mov r3, r0 mov r0, r1 mov r1, r3 mov r3, #1 /* DST_IS_USER */ ldr r4, .L_arm_memcpy mov lr, pc ldr pc, [r4] cmp r0, #0 ldmfd sp!, {r0-r2, r4, lr} moveq r0, #0 RETeq .Lnormale: stmfd sp!, {r10-r11, lr} GET_PCB(r10) ldr r10, [r10] mov r3, #0x00 adr ip, .Lcopyout_fault ldr r11, [r10, #PCB_ONFAULT] str ip, [r10, #PCB_ONFAULT] bl .Lcopyout_guts str r11, [r10, #PCB_ONFAULT] mov r0, #0x00 ldmfd sp!, {r10-r11, pc} .Lcopyout_fault: ldr r0, =EFAULT str r11, [r10, #PCB_ONFAULT] cmp r3, #0x00 - ldmgtfd sp!, {r4-r7} /* r3 > 0 Restore r4-r7 */ - ldmltfd sp!, {r4-r9} /* r3 < 0 Restore r4-r9 */ + ldmfdgt sp!, {r4-r7} /* r3 > 0 Restore r4-r7 */ + ldmfdlt sp!, {r4-r9} /* r3 < 0 Restore r4-r9 */ ldmfd sp!, {r10-r11, pc} .Lcopyout_guts: pld [r0] /* Word-align the destination buffer */ ands ip, r1, #0x03 /* Already word aligned? */ beq .Lcopyout_wordaligned /* Yup */ rsb ip, ip, #0x04 cmp r2, ip /* Enough bytes left to align it? */ blt .Lcopyout_l4_2 /* Nope. Just copy bytewise */ sub r2, r2, ip rsbs ip, ip, #0x03 addne pc, pc, ip, lsl #3 nop ldrb ip, [r0], #0x01 strbt ip, [r1], #0x01 ldrb ip, [r0], #0x01 strbt ip, [r1], #0x01 ldrb ip, [r0], #0x01 strbt ip, [r1], #0x01 cmp r2, #0x00 /* All done? */ RETeq /* Destination buffer is now word aligned */ .Lcopyout_wordaligned: ands ip, r0, #0x03 /* Is src also word-aligned? */ bne .Lcopyout_bad_align /* Nope. Things just got bad */ cmp r2, #0x08 /* Less than 8 bytes remaining? */ blt .Lcopyout_w_less_than8 /* Quad-align the destination buffer */ tst r0, #0x07 /* Already quad aligned? */ ldrne ip, [r0], #0x04 subne r2, r2, #0x04 - strnet ip, [r1], #0x04 + strtne ip, [r1], #0x04 stmfd sp!, {r4-r9} /* Free up some registers */ mov r3, #-1 /* Signal restore r4-r9 */ /* Destination buffer word aligned, source is quad aligned */ subs r2, r2, #0x80 blt .Lcopyout_w_lessthan128 /* Copy 128 bytes at a time */ .Lcopyout_w_loop128: ldrd r4, [r0], #0x08 /* LD:00-07 */ pld [r0, #0x18] /* Prefetch 0x20 */ ldrd r6, [r0], #0x08 /* LD:08-0f */ ldrd r8, [r0], #0x08 /* LD:10-17 */ strt r4, [r1], #0x04 /* ST:00-03 */ strt r5, [r1], #0x04 /* ST:04-07 */ ldrd r4, [r0], #0x08 /* LD:18-1f */ strt r6, [r1], #0x04 /* ST:08-0b */ strt r7, [r1], #0x04 /* ST:0c-0f */ ldrd r6, [r0], #0x08 /* LD:20-27 */ pld [r0, #0x18] /* Prefetch 0x40 */ strt r8, [r1], #0x04 /* ST:10-13 */ strt r9, [r1], #0x04 /* ST:14-17 */ ldrd r8, [r0], #0x08 /* LD:28-2f */ strt r4, [r1], #0x04 /* ST:18-1b */ strt r5, [r1], #0x04 /* ST:1c-1f */ ldrd r4, [r0], #0x08 /* LD:30-37 */ strt r6, [r1], #0x04 /* ST:20-23 */ strt r7, [r1], #0x04 /* ST:24-27 */ ldrd r6, [r0], #0x08 /* LD:38-3f */ strt r8, [r1], #0x04 /* ST:28-2b */ strt r9, [r1], #0x04 /* ST:2c-2f */ ldrd r8, [r0], #0x08 /* LD:40-47 */ pld [r0, #0x18] /* Prefetch 0x60 */ strt r4, [r1], #0x04 /* ST:30-33 */ strt r5, [r1], #0x04 /* ST:34-37 */ ldrd r4, [r0], #0x08 /* LD:48-4f */ strt r6, [r1], #0x04 /* ST:38-3b */ strt r7, [r1], #0x04 /* ST:3c-3f */ ldrd r6, [r0], #0x08 /* LD:50-57 */ strt r8, [r1], #0x04 /* ST:40-43 */ strt r9, [r1], #0x04 /* ST:44-47 */ ldrd r8, [r0], #0x08 /* LD:58-4f */ strt r4, [r1], #0x04 /* ST:48-4b */ strt r5, [r1], #0x04 /* ST:4c-4f */ ldrd r4, [r0], #0x08 /* LD:60-67 */ pld [r0, #0x18] /* Prefetch 0x80 */ strt r6, [r1], #0x04 /* ST:50-53 */ strt r7, [r1], #0x04 /* ST:54-57 */ ldrd r6, [r0], #0x08 /* LD:68-6f */ strt r8, [r1], #0x04 /* ST:58-5b */ strt r9, [r1], #0x04 /* ST:5c-5f */ ldrd r8, [r0], #0x08 /* LD:70-77 */ strt r4, [r1], #0x04 /* ST:60-63 */ strt r5, [r1], #0x04 /* ST:64-67 */ ldrd r4, [r0], #0x08 /* LD:78-7f */ strt r6, [r1], #0x04 /* ST:68-6b */ strt r7, [r1], #0x04 /* ST:6c-6f */ strt r8, [r1], #0x04 /* ST:70-73 */ strt r9, [r1], #0x04 /* ST:74-77 */ subs r2, r2, #0x80 strt r4, [r1], #0x04 /* ST:78-7b */ strt r5, [r1], #0x04 /* ST:7c-7f */ bge .Lcopyout_w_loop128 .Lcopyout_w_lessthan128: adds r2, r2, #0x80 /* Adjust for extra sub */ - ldmeqfd sp!, {r4-r9} + ldmfdeq sp!, {r4-r9} RETeq /* Return now if done */ subs r2, r2, #0x20 blt .Lcopyout_w_lessthan32 /* Copy 32 bytes at a time */ .Lcopyout_w_loop32: ldrd r4, [r0], #0x08 pld [r0, #0x18] ldrd r6, [r0], #0x08 ldrd r8, [r0], #0x08 strt r4, [r1], #0x04 strt r5, [r1], #0x04 ldrd r4, [r0], #0x08 strt r6, [r1], #0x04 strt r7, [r1], #0x04 strt r8, [r1], #0x04 strt r9, [r1], #0x04 subs r2, r2, #0x20 strt r4, [r1], #0x04 strt r5, [r1], #0x04 bge .Lcopyout_w_loop32 .Lcopyout_w_lessthan32: adds r2, r2, #0x20 /* Adjust for extra sub */ - ldmeqfd sp!, {r4-r9} + ldmfdeq sp!, {r4-r9} RETeq /* Return now if done */ and r4, r2, #0x18 rsb r5, r4, #0x18 subs r2, r2, r4 add pc, pc, r5, lsl #1 nop /* At least 24 bytes remaining */ ldrd r4, [r0], #0x08 strt r4, [r1], #0x04 strt r5, [r1], #0x04 nop /* At least 16 bytes remaining */ ldrd r4, [r0], #0x08 strt r4, [r1], #0x04 strt r5, [r1], #0x04 nop /* At least 8 bytes remaining */ ldrd r4, [r0], #0x08 strt r4, [r1], #0x04 strt r5, [r1], #0x04 nop /* Less than 8 bytes remaining */ ldmfd sp!, {r4-r9} RETeq /* Return now if done */ mov r3, #0x00 .Lcopyout_w_less_than8: subs r2, r2, #0x04 ldrge ip, [r0], #0x04 - strget ip, [r1], #0x04 + strtge ip, [r1], #0x04 RETeq /* Return now if done */ addlt r2, r2, #0x04 ldrb ip, [r0], #0x01 cmp r2, #0x02 - ldrgeb r2, [r0], #0x01 + ldrbge r2, [r0], #0x01 strbt ip, [r1], #0x01 - ldrgtb ip, [r0] - strgebt r2, [r1], #0x01 - strgtbt ip, [r1] + ldrbgt ip, [r0] + strbtge r2, [r1], #0x01 + strbtgt ip, [r1] RET /* * At this point, it has not been possible to word align both buffers. * The destination buffer (r1) is word aligned, but the source buffer * (r0) is not. */ .Lcopyout_bad_align: stmfd sp!, {r4-r7} mov r3, #0x01 bic r0, r0, #0x03 cmp ip, #2 ldr ip, [r0], #0x04 bgt .Lcopyout_bad3 beq .Lcopyout_bad2 b .Lcopyout_bad1 .Lcopyout_bad1_loop16: #ifdef __ARMEB__ mov r4, ip, lsl #8 #else mov r4, ip, lsr #8 #endif ldr r5, [r0], #0x04 pld [r0, #0x018] ldr r6, [r0], #0x04 ldr r7, [r0], #0x04 ldr ip, [r0], #0x04 #ifdef __ARMEB__ orr r4, r4, r5, lsr #24 mov r5, r5, lsl #8 orr r5, r5, r6, lsr #24 mov r6, r6, lsl #8 orr r6, r6, r7, lsr #24 mov r7, r7, lsl #8 orr r7, r7, ip, lsr #24 #else orr r4, r4, r5, lsl #24 mov r5, r5, lsr #8 orr r5, r5, r6, lsl #24 mov r6, r6, lsr #8 orr r6, r6, r7, lsl #24 mov r7, r7, lsr #8 orr r7, r7, ip, lsl #24 #endif strt r4, [r1], #0x04 strt r5, [r1], #0x04 strt r6, [r1], #0x04 strt r7, [r1], #0x04 .Lcopyout_bad1: subs r2, r2, #0x10 bge .Lcopyout_bad1_loop16 adds r2, r2, #0x10 - ldmeqfd sp!, {r4-r7} + ldmfdeq sp!, {r4-r7} RETeq /* Return now if done */ subs r2, r2, #0x04 sublt r0, r0, #0x03 blt .Lcopyout_l4 .Lcopyout_bad1_loop4: #ifdef __ARMEB__ mov r4, ip, lsl #8 #else mov r4, ip, lsr #8 #endif ldr ip, [r0], #0x04 subs r2, r2, #0x04 #ifdef __ARMEB__ orr r4, r4, ip, lsr #24 #else orr r4, r4, ip, lsl #24 #endif strt r4, [r1], #0x04 bge .Lcopyout_bad1_loop4 sub r0, r0, #0x03 b .Lcopyout_l4 .Lcopyout_bad2_loop16: #ifdef __ARMEB__ mov r4, ip, lsl #16 #else mov r4, ip, lsr #16 #endif ldr r5, [r0], #0x04 pld [r0, #0x018] ldr r6, [r0], #0x04 ldr r7, [r0], #0x04 ldr ip, [r0], #0x04 #ifdef __ARMEB__ orr r4, r4, r5, lsr #16 mov r5, r5, lsl #16 orr r5, r5, r6, lsr #16 mov r6, r6, lsl #16 orr r6, r6, r7, lsr #16 mov r7, r7, lsl #16 orr r7, r7, ip, lsr #16 #else orr r4, r4, r5, lsl #16 mov r5, r5, lsr #16 orr r5, r5, r6, lsl #16 mov r6, r6, lsr #16 orr r6, r6, r7, lsl #16 mov r7, r7, lsr #16 orr r7, r7, ip, lsl #16 #endif strt r4, [r1], #0x04 strt r5, [r1], #0x04 strt r6, [r1], #0x04 strt r7, [r1], #0x04 .Lcopyout_bad2: subs r2, r2, #0x10 bge .Lcopyout_bad2_loop16 adds r2, r2, #0x10 - ldmeqfd sp!, {r4-r7} + ldmfdeq sp!, {r4-r7} RETeq /* Return now if done */ subs r2, r2, #0x04 sublt r0, r0, #0x02 blt .Lcopyout_l4 .Lcopyout_bad2_loop4: #ifdef __ARMEB__ mov r4, ip, lsl #16 #else mov r4, ip, lsr #16 #endif ldr ip, [r0], #0x04 subs r2, r2, #0x04 #ifdef __ARMEB__ orr r4, r4, ip, lsr #16 #else orr r4, r4, ip, lsl #16 #endif strt r4, [r1], #0x04 bge .Lcopyout_bad2_loop4 sub r0, r0, #0x02 b .Lcopyout_l4 .Lcopyout_bad3_loop16: #ifdef __ARMEB__ mov r4, ip, lsl #24 #else mov r4, ip, lsr #24 #endif ldr r5, [r0], #0x04 pld [r0, #0x018] ldr r6, [r0], #0x04 ldr r7, [r0], #0x04 ldr ip, [r0], #0x04 #ifdef __ARMEB__ orr r4, r4, r5, lsr #8 mov r5, r5, lsl #24 orr r5, r5, r6, lsr #8 mov r6, r6, lsl #24 orr r6, r6, r7, lsr #8 mov r7, r7, lsl #24 orr r7, r7, ip, lsr #8 #else orr r4, r4, r5, lsl #8 mov r5, r5, lsr #24 orr r5, r5, r6, lsl #8 mov r6, r6, lsr #24 orr r6, r6, r7, lsl #8 mov r7, r7, lsr #24 orr r7, r7, ip, lsl #8 #endif strt r4, [r1], #0x04 strt r5, [r1], #0x04 strt r6, [r1], #0x04 strt r7, [r1], #0x04 .Lcopyout_bad3: subs r2, r2, #0x10 bge .Lcopyout_bad3_loop16 adds r2, r2, #0x10 - ldmeqfd sp!, {r4-r7} + ldmfdeq sp!, {r4-r7} RETeq /* Return now if done */ subs r2, r2, #0x04 sublt r0, r0, #0x01 blt .Lcopyout_l4 .Lcopyout_bad3_loop4: #ifdef __ARMEB__ mov r4, ip, lsl #24 #else mov r4, ip, lsr #24 #endif ldr ip, [r0], #0x04 subs r2, r2, #0x04 #ifdef __ARMEB__ orr r4, r4, ip, lsr #8 #else orr r4, r4, ip, lsl #8 #endif strt r4, [r1], #0x04 bge .Lcopyout_bad3_loop4 sub r0, r0, #0x01 .Lcopyout_l4: ldmfd sp!, {r4-r7} mov r3, #0x00 adds r2, r2, #0x04 RETeq .Lcopyout_l4_2: rsbs r2, r2, #0x03 addne pc, pc, r2, lsl #3 nop ldrb ip, [r0], #0x01 strbt ip, [r1], #0x01 ldrb ip, [r0], #0x01 strbt ip, [r1], #0x01 ldrb ip, [r0] strbt ip, [r1] RET END(copyout) Index: head/sys/arm/arm/blockio.S =================================================================== --- head/sys/arm/arm/blockio.S (revision 275417) +++ head/sys/arm/arm/blockio.S (revision 275418) @@ -1,594 +1,596 @@ /* $NetBSD: blockio.S,v 1.5 2002/08/15 01:38:16 briggs Exp $ */ /*- * Copyright (c) 2001 Ben Harris. * Copyright (c) 1994 Mark Brinicombe. * Copyright (c) 1994 Brini. * All rights reserved. * * This code is derived from software written for Brini by Mark Brinicombe * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. 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. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed by Brini. * 4. The name of the company nor the name of the author may be used to * endorse or promote products derived from this software without specific * prior written permission. * * THIS SOFTWARE IS PROVIDED BY BRINI ``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 BRINI 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. * * RiscBSD kernel project * * blockio.S * * optimised block read/write from/to IO routines. * * Created : 08/10/94 * Modified : 22/01/99 -- R.Earnshaw * Faster, and small tweaks for StrongARM */ #include __FBSDID("$FreeBSD$"); + .syntax unified + /* * Read bytes from an I/O address into a block of memory * * r0 = address to read from (IO) * r1 = address to write to (memory) * r2 = length */ /* This code will look very familiar if you've read _memcpy(). */ ENTRY(read_multi_1) mov ip, sp stmfd sp!, {fp, ip, lr, pc} sub fp, ip, #4 subs r2, r2, #4 /* r2 = length - 4 */ blt .Lrm1_l4 /* less than 4 bytes */ ands r12, r1, #3 beq .Lrm1_main /* aligned destination */ rsb r12, r12, #4 cmp r12, #2 ldrb r3, [r0] strb r3, [r1], #1 - ldrgeb r3, [r0] - strgeb r3, [r1], #1 - ldrgtb r3, [r0] - strgtb r3, [r1], #1 + ldrbge r3, [r0] + strbge r3, [r1], #1 + ldrbgt r3, [r0] + strbgt r3, [r1], #1 subs r2, r2, r12 blt .Lrm1_l4 .Lrm1_main: .Lrm1loop: ldrb r3, [r0] ldrb r12, [r0] orr r3, r3, r12, lsl #8 ldrb r12, [r0] orr r3, r3, r12, lsl #16 ldrb r12, [r0] orr r3, r3, r12, lsl #24 str r3, [r1], #4 subs r2, r2, #4 bge .Lrm1loop .Lrm1_l4: adds r2, r2, #4 /* r2 = length again */ - ldmeqdb fp, {fp, sp, pc} + ldmdbeq fp, {fp, sp, pc} RETeq cmp r2, #2 ldrb r3, [r0] strb r3, [r1], #1 - ldrgeb r3, [r0] - strgeb r3, [r1], #1 - ldrgtb r3, [r0] - strgtb r3, [r1], #1 + ldrbge r3, [r0] + strbge r3, [r1], #1 + ldrbgt r3, [r0] + strbgt r3, [r1], #1 ldmdb fp, {fp, sp, pc} END(read_multi_1) /* * Write bytes to an I/O address from a block of memory * * r0 = address to write to (IO) * r1 = address to read from (memory) * r2 = length */ /* This code will look very familiar if you've read _memcpy(). */ ENTRY(write_multi_1) mov ip, sp stmfd sp!, {fp, ip, lr, pc} sub fp, ip, #4 subs r2, r2, #4 /* r2 = length - 4 */ blt .Lwm1_l4 /* less than 4 bytes */ ands r12, r1, #3 beq .Lwm1_main /* aligned source */ rsb r12, r12, #4 cmp r12, #2 ldrb r3, [r1], #1 strb r3, [r0] - ldrgeb r3, [r1], #1 - strgeb r3, [r0] - ldrgtb r3, [r1], #1 - strgtb r3, [r0] + ldrbge r3, [r1], #1 + strbge r3, [r0] + ldrbgt r3, [r1], #1 + strbgt r3, [r0] subs r2, r2, r12 blt .Lwm1_l4 .Lwm1_main: .Lwm1loop: ldr r3, [r1], #4 strb r3, [r0] mov r3, r3, lsr #8 strb r3, [r0] mov r3, r3, lsr #8 strb r3, [r0] mov r3, r3, lsr #8 strb r3, [r0] subs r2, r2, #4 bge .Lwm1loop .Lwm1_l4: adds r2, r2, #4 /* r2 = length again */ - ldmeqdb fp, {fp, sp, pc} + ldmdbeq fp, {fp, sp, pc} cmp r2, #2 ldrb r3, [r1], #1 strb r3, [r0] - ldrgeb r3, [r1], #1 - strgeb r3, [r0] - ldrgtb r3, [r1], #1 - strgtb r3, [r0] + ldrbge r3, [r1], #1 + strbge r3, [r0] + ldrbgt r3, [r1], #1 + strbgt r3, [r0] ldmdb fp, {fp, sp, pc} END(write_multi_1) /* * Reads short ints (16 bits) from an I/O address into a block of memory * * r0 = address to read from (IO) * r1 = address to write to (memory) * r2 = length */ ENTRY(insw) /* Make sure that we have a positive length */ cmp r2, #0x00000000 movle pc, lr /* If the destination address and the size is word aligned, do it fast */ tst r2, #0x00000001 tsteq r1, #0x00000003 beq .Lfastinsw /* Non aligned insw */ .Linswloop: ldr r3, [r0] subs r2, r2, #0x00000001 /* Loop test in load delay slot */ strb r3, [r1], #0x0001 mov r3, r3, lsr #8 strb r3, [r1], #0x0001 bgt .Linswloop RET /* Word aligned insw */ .Lfastinsw: .Lfastinswloop: ldr r3, [r0, #0x0002] /* take advantage of nonaligned * word accesses */ ldr ip, [r0] mov r3, r3, lsr #16 /* Put the two shorts together */ orr r3, r3, ip, lsl #16 str r3, [r1], #0x0004 /* Store */ subs r2, r2, #0x00000002 /* Next */ bgt .Lfastinswloop RET END(insw) /* * Writes short ints (16 bits) from a block of memory to an I/O address * * r0 = address to write to (IO) * r1 = address to read from (memory) * r2 = length */ ENTRY(outsw) /* Make sure that we have a positive length */ cmp r2, #0x00000000 movle pc, lr /* If the destination address and the size is word aligned, do it fast */ tst r2, #0x00000001 tsteq r1, #0x00000003 beq .Lfastoutsw /* Non aligned outsw */ .Loutswloop: ldrb r3, [r1], #0x0001 ldrb ip, [r1], #0x0001 subs r2, r2, #0x00000001 /* Loop test in load delay slot */ orr r3, r3, ip, lsl #8 orr r3, r3, r3, lsl #16 str r3, [r0] bgt .Loutswloop RET /* Word aligned outsw */ .Lfastoutsw: .Lfastoutswloop: ldr r3, [r1], #0x0004 /* r3 = (H)(L) */ subs r2, r2, #0x00000002 /* Loop test in load delay slot */ eor ip, r3, r3, lsr #16 /* ip = (H)(H^L) */ eor r3, r3, ip, lsl #16 /* r3 = (H^H^L)(L) = (L)(L) */ eor ip, ip, r3, lsr #16 /* ip = (H)(H^L^L) = (H)(H) */ str r3, [r0] str ip, [r0] /* mov ip, r3, lsl #16 * orr ip, ip, ip, lsr #16 * str ip, [r0] * * mov ip, r3, lsr #16 * orr ip, ip, ip, lsl #16 * str ip, [r0] */ bgt .Lfastoutswloop RET END(outsw) /* * reads short ints (16 bits) from an I/O address into a block of memory * with a length garenteed to be a multiple of 16 bytes * with a word aligned destination address * * r0 = address to read from (IO) * r1 = address to write to (memory) * r2 = length */ ENTRY(insw16) /* Make sure that we have a positive length */ cmp r2, #0x00000000 movle pc, lr /* If the destination address is word aligned and the size suitably aligned, do it fast */ tst r2, #0x00000007 tsteq r1, #0x00000003 bne _C_LABEL(insw) /* Word aligned insw */ stmfd sp!, {r4,r5,lr} .Linsw16loop: ldr r3, [r0, #0x0002] /* take advantage of nonaligned * word accesses */ ldr lr, [r0] mov r3, r3, lsr #16 /* Put the two shorts together */ orr r3, r3, lr, lsl #16 ldr r4, [r0, #0x0002] /* take advantage of nonaligned * word accesses */ ldr lr, [r0] mov r4, r4, lsr #16 /* Put the two shorts together */ orr r4, r4, lr, lsl #16 ldr r5, [r0, #0x0002] /* take advantage of nonaligned * word accesses */ ldr lr, [r0] mov r5, r5, lsr #16 /* Put the two shorts together */ orr r5, r5, lr, lsl #16 ldr ip, [r0, #0x0002] /* take advantage of nonaligned * word accesses */ ldr lr, [r0] mov ip, ip, lsr #16 /* Put the two shorts together */ orr ip, ip, lr, lsl #16 stmia r1!, {r3-r5,ip} subs r2, r2, #0x00000008 /* Next */ bgt .Linsw16loop ldmfd sp!, {r4,r5,pc} /* Restore regs and go home */ END(insw16) /* * Writes short ints (16 bits) from a block of memory to an I/O address * * r0 = address to write to (IO) * r1 = address to read from (memory) * r2 = length */ ENTRY(outsw16) /* Make sure that we have a positive length */ cmp r2, #0x00000000 movle pc, lr /* If the destination address is word aligned and the size suitably aligned, do it fast */ tst r2, #0x00000007 tsteq r1, #0x00000003 bne _C_LABEL(outsw) /* Word aligned outsw */ stmfd sp!, {r4,r5,lr} .Loutsw16loop: ldmia r1!, {r4,r5,ip,lr} eor r3, r4, r4, lsl #16 /* r3 = (A^B)(B) */ eor r4, r4, r3, lsr #16 /* r4 = (A)(B^A^B) = (A)(A) */ eor r3, r3, r4, lsl #16 /* r3 = (A^B^A)(B) = (B)(B) */ str r3, [r0] str r4, [r0] /* mov r3, r4, lsl #16 * orr r3, r3, r3, lsr #16 * str r3, [r0] * * mov r3, r4, lsr #16 * orr r3, r3, r3, lsl #16 * str r3, [r0] */ eor r3, r5, r5, lsl #16 /* r3 = (A^B)(B) */ eor r5, r5, r3, lsr #16 /* r4 = (A)(B^A^B) = (A)(A) */ eor r3, r3, r5, lsl #16 /* r3 = (A^B^A)(B) = (B)(B) */ str r3, [r0] str r5, [r0] eor r3, ip, ip, lsl #16 /* r3 = (A^B)(B) */ eor ip, ip, r3, lsr #16 /* r4 = (A)(B^A^B) = (A)(A) */ eor r3, r3, ip, lsl #16 /* r3 = (A^B^A)(B) = (B)(B) */ str r3, [r0] str ip, [r0] eor r3, lr, lr, lsl #16 /* r3 = (A^B)(B) */ eor lr, lr, r3, lsr #16 /* r4 = (A)(B^A^B) = (A)(A) */ eor r3, r3, lr, lsl #16 /* r3 = (A^B^A)(B) = (B)(B) */ str r3, [r0] str lr, [r0] subs r2, r2, #0x00000008 bgt .Loutsw16loop ldmfd sp!, {r4,r5,pc} /* and go home */ END(outsw16) /* * reads short ints (16 bits) from an I/O address into a block of memory * The I/O address is assumed to be mapped multiple times in a block of * 8 words. * The destination address should be word aligned. * * r0 = address to read from (IO) * r1 = address to write to (memory) * r2 = length */ ENTRY(inswm8) /* Make sure that we have a positive length */ cmp r2, #0x00000000 movle pc, lr /* If the destination address is word aligned and the size suitably aligned, do it fast */ tst r1, #0x00000003 bne _C_LABEL(insw) /* Word aligned insw */ stmfd sp!, {r4-r9,lr} mov lr, #0xff000000 orr lr, lr, #0x00ff0000 .Linswm8_loop8: cmp r2, #8 bcc .Linswm8_l8 ldmia r0, {r3-r9,ip} bic r3, r3, lr orr r3, r3, r4, lsl #16 bic r5, r5, lr orr r4, r5, r6, lsl #16 bic r7, r7, lr orr r5, r7, r8, lsl #16 bic r9, r9, lr orr r6, r9, ip, lsl #16 stmia r1!, {r3-r6} subs r2, r2, #0x00000008 /* Next */ bne .Linswm8_loop8 beq .Linswm8_l1 .Linswm8_l8: cmp r2, #4 bcc .Linswm8_l4 ldmia r0, {r3-r6} bic r3, r3, lr orr r3, r3, r4, lsl #16 bic r5, r5, lr orr r4, r5, r6, lsl #16 stmia r1!, {r3-r4} subs r2, r2, #0x00000004 beq .Linswm8_l1 .Linswm8_l4: cmp r2, #2 bcc .Linswm8_l2 ldmia r0, {r3-r4} bic r3, r3, lr orr r3, r3, r4, lsl #16 str r3, [r1], #0x0004 subs r2, r2, #0x00000002 beq .Linswm8_l1 .Linswm8_l2: cmp r2, #1 bcc .Linswm8_l1 ldr r3, [r0] subs r2, r2, #0x00000001 /* Test in load delay slot */ /* XXX, why don't we use result? */ strb r3, [r1], #0x0001 mov r3, r3, lsr #8 strb r3, [r1], #0x0001 .Linswm8_l1: ldmfd sp!, {r4-r9,pc} /* And go home */ END(inswm8) /* * write short ints (16 bits) to an I/O address from a block of memory * The I/O address is assumed to be mapped multiple times in a block of * 8 words. * The source address should be word aligned. * * r0 = address to read to (IO) * r1 = address to write from (memory) * r2 = length */ ENTRY(outswm8) /* Make sure that we have a positive length */ cmp r2, #0x00000000 movle pc, lr /* If the destination address is word aligned and the size suitably aligned, do it fast */ tst r1, #0x00000003 bne _C_LABEL(outsw) /* Word aligned outsw */ stmfd sp!, {r4-r8,lr} .Loutswm8_loop8: cmp r2, #8 bcc .Loutswm8_l8 ldmia r1!, {r3,r5,r7,ip} eor r4, r3, r3, lsr #16 /* r4 = (A)(A^B) */ eor r3, r3, r4, lsl #16 /* r3 = (A^A^B)(B) = (B)(B) */ eor r4, r4, r3, lsr #16 /* r4 = (A)(B^A^B) = (A)(A) */ eor r6, r5, r5, lsr #16 /* r6 = (A)(A^B) */ eor r5, r5, r6, lsl #16 /* r5 = (A^A^B)(B) = (B)(B) */ eor r6, r6, r5, lsr #16 /* r6 = (A)(B^A^B) = (A)(A) */ eor r8, r7, r7, lsr #16 /* r8 = (A)(A^B) */ eor r7, r7, r8, lsl #16 /* r7 = (A^A^B)(B) = (B)(B) */ eor r8, r8, r7, lsr #16 /* r8 = (A)(B^A^B) = (A)(A) */ eor lr, ip, ip, lsr #16 /* lr = (A)(A^B) */ eor ip, ip, lr, lsl #16 /* ip = (A^A^B)(B) = (B)(B) */ eor lr, lr, ip, lsr #16 /* lr = (A)(B^A^B) = (A)(A) */ stmia r0, {r3-r8,ip,lr} subs r2, r2, #0x00000008 /* Next */ bne .Loutswm8_loop8 beq .Loutswm8_l1 .Loutswm8_l8: cmp r2, #4 bcc .Loutswm8_l4 ldmia r1!, {r3-r4} eor r6, r3, r3, lsr #16 /* r6 = (A)(A^B) */ eor r5, r3, r6, lsl #16 /* r5 = (A^A^B)(B) = (B)(B) */ eor r6, r6, r5, lsr #16 /* r6 = (A)(B^A^B) = (A)(A) */ eor r8, r4, r4, lsr #16 /* r8 = (A)(A^B) */ eor r7, r4, r8, lsl #16 /* r7 = (A^A^B)(B) = (B)(B) */ eor r8, r8, r7, lsr #16 /* r8 = (A)(B^A^B) = (A)(A) */ stmia r0, {r5-r8} subs r2, r2, #0x00000004 beq .Loutswm8_l1 .Loutswm8_l4: cmp r2, #2 bcc .Loutswm8_l2 ldr r3, [r1], #0x0004 /* r3 = (A)(B) */ subs r2, r2, #0x00000002 /* Done test in Load delay slot */ eor r5, r3, r3, lsr #16 /* r5 = (A)(A^B)*/ eor r4, r3, r5, lsl #16 /* r4 = (A^A^B)(B) = (B)(B) */ eor r5, r5, r4, lsr #16 /* r5 = (A)(B^A^B) = (A)(A) */ stmia r0, {r4, r5} beq .Loutswm8_l1 .Loutswm8_l2: cmp r2, #1 bcc .Loutswm8_l1 ldrb r3, [r1], #0x0001 ldrb r4, [r1], #0x0001 subs r2, r2, #0x00000001 /* Done test in load delay slot */ /* XXX This test isn't used? */ orr r3, r3, r4, lsl #8 orr r3, r3, r3, lsl #16 str r3, [r0] .Loutswm8_l1: ldmfd sp!, {r4-r8,pc} /* And go home */ END(outswm8) Index: head/sys/arm/arm/in_cksum_arm.S =================================================================== --- head/sys/arm/arm/in_cksum_arm.S (revision 275417) +++ head/sys/arm/arm/in_cksum_arm.S (revision 275418) @@ -1,343 +1,344 @@ /* $NetBSD: in_cksum_arm.S,v 1.2 2003/09/23 10:01:36 scw Exp $ */ /*- * Copyright 2003 Wasabi Systems, Inc. * All rights reserved. * * Written by Steve C. Woodford for Wasabi Systems, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. 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. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed for the NetBSD Project by * Wasabi Systems, Inc. * 4. The name of Wasabi Systems, Inc. may not be used to endorse * or promote products derived from this software without specific prior * written permission. * * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``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 WASABI SYSTEMS, INC * 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. * */ /* * Hand-optimised in_cksum() and in4_cksum() implementations for ARM/armv5e */ #include "opt_inet.h" #include #include "assym.s" __FBSDID("$FreeBSD$"); + .syntax unified /* * int in_cksum(struct mbuf *m, int len) * * Entry: * r0 m * r1 len * * NOTE: Assumes 'm' is *never* NULL. */ /* LINTSTUB: Func: int in_cksum(struct mbuf *, int) */ ENTRY(in_cksum) stmfd sp!, {r4-r11,lr} mov r8, #0x00 mov r9, r1 mov r10, #0x00 mov ip, r0 .Lin_cksum_loop: ldr r1, [ip, #(M_LEN)] ldr r0, [ip, #(M_DATA)] ldr ip, [ip, #(M_NEXT)] .Lin_cksum_entry4: cmp r9, r1 movlt r1, r9 sub r9, r9, r1 eor r11, r10, r0 add r10, r10, r1 adds r2, r1, #0x00 blne _ASM_LABEL(L_cksumdata) tst r11, #0x01 movne r2, r2, ror #8 adds r8, r8, r2 adc r8, r8, #0x00 cmp ip, #0x00 bne .Lin_cksum_loop mov r1, #0xff orr r1, r1, #0xff00 and r0, r8, r1 add r0, r0, r8, lsr #16 add r0, r0, r0, lsr #16 and r0, r0, r1 eor r0, r0, r1 ldmfd sp!, {r4-r11,pc} END(in_cksum) ENTRY(do_cksum) stmfd sp!, {r4-r7, lr} bl L_cksumdata mov r0, r2 ldmfd sp!, {r4-r7, pc} END(do_cksum) /* * The main in*_cksum() workhorse... * * Entry parameters: * r0 Pointer to buffer * r1 Buffer length * lr Return address * * Returns: * r2 Accumulated 32-bit sum * * Clobbers: * r0-r7 */ /* LINTSTUB: Ignore */ ASENTRY_NP(L_cksumdata) #ifdef _ARM_ARCH_5E pld [r0] /* Pre-fetch the start of the buffer */ #endif mov r2, #0 /* We first have to word-align the buffer. */ ands r7, r0, #0x03 beq .Lcksumdata_wordaligned rsb r7, r7, #0x04 cmp r1, r7 /* Enough bytes left to make it? */ blt .Lcksumdata_endgame cmp r7, #0x02 ldrb r4, [r0], #0x01 /* Fetch 1st byte */ - ldrgeb r5, [r0], #0x01 /* Fetch 2nd byte */ + ldrbge r5, [r0], #0x01 /* Fetch 2nd byte */ movlt r5, #0x00 - ldrgtb r6, [r0], #0x01 /* Fetch 3rd byte */ + ldrbgt r6, [r0], #0x01 /* Fetch 3rd byte */ movle r6, #0x00 /* Combine the three bytes depending on endianness and alignment */ #ifdef __ARMEB__ orreq r2, r5, r4, lsl #8 orreq r2, r2, r6, lsl #24 orrne r2, r4, r5, lsl #8 orrne r2, r2, r6, lsl #16 #else orreq r2, r4, r5, lsl #8 orreq r2, r2, r6, lsl #16 orrne r2, r5, r4, lsl #8 orrne r2, r2, r6, lsl #24 #endif subs r1, r1, r7 /* Update length */ RETeq /* All done? */ /* Buffer is now word aligned */ .Lcksumdata_wordaligned: #ifdef _ARM_ARCH_5E cmp r1, #0x04 /* Less than 4 bytes left? */ blt .Lcksumdata_endgame /* Yup */ /* Now quad-align, if necessary */ ands r7, r0, #0x04 ldrne r7, [r0], #0x04 subne r1, r1, #0x04 subs r1, r1, #0x40 blt .Lcksumdata_bigloop_end /* Note: C flag clear if branch taken */ /* * Buffer is now quad aligned. Sum 64 bytes at a time. * Note: First ldrd is hoisted above the loop, together with * setting r6 to zero to avoid stalling for results in the * loop. (r7 is live, from above). */ ldrd r4, [r0], #0x08 mov r6, #0x00 .Lcksumdata_bigloop: pld [r0, #0x18] adds r2, r2, r6 adcs r2, r2, r7 ldrd r6, [r0], #0x08 adcs r2, r2, r4 adcs r2, r2, r5 ldrd r4, [r0], #0x08 adcs r2, r2, r6 adcs r2, r2, r7 ldrd r6, [r0], #0x08 adcs r2, r2, r4 adcs r2, r2, r5 ldrd r4, [r0], #0x08 adcs r2, r2, r6 adcs r2, r2, r7 pld [r0, #0x18] ldrd r6, [r0], #0x08 adcs r2, r2, r4 adcs r2, r2, r5 ldrd r4, [r0], #0x08 adcs r2, r2, r6 adcs r2, r2, r7 ldrd r6, [r0], #0x08 adcs r2, r2, r4 adcs r2, r2, r5 adc r2, r2, #0x00 subs r1, r1, #0x40 - ldrged r4, [r0], #0x08 + ldrdge r4, [r0], #0x08 bge .Lcksumdata_bigloop adds r2, r2, r6 /* r6/r7 still need summing */ .Lcksumdata_bigloop_end: adcs r2, r2, r7 adc r2, r2, #0x00 #else /* !_ARM_ARCH_5E */ subs r1, r1, #0x40 blt .Lcksumdata_bigloop_end .Lcksumdata_bigloop: ldmia r0!, {r3, r4, r5, r6} adds r2, r2, r3 adcs r2, r2, r4 adcs r2, r2, r5 ldmia r0!, {r3, r4, r5, r7} adcs r2, r2, r6 adcs r2, r2, r3 adcs r2, r2, r4 adcs r2, r2, r5 ldmia r0!, {r3, r4, r5, r6} adcs r2, r2, r7 adcs r2, r2, r3 adcs r2, r2, r4 adcs r2, r2, r5 ldmia r0!, {r3, r4, r5, r7} adcs r2, r2, r6 adcs r2, r2, r3 adcs r2, r2, r4 adcs r2, r2, r5 adcs r2, r2, r7 adc r2, r2, #0x00 subs r1, r1, #0x40 bge .Lcksumdata_bigloop .Lcksumdata_bigloop_end: #endif adds r1, r1, #0x40 RETeq cmp r1, #0x20 #ifdef _ARM_ARCH_5E - ldrged r4, [r0], #0x08 /* Avoid stalling pld and result */ + ldrdge r4, [r0], #0x08 /* Avoid stalling pld and result */ blt .Lcksumdata_less_than_32 pld [r0, #0x18] ldrd r6, [r0], #0x08 adds r2, r2, r4 adcs r2, r2, r5 ldrd r4, [r0], #0x08 adcs r2, r2, r6 adcs r2, r2, r7 ldrd r6, [r0], #0x08 adcs r2, r2, r4 adcs r2, r2, r5 adcs r2, r2, r6 /* XXX: Unavoidable result stall */ adcs r2, r2, r7 #else blt .Lcksumdata_less_than_32 ldmia r0!, {r3, r4, r5, r6} adds r2, r2, r3 adcs r2, r2, r4 adcs r2, r2, r5 ldmia r0!, {r3, r4, r5, r7} adcs r2, r2, r6 adcs r2, r2, r3 adcs r2, r2, r4 adcs r2, r2, r5 adcs r2, r2, r7 #endif adc r2, r2, #0x00 subs r1, r1, #0x20 RETeq .Lcksumdata_less_than_32: /* There are less than 32 bytes left */ and r3, r1, #0x18 rsb r4, r3, #0x18 sub r1, r1, r3 adds r4, r4, r4, lsr #1 /* Side effect: Clear carry flag */ addne pc, pc, r4 nop /* * Note: We use ldm here, even on armv5e, since the combined issue/result * latencies for ldm and ldrd are the same. Using ldm avoids needless #ifdefs. */ /* At least 24 bytes remaining... */ ldmia r0!, {r4, r5} adcs r2, r2, r4 adcs r2, r2, r5 /* At least 16 bytes remaining... */ ldmia r0!, {r4, r5} adcs r2, r2, r4 adcs r2, r2, r5 /* At least 8 bytes remaining... */ ldmia r0!, {r4, r5} adcs r2, r2, r4 adcs r2, r2, r5 /* Less than 8 bytes remaining... */ adc r2, r2, #0x00 subs r1, r1, #0x04 blt .Lcksumdata_lessthan4 ldr r4, [r0], #0x04 sub r1, r1, #0x04 adds r2, r2, r4 adc r2, r2, #0x00 /* Deal with < 4 bytes remaining */ .Lcksumdata_lessthan4: adds r1, r1, #0x04 RETeq /* Deal with 1 to 3 remaining bytes, possibly misaligned */ .Lcksumdata_endgame: ldrb r3, [r0] /* Fetch first byte */ cmp r1, #0x02 - ldrgeb r4, [r0, #0x01] /* Fetch 2nd and 3rd as necessary */ + ldrbge r4, [r0, #0x01] /* Fetch 2nd and 3rd as necessary */ movlt r4, #0x00 - ldrgtb r5, [r0, #0x02] + ldrbgt r5, [r0, #0x02] movle r5, #0x00 /* Combine the three bytes depending on endianness and alignment */ tst r0, #0x01 #ifdef __ARMEB__ orreq r3, r4, r3, lsl #8 orreq r3, r3, r5, lsl #24 orrne r3, r3, r4, lsl #8 orrne r3, r3, r5, lsl #16 #else orreq r3, r3, r4, lsl #8 orreq r3, r3, r5, lsl #16 orrne r3, r4, r3, lsl #8 orrne r3, r3, r5, lsl #24 #endif adds r2, r2, r3 adc r2, r2, #0x00 RET END(L_cksumdata) Index: head/sys/arm/arm/support.S =================================================================== --- head/sys/arm/arm/support.S (revision 275417) +++ head/sys/arm/arm/support.S (revision 275418) @@ -1,2960 +1,2963 @@ /*- * Copyright (c) 2004 Olivier Houchard * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. 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 AUTHOR 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 AUTHOR 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. */ /* * Copyright 2003 Wasabi Systems, Inc. * All rights reserved. * * Written by Steve C. Woodford for Wasabi Systems, Inc. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. 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. * 3. All advertising materials mentioning features or use of this software * must display the following acknowledgement: * This product includes software developed for the NetBSD Project by * Wasabi Systems, Inc. * 4. The name of Wasabi Systems, Inc. may not be used to endorse * or promote products derived from this software without specific prior * written permission. * * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``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 WASABI SYSTEMS, INC * 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. */ /* * Copyright (c) 1997 The NetBSD Foundation, Inc. * All rights reserved. * * This code is derived from software contributed to The NetBSD Foundation * by Neil A. Carson and Mark Brinicombe * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. 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 NETBSD FOUNDATION, INC. 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 FOUNDATION 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 __FBSDID("$FreeBSD$"); #include "assym.s" + .syntax unified + .L_arm_memcpy: .word _C_LABEL(_arm_memcpy) .L_arm_bzero: .word _C_LABEL(_arm_bzero) .L_min_memcpy_size: .word _C_LABEL(_min_memcpy_size) .L_min_bzero_size: .word _C_LABEL(_min_bzero_size) /* * memset: Sets a block of memory to the specified value * * On entry: * r0 - dest address * r1 - byte to write * r2 - number of bytes to write * * On exit: * r0 - dest address */ /* LINTSTUB: Func: void bzero(void *, size_t) */ ENTRY(bzero) ldr r3, .L_arm_bzero ldr r3, [r3] cmp r3, #0 beq .Lnormal0 ldr r2, .L_min_bzero_size ldr r2, [r2] cmp r1, r2 blt .Lnormal0 stmfd sp!, {r0, r1, lr} mov r2, #0 mov lr, pc mov pc, r3 cmp r0, #0 ldmfd sp!, {r0, r1, lr} RETeq .Lnormal0: mov r3, #0x00 b do_memset END(bzero) /* LINTSTUB: Func: void *memset(void *, int, size_t) */ ENTRY(memset) and r3, r1, #0xff /* We deal with bytes */ mov r1, r2 do_memset: cmp r1, #0x04 /* Do we have less than 4 bytes */ mov ip, r0 blt .Lmemset_lessthanfour /* Ok first we will word align the address */ ands r2, ip, #0x03 /* Get the bottom two bits */ bne .Lmemset_wordunaligned /* The address is not word aligned */ /* We are now word aligned */ .Lmemset_wordaligned: orr r3, r3, r3, lsl #8 /* Extend value to 16-bits */ #ifdef _ARM_ARCH_5E tst ip, #0x04 /* Quad-align for armv5e */ #else cmp r1, #0x10 #endif orr r3, r3, r3, lsl #16 /* Extend value to 32-bits */ #ifdef _ARM_ARCH_5E subne r1, r1, #0x04 /* Quad-align if necessary */ strne r3, [ip], #0x04 cmp r1, #0x10 #endif blt .Lmemset_loop4 /* If less than 16 then use words */ mov r2, r3 /* Duplicate data */ cmp r1, #0x80 /* If < 128 then skip the big loop */ blt .Lmemset_loop32 /* Do 128 bytes at a time */ .Lmemset_loop128: subs r1, r1, #0x80 #ifdef _ARM_ARCH_5E - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 #else - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} #endif bgt .Lmemset_loop128 RETeq /* Zero length so just exit */ add r1, r1, #0x80 /* Adjust for extra sub */ /* Do 32 bytes at a time */ .Lmemset_loop32: subs r1, r1, #0x20 #ifdef _ARM_ARCH_5E - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 #else - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} #endif bgt .Lmemset_loop32 RETeq /* Zero length so just exit */ adds r1, r1, #0x10 /* Partially adjust for extra sub */ /* Deal with 16 bytes or more */ #ifdef _ARM_ARCH_5E - strged r2, [ip], #0x08 - strged r2, [ip], #0x08 + strdge r2, [ip], #0x08 + strdge r2, [ip], #0x08 #else - stmgeia ip!, {r2-r3} - stmgeia ip!, {r2-r3} + stmiage ip!, {r2-r3} + stmiage ip!, {r2-r3} #endif RETeq /* Zero length so just exit */ addlt r1, r1, #0x10 /* Possibly adjust for extra sub */ /* We have at least 4 bytes so copy as words */ .Lmemset_loop4: subs r1, r1, #0x04 strge r3, [ip], #0x04 bgt .Lmemset_loop4 RETeq /* Zero length so just exit */ #ifdef _ARM_ARCH_5E /* Compensate for 64-bit alignment check */ adds r1, r1, #0x04 RETeq cmp r1, #2 #else cmp r1, #-2 #endif strb r3, [ip], #0x01 /* Set 1 byte */ - strgeb r3, [ip], #0x01 /* Set another byte */ - strgtb r3, [ip] /* and a third */ + strbge r3, [ip], #0x01 /* Set another byte */ + strbgt r3, [ip] /* and a third */ RET /* Exit */ .Lmemset_wordunaligned: rsb r2, r2, #0x004 strb r3, [ip], #0x01 /* Set 1 byte */ cmp r2, #0x02 - strgeb r3, [ip], #0x01 /* Set another byte */ + strbge r3, [ip], #0x01 /* Set another byte */ sub r1, r1, r2 - strgtb r3, [ip], #0x01 /* and a third */ + strbgt r3, [ip], #0x01 /* and a third */ cmp r1, #0x04 /* More than 4 bytes left? */ bge .Lmemset_wordaligned /* Yup */ .Lmemset_lessthanfour: cmp r1, #0x00 RETeq /* Zero length so exit */ strb r3, [ip], #0x01 /* Set 1 byte */ cmp r1, #0x02 - strgeb r3, [ip], #0x01 /* Set another byte */ - strgtb r3, [ip] /* and a third */ + strbge r3, [ip], #0x01 /* Set another byte */ + strbgt r3, [ip] /* and a third */ RET /* Exit */ -END(memset) +EEND(memset) +END(bzero) ENTRY(bcmp) mov ip, r0 cmp r2, #0x06 beq .Lmemcmp_6bytes mov r0, #0x00 /* Are both addresses aligned the same way? */ cmp r2, #0x00 - eornes r3, ip, r1 + eorsne r3, ip, r1 RETeq /* len == 0, or same addresses! */ tst r3, #0x03 subne r2, r2, #0x01 bne .Lmemcmp_bytewise2 /* Badly aligned. Do it the slow way */ /* Word-align the addresses, if necessary */ sub r3, r1, #0x05 ands r3, r3, #0x03 add r3, r3, r3, lsl #1 addne pc, pc, r3, lsl #3 nop /* Compare up to 3 bytes */ ldrb r0, [ip], #0x01 ldrb r3, [r1], #0x01 subs r0, r0, r3 RETne subs r2, r2, #0x01 RETeq /* Compare up to 2 bytes */ ldrb r0, [ip], #0x01 ldrb r3, [r1], #0x01 subs r0, r0, r3 RETne subs r2, r2, #0x01 RETeq /* Compare 1 byte */ ldrb r0, [ip], #0x01 ldrb r3, [r1], #0x01 subs r0, r0, r3 RETne subs r2, r2, #0x01 RETeq /* Compare 4 bytes at a time, if possible */ subs r2, r2, #0x04 bcc .Lmemcmp_bytewise .Lmemcmp_word_aligned: ldr r0, [ip], #0x04 ldr r3, [r1], #0x04 subs r2, r2, #0x04 cmpcs r0, r3 beq .Lmemcmp_word_aligned sub r0, r0, r3 /* Correct for extra subtraction, and check if done */ adds r2, r2, #0x04 cmpeq r0, #0x00 /* If done, did all bytes match? */ RETeq /* Yup. Just return */ /* Re-do the final word byte-wise */ sub ip, ip, #0x04 sub r1, r1, #0x04 .Lmemcmp_bytewise: add r2, r2, #0x03 .Lmemcmp_bytewise2: ldrb r0, [ip], #0x01 ldrb r3, [r1], #0x01 subs r2, r2, #0x01 cmpcs r0, r3 beq .Lmemcmp_bytewise2 sub r0, r0, r3 RET /* * 6 byte compares are very common, thanks to the network stack. * This code is hand-scheduled to reduce the number of stalls for * load results. Everything else being equal, this will be ~32% * faster than a byte-wise memcmp. */ .align 5 .Lmemcmp_6bytes: ldrb r3, [r1, #0x00] /* r3 = b2#0 */ ldrb r0, [ip, #0x00] /* r0 = b1#0 */ ldrb r2, [r1, #0x01] /* r2 = b2#1 */ subs r0, r0, r3 /* r0 = b1#0 - b2#0 */ - ldreqb r3, [ip, #0x01] /* r3 = b1#1 */ + ldrbeq r3, [ip, #0x01] /* r3 = b1#1 */ RETne /* Return if mismatch on #0 */ subs r0, r3, r2 /* r0 = b1#1 - b2#1 */ - ldreqb r3, [r1, #0x02] /* r3 = b2#2 */ - ldreqb r0, [ip, #0x02] /* r0 = b1#2 */ + ldrbeq r3, [r1, #0x02] /* r3 = b2#2 */ + ldrbeq r0, [ip, #0x02] /* r0 = b1#2 */ RETne /* Return if mismatch on #1 */ ldrb r2, [r1, #0x03] /* r2 = b2#3 */ subs r0, r0, r3 /* r0 = b1#2 - b2#2 */ - ldreqb r3, [ip, #0x03] /* r3 = b1#3 */ + ldrbeq r3, [ip, #0x03] /* r3 = b1#3 */ RETne /* Return if mismatch on #2 */ subs r0, r3, r2 /* r0 = b1#3 - b2#3 */ - ldreqb r3, [r1, #0x04] /* r3 = b2#4 */ - ldreqb r0, [ip, #0x04] /* r0 = b1#4 */ + ldrbeq r3, [r1, #0x04] /* r3 = b2#4 */ + ldrbeq r0, [ip, #0x04] /* r0 = b1#4 */ RETne /* Return if mismatch on #3 */ ldrb r2, [r1, #0x05] /* r2 = b2#5 */ subs r0, r0, r3 /* r0 = b1#4 - b2#4 */ - ldreqb r3, [ip, #0x05] /* r3 = b1#5 */ + ldrbeq r3, [ip, #0x05] /* r3 = b1#5 */ RETne /* Return if mismatch on #4 */ sub r0, r3, r2 /* r0 = b1#5 - b2#5 */ RET END(bcmp) ENTRY(bcopy) /* switch the source and destination registers */ eor r0, r1, r0 eor r1, r0, r1 eor r0, r1, r0 EENTRY(memmove) /* Do the buffers overlap? */ cmp r0, r1 RETeq /* Bail now if src/dst are the same */ subcc r3, r0, r1 /* if (dst > src) r3 = dst - src */ subcs r3, r1, r0 /* if (src > dsr) r3 = src - dst */ cmp r3, r2 /* if (r3 < len) we have an overlap */ bcc PIC_SYM(_C_LABEL(memcpy), PLT) /* Determine copy direction */ cmp r1, r0 bcc .Lmemmove_backwards moveq r0, #0 /* Quick abort for len=0 */ RETeq stmdb sp!, {r0, lr} /* memmove() returns dest addr */ subs r2, r2, #4 blt .Lmemmove_fl4 /* less than 4 bytes */ ands r12, r0, #3 bne .Lmemmove_fdestul /* oh unaligned destination addr */ ands r12, r1, #3 bne .Lmemmove_fsrcul /* oh unaligned source addr */ .Lmemmove_ft8: /* We have aligned source and destination */ subs r2, r2, #8 blt .Lmemmove_fl12 /* less than 12 bytes (4 from above) */ subs r2, r2, #0x14 blt .Lmemmove_fl32 /* less than 32 bytes (12 from above) */ stmdb sp!, {r4} /* borrow r4 */ /* blat 32 bytes at a time */ /* XXX for really big copies perhaps we should use more registers */ .Lmemmove_floop32: ldmia r1!, {r3, r4, r12, lr} stmia r0!, {r3, r4, r12, lr} ldmia r1!, {r3, r4, r12, lr} stmia r0!, {r3, r4, r12, lr} subs r2, r2, #0x20 bge .Lmemmove_floop32 cmn r2, #0x10 - ldmgeia r1!, {r3, r4, r12, lr} /* blat a remaining 16 bytes */ - stmgeia r0!, {r3, r4, r12, lr} + ldmiage r1!, {r3, r4, r12, lr} /* blat a remaining 16 bytes */ + stmiage r0!, {r3, r4, r12, lr} subge r2, r2, #0x10 ldmia sp!, {r4} /* return r4 */ .Lmemmove_fl32: adds r2, r2, #0x14 /* blat 12 bytes at a time */ .Lmemmove_floop12: - ldmgeia r1!, {r3, r12, lr} - stmgeia r0!, {r3, r12, lr} - subges r2, r2, #0x0c + ldmiage r1!, {r3, r12, lr} + stmiage r0!, {r3, r12, lr} + subsge r2, r2, #0x0c bge .Lmemmove_floop12 .Lmemmove_fl12: adds r2, r2, #8 blt .Lmemmove_fl4 subs r2, r2, #4 ldrlt r3, [r1], #4 strlt r3, [r0], #4 - ldmgeia r1!, {r3, r12} - stmgeia r0!, {r3, r12} + ldmiage r1!, {r3, r12} + stmiage r0!, {r3, r12} subge r2, r2, #4 .Lmemmove_fl4: /* less than 4 bytes to go */ adds r2, r2, #4 - ldmeqia sp!, {r0, pc} /* done */ + ldmiaeq sp!, {r0, pc} /* done */ /* copy the crud byte at a time */ cmp r2, #2 ldrb r3, [r1], #1 strb r3, [r0], #1 - ldrgeb r3, [r1], #1 - strgeb r3, [r0], #1 - ldrgtb r3, [r1], #1 - strgtb r3, [r0], #1 + ldrbge r3, [r1], #1 + strbge r3, [r0], #1 + ldrbgt r3, [r1], #1 + strbgt r3, [r0], #1 ldmia sp!, {r0, pc} /* erg - unaligned destination */ .Lmemmove_fdestul: rsb r12, r12, #4 cmp r12, #2 /* align destination with byte copies */ ldrb r3, [r1], #1 strb r3, [r0], #1 - ldrgeb r3, [r1], #1 - strgeb r3, [r0], #1 - ldrgtb r3, [r1], #1 - strgtb r3, [r0], #1 + ldrbge r3, [r1], #1 + strbge r3, [r0], #1 + ldrbgt r3, [r1], #1 + strbgt r3, [r0], #1 subs r2, r2, r12 blt .Lmemmove_fl4 /* less the 4 bytes */ ands r12, r1, #3 beq .Lmemmove_ft8 /* we have an aligned source */ /* erg - unaligned source */ /* This is where it gets nasty ... */ .Lmemmove_fsrcul: bic r1, r1, #3 ldr lr, [r1], #4 cmp r12, #2 bgt .Lmemmove_fsrcul3 beq .Lmemmove_fsrcul2 cmp r2, #0x0c blt .Lmemmove_fsrcul1loop4 sub r2, r2, #0x0c stmdb sp!, {r4, r5} .Lmemmove_fsrcul1loop16: #ifdef __ARMEB__ mov r3, lr, lsl #8 #else mov r3, lr, lsr #8 #endif ldmia r1!, {r4, r5, r12, lr} #ifdef __ARMEB__ orr r3, r3, r4, lsr #24 mov r4, r4, lsl #8 orr r4, r4, r5, lsr #24 mov r5, r5, lsl #8 orr r5, r5, r12, lsr #24 mov r12, r12, lsl #8 orr r12, r12, lr, lsr #24 #else orr r3, r3, r4, lsl #24 mov r4, r4, lsr #8 orr r4, r4, r5, lsl #24 mov r5, r5, lsr #8 orr r5, r5, r12, lsl #24 mov r12, r12, lsr #8 orr r12, r12, lr, lsl #24 #endif stmia r0!, {r3-r5, r12} subs r2, r2, #0x10 bge .Lmemmove_fsrcul1loop16 ldmia sp!, {r4, r5} adds r2, r2, #0x0c blt .Lmemmove_fsrcul1l4 .Lmemmove_fsrcul1loop4: #ifdef __ARMEB__ mov r12, lr, lsl #8 #else mov r12, lr, lsr #8 #endif ldr lr, [r1], #4 #ifdef __ARMEB__ orr r12, r12, lr, lsr #24 #else orr r12, r12, lr, lsl #24 #endif str r12, [r0], #4 subs r2, r2, #4 bge .Lmemmove_fsrcul1loop4 .Lmemmove_fsrcul1l4: sub r1, r1, #3 b .Lmemmove_fl4 .Lmemmove_fsrcul2: cmp r2, #0x0c blt .Lmemmove_fsrcul2loop4 sub r2, r2, #0x0c stmdb sp!, {r4, r5} .Lmemmove_fsrcul2loop16: #ifdef __ARMEB__ mov r3, lr, lsl #16 #else mov r3, lr, lsr #16 #endif ldmia r1!, {r4, r5, r12, lr} #ifdef __ARMEB__ orr r3, r3, r4, lsr #16 mov r4, r4, lsl #16 orr r4, r4, r5, lsr #16 mov r5, r5, lsl #16 orr r5, r5, r12, lsr #16 mov r12, r12, lsl #16 orr r12, r12, lr, lsr #16 #else orr r3, r3, r4, lsl #16 mov r4, r4, lsr #16 orr r4, r4, r5, lsl #16 mov r5, r5, lsr #16 orr r5, r5, r12, lsl #16 mov r12, r12, lsr #16 orr r12, r12, lr, lsl #16 #endif stmia r0!, {r3-r5, r12} subs r2, r2, #0x10 bge .Lmemmove_fsrcul2loop16 ldmia sp!, {r4, r5} adds r2, r2, #0x0c blt .Lmemmove_fsrcul2l4 .Lmemmove_fsrcul2loop4: #ifdef __ARMEB__ mov r12, lr, lsl #16 #else mov r12, lr, lsr #16 #endif ldr lr, [r1], #4 #ifdef __ARMEB__ orr r12, r12, lr, lsr #16 #else orr r12, r12, lr, lsl #16 #endif str r12, [r0], #4 subs r2, r2, #4 bge .Lmemmove_fsrcul2loop4 .Lmemmove_fsrcul2l4: sub r1, r1, #2 b .Lmemmove_fl4 .Lmemmove_fsrcul3: cmp r2, #0x0c blt .Lmemmove_fsrcul3loop4 sub r2, r2, #0x0c stmdb sp!, {r4, r5} .Lmemmove_fsrcul3loop16: #ifdef __ARMEB__ mov r3, lr, lsl #24 #else mov r3, lr, lsr #24 #endif ldmia r1!, {r4, r5, r12, lr} #ifdef __ARMEB__ orr r3, r3, r4, lsr #8 mov r4, r4, lsl #24 orr r4, r4, r5, lsr #8 mov r5, r5, lsl #24 orr r5, r5, r12, lsr #8 mov r12, r12, lsl #24 orr r12, r12, lr, lsr #8 #else orr r3, r3, r4, lsl #8 mov r4, r4, lsr #24 orr r4, r4, r5, lsl #8 mov r5, r5, lsr #24 orr r5, r5, r12, lsl #8 mov r12, r12, lsr #24 orr r12, r12, lr, lsl #8 #endif stmia r0!, {r3-r5, r12} subs r2, r2, #0x10 bge .Lmemmove_fsrcul3loop16 ldmia sp!, {r4, r5} adds r2, r2, #0x0c blt .Lmemmove_fsrcul3l4 .Lmemmove_fsrcul3loop4: #ifdef __ARMEB__ mov r12, lr, lsl #24 #else mov r12, lr, lsr #24 #endif ldr lr, [r1], #4 #ifdef __ARMEB__ orr r12, r12, lr, lsr #8 #else orr r12, r12, lr, lsl #8 #endif str r12, [r0], #4 subs r2, r2, #4 bge .Lmemmove_fsrcul3loop4 .Lmemmove_fsrcul3l4: sub r1, r1, #1 b .Lmemmove_fl4 .Lmemmove_backwards: add r1, r1, r2 add r0, r0, r2 subs r2, r2, #4 blt .Lmemmove_bl4 /* less than 4 bytes */ ands r12, r0, #3 bne .Lmemmove_bdestul /* oh unaligned destination addr */ ands r12, r1, #3 bne .Lmemmove_bsrcul /* oh unaligned source addr */ .Lmemmove_bt8: /* We have aligned source and destination */ subs r2, r2, #8 blt .Lmemmove_bl12 /* less than 12 bytes (4 from above) */ stmdb sp!, {r4, lr} subs r2, r2, #0x14 /* less than 32 bytes (12 from above) */ blt .Lmemmove_bl32 /* blat 32 bytes at a time */ /* XXX for really big copies perhaps we should use more registers */ .Lmemmove_bloop32: ldmdb r1!, {r3, r4, r12, lr} stmdb r0!, {r3, r4, r12, lr} ldmdb r1!, {r3, r4, r12, lr} stmdb r0!, {r3, r4, r12, lr} subs r2, r2, #0x20 bge .Lmemmove_bloop32 .Lmemmove_bl32: cmn r2, #0x10 - ldmgedb r1!, {r3, r4, r12, lr} /* blat a remaining 16 bytes */ - stmgedb r0!, {r3, r4, r12, lr} + ldmdbge r1!, {r3, r4, r12, lr} /* blat a remaining 16 bytes */ + stmdbge r0!, {r3, r4, r12, lr} subge r2, r2, #0x10 adds r2, r2, #0x14 - ldmgedb r1!, {r3, r12, lr} /* blat a remaining 12 bytes */ - stmgedb r0!, {r3, r12, lr} + ldmdbge r1!, {r3, r12, lr} /* blat a remaining 12 bytes */ + stmdbge r0!, {r3, r12, lr} subge r2, r2, #0x0c ldmia sp!, {r4, lr} .Lmemmove_bl12: adds r2, r2, #8 blt .Lmemmove_bl4 subs r2, r2, #4 ldrlt r3, [r1, #-4]! strlt r3, [r0, #-4]! - ldmgedb r1!, {r3, r12} - stmgedb r0!, {r3, r12} + ldmdbge r1!, {r3, r12} + stmdbge r0!, {r3, r12} subge r2, r2, #4 .Lmemmove_bl4: /* less than 4 bytes to go */ adds r2, r2, #4 RETeq /* done */ /* copy the crud byte at a time */ cmp r2, #2 ldrb r3, [r1, #-1]! strb r3, [r0, #-1]! - ldrgeb r3, [r1, #-1]! - strgeb r3, [r0, #-1]! - ldrgtb r3, [r1, #-1]! - strgtb r3, [r0, #-1]! + ldrbge r3, [r1, #-1]! + strbge r3, [r0, #-1]! + ldrbgt r3, [r1, #-1]! + strbgt r3, [r0, #-1]! RET /* erg - unaligned destination */ .Lmemmove_bdestul: cmp r12, #2 /* align destination with byte copies */ ldrb r3, [r1, #-1]! strb r3, [r0, #-1]! - ldrgeb r3, [r1, #-1]! - strgeb r3, [r0, #-1]! - ldrgtb r3, [r1, #-1]! - strgtb r3, [r0, #-1]! + ldrbge r3, [r1, #-1]! + strbge r3, [r0, #-1]! + ldrbgt r3, [r1, #-1]! + strbgt r3, [r0, #-1]! subs r2, r2, r12 blt .Lmemmove_bl4 /* less than 4 bytes to go */ ands r12, r1, #3 beq .Lmemmove_bt8 /* we have an aligned source */ /* erg - unaligned source */ /* This is where it gets nasty ... */ .Lmemmove_bsrcul: bic r1, r1, #3 ldr r3, [r1, #0] cmp r12, #2 blt .Lmemmove_bsrcul1 beq .Lmemmove_bsrcul2 cmp r2, #0x0c blt .Lmemmove_bsrcul3loop4 sub r2, r2, #0x0c stmdb sp!, {r4, r5, lr} .Lmemmove_bsrcul3loop16: #ifdef __ARMEB__ mov lr, r3, lsr #8 #else mov lr, r3, lsl #8 #endif ldmdb r1!, {r3-r5, r12} #ifdef __ARMEB__ orr lr, lr, r12, lsl #24 mov r12, r12, lsr #8 orr r12, r12, r5, lsl #24 mov r5, r5, lsr #8 orr r5, r5, r4, lsl #24 mov r4, r4, lsr #8 orr r4, r4, r3, lsl #24 #else orr lr, lr, r12, lsr #24 mov r12, r12, lsl #8 orr r12, r12, r5, lsr #24 mov r5, r5, lsl #8 orr r5, r5, r4, lsr #24 mov r4, r4, lsl #8 orr r4, r4, r3, lsr #24 #endif stmdb r0!, {r4, r5, r12, lr} subs r2, r2, #0x10 bge .Lmemmove_bsrcul3loop16 ldmia sp!, {r4, r5, lr} adds r2, r2, #0x0c blt .Lmemmove_bsrcul3l4 .Lmemmove_bsrcul3loop4: #ifdef __ARMEB__ mov r12, r3, lsr #8 #else mov r12, r3, lsl #8 #endif ldr r3, [r1, #-4]! #ifdef __ARMEB__ orr r12, r12, r3, lsl #24 #else orr r12, r12, r3, lsr #24 #endif str r12, [r0, #-4]! subs r2, r2, #4 bge .Lmemmove_bsrcul3loop4 .Lmemmove_bsrcul3l4: add r1, r1, #3 b .Lmemmove_bl4 .Lmemmove_bsrcul2: cmp r2, #0x0c blt .Lmemmove_bsrcul2loop4 sub r2, r2, #0x0c stmdb sp!, {r4, r5, lr} .Lmemmove_bsrcul2loop16: #ifdef __ARMEB__ mov lr, r3, lsr #16 #else mov lr, r3, lsl #16 #endif ldmdb r1!, {r3-r5, r12} #ifdef __ARMEB__ orr lr, lr, r12, lsl #16 mov r12, r12, lsr #16 orr r12, r12, r5, lsl #16 mov r5, r5, lsr #16 orr r5, r5, r4, lsl #16 mov r4, r4, lsr #16 orr r4, r4, r3, lsl #16 #else orr lr, lr, r12, lsr #16 mov r12, r12, lsl #16 orr r12, r12, r5, lsr #16 mov r5, r5, lsl #16 orr r5, r5, r4, lsr #16 mov r4, r4, lsl #16 orr r4, r4, r3, lsr #16 #endif stmdb r0!, {r4, r5, r12, lr} subs r2, r2, #0x10 bge .Lmemmove_bsrcul2loop16 ldmia sp!, {r4, r5, lr} adds r2, r2, #0x0c blt .Lmemmove_bsrcul2l4 .Lmemmove_bsrcul2loop4: #ifdef __ARMEB__ mov r12, r3, lsr #16 #else mov r12, r3, lsl #16 #endif ldr r3, [r1, #-4]! #ifdef __ARMEB__ orr r12, r12, r3, lsl #16 #else orr r12, r12, r3, lsr #16 #endif str r12, [r0, #-4]! subs r2, r2, #4 bge .Lmemmove_bsrcul2loop4 .Lmemmove_bsrcul2l4: add r1, r1, #2 b .Lmemmove_bl4 .Lmemmove_bsrcul1: cmp r2, #0x0c blt .Lmemmove_bsrcul1loop4 sub r2, r2, #0x0c stmdb sp!, {r4, r5, lr} .Lmemmove_bsrcul1loop32: #ifdef __ARMEB__ mov lr, r3, lsr #24 #else mov lr, r3, lsl #24 #endif ldmdb r1!, {r3-r5, r12} #ifdef __ARMEB__ orr lr, lr, r12, lsl #8 mov r12, r12, lsr #24 orr r12, r12, r5, lsl #8 mov r5, r5, lsr #24 orr r5, r5, r4, lsl #8 mov r4, r4, lsr #24 orr r4, r4, r3, lsl #8 #else orr lr, lr, r12, lsr #8 mov r12, r12, lsl #24 orr r12, r12, r5, lsr #8 mov r5, r5, lsl #24 orr r5, r5, r4, lsr #8 mov r4, r4, lsl #24 orr r4, r4, r3, lsr #8 #endif stmdb r0!, {r4, r5, r12, lr} subs r2, r2, #0x10 bge .Lmemmove_bsrcul1loop32 ldmia sp!, {r4, r5, lr} adds r2, r2, #0x0c blt .Lmemmove_bsrcul1l4 .Lmemmove_bsrcul1loop4: #ifdef __ARMEB__ mov r12, r3, lsr #24 #else mov r12, r3, lsl #24 #endif ldr r3, [r1, #-4]! #ifdef __ARMEB__ orr r12, r12, r3, lsl #8 #else orr r12, r12, r3, lsr #8 #endif str r12, [r0, #-4]! subs r2, r2, #4 bge .Lmemmove_bsrcul1loop4 .Lmemmove_bsrcul1l4: add r1, r1, #1 b .Lmemmove_bl4 EEND(memmove) END(bcopy) #if !defined(_ARM_ARCH_5E) ENTRY(memcpy) /* save leaf functions having to store this away */ /* Do not check arm_memcpy if we're running from flash */ #if defined(FLASHADDR) && defined(PHYSADDR) #if FLASHADDR > PHYSADDR ldr r3, =FLASHADDR cmp r3, pc bls .Lnormal #else ldr r3, =FLASHADDR cmp r3, pc bhi .Lnormal #endif #endif ldr r3, .L_arm_memcpy ldr r3, [r3] cmp r3, #0 beq .Lnormal ldr r3, .L_min_memcpy_size ldr r3, [r3] cmp r2, r3 blt .Lnormal stmfd sp!, {r0-r2, r4, lr} mov r3, #0 ldr r4, .L_arm_memcpy mov lr, pc ldr pc, [r4] cmp r0, #0 ldmfd sp!, {r0-r2, r4, lr} RETeq .Lnormal: stmdb sp!, {r0, lr} /* memcpy() returns dest addr */ subs r2, r2, #4 blt .Lmemcpy_l4 /* less than 4 bytes */ ands r12, r0, #3 bne .Lmemcpy_destul /* oh unaligned destination addr */ ands r12, r1, #3 bne .Lmemcpy_srcul /* oh unaligned source addr */ .Lmemcpy_t8: /* We have aligned source and destination */ subs r2, r2, #8 blt .Lmemcpy_l12 /* less than 12 bytes (4 from above) */ subs r2, r2, #0x14 blt .Lmemcpy_l32 /* less than 32 bytes (12 from above) */ stmdb sp!, {r4} /* borrow r4 */ /* blat 32 bytes at a time */ /* XXX for really big copies perhaps we should use more registers */ .Lmemcpy_loop32: ldmia r1!, {r3, r4, r12, lr} stmia r0!, {r3, r4, r12, lr} ldmia r1!, {r3, r4, r12, lr} stmia r0!, {r3, r4, r12, lr} subs r2, r2, #0x20 bge .Lmemcpy_loop32 cmn r2, #0x10 - ldmgeia r1!, {r3, r4, r12, lr} /* blat a remaining 16 bytes */ - stmgeia r0!, {r3, r4, r12, lr} + ldmiage r1!, {r3, r4, r12, lr} /* blat a remaining 16 bytes */ + stmiage r0!, {r3, r4, r12, lr} subge r2, r2, #0x10 ldmia sp!, {r4} /* return r4 */ .Lmemcpy_l32: adds r2, r2, #0x14 /* blat 12 bytes at a time */ .Lmemcpy_loop12: - ldmgeia r1!, {r3, r12, lr} - stmgeia r0!, {r3, r12, lr} - subges r2, r2, #0x0c + ldmiage r1!, {r3, r12, lr} + stmiage r0!, {r3, r12, lr} + subsge r2, r2, #0x0c bge .Lmemcpy_loop12 .Lmemcpy_l12: adds r2, r2, #8 blt .Lmemcpy_l4 subs r2, r2, #4 ldrlt r3, [r1], #4 strlt r3, [r0], #4 - ldmgeia r1!, {r3, r12} - stmgeia r0!, {r3, r12} + ldmiage r1!, {r3, r12} + stmiage r0!, {r3, r12} subge r2, r2, #4 .Lmemcpy_l4: /* less than 4 bytes to go */ adds r2, r2, #4 #ifdef __APCS_26_ - ldmeqia sp!, {r0, pc}^ /* done */ + ldmiaeq sp!, {r0, pc}^ /* done */ #else - ldmeqia sp!, {r0, pc} /* done */ + ldmiaeq sp!, {r0, pc} /* done */ #endif /* copy the crud byte at a time */ cmp r2, #2 ldrb r3, [r1], #1 strb r3, [r0], #1 - ldrgeb r3, [r1], #1 - strgeb r3, [r0], #1 - ldrgtb r3, [r1], #1 - strgtb r3, [r0], #1 + ldrbge r3, [r1], #1 + strbge r3, [r0], #1 + ldrbgt r3, [r1], #1 + strbgt r3, [r0], #1 ldmia sp!, {r0, pc} /* erg - unaligned destination */ .Lmemcpy_destul: rsb r12, r12, #4 cmp r12, #2 /* align destination with byte copies */ ldrb r3, [r1], #1 strb r3, [r0], #1 - ldrgeb r3, [r1], #1 - strgeb r3, [r0], #1 - ldrgtb r3, [r1], #1 - strgtb r3, [r0], #1 + ldrbge r3, [r1], #1 + strbge r3, [r0], #1 + ldrbgt r3, [r1], #1 + strbgt r3, [r0], #1 subs r2, r2, r12 blt .Lmemcpy_l4 /* less the 4 bytes */ ands r12, r1, #3 beq .Lmemcpy_t8 /* we have an aligned source */ /* erg - unaligned source */ /* This is where it gets nasty ... */ .Lmemcpy_srcul: bic r1, r1, #3 ldr lr, [r1], #4 cmp r12, #2 bgt .Lmemcpy_srcul3 beq .Lmemcpy_srcul2 cmp r2, #0x0c blt .Lmemcpy_srcul1loop4 sub r2, r2, #0x0c stmdb sp!, {r4, r5} .Lmemcpy_srcul1loop16: mov r3, lr, lsr #8 ldmia r1!, {r4, r5, r12, lr} orr r3, r3, r4, lsl #24 mov r4, r4, lsr #8 orr r4, r4, r5, lsl #24 mov r5, r5, lsr #8 orr r5, r5, r12, lsl #24 mov r12, r12, lsr #8 orr r12, r12, lr, lsl #24 stmia r0!, {r3-r5, r12} subs r2, r2, #0x10 bge .Lmemcpy_srcul1loop16 ldmia sp!, {r4, r5} adds r2, r2, #0x0c blt .Lmemcpy_srcul1l4 .Lmemcpy_srcul1loop4: mov r12, lr, lsr #8 ldr lr, [r1], #4 orr r12, r12, lr, lsl #24 str r12, [r0], #4 subs r2, r2, #4 bge .Lmemcpy_srcul1loop4 .Lmemcpy_srcul1l4: sub r1, r1, #3 b .Lmemcpy_l4 .Lmemcpy_srcul2: cmp r2, #0x0c blt .Lmemcpy_srcul2loop4 sub r2, r2, #0x0c stmdb sp!, {r4, r5} .Lmemcpy_srcul2loop16: mov r3, lr, lsr #16 ldmia r1!, {r4, r5, r12, lr} orr r3, r3, r4, lsl #16 mov r4, r4, lsr #16 orr r4, r4, r5, lsl #16 mov r5, r5, lsr #16 orr r5, r5, r12, lsl #16 mov r12, r12, lsr #16 orr r12, r12, lr, lsl #16 stmia r0!, {r3-r5, r12} subs r2, r2, #0x10 bge .Lmemcpy_srcul2loop16 ldmia sp!, {r4, r5} adds r2, r2, #0x0c blt .Lmemcpy_srcul2l4 .Lmemcpy_srcul2loop4: mov r12, lr, lsr #16 ldr lr, [r1], #4 orr r12, r12, lr, lsl #16 str r12, [r0], #4 subs r2, r2, #4 bge .Lmemcpy_srcul2loop4 .Lmemcpy_srcul2l4: sub r1, r1, #2 b .Lmemcpy_l4 .Lmemcpy_srcul3: cmp r2, #0x0c blt .Lmemcpy_srcul3loop4 sub r2, r2, #0x0c stmdb sp!, {r4, r5} .Lmemcpy_srcul3loop16: mov r3, lr, lsr #24 ldmia r1!, {r4, r5, r12, lr} orr r3, r3, r4, lsl #8 mov r4, r4, lsr #24 orr r4, r4, r5, lsl #8 mov r5, r5, lsr #24 orr r5, r5, r12, lsl #8 mov r12, r12, lsr #24 orr r12, r12, lr, lsl #8 stmia r0!, {r3-r5, r12} subs r2, r2, #0x10 bge .Lmemcpy_srcul3loop16 ldmia sp!, {r4, r5} adds r2, r2, #0x0c blt .Lmemcpy_srcul3l4 .Lmemcpy_srcul3loop4: mov r12, lr, lsr #24 ldr lr, [r1], #4 orr r12, r12, lr, lsl #8 str r12, [r0], #4 subs r2, r2, #4 bge .Lmemcpy_srcul3loop4 .Lmemcpy_srcul3l4: sub r1, r1, #1 b .Lmemcpy_l4 END(memcpy) #else /* LINTSTUB: Func: void *memcpy(void *dst, const void *src, size_t len) */ ENTRY(memcpy) pld [r1] cmp r2, #0x0c ble .Lmemcpy_short /* <= 12 bytes */ #ifdef FLASHADDR #if FLASHADDR > PHYSADDR ldr r3, =FLASHADDR cmp r3, pc bls .Lnormal #else ldr r3, =FLASHADDR cmp r3, pc bhi .Lnormal #endif #endif ldr r3, .L_arm_memcpy ldr r3, [r3] cmp r3, #0 beq .Lnormal ldr r3, .L_min_memcpy_size ldr r3, [r3] cmp r2, r3 blt .Lnormal stmfd sp!, {r0-r2, r4, lr} mov r3, #0 ldr r4, .L_arm_memcpy mov lr, pc ldr pc, [r4] cmp r0, #0 ldmfd sp!, {r0-r2, r4, lr} RETeq .Lnormal: mov r3, r0 /* We must not clobber r0 */ /* Word-align the destination buffer */ ands ip, r3, #0x03 /* Already word aligned? */ beq .Lmemcpy_wordaligned /* Yup */ cmp ip, #0x02 ldrb ip, [r1], #0x01 sub r2, r2, #0x01 strb ip, [r3], #0x01 - ldrleb ip, [r1], #0x01 + ldrble ip, [r1], #0x01 suble r2, r2, #0x01 - strleb ip, [r3], #0x01 - ldrltb ip, [r1], #0x01 + strble ip, [r3], #0x01 + ldrblt ip, [r1], #0x01 sublt r2, r2, #0x01 - strltb ip, [r3], #0x01 + strblt ip, [r3], #0x01 /* Destination buffer is now word aligned */ .Lmemcpy_wordaligned: ands ip, r1, #0x03 /* Is src also word-aligned? */ bne .Lmemcpy_bad_align /* Nope. Things just got bad */ /* Quad-align the destination buffer */ tst r3, #0x07 /* Already quad aligned? */ ldrne ip, [r1], #0x04 stmfd sp!, {r4-r9} /* Free up some registers */ subne r2, r2, #0x04 strne ip, [r3], #0x04 /* Destination buffer quad aligned, source is at least word aligned */ subs r2, r2, #0x80 blt .Lmemcpy_w_lessthan128 /* Copy 128 bytes at a time */ .Lmemcpy_w_loop128: ldr r4, [r1], #0x04 /* LD:00-03 */ ldr r5, [r1], #0x04 /* LD:04-07 */ pld [r1, #0x18] /* Prefetch 0x20 */ ldr r6, [r1], #0x04 /* LD:08-0b */ ldr r7, [r1], #0x04 /* LD:0c-0f */ ldr r8, [r1], #0x04 /* LD:10-13 */ ldr r9, [r1], #0x04 /* LD:14-17 */ strd r4, [r3], #0x08 /* ST:00-07 */ ldr r4, [r1], #0x04 /* LD:18-1b */ ldr r5, [r1], #0x04 /* LD:1c-1f */ strd r6, [r3], #0x08 /* ST:08-0f */ ldr r6, [r1], #0x04 /* LD:20-23 */ ldr r7, [r1], #0x04 /* LD:24-27 */ pld [r1, #0x18] /* Prefetch 0x40 */ strd r8, [r3], #0x08 /* ST:10-17 */ ldr r8, [r1], #0x04 /* LD:28-2b */ ldr r9, [r1], #0x04 /* LD:2c-2f */ strd r4, [r3], #0x08 /* ST:18-1f */ ldr r4, [r1], #0x04 /* LD:30-33 */ ldr r5, [r1], #0x04 /* LD:34-37 */ strd r6, [r3], #0x08 /* ST:20-27 */ ldr r6, [r1], #0x04 /* LD:38-3b */ ldr r7, [r1], #0x04 /* LD:3c-3f */ strd r8, [r3], #0x08 /* ST:28-2f */ ldr r8, [r1], #0x04 /* LD:40-43 */ ldr r9, [r1], #0x04 /* LD:44-47 */ pld [r1, #0x18] /* Prefetch 0x60 */ strd r4, [r3], #0x08 /* ST:30-37 */ ldr r4, [r1], #0x04 /* LD:48-4b */ ldr r5, [r1], #0x04 /* LD:4c-4f */ strd r6, [r3], #0x08 /* ST:38-3f */ ldr r6, [r1], #0x04 /* LD:50-53 */ ldr r7, [r1], #0x04 /* LD:54-57 */ strd r8, [r3], #0x08 /* ST:40-47 */ ldr r8, [r1], #0x04 /* LD:58-5b */ ldr r9, [r1], #0x04 /* LD:5c-5f */ strd r4, [r3], #0x08 /* ST:48-4f */ ldr r4, [r1], #0x04 /* LD:60-63 */ ldr r5, [r1], #0x04 /* LD:64-67 */ pld [r1, #0x18] /* Prefetch 0x80 */ strd r6, [r3], #0x08 /* ST:50-57 */ ldr r6, [r1], #0x04 /* LD:68-6b */ ldr r7, [r1], #0x04 /* LD:6c-6f */ strd r8, [r3], #0x08 /* ST:58-5f */ ldr r8, [r1], #0x04 /* LD:70-73 */ ldr r9, [r1], #0x04 /* LD:74-77 */ strd r4, [r3], #0x08 /* ST:60-67 */ ldr r4, [r1], #0x04 /* LD:78-7b */ ldr r5, [r1], #0x04 /* LD:7c-7f */ strd r6, [r3], #0x08 /* ST:68-6f */ strd r8, [r3], #0x08 /* ST:70-77 */ subs r2, r2, #0x80 strd r4, [r3], #0x08 /* ST:78-7f */ bge .Lmemcpy_w_loop128 .Lmemcpy_w_lessthan128: adds r2, r2, #0x80 /* Adjust for extra sub */ - ldmeqfd sp!, {r4-r9} + ldmfdeq sp!, {r4-r9} RETeq /* Return now if done */ subs r2, r2, #0x20 blt .Lmemcpy_w_lessthan32 /* Copy 32 bytes at a time */ .Lmemcpy_w_loop32: ldr r4, [r1], #0x04 ldr r5, [r1], #0x04 pld [r1, #0x18] ldr r6, [r1], #0x04 ldr r7, [r1], #0x04 ldr r8, [r1], #0x04 ldr r9, [r1], #0x04 strd r4, [r3], #0x08 ldr r4, [r1], #0x04 ldr r5, [r1], #0x04 strd r6, [r3], #0x08 strd r8, [r3], #0x08 subs r2, r2, #0x20 strd r4, [r3], #0x08 bge .Lmemcpy_w_loop32 .Lmemcpy_w_lessthan32: adds r2, r2, #0x20 /* Adjust for extra sub */ - ldmeqfd sp!, {r4-r9} + ldmfdeq sp!, {r4-r9} RETeq /* Return now if done */ and r4, r2, #0x18 rsbs r4, r4, #0x18 addne pc, pc, r4, lsl #1 nop /* At least 24 bytes remaining */ ldr r4, [r1], #0x04 ldr r5, [r1], #0x04 sub r2, r2, #0x08 strd r4, [r3], #0x08 /* At least 16 bytes remaining */ ldr r4, [r1], #0x04 ldr r5, [r1], #0x04 sub r2, r2, #0x08 strd r4, [r3], #0x08 /* At least 8 bytes remaining */ ldr r4, [r1], #0x04 ldr r5, [r1], #0x04 subs r2, r2, #0x08 strd r4, [r3], #0x08 /* Less than 8 bytes remaining */ ldmfd sp!, {r4-r9} RETeq /* Return now if done */ subs r2, r2, #0x04 ldrge ip, [r1], #0x04 strge ip, [r3], #0x04 RETeq /* Return now if done */ addlt r2, r2, #0x04 ldrb ip, [r1], #0x01 cmp r2, #0x02 - ldrgeb r2, [r1], #0x01 + ldrbge r2, [r1], #0x01 strb ip, [r3], #0x01 - ldrgtb ip, [r1] - strgeb r2, [r3], #0x01 - strgtb ip, [r3] + ldrbgt ip, [r1] + strbge r2, [r3], #0x01 + strbgt ip, [r3] RET /* * At this point, it has not been possible to word align both buffers. * The destination buffer is word aligned, but the source buffer is not. */ .Lmemcpy_bad_align: stmfd sp!, {r4-r7} bic r1, r1, #0x03 cmp ip, #2 ldr ip, [r1], #0x04 bgt .Lmemcpy_bad3 beq .Lmemcpy_bad2 b .Lmemcpy_bad1 .Lmemcpy_bad1_loop16: #ifdef __ARMEB__ mov r4, ip, lsl #8 #else mov r4, ip, lsr #8 #endif ldr r5, [r1], #0x04 pld [r1, #0x018] ldr r6, [r1], #0x04 ldr r7, [r1], #0x04 ldr ip, [r1], #0x04 #ifdef __ARMEB__ orr r4, r4, r5, lsr #24 mov r5, r5, lsl #8 orr r5, r5, r6, lsr #24 mov r6, r6, lsl #8 orr r6, r6, r7, lsr #24 mov r7, r7, lsl #8 orr r7, r7, ip, lsr #24 #else orr r4, r4, r5, lsl #24 mov r5, r5, lsr #8 orr r5, r5, r6, lsl #24 mov r6, r6, lsr #8 orr r6, r6, r7, lsl #24 mov r7, r7, lsr #8 orr r7, r7, ip, lsl #24 #endif str r4, [r3], #0x04 str r5, [r3], #0x04 str r6, [r3], #0x04 str r7, [r3], #0x04 .Lmemcpy_bad1: subs r2, r2, #0x10 bge .Lmemcpy_bad1_loop16 adds r2, r2, #0x10 - ldmeqfd sp!, {r4-r7} + ldmfdeq sp!, {r4-r7} RETeq /* Return now if done */ subs r2, r2, #0x04 sublt r1, r1, #0x03 blt .Lmemcpy_bad_done .Lmemcpy_bad1_loop4: #ifdef __ARMEB__ mov r4, ip, lsl #8 #else mov r4, ip, lsr #8 #endif ldr ip, [r1], #0x04 subs r2, r2, #0x04 #ifdef __ARMEB__ orr r4, r4, ip, lsr #24 #else orr r4, r4, ip, lsl #24 #endif str r4, [r3], #0x04 bge .Lmemcpy_bad1_loop4 sub r1, r1, #0x03 b .Lmemcpy_bad_done .Lmemcpy_bad2_loop16: #ifdef __ARMEB__ mov r4, ip, lsl #16 #else mov r4, ip, lsr #16 #endif ldr r5, [r1], #0x04 pld [r1, #0x018] ldr r6, [r1], #0x04 ldr r7, [r1], #0x04 ldr ip, [r1], #0x04 #ifdef __ARMEB__ orr r4, r4, r5, lsr #16 mov r5, r5, lsl #16 orr r5, r5, r6, lsr #16 mov r6, r6, lsl #16 orr r6, r6, r7, lsr #16 mov r7, r7, lsl #16 orr r7, r7, ip, lsr #16 #else orr r4, r4, r5, lsl #16 mov r5, r5, lsr #16 orr r5, r5, r6, lsl #16 mov r6, r6, lsr #16 orr r6, r6, r7, lsl #16 mov r7, r7, lsr #16 orr r7, r7, ip, lsl #16 #endif str r4, [r3], #0x04 str r5, [r3], #0x04 str r6, [r3], #0x04 str r7, [r3], #0x04 .Lmemcpy_bad2: subs r2, r2, #0x10 bge .Lmemcpy_bad2_loop16 adds r2, r2, #0x10 - ldmeqfd sp!, {r4-r7} + ldmfdeq sp!, {r4-r7} RETeq /* Return now if done */ subs r2, r2, #0x04 sublt r1, r1, #0x02 blt .Lmemcpy_bad_done .Lmemcpy_bad2_loop4: #ifdef __ARMEB__ mov r4, ip, lsl #16 #else mov r4, ip, lsr #16 #endif ldr ip, [r1], #0x04 subs r2, r2, #0x04 #ifdef __ARMEB__ orr r4, r4, ip, lsr #16 #else orr r4, r4, ip, lsl #16 #endif str r4, [r3], #0x04 bge .Lmemcpy_bad2_loop4 sub r1, r1, #0x02 b .Lmemcpy_bad_done .Lmemcpy_bad3_loop16: #ifdef __ARMEB__ mov r4, ip, lsl #24 #else mov r4, ip, lsr #24 #endif ldr r5, [r1], #0x04 pld [r1, #0x018] ldr r6, [r1], #0x04 ldr r7, [r1], #0x04 ldr ip, [r1], #0x04 #ifdef __ARMEB__ orr r4, r4, r5, lsr #8 mov r5, r5, lsl #24 orr r5, r5, r6, lsr #8 mov r6, r6, lsl #24 orr r6, r6, r7, lsr #8 mov r7, r7, lsl #24 orr r7, r7, ip, lsr #8 #else orr r4, r4, r5, lsl #8 mov r5, r5, lsr #24 orr r5, r5, r6, lsl #8 mov r6, r6, lsr #24 orr r6, r6, r7, lsl #8 mov r7, r7, lsr #24 orr r7, r7, ip, lsl #8 #endif str r4, [r3], #0x04 str r5, [r3], #0x04 str r6, [r3], #0x04 str r7, [r3], #0x04 .Lmemcpy_bad3: subs r2, r2, #0x10 bge .Lmemcpy_bad3_loop16 adds r2, r2, #0x10 - ldmeqfd sp!, {r4-r7} + ldmfdeq sp!, {r4-r7} RETeq /* Return now if done */ subs r2, r2, #0x04 sublt r1, r1, #0x01 blt .Lmemcpy_bad_done .Lmemcpy_bad3_loop4: #ifdef __ARMEB__ mov r4, ip, lsl #24 #else mov r4, ip, lsr #24 #endif ldr ip, [r1], #0x04 subs r2, r2, #0x04 #ifdef __ARMEB__ orr r4, r4, ip, lsr #8 #else orr r4, r4, ip, lsl #8 #endif str r4, [r3], #0x04 bge .Lmemcpy_bad3_loop4 sub r1, r1, #0x01 .Lmemcpy_bad_done: ldmfd sp!, {r4-r7} adds r2, r2, #0x04 RETeq ldrb ip, [r1], #0x01 cmp r2, #0x02 - ldrgeb r2, [r1], #0x01 + ldrbge r2, [r1], #0x01 strb ip, [r3], #0x01 - ldrgtb ip, [r1] - strgeb r2, [r3], #0x01 - strgtb ip, [r3] + ldrbgt ip, [r1] + strbge r2, [r3], #0x01 + strbgt ip, [r3] RET /* * Handle short copies (less than 16 bytes), possibly misaligned. * Some of these are *very* common, thanks to the network stack, * and so are handled specially. */ .Lmemcpy_short: add pc, pc, r2, lsl #2 nop RET /* 0x00 */ b .Lmemcpy_bytewise /* 0x01 */ b .Lmemcpy_bytewise /* 0x02 */ b .Lmemcpy_bytewise /* 0x03 */ b .Lmemcpy_4 /* 0x04 */ b .Lmemcpy_bytewise /* 0x05 */ b .Lmemcpy_6 /* 0x06 */ b .Lmemcpy_bytewise /* 0x07 */ b .Lmemcpy_8 /* 0x08 */ b .Lmemcpy_bytewise /* 0x09 */ b .Lmemcpy_bytewise /* 0x0a */ b .Lmemcpy_bytewise /* 0x0b */ b .Lmemcpy_c /* 0x0c */ .Lmemcpy_bytewise: mov r3, r0 /* We must not clobber r0 */ ldrb ip, [r1], #0x01 1: subs r2, r2, #0x01 strb ip, [r3], #0x01 - ldrneb ip, [r1], #0x01 + ldrbne ip, [r1], #0x01 bne 1b RET /****************************************************************************** * Special case for 4 byte copies */ #define LMEMCPY_4_LOG2 6 /* 64 bytes */ #define LMEMCPY_4_PAD .align LMEMCPY_4_LOG2 LMEMCPY_4_PAD .Lmemcpy_4: and r2, r1, #0x03 orr r2, r2, r0, lsl #2 ands r2, r2, #0x0f sub r3, pc, #0x14 addne pc, r3, r2, lsl #LMEMCPY_4_LOG2 /* * 0000: dst is 32-bit aligned, src is 32-bit aligned */ ldr r2, [r1] str r2, [r0] RET LMEMCPY_4_PAD /* * 0001: dst is 32-bit aligned, src is 8-bit aligned */ ldr r3, [r1, #-1] /* BE:r3 = x012 LE:r3 = 210x */ ldr r2, [r1, #3] /* BE:r2 = 3xxx LE:r2 = xxx3 */ #ifdef __ARMEB__ mov r3, r3, lsl #8 /* r3 = 012. */ orr r3, r3, r2, lsr #24 /* r3 = 0123 */ #else mov r3, r3, lsr #8 /* r3 = .210 */ orr r3, r3, r2, lsl #24 /* r3 = 3210 */ #endif str r3, [r0] RET LMEMCPY_4_PAD /* * 0010: dst is 32-bit aligned, src is 16-bit aligned */ #ifdef __ARMEB__ ldrh r3, [r1] ldrh r2, [r1, #0x02] #else ldrh r3, [r1, #0x02] ldrh r2, [r1] #endif orr r3, r2, r3, lsl #16 str r3, [r0] RET LMEMCPY_4_PAD /* * 0011: dst is 32-bit aligned, src is 8-bit aligned */ ldr r3, [r1, #-3] /* BE:r3 = xxx0 LE:r3 = 0xxx */ ldr r2, [r1, #1] /* BE:r2 = 123x LE:r2 = x321 */ #ifdef __ARMEB__ mov r3, r3, lsl #24 /* r3 = 0... */ orr r3, r3, r2, lsr #8 /* r3 = 0123 */ #else mov r3, r3, lsr #24 /* r3 = ...0 */ orr r3, r3, r2, lsl #8 /* r3 = 3210 */ #endif str r3, [r0] RET LMEMCPY_4_PAD /* * 0100: dst is 8-bit aligned, src is 32-bit aligned */ ldr r2, [r1] #ifdef __ARMEB__ strb r2, [r0, #0x03] mov r3, r2, lsr #8 mov r1, r2, lsr #24 strb r1, [r0] #else strb r2, [r0] mov r3, r2, lsr #8 mov r1, r2, lsr #24 strb r1, [r0, #0x03] #endif strh r3, [r0, #0x01] RET LMEMCPY_4_PAD /* * 0101: dst is 8-bit aligned, src is 8-bit aligned */ ldrb r2, [r1] ldrh r3, [r1, #0x01] ldrb r1, [r1, #0x03] strb r2, [r0] strh r3, [r0, #0x01] strb r1, [r0, #0x03] RET LMEMCPY_4_PAD /* * 0110: dst is 8-bit aligned, src is 16-bit aligned */ ldrh r2, [r1] /* BE:r2 = ..01 LE:r2 = ..10 */ ldrh r3, [r1, #0x02] /* LE:r3 = ..23 LE:r3 = ..32 */ #ifdef __ARMEB__ mov r1, r2, lsr #8 /* r1 = ...0 */ strb r1, [r0] mov r2, r2, lsl #8 /* r2 = .01. */ orr r2, r2, r3, lsr #8 /* r2 = .012 */ #else strb r2, [r0] mov r2, r2, lsr #8 /* r2 = ...1 */ orr r2, r2, r3, lsl #8 /* r2 = .321 */ mov r3, r3, lsr #8 /* r3 = ...3 */ #endif strh r2, [r0, #0x01] strb r3, [r0, #0x03] RET LMEMCPY_4_PAD /* * 0111: dst is 8-bit aligned, src is 8-bit aligned */ ldrb r2, [r1] ldrh r3, [r1, #0x01] ldrb r1, [r1, #0x03] strb r2, [r0] strh r3, [r0, #0x01] strb r1, [r0, #0x03] RET LMEMCPY_4_PAD /* * 1000: dst is 16-bit aligned, src is 32-bit aligned */ ldr r2, [r1] #ifdef __ARMEB__ strh r2, [r0, #0x02] mov r3, r2, lsr #16 strh r3, [r0] #else strh r2, [r0] mov r3, r2, lsr #16 strh r3, [r0, #0x02] #endif RET LMEMCPY_4_PAD /* * 1001: dst is 16-bit aligned, src is 8-bit aligned */ ldr r2, [r1, #-1] /* BE:r2 = x012 LE:r2 = 210x */ ldr r3, [r1, #3] /* BE:r3 = 3xxx LE:r3 = xxx3 */ mov r1, r2, lsr #8 /* BE:r1 = .x01 LE:r1 = .210 */ strh r1, [r0] #ifdef __ARMEB__ mov r2, r2, lsl #8 /* r2 = 012. */ orr r2, r2, r3, lsr #24 /* r2 = 0123 */ #else mov r2, r2, lsr #24 /* r2 = ...2 */ orr r2, r2, r3, lsl #8 /* r2 = xx32 */ #endif strh r2, [r0, #0x02] RET LMEMCPY_4_PAD /* * 1010: dst is 16-bit aligned, src is 16-bit aligned */ ldrh r2, [r1] ldrh r3, [r1, #0x02] strh r2, [r0] strh r3, [r0, #0x02] RET LMEMCPY_4_PAD /* * 1011: dst is 16-bit aligned, src is 8-bit aligned */ ldr r3, [r1, #1] /* BE:r3 = 123x LE:r3 = x321 */ ldr r2, [r1, #-3] /* BE:r2 = xxx0 LE:r2 = 0xxx */ mov r1, r3, lsr #8 /* BE:r1 = .123 LE:r1 = .x32 */ strh r1, [r0, #0x02] #ifdef __ARMEB__ mov r3, r3, lsr #24 /* r3 = ...1 */ orr r3, r3, r2, lsl #8 /* r3 = xx01 */ #else mov r3, r3, lsl #8 /* r3 = 321. */ orr r3, r3, r2, lsr #24 /* r3 = 3210 */ #endif strh r3, [r0] RET LMEMCPY_4_PAD /* * 1100: dst is 8-bit aligned, src is 32-bit aligned */ ldr r2, [r1] /* BE:r2 = 0123 LE:r2 = 3210 */ #ifdef __ARMEB__ strb r2, [r0, #0x03] mov r3, r2, lsr #8 mov r1, r2, lsr #24 strh r3, [r0, #0x01] strb r1, [r0] #else strb r2, [r0] mov r3, r2, lsr #8 mov r1, r2, lsr #24 strh r3, [r0, #0x01] strb r1, [r0, #0x03] #endif RET LMEMCPY_4_PAD /* * 1101: dst is 8-bit aligned, src is 8-bit aligned */ ldrb r2, [r1] ldrh r3, [r1, #0x01] ldrb r1, [r1, #0x03] strb r2, [r0] strh r3, [r0, #0x01] strb r1, [r0, #0x03] RET LMEMCPY_4_PAD /* * 1110: dst is 8-bit aligned, src is 16-bit aligned */ #ifdef __ARMEB__ ldrh r3, [r1, #0x02] /* BE:r3 = ..23 LE:r3 = ..32 */ ldrh r2, [r1] /* BE:r2 = ..01 LE:r2 = ..10 */ strb r3, [r0, #0x03] mov r3, r3, lsr #8 /* r3 = ...2 */ orr r3, r3, r2, lsl #8 /* r3 = ..12 */ strh r3, [r0, #0x01] mov r2, r2, lsr #8 /* r2 = ...0 */ strb r2, [r0] #else ldrh r2, [r1] /* BE:r2 = ..01 LE:r2 = ..10 */ ldrh r3, [r1, #0x02] /* BE:r3 = ..23 LE:r3 = ..32 */ strb r2, [r0] mov r2, r2, lsr #8 /* r2 = ...1 */ orr r2, r2, r3, lsl #8 /* r2 = .321 */ strh r2, [r0, #0x01] mov r3, r3, lsr #8 /* r3 = ...3 */ strb r3, [r0, #0x03] #endif RET LMEMCPY_4_PAD /* * 1111: dst is 8-bit aligned, src is 8-bit aligned */ ldrb r2, [r1] ldrh r3, [r1, #0x01] ldrb r1, [r1, #0x03] strb r2, [r0] strh r3, [r0, #0x01] strb r1, [r0, #0x03] RET LMEMCPY_4_PAD /****************************************************************************** * Special case for 6 byte copies */ #define LMEMCPY_6_LOG2 6 /* 64 bytes */ #define LMEMCPY_6_PAD .align LMEMCPY_6_LOG2 LMEMCPY_6_PAD .Lmemcpy_6: and r2, r1, #0x03 orr r2, r2, r0, lsl #2 ands r2, r2, #0x0f sub r3, pc, #0x14 addne pc, r3, r2, lsl #LMEMCPY_6_LOG2 /* * 0000: dst is 32-bit aligned, src is 32-bit aligned */ ldr r2, [r1] ldrh r3, [r1, #0x04] str r2, [r0] strh r3, [r0, #0x04] RET LMEMCPY_6_PAD /* * 0001: dst is 32-bit aligned, src is 8-bit aligned */ ldr r2, [r1, #-1] /* BE:r2 = x012 LE:r2 = 210x */ ldr r3, [r1, #0x03] /* BE:r3 = 345x LE:r3 = x543 */ #ifdef __ARMEB__ mov r2, r2, lsl #8 /* r2 = 012. */ orr r2, r2, r3, lsr #24 /* r2 = 0123 */ #else mov r2, r2, lsr #8 /* r2 = .210 */ orr r2, r2, r3, lsl #24 /* r2 = 3210 */ #endif mov r3, r3, lsr #8 /* BE:r3 = .345 LE:r3 = .x54 */ str r2, [r0] strh r3, [r0, #0x04] RET LMEMCPY_6_PAD /* * 0010: dst is 32-bit aligned, src is 16-bit aligned */ ldr r3, [r1, #0x02] /* BE:r3 = 2345 LE:r3 = 5432 */ ldrh r2, [r1] /* BE:r2 = ..01 LE:r2 = ..10 */ #ifdef __ARMEB__ mov r1, r3, lsr #16 /* r1 = ..23 */ orr r1, r1, r2, lsl #16 /* r1 = 0123 */ str r1, [r0] strh r3, [r0, #0x04] #else mov r1, r3, lsr #16 /* r1 = ..54 */ orr r2, r2, r3, lsl #16 /* r2 = 3210 */ str r2, [r0] strh r1, [r0, #0x04] #endif RET LMEMCPY_6_PAD /* * 0011: dst is 32-bit aligned, src is 8-bit aligned */ ldr r2, [r1, #-3] /* BE:r2 = xxx0 LE:r2 = 0xxx */ ldr r3, [r1, #1] /* BE:r3 = 1234 LE:r3 = 4321 */ ldr r1, [r1, #5] /* BE:r1 = 5xxx LE:r3 = xxx5 */ #ifdef __ARMEB__ mov r2, r2, lsl #24 /* r2 = 0... */ orr r2, r2, r3, lsr #8 /* r2 = 0123 */ mov r3, r3, lsl #8 /* r3 = 234. */ orr r1, r3, r1, lsr #24 /* r1 = 2345 */ #else mov r2, r2, lsr #24 /* r2 = ...0 */ orr r2, r2, r3, lsl #8 /* r2 = 3210 */ mov r1, r1, lsl #8 /* r1 = xx5. */ orr r1, r1, r3, lsr #24 /* r1 = xx54 */ #endif str r2, [r0] strh r1, [r0, #0x04] RET LMEMCPY_6_PAD /* * 0100: dst is 8-bit aligned, src is 32-bit aligned */ ldr r3, [r1] /* BE:r3 = 0123 LE:r3 = 3210 */ ldrh r2, [r1, #0x04] /* BE:r2 = ..45 LE:r2 = ..54 */ mov r1, r3, lsr #8 /* BE:r1 = .012 LE:r1 = .321 */ strh r1, [r0, #0x01] #ifdef __ARMEB__ mov r1, r3, lsr #24 /* r1 = ...0 */ strb r1, [r0] mov r3, r3, lsl #8 /* r3 = 123. */ orr r3, r3, r2, lsr #8 /* r3 = 1234 */ #else strb r3, [r0] mov r3, r3, lsr #24 /* r3 = ...3 */ orr r3, r3, r2, lsl #8 /* r3 = .543 */ mov r2, r2, lsr #8 /* r2 = ...5 */ #endif strh r3, [r0, #0x03] strb r2, [r0, #0x05] RET LMEMCPY_6_PAD /* * 0101: dst is 8-bit aligned, src is 8-bit aligned */ ldrb r2, [r1] ldrh r3, [r1, #0x01] ldrh ip, [r1, #0x03] ldrb r1, [r1, #0x05] strb r2, [r0] strh r3, [r0, #0x01] strh ip, [r0, #0x03] strb r1, [r0, #0x05] RET LMEMCPY_6_PAD /* * 0110: dst is 8-bit aligned, src is 16-bit aligned */ ldrh r2, [r1] /* BE:r2 = ..01 LE:r2 = ..10 */ ldr r1, [r1, #0x02] /* BE:r1 = 2345 LE:r1 = 5432 */ #ifdef __ARMEB__ mov r3, r2, lsr #8 /* r3 = ...0 */ strb r3, [r0] strb r1, [r0, #0x05] mov r3, r1, lsr #8 /* r3 = .234 */ strh r3, [r0, #0x03] mov r3, r2, lsl #8 /* r3 = .01. */ orr r3, r3, r1, lsr #24 /* r3 = .012 */ strh r3, [r0, #0x01] #else strb r2, [r0] mov r3, r1, lsr #24 strb r3, [r0, #0x05] mov r3, r1, lsr #8 /* r3 = .543 */ strh r3, [r0, #0x03] mov r3, r2, lsr #8 /* r3 = ...1 */ orr r3, r3, r1, lsl #8 /* r3 = 4321 */ strh r3, [r0, #0x01] #endif RET LMEMCPY_6_PAD /* * 0111: dst is 8-bit aligned, src is 8-bit aligned */ ldrb r2, [r1] ldrh r3, [r1, #0x01] ldrh ip, [r1, #0x03] ldrb r1, [r1, #0x05] strb r2, [r0] strh r3, [r0, #0x01] strh ip, [r0, #0x03] strb r1, [r0, #0x05] RET LMEMCPY_6_PAD /* * 1000: dst is 16-bit aligned, src is 32-bit aligned */ #ifdef __ARMEB__ ldr r2, [r1] /* r2 = 0123 */ ldrh r3, [r1, #0x04] /* r3 = ..45 */ mov r1, r2, lsr #16 /* r1 = ..01 */ orr r3, r3, r2, lsl#16 /* r3 = 2345 */ strh r1, [r0] str r3, [r0, #0x02] #else ldrh r2, [r1, #0x04] /* r2 = ..54 */ ldr r3, [r1] /* r3 = 3210 */ mov r2, r2, lsl #16 /* r2 = 54.. */ orr r2, r2, r3, lsr #16 /* r2 = 5432 */ strh r3, [r0] str r2, [r0, #0x02] #endif RET LMEMCPY_6_PAD /* * 1001: dst is 16-bit aligned, src is 8-bit aligned */ ldr r3, [r1, #-1] /* BE:r3 = x012 LE:r3 = 210x */ ldr r2, [r1, #3] /* BE:r2 = 345x LE:r2 = x543 */ mov r1, r3, lsr #8 /* BE:r1 = .x01 LE:r1 = .210 */ #ifdef __ARMEB__ mov r2, r2, lsr #8 /* r2 = .345 */ orr r2, r2, r3, lsl #24 /* r2 = 2345 */ #else mov r2, r2, lsl #8 /* r2 = 543. */ orr r2, r2, r3, lsr #24 /* r2 = 5432 */ #endif strh r1, [r0] str r2, [r0, #0x02] RET LMEMCPY_6_PAD /* * 1010: dst is 16-bit aligned, src is 16-bit aligned */ ldrh r2, [r1] ldr r3, [r1, #0x02] strh r2, [r0] str r3, [r0, #0x02] RET LMEMCPY_6_PAD /* * 1011: dst is 16-bit aligned, src is 8-bit aligned */ ldrb r3, [r1] /* r3 = ...0 */ ldr r2, [r1, #0x01] /* BE:r2 = 1234 LE:r2 = 4321 */ ldrb r1, [r1, #0x05] /* r1 = ...5 */ #ifdef __ARMEB__ mov r3, r3, lsl #8 /* r3 = ..0. */ orr r3, r3, r2, lsr #24 /* r3 = ..01 */ orr r1, r1, r2, lsl #8 /* r1 = 2345 */ #else orr r3, r3, r2, lsl #8 /* r3 = 3210 */ mov r1, r1, lsl #24 /* r1 = 5... */ orr r1, r1, r2, lsr #8 /* r1 = 5432 */ #endif strh r3, [r0] str r1, [r0, #0x02] RET LMEMCPY_6_PAD /* * 1100: dst is 8-bit aligned, src is 32-bit aligned */ ldr r2, [r1] /* BE:r2 = 0123 LE:r2 = 3210 */ ldrh r1, [r1, #0x04] /* BE:r1 = ..45 LE:r1 = ..54 */ #ifdef __ARMEB__ mov r3, r2, lsr #24 /* r3 = ...0 */ strb r3, [r0] mov r2, r2, lsl #8 /* r2 = 123. */ orr r2, r2, r1, lsr #8 /* r2 = 1234 */ #else strb r2, [r0] mov r2, r2, lsr #8 /* r2 = .321 */ orr r2, r2, r1, lsl #24 /* r2 = 4321 */ mov r1, r1, lsr #8 /* r1 = ...5 */ #endif str r2, [r0, #0x01] strb r1, [r0, #0x05] RET LMEMCPY_6_PAD /* * 1101: dst is 8-bit aligned, src is 8-bit aligned */ ldrb r2, [r1] ldrh r3, [r1, #0x01] ldrh ip, [r1, #0x03] ldrb r1, [r1, #0x05] strb r2, [r0] strh r3, [r0, #0x01] strh ip, [r0, #0x03] strb r1, [r0, #0x05] RET LMEMCPY_6_PAD /* * 1110: dst is 8-bit aligned, src is 16-bit aligned */ ldrh r2, [r1] /* BE:r2 = ..01 LE:r2 = ..10 */ ldr r1, [r1, #0x02] /* BE:r1 = 2345 LE:r1 = 5432 */ #ifdef __ARMEB__ mov r3, r2, lsr #8 /* r3 = ...0 */ strb r3, [r0] mov r2, r2, lsl #24 /* r2 = 1... */ orr r2, r2, r1, lsr #8 /* r2 = 1234 */ #else strb r2, [r0] mov r2, r2, lsr #8 /* r2 = ...1 */ orr r2, r2, r1, lsl #8 /* r2 = 4321 */ mov r1, r1, lsr #24 /* r1 = ...5 */ #endif str r2, [r0, #0x01] strb r1, [r0, #0x05] RET LMEMCPY_6_PAD /* * 1111: dst is 8-bit aligned, src is 8-bit aligned */ ldrb r2, [r1] ldr r3, [r1, #0x01] ldrb r1, [r1, #0x05] strb r2, [r0] str r3, [r0, #0x01] strb r1, [r0, #0x05] RET LMEMCPY_6_PAD /****************************************************************************** * Special case for 8 byte copies */ #define LMEMCPY_8_LOG2 6 /* 64 bytes */ #define LMEMCPY_8_PAD .align LMEMCPY_8_LOG2 LMEMCPY_8_PAD .Lmemcpy_8: and r2, r1, #0x03 orr r2, r2, r0, lsl #2 ands r2, r2, #0x0f sub r3, pc, #0x14 addne pc, r3, r2, lsl #LMEMCPY_8_LOG2 /* * 0000: dst is 32-bit aligned, src is 32-bit aligned */ ldr r2, [r1] ldr r3, [r1, #0x04] str r2, [r0] str r3, [r0, #0x04] RET LMEMCPY_8_PAD /* * 0001: dst is 32-bit aligned, src is 8-bit aligned */ ldr r3, [r1, #-1] /* BE:r3 = x012 LE:r3 = 210x */ ldr r2, [r1, #0x03] /* BE:r2 = 3456 LE:r2 = 6543 */ ldrb r1, [r1, #0x07] /* r1 = ...7 */ #ifdef __ARMEB__ mov r3, r3, lsl #8 /* r3 = 012. */ orr r3, r3, r2, lsr #24 /* r3 = 0123 */ orr r2, r1, r2, lsl #8 /* r2 = 4567 */ #else mov r3, r3, lsr #8 /* r3 = .210 */ orr r3, r3, r2, lsl #24 /* r3 = 3210 */ mov r1, r1, lsl #24 /* r1 = 7... */ orr r2, r1, r2, lsr #8 /* r2 = 7654 */ #endif str r3, [r0] str r2, [r0, #0x04] RET LMEMCPY_8_PAD /* * 0010: dst is 32-bit aligned, src is 16-bit aligned */ ldrh r2, [r1] /* BE:r2 = ..01 LE:r2 = ..10 */ ldr r3, [r1, #0x02] /* BE:r3 = 2345 LE:r3 = 5432 */ ldrh r1, [r1, #0x06] /* BE:r1 = ..67 LE:r1 = ..76 */ #ifdef __ARMEB__ mov r2, r2, lsl #16 /* r2 = 01.. */ orr r2, r2, r3, lsr #16 /* r2 = 0123 */ orr r3, r1, r3, lsl #16 /* r3 = 4567 */ #else orr r2, r2, r3, lsl #16 /* r2 = 3210 */ mov r3, r3, lsr #16 /* r3 = ..54 */ orr r3, r3, r1, lsl #16 /* r3 = 7654 */ #endif str r2, [r0] str r3, [r0, #0x04] RET LMEMCPY_8_PAD /* * 0011: dst is 32-bit aligned, src is 8-bit aligned */ ldrb r3, [r1] /* r3 = ...0 */ ldr r2, [r1, #0x01] /* BE:r2 = 1234 LE:r2 = 4321 */ ldr r1, [r1, #0x05] /* BE:r1 = 567x LE:r1 = x765 */ #ifdef __ARMEB__ mov r3, r3, lsl #24 /* r3 = 0... */ orr r3, r3, r2, lsr #8 /* r3 = 0123 */ mov r2, r2, lsl #24 /* r2 = 4... */ orr r2, r2, r1, lsr #8 /* r2 = 4567 */ #else orr r3, r3, r2, lsl #8 /* r3 = 3210 */ mov r2, r2, lsr #24 /* r2 = ...4 */ orr r2, r2, r1, lsl #8 /* r2 = 7654 */ #endif str r3, [r0] str r2, [r0, #0x04] RET LMEMCPY_8_PAD /* * 0100: dst is 8-bit aligned, src is 32-bit aligned */ ldr r3, [r1] /* BE:r3 = 0123 LE:r3 = 3210 */ ldr r2, [r1, #0x04] /* BE:r2 = 4567 LE:r2 = 7654 */ #ifdef __ARMEB__ mov r1, r3, lsr #24 /* r1 = ...0 */ strb r1, [r0] mov r1, r3, lsr #8 /* r1 = .012 */ strb r2, [r0, #0x07] mov r3, r3, lsl #24 /* r3 = 3... */ orr r3, r3, r2, lsr #8 /* r3 = 3456 */ #else strb r3, [r0] mov r1, r2, lsr #24 /* r1 = ...7 */ strb r1, [r0, #0x07] mov r1, r3, lsr #8 /* r1 = .321 */ mov r3, r3, lsr #24 /* r3 = ...3 */ orr r3, r3, r2, lsl #8 /* r3 = 6543 */ #endif strh r1, [r0, #0x01] str r3, [r0, #0x03] RET LMEMCPY_8_PAD /* * 0101: dst is 8-bit aligned, src is 8-bit aligned */ ldrb r2, [r1] ldrh r3, [r1, #0x01] ldr ip, [r1, #0x03] ldrb r1, [r1, #0x07] strb r2, [r0] strh r3, [r0, #0x01] str ip, [r0, #0x03] strb r1, [r0, #0x07] RET LMEMCPY_8_PAD /* * 0110: dst is 8-bit aligned, src is 16-bit aligned */ ldrh r2, [r1] /* BE:r2 = ..01 LE:r2 = ..10 */ ldr r3, [r1, #0x02] /* BE:r3 = 2345 LE:r3 = 5432 */ ldrh r1, [r1, #0x06] /* BE:r1 = ..67 LE:r1 = ..76 */ #ifdef __ARMEB__ mov ip, r2, lsr #8 /* ip = ...0 */ strb ip, [r0] mov ip, r2, lsl #8 /* ip = .01. */ orr ip, ip, r3, lsr #24 /* ip = .012 */ strb r1, [r0, #0x07] mov r3, r3, lsl #8 /* r3 = 345. */ orr r3, r3, r1, lsr #8 /* r3 = 3456 */ #else strb r2, [r0] /* 0 */ mov ip, r1, lsr #8 /* ip = ...7 */ strb ip, [r0, #0x07] /* 7 */ mov ip, r2, lsr #8 /* ip = ...1 */ orr ip, ip, r3, lsl #8 /* ip = 4321 */ mov r3, r3, lsr #8 /* r3 = .543 */ orr r3, r3, r1, lsl #24 /* r3 = 6543 */ #endif strh ip, [r0, #0x01] str r3, [r0, #0x03] RET LMEMCPY_8_PAD /* * 0111: dst is 8-bit aligned, src is 8-bit aligned */ ldrb r3, [r1] /* r3 = ...0 */ ldr ip, [r1, #0x01] /* BE:ip = 1234 LE:ip = 4321 */ ldrh r2, [r1, #0x05] /* BE:r2 = ..56 LE:r2 = ..65 */ ldrb r1, [r1, #0x07] /* r1 = ...7 */ strb r3, [r0] mov r3, ip, lsr #16 /* BE:r3 = ..12 LE:r3 = ..43 */ #ifdef __ARMEB__ strh r3, [r0, #0x01] orr r2, r2, ip, lsl #16 /* r2 = 3456 */ #else strh ip, [r0, #0x01] orr r2, r3, r2, lsl #16 /* r2 = 6543 */ #endif str r2, [r0, #0x03] strb r1, [r0, #0x07] RET LMEMCPY_8_PAD /* * 1000: dst is 16-bit aligned, src is 32-bit aligned */ ldr r2, [r1] /* BE:r2 = 0123 LE:r2 = 3210 */ ldr r3, [r1, #0x04] /* BE:r3 = 4567 LE:r3 = 7654 */ mov r1, r2, lsr #16 /* BE:r1 = ..01 LE:r1 = ..32 */ #ifdef __ARMEB__ strh r1, [r0] mov r1, r3, lsr #16 /* r1 = ..45 */ orr r2, r1 ,r2, lsl #16 /* r2 = 2345 */ #else strh r2, [r0] orr r2, r1, r3, lsl #16 /* r2 = 5432 */ mov r3, r3, lsr #16 /* r3 = ..76 */ #endif str r2, [r0, #0x02] strh r3, [r0, #0x06] RET LMEMCPY_8_PAD /* * 1001: dst is 16-bit aligned, src is 8-bit aligned */ ldr r2, [r1, #-1] /* BE:r2 = x012 LE:r2 = 210x */ ldr r3, [r1, #0x03] /* BE:r3 = 3456 LE:r3 = 6543 */ ldrb ip, [r1, #0x07] /* ip = ...7 */ mov r1, r2, lsr #8 /* BE:r1 = .x01 LE:r1 = .210 */ strh r1, [r0] #ifdef __ARMEB__ mov r1, r2, lsl #24 /* r1 = 2... */ orr r1, r1, r3, lsr #8 /* r1 = 2345 */ orr r3, ip, r3, lsl #8 /* r3 = 4567 */ #else mov r1, r2, lsr #24 /* r1 = ...2 */ orr r1, r1, r3, lsl #8 /* r1 = 5432 */ mov r3, r3, lsr #24 /* r3 = ...6 */ orr r3, r3, ip, lsl #8 /* r3 = ..76 */ #endif str r1, [r0, #0x02] strh r3, [r0, #0x06] RET LMEMCPY_8_PAD /* * 1010: dst is 16-bit aligned, src is 16-bit aligned */ ldrh r2, [r1] ldr ip, [r1, #0x02] ldrh r3, [r1, #0x06] strh r2, [r0] str ip, [r0, #0x02] strh r3, [r0, #0x06] RET LMEMCPY_8_PAD /* * 1011: dst is 16-bit aligned, src is 8-bit aligned */ ldr r3, [r1, #0x05] /* BE:r3 = 567x LE:r3 = x765 */ ldr r2, [r1, #0x01] /* BE:r2 = 1234 LE:r2 = 4321 */ ldrb ip, [r1] /* ip = ...0 */ mov r1, r3, lsr #8 /* BE:r1 = .567 LE:r1 = .x76 */ strh r1, [r0, #0x06] #ifdef __ARMEB__ mov r3, r3, lsr #24 /* r3 = ...5 */ orr r3, r3, r2, lsl #8 /* r3 = 2345 */ mov r2, r2, lsr #24 /* r2 = ...1 */ orr r2, r2, ip, lsl #8 /* r2 = ..01 */ #else mov r3, r3, lsl #24 /* r3 = 5... */ orr r3, r3, r2, lsr #8 /* r3 = 5432 */ orr r2, ip, r2, lsl #8 /* r2 = 3210 */ #endif str r3, [r0, #0x02] strh r2, [r0] RET LMEMCPY_8_PAD /* * 1100: dst is 8-bit aligned, src is 32-bit aligned */ ldr r3, [r1, #0x04] /* BE:r3 = 4567 LE:r3 = 7654 */ ldr r2, [r1] /* BE:r2 = 0123 LE:r2 = 3210 */ mov r1, r3, lsr #8 /* BE:r1 = .456 LE:r1 = .765 */ strh r1, [r0, #0x05] #ifdef __ARMEB__ strb r3, [r0, #0x07] mov r1, r2, lsr #24 /* r1 = ...0 */ strb r1, [r0] mov r2, r2, lsl #8 /* r2 = 123. */ orr r2, r2, r3, lsr #24 /* r2 = 1234 */ str r2, [r0, #0x01] #else strb r2, [r0] mov r1, r3, lsr #24 /* r1 = ...7 */ strb r1, [r0, #0x07] mov r2, r2, lsr #8 /* r2 = .321 */ orr r2, r2, r3, lsl #24 /* r2 = 4321 */ str r2, [r0, #0x01] #endif RET LMEMCPY_8_PAD /* * 1101: dst is 8-bit aligned, src is 8-bit aligned */ ldrb r3, [r1] /* r3 = ...0 */ ldrh r2, [r1, #0x01] /* BE:r2 = ..12 LE:r2 = ..21 */ ldr ip, [r1, #0x03] /* BE:ip = 3456 LE:ip = 6543 */ ldrb r1, [r1, #0x07] /* r1 = ...7 */ strb r3, [r0] mov r3, ip, lsr #16 /* BE:r3 = ..34 LE:r3 = ..65 */ #ifdef __ARMEB__ strh ip, [r0, #0x05] orr r2, r3, r2, lsl #16 /* r2 = 1234 */ #else strh r3, [r0, #0x05] orr r2, r2, ip, lsl #16 /* r2 = 4321 */ #endif str r2, [r0, #0x01] strb r1, [r0, #0x07] RET LMEMCPY_8_PAD /* * 1110: dst is 8-bit aligned, src is 16-bit aligned */ ldrh r2, [r1] /* BE:r2 = ..01 LE:r2 = ..10 */ ldr r3, [r1, #0x02] /* BE:r3 = 2345 LE:r3 = 5432 */ ldrh r1, [r1, #0x06] /* BE:r1 = ..67 LE:r1 = ..76 */ #ifdef __ARMEB__ mov ip, r2, lsr #8 /* ip = ...0 */ strb ip, [r0] mov ip, r2, lsl #24 /* ip = 1... */ orr ip, ip, r3, lsr #8 /* ip = 1234 */ strb r1, [r0, #0x07] mov r1, r1, lsr #8 /* r1 = ...6 */ orr r1, r1, r3, lsl #8 /* r1 = 3456 */ #else strb r2, [r0] mov ip, r2, lsr #8 /* ip = ...1 */ orr ip, ip, r3, lsl #8 /* ip = 4321 */ mov r2, r1, lsr #8 /* r2 = ...7 */ strb r2, [r0, #0x07] mov r1, r1, lsl #8 /* r1 = .76. */ orr r1, r1, r3, lsr #24 /* r1 = .765 */ #endif str ip, [r0, #0x01] strh r1, [r0, #0x05] RET LMEMCPY_8_PAD /* * 1111: dst is 8-bit aligned, src is 8-bit aligned */ ldrb r2, [r1] ldr ip, [r1, #0x01] ldrh r3, [r1, #0x05] ldrb r1, [r1, #0x07] strb r2, [r0] str ip, [r0, #0x01] strh r3, [r0, #0x05] strb r1, [r0, #0x07] RET LMEMCPY_8_PAD /****************************************************************************** * Special case for 12 byte copies */ #define LMEMCPY_C_LOG2 7 /* 128 bytes */ #define LMEMCPY_C_PAD .align LMEMCPY_C_LOG2 LMEMCPY_C_PAD .Lmemcpy_c: and r2, r1, #0x03 orr r2, r2, r0, lsl #2 ands r2, r2, #0x0f sub r3, pc, #0x14 addne pc, r3, r2, lsl #LMEMCPY_C_LOG2 /* * 0000: dst is 32-bit aligned, src is 32-bit aligned */ ldr r2, [r1] ldr r3, [r1, #0x04] ldr r1, [r1, #0x08] str r2, [r0] str r3, [r0, #0x04] str r1, [r0, #0x08] RET LMEMCPY_C_PAD /* * 0001: dst is 32-bit aligned, src is 8-bit aligned */ ldrb r2, [r1, #0xb] /* r2 = ...B */ ldr ip, [r1, #0x07] /* BE:ip = 789A LE:ip = A987 */ ldr r3, [r1, #0x03] /* BE:r3 = 3456 LE:r3 = 6543 */ ldr r1, [r1, #-1] /* BE:r1 = x012 LE:r1 = 210x */ #ifdef __ARMEB__ orr r2, r2, ip, lsl #8 /* r2 = 89AB */ str r2, [r0, #0x08] mov r2, ip, lsr #24 /* r2 = ...7 */ orr r2, r2, r3, lsl #8 /* r2 = 4567 */ mov r1, r1, lsl #8 /* r1 = 012. */ orr r1, r1, r3, lsr #24 /* r1 = 0123 */ #else mov r2, r2, lsl #24 /* r2 = B... */ orr r2, r2, ip, lsr #8 /* r2 = BA98 */ str r2, [r0, #0x08] mov r2, ip, lsl #24 /* r2 = 7... */ orr r2, r2, r3, lsr #8 /* r2 = 7654 */ mov r1, r1, lsr #8 /* r1 = .210 */ orr r1, r1, r3, lsl #24 /* r1 = 3210 */ #endif str r2, [r0, #0x04] str r1, [r0] RET LMEMCPY_C_PAD /* * 0010: dst is 32-bit aligned, src is 16-bit aligned */ ldrh r2, [r1] /* BE:r2 = ..01 LE:r2 = ..10 */ ldr r3, [r1, #0x02] /* BE:r3 = 2345 LE:r3 = 5432 */ ldr ip, [r1, #0x06] /* BE:ip = 6789 LE:ip = 9876 */ ldrh r1, [r1, #0x0a] /* BE:r1 = ..AB LE:r1 = ..BA */ #ifdef __ARMEB__ mov r2, r2, lsl #16 /* r2 = 01.. */ orr r2, r2, r3, lsr #16 /* r2 = 0123 */ str r2, [r0] mov r3, r3, lsl #16 /* r3 = 45.. */ orr r3, r3, ip, lsr #16 /* r3 = 4567 */ orr r1, r1, ip, lsl #16 /* r1 = 89AB */ #else orr r2, r2, r3, lsl #16 /* r2 = 3210 */ str r2, [r0] mov r3, r3, lsr #16 /* r3 = ..54 */ orr r3, r3, ip, lsl #16 /* r3 = 7654 */ mov r1, r1, lsl #16 /* r1 = BA.. */ orr r1, r1, ip, lsr #16 /* r1 = BA98 */ #endif str r3, [r0, #0x04] str r1, [r0, #0x08] RET LMEMCPY_C_PAD /* * 0011: dst is 32-bit aligned, src is 8-bit aligned */ ldrb r2, [r1] /* r2 = ...0 */ ldr r3, [r1, #0x01] /* BE:r3 = 1234 LE:r3 = 4321 */ ldr ip, [r1, #0x05] /* BE:ip = 5678 LE:ip = 8765 */ ldr r1, [r1, #0x09] /* BE:r1 = 9ABx LE:r1 = xBA9 */ #ifdef __ARMEB__ mov r2, r2, lsl #24 /* r2 = 0... */ orr r2, r2, r3, lsr #8 /* r2 = 0123 */ str r2, [r0] mov r3, r3, lsl #24 /* r3 = 4... */ orr r3, r3, ip, lsr #8 /* r3 = 4567 */ mov r1, r1, lsr #8 /* r1 = .9AB */ orr r1, r1, ip, lsl #24 /* r1 = 89AB */ #else orr r2, r2, r3, lsl #8 /* r2 = 3210 */ str r2, [r0] mov r3, r3, lsr #24 /* r3 = ...4 */ orr r3, r3, ip, lsl #8 /* r3 = 7654 */ mov r1, r1, lsl #8 /* r1 = BA9. */ orr r1, r1, ip, lsr #24 /* r1 = BA98 */ #endif str r3, [r0, #0x04] str r1, [r0, #0x08] RET LMEMCPY_C_PAD /* * 0100: dst is 8-bit aligned (byte 1), src is 32-bit aligned */ ldr r2, [r1] /* BE:r2 = 0123 LE:r2 = 3210 */ ldr r3, [r1, #0x04] /* BE:r3 = 4567 LE:r3 = 7654 */ ldr ip, [r1, #0x08] /* BE:ip = 89AB LE:ip = BA98 */ mov r1, r2, lsr #8 /* BE:r1 = .012 LE:r1 = .321 */ strh r1, [r0, #0x01] #ifdef __ARMEB__ mov r1, r2, lsr #24 /* r1 = ...0 */ strb r1, [r0] mov r1, r2, lsl #24 /* r1 = 3... */ orr r2, r1, r3, lsr #8 /* r1 = 3456 */ mov r1, r3, lsl #24 /* r1 = 7... */ orr r1, r1, ip, lsr #8 /* r1 = 789A */ #else strb r2, [r0] mov r1, r2, lsr #24 /* r1 = ...3 */ orr r2, r1, r3, lsl #8 /* r1 = 6543 */ mov r1, r3, lsr #24 /* r1 = ...7 */ orr r1, r1, ip, lsl #8 /* r1 = A987 */ mov ip, ip, lsr #24 /* ip = ...B */ #endif str r2, [r0, #0x03] str r1, [r0, #0x07] strb ip, [r0, #0x0b] RET LMEMCPY_C_PAD /* * 0101: dst is 8-bit aligned (byte 1), src is 8-bit aligned (byte 1) */ ldrb r2, [r1] ldrh r3, [r1, #0x01] ldr ip, [r1, #0x03] strb r2, [r0] ldr r2, [r1, #0x07] ldrb r1, [r1, #0x0b] strh r3, [r0, #0x01] str ip, [r0, #0x03] str r2, [r0, #0x07] strb r1, [r0, #0x0b] RET LMEMCPY_C_PAD /* * 0110: dst is 8-bit aligned (byte 1), src is 16-bit aligned */ ldrh r2, [r1] /* BE:r2 = ..01 LE:r2 = ..10 */ ldr r3, [r1, #0x02] /* BE:r3 = 2345 LE:r3 = 5432 */ ldr ip, [r1, #0x06] /* BE:ip = 6789 LE:ip = 9876 */ ldrh r1, [r1, #0x0a] /* BE:r1 = ..AB LE:r1 = ..BA */ #ifdef __ARMEB__ mov r2, r2, ror #8 /* r2 = 1..0 */ strb r2, [r0] mov r2, r2, lsr #16 /* r2 = ..1. */ orr r2, r2, r3, lsr #24 /* r2 = ..12 */ strh r2, [r0, #0x01] mov r2, r3, lsl #8 /* r2 = 345. */ orr r3, r2, ip, lsr #24 /* r3 = 3456 */ mov r2, ip, lsl #8 /* r2 = 789. */ orr r2, r2, r1, lsr #8 /* r2 = 789A */ #else strb r2, [r0] mov r2, r2, lsr #8 /* r2 = ...1 */ orr r2, r2, r3, lsl #8 /* r2 = 4321 */ strh r2, [r0, #0x01] mov r2, r3, lsr #8 /* r2 = .543 */ orr r3, r2, ip, lsl #24 /* r3 = 6543 */ mov r2, ip, lsr #8 /* r2 = .987 */ orr r2, r2, r1, lsl #24 /* r2 = A987 */ mov r1, r1, lsr #8 /* r1 = ...B */ #endif str r3, [r0, #0x03] str r2, [r0, #0x07] strb r1, [r0, #0x0b] RET LMEMCPY_C_PAD /* * 0111: dst is 8-bit aligned (byte 1), src is 8-bit aligned (byte 3) */ ldrb r2, [r1] ldr r3, [r1, #0x01] /* BE:r3 = 1234 LE:r3 = 4321 */ ldr ip, [r1, #0x05] /* BE:ip = 5678 LE:ip = 8765 */ ldr r1, [r1, #0x09] /* BE:r1 = 9ABx LE:r1 = xBA9 */ strb r2, [r0] #ifdef __ARMEB__ mov r2, r3, lsr #16 /* r2 = ..12 */ strh r2, [r0, #0x01] mov r3, r3, lsl #16 /* r3 = 34.. */ orr r3, r3, ip, lsr #16 /* r3 = 3456 */ mov ip, ip, lsl #16 /* ip = 78.. */ orr ip, ip, r1, lsr #16 /* ip = 789A */ mov r1, r1, lsr #8 /* r1 = .9AB */ #else strh r3, [r0, #0x01] mov r3, r3, lsr #16 /* r3 = ..43 */ orr r3, r3, ip, lsl #16 /* r3 = 6543 */ mov ip, ip, lsr #16 /* ip = ..87 */ orr ip, ip, r1, lsl #16 /* ip = A987 */ mov r1, r1, lsr #16 /* r1 = ..xB */ #endif str r3, [r0, #0x03] str ip, [r0, #0x07] strb r1, [r0, #0x0b] RET LMEMCPY_C_PAD /* * 1000: dst is 16-bit aligned, src is 32-bit aligned */ ldr ip, [r1] /* BE:ip = 0123 LE:ip = 3210 */ ldr r3, [r1, #0x04] /* BE:r3 = 4567 LE:r3 = 7654 */ ldr r2, [r1, #0x08] /* BE:r2 = 89AB LE:r2 = BA98 */ mov r1, ip, lsr #16 /* BE:r1 = ..01 LE:r1 = ..32 */ #ifdef __ARMEB__ strh r1, [r0] mov r1, ip, lsl #16 /* r1 = 23.. */ orr r1, r1, r3, lsr #16 /* r1 = 2345 */ mov r3, r3, lsl #16 /* r3 = 67.. */ orr r3, r3, r2, lsr #16 /* r3 = 6789 */ #else strh ip, [r0] orr r1, r1, r3, lsl #16 /* r1 = 5432 */ mov r3, r3, lsr #16 /* r3 = ..76 */ orr r3, r3, r2, lsl #16 /* r3 = 9876 */ mov r2, r2, lsr #16 /* r2 = ..BA */ #endif str r1, [r0, #0x02] str r3, [r0, #0x06] strh r2, [r0, #0x0a] RET LMEMCPY_C_PAD /* * 1001: dst is 16-bit aligned, src is 8-bit aligned (byte 1) */ ldr r2, [r1, #-1] /* BE:r2 = x012 LE:r2 = 210x */ ldr r3, [r1, #0x03] /* BE:r3 = 3456 LE:r3 = 6543 */ mov ip, r2, lsr #8 /* BE:ip = .x01 LE:ip = .210 */ strh ip, [r0] ldr ip, [r1, #0x07] /* BE:ip = 789A LE:ip = A987 */ ldrb r1, [r1, #0x0b] /* r1 = ...B */ #ifdef __ARMEB__ mov r2, r2, lsl #24 /* r2 = 2... */ orr r2, r2, r3, lsr #8 /* r2 = 2345 */ mov r3, r3, lsl #24 /* r3 = 6... */ orr r3, r3, ip, lsr #8 /* r3 = 6789 */ orr r1, r1, ip, lsl #8 /* r1 = 89AB */ #else mov r2, r2, lsr #24 /* r2 = ...2 */ orr r2, r2, r3, lsl #8 /* r2 = 5432 */ mov r3, r3, lsr #24 /* r3 = ...6 */ orr r3, r3, ip, lsl #8 /* r3 = 9876 */ mov r1, r1, lsl #8 /* r1 = ..B. */ orr r1, r1, ip, lsr #24 /* r1 = ..BA */ #endif str r2, [r0, #0x02] str r3, [r0, #0x06] strh r1, [r0, #0x0a] RET LMEMCPY_C_PAD /* * 1010: dst is 16-bit aligned, src is 16-bit aligned */ ldrh r2, [r1] ldr r3, [r1, #0x02] ldr ip, [r1, #0x06] ldrh r1, [r1, #0x0a] strh r2, [r0] str r3, [r0, #0x02] str ip, [r0, #0x06] strh r1, [r0, #0x0a] RET LMEMCPY_C_PAD /* * 1011: dst is 16-bit aligned, src is 8-bit aligned (byte 3) */ ldr r2, [r1, #0x09] /* BE:r2 = 9ABx LE:r2 = xBA9 */ ldr r3, [r1, #0x05] /* BE:r3 = 5678 LE:r3 = 8765 */ mov ip, r2, lsr #8 /* BE:ip = .9AB LE:ip = .xBA */ strh ip, [r0, #0x0a] ldr ip, [r1, #0x01] /* BE:ip = 1234 LE:ip = 4321 */ ldrb r1, [r1] /* r1 = ...0 */ #ifdef __ARMEB__ mov r2, r2, lsr #24 /* r2 = ...9 */ orr r2, r2, r3, lsl #8 /* r2 = 6789 */ mov r3, r3, lsr #24 /* r3 = ...5 */ orr r3, r3, ip, lsl #8 /* r3 = 2345 */ mov r1, r1, lsl #8 /* r1 = ..0. */ orr r1, r1, ip, lsr #24 /* r1 = ..01 */ #else mov r2, r2, lsl #24 /* r2 = 9... */ orr r2, r2, r3, lsr #8 /* r2 = 9876 */ mov r3, r3, lsl #24 /* r3 = 5... */ orr r3, r3, ip, lsr #8 /* r3 = 5432 */ orr r1, r1, ip, lsl #8 /* r1 = 3210 */ #endif str r2, [r0, #0x06] str r3, [r0, #0x02] strh r1, [r0] RET LMEMCPY_C_PAD /* * 1100: dst is 8-bit aligned (byte 3), src is 32-bit aligned */ ldr r2, [r1] /* BE:r2 = 0123 LE:r2 = 3210 */ ldr ip, [r1, #0x04] /* BE:ip = 4567 LE:ip = 7654 */ ldr r1, [r1, #0x08] /* BE:r1 = 89AB LE:r1 = BA98 */ #ifdef __ARMEB__ mov r3, r2, lsr #24 /* r3 = ...0 */ strb r3, [r0] mov r2, r2, lsl #8 /* r2 = 123. */ orr r2, r2, ip, lsr #24 /* r2 = 1234 */ str r2, [r0, #0x01] mov r2, ip, lsl #8 /* r2 = 567. */ orr r2, r2, r1, lsr #24 /* r2 = 5678 */ str r2, [r0, #0x05] mov r2, r1, lsr #8 /* r2 = ..9A */ strh r2, [r0, #0x09] strb r1, [r0, #0x0b] #else strb r2, [r0] mov r3, r2, lsr #8 /* r3 = .321 */ orr r3, r3, ip, lsl #24 /* r3 = 4321 */ str r3, [r0, #0x01] mov r3, ip, lsr #8 /* r3 = .765 */ orr r3, r3, r1, lsl #24 /* r3 = 8765 */ str r3, [r0, #0x05] mov r1, r1, lsr #8 /* r1 = .BA9 */ strh r1, [r0, #0x09] mov r1, r1, lsr #16 /* r1 = ...B */ strb r1, [r0, #0x0b] #endif RET LMEMCPY_C_PAD /* * 1101: dst is 8-bit aligned (byte 3), src is 8-bit aligned (byte 1) */ ldrb r2, [r1, #0x0b] /* r2 = ...B */ ldr r3, [r1, #0x07] /* BE:r3 = 789A LE:r3 = A987 */ ldr ip, [r1, #0x03] /* BE:ip = 3456 LE:ip = 6543 */ ldr r1, [r1, #-1] /* BE:r1 = x012 LE:r1 = 210x */ strb r2, [r0, #0x0b] #ifdef __ARMEB__ strh r3, [r0, #0x09] mov r3, r3, lsr #16 /* r3 = ..78 */ orr r3, r3, ip, lsl #16 /* r3 = 5678 */ mov ip, ip, lsr #16 /* ip = ..34 */ orr ip, ip, r1, lsl #16 /* ip = 1234 */ mov r1, r1, lsr #16 /* r1 = ..x0 */ #else mov r2, r3, lsr #16 /* r2 = ..A9 */ strh r2, [r0, #0x09] mov r3, r3, lsl #16 /* r3 = 87.. */ orr r3, r3, ip, lsr #16 /* r3 = 8765 */ mov ip, ip, lsl #16 /* ip = 43.. */ orr ip, ip, r1, lsr #16 /* ip = 4321 */ mov r1, r1, lsr #8 /* r1 = .210 */ #endif str r3, [r0, #0x05] str ip, [r0, #0x01] strb r1, [r0] RET LMEMCPY_C_PAD /* * 1110: dst is 8-bit aligned (byte 3), src is 16-bit aligned */ #ifdef __ARMEB__ ldrh r2, [r1, #0x0a] /* r2 = ..AB */ ldr ip, [r1, #0x06] /* ip = 6789 */ ldr r3, [r1, #0x02] /* r3 = 2345 */ ldrh r1, [r1] /* r1 = ..01 */ strb r2, [r0, #0x0b] mov r2, r2, lsr #8 /* r2 = ...A */ orr r2, r2, ip, lsl #8 /* r2 = 789A */ mov ip, ip, lsr #8 /* ip = .678 */ orr ip, ip, r3, lsl #24 /* ip = 5678 */ mov r3, r3, lsr #8 /* r3 = .234 */ orr r3, r3, r1, lsl #24 /* r3 = 1234 */ mov r1, r1, lsr #8 /* r1 = ...0 */ strb r1, [r0] str r3, [r0, #0x01] str ip, [r0, #0x05] strh r2, [r0, #0x09] #else ldrh r2, [r1] /* r2 = ..10 */ ldr r3, [r1, #0x02] /* r3 = 5432 */ ldr ip, [r1, #0x06] /* ip = 9876 */ ldrh r1, [r1, #0x0a] /* r1 = ..BA */ strb r2, [r0] mov r2, r2, lsr #8 /* r2 = ...1 */ orr r2, r2, r3, lsl #8 /* r2 = 4321 */ mov r3, r3, lsr #24 /* r3 = ...5 */ orr r3, r3, ip, lsl #8 /* r3 = 8765 */ mov ip, ip, lsr #24 /* ip = ...9 */ orr ip, ip, r1, lsl #8 /* ip = .BA9 */ mov r1, r1, lsr #8 /* r1 = ...B */ str r2, [r0, #0x01] str r3, [r0, #0x05] strh ip, [r0, #0x09] strb r1, [r0, #0x0b] #endif RET LMEMCPY_C_PAD /* * 1111: dst is 8-bit aligned (byte 3), src is 8-bit aligned (byte 3) */ ldrb r2, [r1] ldr r3, [r1, #0x01] ldr ip, [r1, #0x05] strb r2, [r0] ldrh r2, [r1, #0x09] ldrb r1, [r1, #0x0b] str r3, [r0, #0x01] str ip, [r0, #0x05] strh r2, [r0, #0x09] strb r1, [r0, #0x0b] RET END(memcpy) #endif /* _ARM_ARCH_5E */ #ifdef GPROF ENTRY(user) nop END(user) ENTRY(btrap) nop END(btrap) ENTRY(etrap) nop END(etrap) ENTRY(bintr) nop END(bintr) ENTRY(eintr) nop END(eintr) #endif