1#include "arm_asm.h" 2#include "arm_arch.h" 3 4#if defined(__thumb2__) 5.syntax unified 6.thumb 7#else 8.code 32 9#endif 10 11.text 12.type mul_1x1_ialu,%function 13.align 5 14mul_1x1_ialu: 15 mov r4,#0 16 bic r5,r1,#3<<30 @ a1=a&0x3fffffff 17 str r4,[sp,#0] @ tab[0]=0 18 add r6,r5,r5 @ a2=a1<<1 19 str r5,[sp,#4] @ tab[1]=a1 20 eor r7,r5,r6 @ a1^a2 21 str r6,[sp,#8] @ tab[2]=a2 22 mov r8,r5,lsl#2 @ a4=a1<<2 23 str r7,[sp,#12] @ tab[3]=a1^a2 24 eor r9,r5,r8 @ a1^a4 25 str r8,[sp,#16] @ tab[4]=a4 26 eor r4,r6,r8 @ a2^a4 27 str r9,[sp,#20] @ tab[5]=a1^a4 28 eor r7,r7,r8 @ a1^a2^a4 29 str r4,[sp,#24] @ tab[6]=a2^a4 30 and r8,r12,r0,lsl#2 31 str r7,[sp,#28] @ tab[7]=a1^a2^a4 32 33 and r9,r12,r0,lsr#1 34 ldr r5,[sp,r8] @ tab[b & 0x7] 35 and r8,r12,r0,lsr#4 36 ldr r7,[sp,r9] @ tab[b >> 3 & 0x7] 37 and r9,r12,r0,lsr#7 38 ldr r6,[sp,r8] @ tab[b >> 6 & 0x7] 39 eor r5,r5,r7,lsl#3 @ stall 40 mov r4,r7,lsr#29 41 ldr r7,[sp,r9] @ tab[b >> 9 & 0x7] 42 43 and r8,r12,r0,lsr#10 44 eor r5,r5,r6,lsl#6 45 eor r4,r4,r6,lsr#26 46 ldr r6,[sp,r8] @ tab[b >> 12 & 0x7] 47 48 and r9,r12,r0,lsr#13 49 eor r5,r5,r7,lsl#9 50 eor r4,r4,r7,lsr#23 51 ldr r7,[sp,r9] @ tab[b >> 15 & 0x7] 52 53 and r8,r12,r0,lsr#16 54 eor r5,r5,r6,lsl#12 55 eor r4,r4,r6,lsr#20 56 ldr r6,[sp,r8] @ tab[b >> 18 & 0x7] 57 58 and r9,r12,r0,lsr#19 59 eor r5,r5,r7,lsl#15 60 eor r4,r4,r7,lsr#17 61 ldr r7,[sp,r9] @ tab[b >> 21 & 0x7] 62 63 and r8,r12,r0,lsr#22 64 eor r5,r5,r6,lsl#18 65 eor r4,r4,r6,lsr#14 66 ldr r6,[sp,r8] @ tab[b >> 24 & 0x7] 67 68 and r9,r12,r0,lsr#25 69 eor r5,r5,r7,lsl#21 70 eor r4,r4,r7,lsr#11 71 ldr r7,[sp,r9] @ tab[b >> 27 & 0x7] 72 73 tst r1,#1<<30 74 and r8,r12,r0,lsr#28 75 eor r5,r5,r6,lsl#24 76 eor r4,r4,r6,lsr#8 77 ldr r6,[sp,r8] @ tab[b >> 30 ] 78 79#ifdef __thumb2__ 80 itt ne 81#endif 82 eorne r5,r5,r0,lsl#30 83 eorne r4,r4,r0,lsr#2 84 tst r1,#1<<31 85 eor r5,r5,r7,lsl#27 86 eor r4,r4,r7,lsr#5 87#ifdef __thumb2__ 88 itt ne 89#endif 90 eorne r5,r5,r0,lsl#31 91 eorne r4,r4,r0,lsr#1 92 eor r5,r5,r6,lsl#30 93 eor r4,r4,r6,lsr#2 94 95 mov pc,lr 96.size mul_1x1_ialu,.-mul_1x1_ialu 97.globl bn_GF2m_mul_2x2 98.type bn_GF2m_mul_2x2,%function 99.align 5 100bn_GF2m_mul_2x2: 101#if __ARM_MAX_ARCH__>=7 102 stmdb sp!,{r10,lr} 103 ldr r12,.LOPENSSL_armcap 104# if !defined(_WIN32) 105 adr r10,.LOPENSSL_armcap 106 ldr r12,[r12,r10] 107# endif 108# if defined(__APPLE__) || defined(_WIN32) 109 ldr r12,[r12] 110# endif 111 tst r12,#ARMV7_NEON 112 itt ne 113 ldrne r10,[sp],#8 114 bne .LNEON 115 stmdb sp!,{r4,r5,r6,r7,r8,r9} 116#else 117 stmdb sp!,{r4,r5,r6,r7,r8,r9,r10,lr} 118#endif 119 mov r10,r0 @ reassign 1st argument 120 mov r0,r3 @ r0=b1 121 sub r7,sp,#36 122 mov r8,sp 123 and r7,r7,#-32 124 ldr r3,[sp,#32] @ load b0 125 mov r12,#7<<2 126 mov sp,r7 @ allocate tab[8] 127 str r8,[r7,#32] 128 129 bl mul_1x1_ialu @ a1·b1 130 str r5,[r10,#8] 131 str r4,[r10,#12] 132 133 eor r0,r0,r3 @ flip b0 and b1 134 eor r1,r1,r2 @ flip a0 and a1 135 eor r3,r3,r0 136 eor r2,r2,r1 137 eor r0,r0,r3 138 eor r1,r1,r2 139 bl mul_1x1_ialu @ a0·b0 140 str r5,[r10] 141 str r4,[r10,#4] 142 143 eor r1,r1,r2 144 eor r0,r0,r3 145 bl mul_1x1_ialu @ (a1+a0)·(b1+b0) 146 ldmia r10,{r6,r7,r8,r9} 147 eor r5,r5,r4 148 ldr sp,[sp,#32] @ destroy tab[8] 149 eor r4,r4,r7 150 eor r5,r5,r6 151 eor r4,r4,r8 152 eor r5,r5,r9 153 eor r4,r4,r9 154 str r4,[r10,#8] 155 eor r5,r5,r4 156 str r5,[r10,#4] 157 158#if __ARM_ARCH__>=5 159 ldmia sp!,{r4,r5,r6,r7,r8,r9,r10,pc} 160#else 161 ldmia sp!,{r4,r5,r6,r7,r8,r9,r10,lr} 162 tst lr,#1 163 moveq pc,lr @ be binary compatible with V4, yet 164.word 0xe12fff1e @ interoperable with Thumb ISA:-) 165#endif 166#if __ARM_MAX_ARCH__>=7 167.arch armv7-a 168.fpu neon 169 170.align 5 171.LNEON: 172 ldr r12, [sp] @ 5th argument 173 vmov d26, r2, r1 174 vmov d27, r12, r3 175 vmov.i64 d28, #0x0000ffffffffffff 176 vmov.i64 d29, #0x00000000ffffffff 177 vmov.i64 d30, #0x000000000000ffff 178 179 vext.8 d2, d26, d26, #1 @ A1 180 vmull.p8 q1, d2, d27 @ F = A1*B 181 vext.8 d0, d27, d27, #1 @ B1 182 vmull.p8 q0, d26, d0 @ E = A*B1 183 vext.8 d4, d26, d26, #2 @ A2 184 vmull.p8 q2, d4, d27 @ H = A2*B 185 vext.8 d16, d27, d27, #2 @ B2 186 vmull.p8 q8, d26, d16 @ G = A*B2 187 vext.8 d6, d26, d26, #3 @ A3 188 veor q1, q1, q0 @ L = E + F 189 vmull.p8 q3, d6, d27 @ J = A3*B 190 vext.8 d0, d27, d27, #3 @ B3 191 veor q2, q2, q8 @ M = G + H 192 vmull.p8 q0, d26, d0 @ I = A*B3 193 veor d2, d2, d3 @ t0 = (L) (P0 + P1) << 8 194 vand d3, d3, d28 195 vext.8 d16, d27, d27, #4 @ B4 196 veor d4, d4, d5 @ t1 = (M) (P2 + P3) << 16 197 vand d5, d5, d29 198 vmull.p8 q8, d26, d16 @ K = A*B4 199 veor q3, q3, q0 @ N = I + J 200 veor d2, d2, d3 201 veor d4, d4, d5 202 veor d6, d6, d7 @ t2 = (N) (P4 + P5) << 24 203 vand d7, d7, d30 204 vext.8 q1, q1, q1, #15 205 veor d16, d16, d17 @ t3 = (K) (P6 + P7) << 32 206 vmov.i64 d17, #0 207 vext.8 q2, q2, q2, #14 208 veor d6, d6, d7 209 vmull.p8 q0, d26, d27 @ D = A*B 210 vext.8 q8, q8, q8, #12 211 vext.8 q3, q3, q3, #13 212 veor q1, q1, q2 213 veor q3, q3, q8 214 veor q0, q0, q1 215 veor q0, q0, q3 216 217 vst1.32 {q0}, [r0] 218 RET @ bx lr 219#endif 220.size bn_GF2m_mul_2x2,.-bn_GF2m_mul_2x2 221#if __ARM_MAX_ARCH__>=7 222.align 5 223.LOPENSSL_armcap: 224# ifdef _WIN32 225.word OPENSSL_armcap_P 226# else 227.word OPENSSL_armcap_P-. 228# endif 229#endif 230.byte 71,70,40,50,94,109,41,32,77,117,108,116,105,112,108,105,99,97,116,105,111,110,32,102,111,114,32,65,82,77,118,52,47,78,69,79,78,44,32,67,82,89,80,84,79,71,65,77,83,32,98,121,32,60,97,112,112,114,111,64,111,112,101,110,115,115,108,46,111,114,103,62,0 231.align 2 232.align 5 233 234#if __ARM_MAX_ARCH__>=7 235.comm OPENSSL_armcap_P,4,4 236#endif 237