/* * Cast5 Cipher 16-way parallel algorithm (AVX/x86_64) * * Copyright (C) 2012 Johannes Goetzfried * * * Copyright © 2012 Jussi Kivilinna * * This program is free software; you can redistribute it and/or modify * it under the terms of the GNU General Public License as published by * the Free Software Foundation; either version 2 of the License, or * (at your option) any later version. * * This program is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the * GNU General Public License for more details. * * You should have received a copy of the GNU General Public License * along with this program; if not, write to the Free Software * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 * USA * */ #include .file "cast5-avx-x86_64-asm_64.S" .extern cast_s1 .extern cast_s2 .extern cast_s3 .extern cast_s4 /* structure of crypto context */ #define km 0 #define kr (16*4) #define rr ((16*4)+16) /* s-boxes */ #define s1 cast_s1 #define s2 cast_s2 #define s3 cast_s3 #define s4 cast_s4 /********************************************************************** 16-way AVX cast5 **********************************************************************/ #define CTX %rdi #define RL1 %xmm0 #define RR1 %xmm1 #define RL2 %xmm2 #define RR2 %xmm3 #define RL3 %xmm4 #define RR3 %xmm5 #define RL4 %xmm6 #define RR4 %xmm7 #define RX %xmm8 #define RKM %xmm9 #define RKR %xmm10 #define RKRF %xmm11 #define RKRR %xmm12 #define R32 %xmm13 #define R1ST %xmm14 #define RTMP %xmm15 #define RID1 %rbp #define RID1d %ebp #define RID2 %rsi #define RID2d %esi #define RGI1 %rdx #define RGI1bl %dl #define RGI1bh %dh #define RGI2 %rcx #define RGI2bl %cl #define RGI2bh %ch #define RGI3 %rax #define RGI3bl %al #define RGI3bh %ah #define RGI4 %rbx #define RGI4bl %bl #define RGI4bh %bh #define RFS1 %r8 #define RFS1d %r8d #define RFS2 %r9 #define RFS2d %r9d #define RFS3 %r10 #define RFS3d %r10d #define lookup_32bit(src, dst, op1, op2, op3, interleave_op, il_reg) \ movzbl src ## bh, RID1d; \ movzbl src ## bl, RID2d; \ shrq $16, src; \ movl s1(, RID1, 4), dst ## d; \ op1 s2(, RID2, 4), dst ## d; \ movzbl src ## bh, RID1d; \ movzbl src ## bl, RID2d; \ interleave_op(il_reg); \ op2 s3(, RID1, 4), dst ## d; \ op3 s4(, RID2, 4), dst ## d; #define dummy(d) /* do nothing */ #define shr_next(reg) \ shrq $16, reg; #define F_head(a, x, gi1, gi2, op0) \ op0 a, RKM, x; \ vpslld RKRF, x, RTMP; \ vpsrld RKRR, x, x; \ vpor RTMP, x, x; \ \ vmovq x, gi1; \ vpextrq $1, x, gi2; #define F_tail(a, x, gi1, gi2, op1, op2, op3) \ lookup_32bit(##gi1, RFS1, op1, op2, op3, shr_next, ##gi1); \ lookup_32bit(##gi2, RFS3, op1, op2, op3, shr_next, ##gi2); \ \ lookup_32bit(##gi1, RFS2, op1, op2, op3, dummy, none); \ shlq $32, RFS2; \ orq RFS1, RFS2; \ lookup_32bit(##gi2, RFS1, op1, op2, op3, dummy, none); \ shlq $32, RFS1; \ orq RFS1, RFS3; \ \ vmovq RFS2, x; \ vpinsrq $1, RFS3, x, x; #define F_2(a1, b1, a2, b2, op0, op1, op2, op3) \ F_head(b1, RX, RGI1, RGI2, op0); \ F_head(b2, RX, RGI3, RGI4, op0); \ \ F_tail(b1, RX, RGI1, RGI2, op1, op2, op3); \ F_tail(b2, RTMP, RGI3, RGI4, op1, op2, op3); \ \ vpxor a1, RX, a1; \ vpxor a2, RTMP, a2; #define F1_2(a1, b1, a2, b2) \ F_2(a1, b1, a2, b2, vpaddd, xorl, subl, addl) #define F2_2(a1, b1, a2, b2) \ F_2(a1, b1, a2, b2, vpxor, subl, addl, xorl) #define F3_2(a1, b1, a2, b2) \ F_2(a1, b1, a2, b2, vpsubd, addl, xorl, subl) #define subround(a1, b1, a2, b2, f) \ F ## f ## _2(a1, b1, a2, b2); #define round(l, r, n, f) \ vbroadcastss (km+(4*n))(CTX), RKM; \ vpand R1ST, RKR, RKRF; \ vpsubq RKRF, R32, RKRR; \ vpsrldq $1, RKR, RKR; \ subround(l ## 1, r ## 1, l ## 2, r ## 2, f); \ subround(l ## 3, r ## 3, l ## 4, r ## 4, f); #define enc_preload_rkr() \ vbroadcastss .L16_mask, RKR; \ /* add 16-bit rotation to key rotations (mod 32) */ \ vpxor kr(CTX), RKR, RKR; #define dec_preload_rkr() \ vbroadcastss .L16_mask, RKR; \ /* add 16-bit rotation to key rotations (mod 32) */ \ vpxor kr(CTX), RKR, RKR; \ vpshufb .Lbswap128_mask, RKR, RKR; #define transpose_2x4(x0, x1, t0, t1) \ vpunpckldq x1, x0, t0; \ vpunpckhdq x1, x0, t1; \ \ vpunpcklqdq t1, t0, x0; \ vpunpckhqdq t1, t0, x1; #define inpack_blocks(x0, x1, t0, t1, rmask) \ vpshufb rmask, x0, x0; \ vpshufb rmask, x1, x1; \ \ transpose_2x4(x0, x1, t0, t1) #define outunpack_blocks(x0, x1, t0, t1, rmask) \ transpose_2x4(x0, x1, t0, t1) \ \ vpshufb rmask, x0, x0; \ vpshufb rmask, x1, x1; .data .align 16 .Lbswap_mask: .byte 3, 2, 1, 0, 7, 6, 5, 4, 11, 10, 9, 8, 15, 14, 13, 12 .Lbswap128_mask: .byte 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0 .Lbswap_iv_mask: .byte 7, 6, 5, 4, 3, 2, 1, 0, 7, 6, 5, 4, 3, 2, 1, 0 .L16_mask: .byte 16, 16, 16, 16 .L32_mask: .byte 32, 0, 0, 0 .Lfirst_mask: .byte 0x1f, 0, 0, 0 .text .align 16 __cast5_enc_blk16: /* input: * %rdi: ctx, CTX * RL1: blocks 1 and 2 * RR1: blocks 3 and 4 * RL2: blocks 5 and 6 * RR2: blocks 7 and 8 * RL3: blocks 9 and 10 * RR3: blocks 11 and 12 * RL4: blocks 13 and 14 * RR4: blocks 15 and 16 * output: * RL1: encrypted blocks 1 and 2 * RR1: encrypted blocks 3 and 4 * RL2: encrypted blocks 5 and 6 * RR2: encrypted blocks 7 and 8 * RL3: encrypted blocks 9 and 10 * RR3: encrypted blocks 11 and 12 * RL4: encrypted blocks 13 and 14 * RR4: encrypted blocks 15 and 16 */ pushq %rbp; pushq %rbx; vmovdqa .Lbswap_mask, RKM; vmovd .Lfirst_mask, R1ST; vmovd .L32_mask, R32; enc_preload_rkr(); inpack_blocks(RL1, RR1, RTMP, RX, RKM); inpack_blocks(RL2, RR2, RTMP, RX, RKM); inpack_blocks(RL3, RR3, RTMP, RX, RKM); inpack_blocks(RL4, RR4, RTMP, RX, RKM); round(RL, RR, 0, 1); round(RR, RL, 1, 2); round(RL, RR, 2, 3); round(RR, RL, 3, 1); round(RL, RR, 4, 2); round(RR, RL, 5, 3); round(RL, RR, 6, 1); round(RR, RL, 7, 2); round(RL, RR, 8, 3); round(RR, RL, 9, 1); round(RL, RR, 10, 2); round(RR, RL, 11, 3); movzbl rr(CTX), %eax; testl %eax, %eax; jnz .L__skip_enc; round(RL, RR, 12, 1); round(RR, RL, 13, 2); round(RL, RR, 14, 3); round(RR, RL, 15, 1); .L__skip_enc: popq %rbx; popq %rbp; vmovdqa .Lbswap_mask, RKM; outunpack_blocks(RR1, RL1, RTMP, RX, RKM); outunpack_blocks(RR2, RL2, RTMP, RX, RKM); outunpack_blocks(RR3, RL3, RTMP, RX, RKM); outunpack_blocks(RR4, RL4, RTMP, RX, RKM); ret; ENDPROC(__cast5_enc_blk16) .align 16 __cast5_dec_blk16: /* input: * %rdi: ctx, CTX * RL1: encrypted blocks 1 and 2 * RR1: encrypted blocks 3 and 4 * RL2: encrypted blocks 5 and 6 * RR2: encrypted blocks 7 and 8 * RL3: encrypted blocks 9 and 10 * RR3: encrypted blocks 11 and 12 * RL4: encrypted blocks 13 and 14 * RR4: encrypted blocks 15 and 16 * output: * RL1: decrypted blocks 1 and 2 * RR1: decrypted blocks 3 and 4 * RL2: decrypted blocks 5 and 6 * RR2: decrypted blocks 7 and 8 * RL3: decrypted blocks 9 and 10 * RR3: decrypted blocks 11 and 12 * RL4: decrypted blocks 13 and 14 * RR4: decrypted blocks 15 and 16 */ pushq %rbp; pushq %rbx; vmovdqa .Lbswap_mask, RKM; vmovd .Lfirst_mask, R1ST; vmovd .L32_mask, R32; dec_preload_rkr(); inpack_blocks(RL1, RR1, RTMP, RX, RKM); inpack_blocks(RL2, RR2, RTMP, RX, RKM); inpack_blocks(RL3, RR3, RTMP, RX, RKM); inpack_blocks(RL4, RR4, RTMP, RX, RKM); movzbl rr(CTX), %eax; testl %eax, %eax; jnz .L__skip_dec; round(RL, RR, 15, 1); round(RR, RL, 14, 3); round(RL, RR, 13, 2); round(RR, RL, 12, 1); .L__dec_tail: round(RL, RR, 11, 3); round(RR, RL, 10, 2); round(RL, RR, 9, 1); round(RR, RL, 8, 3); round(RL, RR, 7, 2); round(RR, RL, 6, 1); round(RL, RR, 5, 3); round(RR, RL, 4, 2); round(RL, RR, 3, 1); round(RR, RL, 2, 3); round(RL, RR, 1, 2); round(RR, RL, 0, 1); vmovdqa .Lbswap_mask, RKM; popq %rbx; popq %rbp; outunpack_blocks(RR1, RL1, RTMP, RX, RKM); outunpack_blocks(RR2, RL2, RTMP, RX, RKM); outunpack_blocks(RR3, RL3, RTMP, RX, RKM); outunpack_blocks(RR4, RL4, RTMP, RX, RKM); ret; .L__skip_dec: vpsrldq $4, RKR, RKR; jmp .L__dec_tail; ENDPROC(__cast5_dec_blk16) ENTRY(cast5_ecb_enc_16way) /* input: * %rdi: ctx, CTX * %rsi: dst * %rdx: src */ movq %rsi, %r11; vmovdqu (0*4*4)(%rdx), RL1; vmovdqu (1*4*4)(%rdx), RR1; vmovdqu (2*4*4)(%rdx), RL2; vmovdqu (3*4*4)(%rdx), RR2; vmovdqu (4*4*4)(%rdx), RL3; vmovdqu (5*4*4)(%rdx), RR3; vmovdqu (6*4*4)(%rdx), RL4; vmovdqu (7*4*4)(%rdx), RR4; call __cast5_enc_blk16; vmovdqu RR1, (0*4*4)(%r11); vmovdqu RL1, (1*4*4)(%r11); vmovdqu RR2, (2*4*4)(%r11); vmovdqu RL2, (3*4*4)(%r11); vmovdqu RR3, (4*4*4)(%r11); vmovdqu RL3, (5*4*4)(%r11); vmovdqu RR4, (6*4*4)(%r11); vmovdqu RL4, (7*4*4)(%r11); ret; ENDPROC(cast5_ecb_enc_16way) ENTRY(cast5_ecb_dec_16way) /* input: * %rdi: ctx, CTX * %rsi: dst * %rdx: src */ movq %rsi, %r11; vmovdqu (0*4*4)(%rdx), RL1; vmovdqu (1*4*4)(%rdx), RR1; vmovdqu (2*4*4)(%rdx), RL2; vmovdqu (3*4*4)(%rdx), RR2; vmovdqu (4*4*4)(%rdx), RL3; vmovdqu (5*4*4)(%rdx), RR3; vmovdqu (6*4*4)(%rdx), RL4; vmovdqu (7*4*4)(%rdx), RR4; call __cast5_dec_blk16; vmovdqu RR1, (0*4*4)(%r11); vmovdqu RL1, (1*4*4)(%r11); vmovdqu RR2, (2*4*4)(%r11); vmovdqu RL2, (3*4*4)(%r11); vmovdqu RR3, (4*4*4)(%r11); vmovdqu RL3, (5*4*4)(%r11); vmovdqu RR4, (6*4*4)(%r11); vmovdqu RL4, (7*4*4)(%r11); ret; ENDPROC(cast5_ecb_dec_16way) ENTRY(cast5_cbc_dec_16way) /* input: * %rdi: ctx, CTX * %rsi: dst * %rdx: src */ pushq %r12; movq %rsi, %r11; movq %rdx, %r12; vmovdqu (0*16)(%rdx), RL1; vmovdqu (1*16)(%rdx), RR1; vmovdqu (2*16)(%rdx), RL2; vmovdqu (3*16)(%rdx), RR2; vmovdqu (4*16)(%rdx), RL3; vmovdqu (5*16)(%rdx), RR3; vmovdqu (6*16)(%rdx), RL4; vmovdqu (7*16)(%rdx), RR4; call __cast5_dec_blk16; /* xor with src */ vmovq (%r12), RX; vpshufd $0x4f, RX, RX; vpxor RX, RR1, RR1; vpxor 0*16+8(%r12), RL1, RL1; vpxor 1*16+8(%r12), RR2, RR2; vpxor 2*16+8(%r12), RL2, RL2; vpxor 3*16+8(%r12), RR3, RR3; vpxor 4*16+8(%r12), RL3, RL3; vpxor 5*16+8(%r12), RR4, RR4; vpxor 6*16+8(%r12), RL4, RL4; vmovdqu RR1, (0*16)(%r11); vmovdqu RL1, (1*16)(%r11); vmovdqu RR2, (2*16)(%r11); vmovdqu RL2, (3*16)(%r11); vmovdqu RR3, (4*16)(%r11); vmovdqu RL3, (5*16)(%r11); vmovdqu RR4, (6*16)(%r11); vmovdqu RL4, (7*16)(%r11); popq %r12; ret; ENDPROC(cast5_cbc_dec_16way) ENTRY(cast5_ctr_16way) /* input: * %rdi: ctx, CTX * %rsi: dst * %rdx: src * %rcx: iv (big endian, 64bit) */ pushq %r12; movq %rsi, %r11; movq %rdx, %r12; vpcmpeqd RTMP, RTMP, RTMP; vpsrldq $8, RTMP, RTMP; /* low: -1, high: 0 */ vpcmpeqd RKR, RKR, RKR; vpaddq RKR, RKR, RKR; /* low: -2, high: -2 */ vmovdqa .Lbswap_iv_mask, R1ST; vmovdqa .Lbswap128_mask, RKM; /* load IV and byteswap */ vmovq (%rcx), RX; vpshufb R1ST, RX, RX; /* construct IVs */ vpsubq RTMP, RX, RX; /* le: IV1, IV0 */ vpshufb RKM, RX, RL1; /* be: IV0, IV1 */ vpsubq RKR, RX, RX; vpshufb RKM, RX, RR1; /* be: IV2, IV3 */ vpsubq RKR, RX, RX; vpshufb RKM, RX, RL2; /* be: IV4, IV5 */ vpsubq RKR, RX, RX; vpshufb RKM, RX, RR2; /* be: IV6, IV7 */ vpsubq RKR, RX, RX; vpshufb RKM, RX, RL3; /* be: IV8, IV9 */ vpsubq RKR, RX, RX; vpshufb RKM, RX, RR3; /* be: IV10, IV11 */ vpsubq RKR, RX, RX; vpshufb RKM, RX, RL4; /* be: IV12, IV13 */ vpsubq RKR, RX, RX; vpshufb RKM, RX, RR4; /* be: IV14, IV15 */ /* store last IV */ vpsubq RTMP, RX, RX; /* le: IV16, IV14 */ vpshufb R1ST, RX, RX; /* be: IV16, IV16 */ vmovq RX, (%rcx); call __cast5_enc_blk16; /* dst = src ^ iv */ vpxor (0*16)(%r12), RR1, RR1; vpxor (1*16)(%r12), RL1, RL1; vpxor (2*16)(%r12), RR2, RR2; vpxor (3*16)(%r12), RL2, RL2; vpxor (4*16)(%r12), RR3, RR3; vpxor (5*16)(%r12), RL3, RL3; vpxor (6*16)(%r12), RR4, RR4; vpxor (7*16)(%r12), RL4, RL4; vmovdqu RR1, (0*16)(%r11); vmovdqu RL1, (1*16)(%r11); vmovdqu RR2, (2*16)(%r11); vmovdqu RL2, (3*16)(%r11); vmovdqu RR3, (4*16)(%r11); vmovdqu RL3, (5*16)(%r11); vmovdqu RR4, (6*16)(%r11); vmovdqu RL4, (7*16)(%r11); popq %r12; ret; ENDPROC(cast5_ctr_16way)