Commit 64b94cea authored by Jussi Kivilinna's avatar Jussi Kivilinna Committed by Herbert Xu

crypto: blowfish - add x86_64 assembly implementation

Patch adds x86_64 assembly implementation of blowfish. Two set of assembler
functions are provided. First set is regular 'one-block at time'
encrypt/decrypt functions. Second is 'four-block at time' functions that
gain performance increase on out-of-order CPUs. Performance of 4-way
functions should be equal to 1-way functions with in-order CPUs.

Summary of the tcrypt benchmarks:

Blowfish assembler vs blowfish C (256bit 8kb block ECB)
encrypt: 2.2x speed
decrypt: 2.3x speed

Blowfish assembler vs blowfish C (256bit 8kb block CBC)
encrypt: 1.12x speed
decrypt: 2.5x speed

Blowfish assembler vs blowfish C (256bit 8kb block CTR)
encrypt: 2.5x speed

Full output:
http://koti.mbnet.fi/axh/kernel/crypto/tcrypt-speed-blowfish-asm-x86_64.txt
http://koti.mbnet.fi/axh/kernel/crypto/tcrypt-speed-blowfish-c-x86_64.txt

Tests were run on:
 vendor_id	: AuthenticAMD
 cpu family	: 16
 model		: 10
 model name	: AMD Phenom(tm) II X6 1055T Processor
 stepping	: 0
Signed-off-by: default avatarJussi Kivilinna <jussi.kivilinna@mbnet.fi>
Signed-off-by: default avatarHerbert Xu <herbert@gondor.apana.org.au>
parent 7d47b86c
...@@ -7,6 +7,7 @@ obj-$(CONFIG_CRYPTO_TWOFISH_586) += twofish-i586.o ...@@ -7,6 +7,7 @@ obj-$(CONFIG_CRYPTO_TWOFISH_586) += twofish-i586.o
obj-$(CONFIG_CRYPTO_SALSA20_586) += salsa20-i586.o obj-$(CONFIG_CRYPTO_SALSA20_586) += salsa20-i586.o
obj-$(CONFIG_CRYPTO_AES_X86_64) += aes-x86_64.o obj-$(CONFIG_CRYPTO_AES_X86_64) += aes-x86_64.o
obj-$(CONFIG_CRYPTO_BLOWFISH_X86_64) += blowfish-x86_64.o
obj-$(CONFIG_CRYPTO_TWOFISH_X86_64) += twofish-x86_64.o obj-$(CONFIG_CRYPTO_TWOFISH_X86_64) += twofish-x86_64.o
obj-$(CONFIG_CRYPTO_SALSA20_X86_64) += salsa20-x86_64.o obj-$(CONFIG_CRYPTO_SALSA20_X86_64) += salsa20-x86_64.o
obj-$(CONFIG_CRYPTO_AES_NI_INTEL) += aesni-intel.o obj-$(CONFIG_CRYPTO_AES_NI_INTEL) += aesni-intel.o
...@@ -20,6 +21,7 @@ twofish-i586-y := twofish-i586-asm_32.o twofish_glue.o ...@@ -20,6 +21,7 @@ twofish-i586-y := twofish-i586-asm_32.o twofish_glue.o
salsa20-i586-y := salsa20-i586-asm_32.o salsa20_glue.o salsa20-i586-y := salsa20-i586-asm_32.o salsa20_glue.o
aes-x86_64-y := aes-x86_64-asm_64.o aes_glue.o aes-x86_64-y := aes-x86_64-asm_64.o aes_glue.o
blowfish-x86_64-y := blowfish-x86_64-asm_64.o blowfish_glue.o
twofish-x86_64-y := twofish-x86_64-asm_64.o twofish_glue.o twofish-x86_64-y := twofish-x86_64-asm_64.o twofish_glue.o
salsa20-x86_64-y := salsa20-x86_64-asm_64.o salsa20_glue.o salsa20-x86_64-y := salsa20-x86_64-asm_64.o salsa20_glue.o
......
/*
* Blowfish Cipher Algorithm (x86_64)
*
* Copyright (C) 2011 Jussi Kivilinna <jussi.kivilinna@mbnet.fi>
*
* 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
*
*/
.file "blowfish-x86_64-asm.S"
.text
/* structure of crypto context */
#define p 0
#define s0 ((16 + 2) * 4)
#define s1 ((16 + 2 + (1 * 256)) * 4)
#define s2 ((16 + 2 + (2 * 256)) * 4)
#define s3 ((16 + 2 + (3 * 256)) * 4)
/* register macros */
#define CTX %rdi
#define RIO %rsi
#define RX0 %rax
#define RX1 %rbx
#define RX2 %rcx
#define RX3 %rdx
#define RX0d %eax
#define RX1d %ebx
#define RX2d %ecx
#define RX3d %edx
#define RX0bl %al
#define RX1bl %bl
#define RX2bl %cl
#define RX3bl %dl
#define RX0bh %ah
#define RX1bh %bh
#define RX2bh %ch
#define RX3bh %dh
#define RT0 %rbp
#define RT1 %rsi
#define RT0d %ebp
#define RT1d %esi
#define RK0 %r8
#define RK1 %r9
#define RK2 %r10
#define RK3 %r11
#define RK0d %r8d
#define RK1d %r9d
#define RK2d %r10d
#define RK3d %r11d
#define RKEY %r12
/***********************************************************************
* 1-way blowfish
***********************************************************************/
#define F(x, k) \
rorq $16, x; \
movzbl x ## bh, RT0d; \
movzbl x ## bl, RT1d; \
rolq $16, x; \
movl s0(CTX,RT0,4), k ## d; \
addl s1(CTX,RT1,4), k ## d; \
movzbl x ## bh, RT0d; \
movzbl x ## bl, RT1d; \
rolq $32, x; \
xorl s2(CTX,RT0,4), k ## d; \
addl s3(CTX,RT1,4), k ## d; \
xorq k, x;
#define add_roundkey_enc(n) \
xorq p+4*(n)(CTX), RX0;
#define round_enc(n) \
add_roundkey_enc(n); \
\
F(RX0, RK0); \
F(RX0, RK0);
#define round_final_enc(n) \
xorq p+4*(n)(CTX), RX0;
#define add_roundkey_dec(n) \
movq p+4*(n-1)(CTX), RT0; \
rorq $32, RT0; \
xorq RT0, RX0;
#define round_dec(n) \
add_roundkey_dec(n); \
\
F(RX0, RK0); \
F(RX0, RK0); \
#define read_block() \
movq (RIO), RX0; \
rorq $32, RX0; \
bswapq RX0;
#define write_block() \
bswapq RX0; \
movq RX0, (RIO);
#define xor_block() \
bswapq RX0; \
xorq RX0, (RIO);
.align 8
.global __blowfish_enc_blk
.type __blowfish_enc_blk,@function;
__blowfish_enc_blk:
// input:
// %rdi: ctx, CTX
// %rsi: dst
// %rdx: src
// %rcx: bool xor
pushq %rbp;
pushq %rbx;
pushq %rsi;
pushq %rcx;
movq %rdx, RIO;
read_block();
round_enc(0);
round_enc(2);
round_enc(4);
round_enc(6);
round_enc(8);
round_enc(10);
round_enc(12);
round_enc(14);
add_roundkey_enc(16);
popq %rbp;
popq RIO;
test %bpl, %bpl;
jnz __enc_xor;
write_block();
__enc_ret:
popq %rbx;
popq %rbp;
ret;
__enc_xor:
xor_block();
jmp __enc_ret;
.align 8
.global blowfish_dec_blk
.type blowfish_dec_blk,@function;
blowfish_dec_blk:
// input:
// %rdi: ctx, CTX
// %rsi: dst
// %rdx: src
pushq %rbp;
pushq %rbx;
pushq %rsi;
movq %rdx, RIO;
read_block();
round_dec(17);
round_dec(15);
round_dec(13);
round_dec(11);
round_dec(9);
round_dec(7);
round_dec(5);
round_dec(3);
add_roundkey_dec(1);
popq RIO;
write_block();
popq %rbx;
popq %rbp;
ret;
/**********************************************************************
4-way blowfish, four blocks parallel
**********************************************************************/
#define add_preloaded_roundkey4() \
xorq RKEY, RX0; \
xorq RKEY, RX1; \
xorq RKEY, RX2; \
xorq RKEY, RX3;
#define preload_roundkey_enc(n) \
movq p+4*(n)(CTX), RKEY;
#define add_roundkey_enc4(n) \
add_preloaded_roundkey4(); \
preload_roundkey_enc(n + 2);
#define round_enc4(n) \
add_roundkey_enc4(n); \
\
F(RX0, RK0); \
F(RX1, RK1); \
F(RX2, RK2); \
F(RX3, RK3); \
\
F(RX0, RK0); \
F(RX1, RK1); \
F(RX2, RK2); \
F(RX3, RK3);
#define preload_roundkey_dec(n) \
movq p+4*((n)-1)(CTX), RKEY; \
rorq $32, RKEY;
#define add_roundkey_dec4(n) \
add_preloaded_roundkey4(); \
preload_roundkey_dec(n - 2);
#define round_dec4(n) \
add_roundkey_dec4(n); \
\
F(RX0, RK0); \
F(RX1, RK1); \
F(RX2, RK2); \
F(RX3, RK3); \
\
F(RX0, RK0); \
F(RX1, RK1); \
F(RX2, RK2); \
F(RX3, RK3);
#define read_block4() \
movq (RIO), RX0; \
rorq $32, RX0; \
bswapq RX0; \
\
movq 8(RIO), RX1; \
rorq $32, RX1; \
bswapq RX1; \
\
movq 16(RIO), RX2; \
rorq $32, RX2; \
bswapq RX2; \
\
movq 24(RIO), RX3; \
rorq $32, RX3; \
bswapq RX3;
#define write_block4() \
bswapq RX0; \
movq RX0, (RIO); \
\
bswapq RX1; \
movq RX1, 8(RIO); \
\
bswapq RX2; \
movq RX2, 16(RIO); \
\
bswapq RX3; \
movq RX3, 24(RIO);
#define xor_block4() \
bswapq RX0; \
xorq RX0, (RIO); \
\
bswapq RX1; \
xorq RX1, 8(RIO); \
\
bswapq RX2; \
xorq RX2, 16(RIO); \
\
bswapq RX3; \
xorq RX3, 24(RIO);
.align 8
.global __blowfish_enc_blk_4way
.type __blowfish_enc_blk_4way,@function;
__blowfish_enc_blk_4way:
// input:
// %rdi: ctx, CTX
// %rsi: dst
// %rdx: src
// %rcx: bool xor
pushq %rbp;
pushq %rbx;
pushq RKEY;
preload_roundkey_enc(0);
pushq %rsi;
pushq %rcx;
movq %rdx, RIO;
read_block4();
round_enc4(0);
round_enc4(2);
round_enc4(4);
round_enc4(6);
round_enc4(8);
round_enc4(10);
round_enc4(12);
round_enc4(14);
add_preloaded_roundkey4();
popq %rbp;
popq RIO;
test %bpl, %bpl;
jnz __enc_xor4;
write_block4();
__enc_ret4:
popq RKEY;
popq %rbx;
popq %rbp;
ret;
__enc_xor4:
xor_block4();
jmp __enc_ret4;
.align 8
.global blowfish_dec_blk_4way
.type blowfish_dec_blk_4way,@function;
blowfish_dec_blk_4way:
// input:
// %rdi: ctx, CTX
// %rsi: dst
// %rdx: src
pushq %rbp;
pushq %rbx;
pushq RKEY;
preload_roundkey_dec(17);
pushq %rsi;
movq %rdx, RIO;
read_block4();
round_dec4(17);
round_dec4(15);
round_dec4(13);
round_dec4(11);
round_dec4(9);
round_dec4(7);
round_dec4(5);
round_dec4(3);
add_preloaded_roundkey4();
popq RIO;
write_block4();
popq RKEY;
popq %rbx;
popq %rbp;
ret;
This diff is collapsed.
...@@ -620,6 +620,21 @@ config CRYPTO_BLOWFISH_COMMON ...@@ -620,6 +620,21 @@ config CRYPTO_BLOWFISH_COMMON
See also: See also:
<http://www.schneier.com/blowfish.html> <http://www.schneier.com/blowfish.html>
config CRYPTO_BLOWFISH_X86_64
tristate "Blowfish cipher algorithm (x86_64)"
depends on (X86 || UML_X86) && 64BIT
select CRYPTO_ALGAPI
select CRYPTO_BLOWFISH_COMMON
help
Blowfish cipher algorithm (x86_64), by Bruce Schneier.
This is a variable key length cipher which can use keys from 32
bits to 448 bits in length. It's fast, simple and specifically
designed for use on "large microprocessors".
See also:
<http://www.schneier.com/blowfish.html>
config CRYPTO_CAMELLIA config CRYPTO_CAMELLIA
tristate "Camellia cipher algorithms" tristate "Camellia cipher algorithms"
depends on CRYPTO depends on CRYPTO
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment