From 1a63112faf5af60e0ebcc60654eef806e7f6f11a Mon Sep 17 00:00:00 2001 From: Jeffrey Walton Date: Sat, 18 May 2019 23:07:17 -0400 Subject: [PATCH] Add ARM SHA1 asm implementation from Cryptogams (GH #837, PR #838) Add ARM SHA1 asm implementation from Cryptogams. Cryptogams is Andy Polyakov's project used to create high speed crypto algorithms and share them with other developers. Cryptogams has a dual license. First is the OpenSSL license because Andy contributes to OpenSSL. Second is a BSD license for those who want a more permissive license. Andy's implementation runs about 30% faster than C/C++ code. Testing on a 1 GHz Cortex-A7 shows Cryptograms at 16 cpb, and C++ at 23 cpb. The integration instructions are documented at [Cryptogams SHA](https://wiki.openssl.org/index.php/Cryptogams_SHA) on the OpenSSL wiki. --- Filelist.txt | 2 + GNUmakefile | 12 +- GNUmakefile-cross | 8 +- License.txt | 4 +- config_asm.h | 17 +- sha.cpp | 32 + sha1_armv4.S | 1429 +++++++++++++++++++++++++++++++++++++++++++++ sha1_armv4.h | 21 + 8 files changed, 1508 insertions(+), 17 deletions(-) create mode 100644 sha1_armv4.S create mode 100644 sha1_armv4.h diff --git a/Filelist.txt b/Filelist.txt index 56f3c37a..7c654b7e 100644 --- a/Filelist.txt +++ b/Filelist.txt @@ -314,6 +314,8 @@ serpentp.h sha.cpp sha_simd.cpp sha.h +sha1_armv4.h +sha1_armv4.S sha3.cpp sha3.h shacal2.cpp diff --git a/GNUmakefile b/GNUmakefile index d98d3b24..b2a233b2 100755 --- a/GNUmakefile +++ b/GNUmakefile @@ -1057,9 +1057,9 @@ endif # Cryptogams AES for ARMv4 and above. We couple to ARMv7. # Avoid iOS. It cannot consume the assembly. ifeq ($(IS_ARM32),1) - CRYPTOGAMS_AES_FLAG = -march=armv7-a - CRYPTOGAMS_AES_FLAG += -Wa,--noexecstack - SRCS += aes_armv4.S + CRYPTOGAMS_ARCH_FLAG = -march=armv7-a + CRYPTOGAMS_ARCH_FLAG += -Wa,--noexecstack + SRCS += aes_armv4.S sha1_armv4.S endif # List cryptlib.cpp first, then cpu.cpp, then integer.cpp to tame C++ static initialization problems. @@ -1423,7 +1423,7 @@ endif # Dependencies # Cryptogams ARM asm implementation. aes_armv4.o : aes_armv4.S - $(CC) $(strip $(CXXFLAGS) $(CRYPTOGAMS_AES_FLAG) -c) $< + $(CC) $(strip $(CXXFLAGS) $(CRYPTOGAMS_ARCH_FLAG) -c) $< # SSSE3 or NEON available aria_simd.o : aria_simd.cpp @@ -1505,6 +1505,10 @@ rijndael_simd.o : rijndael_simd.cpp sha_simd.o : sha_simd.cpp $(CXX) $(strip $(CXXFLAGS) $(SHA_FLAG) -c) $< +# Cryptogams ARM asm implementation. +sha1_armv4.o : sha1_armv4.S + $(CC) $(strip $(CXXFLAGS) $(CRYPTOGAMS_ARCH_FLAG) -c) $< + sha3_simd.o : sha3_simd.cpp $(CXX) $(strip $(CXXFLAGS) $(SHA3_FLAG) -c) $< diff --git a/GNUmakefile-cross b/GNUmakefile-cross index 6cfcfe08..2a16b8f8 100755 --- a/GNUmakefile-cross +++ b/GNUmakefile-cross @@ -592,9 +592,9 @@ INCL := $(filter-out resource.h,$(sort $(wildcard *.h))) # Avoid iOS. It cannot consume the assembly. ifeq ($(IS_ARM32),1) ifneq ($(IS_IOS),1) - CRYPTOGAMS_AES_FLAG = -march=armv7-a - CRYPTOGAMS_AES_FLAG += -Wa,--noexecstack - SRCS += aes_armv4.S + CRYPTOGAMS_ARCH_FLAG = -march=armv7-a + CRYPTOGAMS_ARCH_FLAG += -Wa,--noexecstack + SRCS += aes_armv4.S sha1_armv4.S endif endif @@ -800,7 +800,7 @@ endif # Dependencies # Cryptogams ARM asm implementation. aes_armv4.o : aes_armv4.S - $(CXX) $(strip $(CXXFLAGS) -fpermissive $(CRYPTOGAMS_AES_FLAG) -c) $< + $(CXX) $(strip $(CXXFLAGS) -fpermissive $(CRYPTOGAMS_ARCH_FLAG) -c) $< cpu-features.o: cpu-features.h cpu-features.c $(CXX) $(strip $(CXXFLAGS) -fpermissive -c) cpu-features.c diff --git a/License.txt b/License.txt index d48cdbb4..629ecc61 100644 --- a/License.txt +++ b/License.txt @@ -29,8 +29,8 @@ Daniel J. Bernstein, Jack Lloyd - chacha.cpp, chacha_simd.cpp, chacha_avx.cpp Andrew Moon - ed25519, x25519, donna_32.cpp, donna_64.cpp, donna_sse.cpp The Crypto++ Library uses portions of Andy Polyakov's CRYPTOGAMS for Poly1305 -scalar multiplication and aes_armv4.S. CRYPTOGAMS is dual licensed with a -permissive BSD-style license. The CRYPTOGAMS license is reproduced below. +scalar multiplication, aes_armv4.S and sha1_armv4.S. CRYPTOGAMS is dual licensed +with a permissive BSD-style license. The CRYPTOGAMS license is reproduced below. The Crypto++ Library uses portions of Jack Lloyd's Botan for ChaCha SSE2 and AVX. Botan placed the code in public domain for Crypto++ to use. diff --git a/config_asm.h b/config_asm.h index 7a798f71..0e23a8cf 100644 --- a/config_asm.h +++ b/config_asm.h @@ -354,15 +354,18 @@ # undef CRYPTOPP_ARM_SHA2_AVAILABLE #endif -// Cryptogams offers an ARM asm AES implementation. Crypto++ does -// not provide an asm implementation. The Cryptogams implementation -// is about 2x faster than C/C++. Define this to use the Cryptogams -// AES implementation on GNU Linux systems. When defined, Crypto++ -// will use aes_armv4.S. LLVM miscompiles aes_armv4.S so disable -// under Clang. See https://bugs.llvm.org/show_bug.cgi?id=38133. +// Cryptogams offers an ARM asm implementations for AES and SHA. Crypto++ does +// not provide an asm implementation. The Cryptogams AES implementation is +// about 50% faster than C/C++, and SHA implementation is about 30% faster +// than C/C++. Define this to use the Cryptogams AES and SHA implementations +// on GNU Linux systems. When defined, Crypto++ will use aes_armv4.S and +// sha1_armv4.S. LLVM miscompiles aes_armv4.S so disable under Clang. Also see +// https://bugs.llvm.org/show_bug.cgi?id=38133. #if !defined(CRYPTOPP_DISABLE_ASM) && defined(__arm__) # if defined(__GNUC__) && !defined(__clang__) -# define CRYPTOGAMS_ARM_AES 1 +# define CRYPTOGAMS_ARM_AES 1 +# define CRYPTOGAMS_ARM_SHA1 1 +# define CRYPTOGAMS_ARM_SHA256 1 # endif #endif diff --git a/sha.cpp b/sha.cpp index 87cf749a..70b6d48a 100644 --- a/sha.cpp +++ b/sha.cpp @@ -56,6 +56,11 @@ extern void SHA1_HashMultipleBlocks_SHANI(word32 *state, const word32 *data, siz extern void SHA256_HashMultipleBlocks_SHANI(word32 *state, const word32 *data, size_t length, ByteOrder order); #endif +#if (CRYPTOGAMS_ARM_SHA1) +extern "C" unsigned int CRYPTOGAMS_armcaps; +extern "C" int sha1_block_data_order(word32* state, const word32 *data, size_t blocks); +#endif + #if CRYPTOPP_ARM_SHA1_AVAILABLE extern void SHA1_HashMultipleBlocks_ARMV8(word32 *state, const word32 *data, size_t length, ByteOrder order); #endif @@ -232,6 +237,12 @@ std::string SHA1::AlgorithmProvider() const if (HasSSE2()) return "SSE2"; #endif +#if CRYPTOGAMS_ARM_SHA1 + if (HasNEON()) + return "NEON"; + if (HasARMv7()) + return "ARMv7"; +#endif #if CRYPTOPP_ARM_SHA1_AVAILABLE if (HasSHA1()) return "ARMv8"; @@ -260,6 +271,14 @@ void SHA1::Transform(word32 *state, const word32 *data) return; } #endif +#if CRYPTOGAMS_ARM_SHA1 && 0 + // TODO: convert LE to BE and use Cryptogams code + if (HasARMv7()) + { + sha1_block_data_order(state, data, 1); + return; + } +#endif #if CRYPTOPP_ARM_SHA1_AVAILABLE if (HasSHA1()) { @@ -283,6 +302,19 @@ size_t SHA1::HashMultipleBlocks(const word32 *input, size_t length) return length & (SHA1::BLOCKSIZE - 1); } #endif +#if CRYPTOGAMS_ARM_SHA1 + if (HasARMv7()) + { + // The Cryptogams code uses a global variable named CRYPTOGAMS_armcaps + // for capabilities like ARMv7 and NEON. Storage is allocated in the + // module. We still need to set CRYPTOGAMS_armcaps accordingly. + // The Cryptogams code defines NEON as 1<<0; see ARMV7_NEON. + static unsigned int unused = CRYPTOGAMS_armcaps = HasNEON() ? (1<<0) : 0; + + sha1_block_data_order(m_state, input, length / SHA1::BLOCKSIZE); + return length & (SHA1::BLOCKSIZE - 1); + } +#endif #if CRYPTOPP_ARM_SHA1_AVAILABLE if (HasSHA1()) { diff --git a/sha1_armv4.S b/sha1_armv4.S new file mode 100644 index 00000000..bfbbbfde --- /dev/null +++ b/sha1_armv4.S @@ -0,0 +1,1429 @@ +@ Copyright 2007-2018 The OpenSSL Project Authors. All Rights Reserved. +@ +@ ==================================================================== +@ Written by Andy Polyakov for the OpenSSL +@ project. The module is, however, dual licensed under OpenSSL and +@ CRYPTOGAMS licenses depending on where you obtain it. For further +@ details see http://www.openssl.org/~appro/cryptogams/. +@ ==================================================================== + +@ JW, MAY 2019: Begin defines from taken from arm_arch.h +@ The defines were included through the header. + +# if !defined(__ARM_ARCH__) +# if defined(__CC_ARM) +# define __ARM_ARCH__ __TARGET_ARCH_ARM +# if defined(__BIG_ENDIAN) +# define __ARMEB__ +# else +# define __ARMEL__ +# endif +# elif defined(__GNUC__) +# if defined(__aarch64__) +# define __ARM_ARCH__ 8 +# if __BYTE_ORDER__==__ORDER_BIG_ENDIAN__ +# define __ARMEB__ +# else +# define __ARMEL__ +# endif + +# elif defined(__ARM_ARCH) +# define __ARM_ARCH__ __ARM_ARCH +# elif defined(__ARM_ARCH_8A__) +# define __ARM_ARCH__ 8 +# elif defined(__ARM_ARCH_7__) || defined(__ARM_ARCH_7A__) || \ + defined(__ARM_ARCH_7R__)|| defined(__ARM_ARCH_7M__) || \ + defined(__ARM_ARCH_7EM__) +# define __ARM_ARCH__ 7 +# elif defined(__ARM_ARCH_6__) || defined(__ARM_ARCH_6J__) || \ + defined(__ARM_ARCH_6K__)|| defined(__ARM_ARCH_6M__) || \ + defined(__ARM_ARCH_6Z__)|| defined(__ARM_ARCH_6ZK__) || \ + defined(__ARM_ARCH_6T2__) +# define __ARM_ARCH__ 6 +# elif defined(__ARM_ARCH_5__) || defined(__ARM_ARCH_5T__) || \ + defined(__ARM_ARCH_5E__)|| defined(__ARM_ARCH_5TE__) || \ + defined(__ARM_ARCH_5TEJ__) +# define __ARM_ARCH__ 5 +# elif defined(__ARM_ARCH_4__) || defined(__ARM_ARCH_4T__) +# define __ARM_ARCH__ 4 +# else +# error "unsupported ARM architecture" +# endif +# endif +# endif + +# if !defined(__ARM_MAX_ARCH__) +# define __ARM_MAX_ARCH__ __ARM_ARCH__ +# endif + +# if __ARM_MAX_ARCH__<__ARM_ARCH__ +# error "__ARM_MAX_ARCH__ can't be less than __ARM_ARCH__" +# elif __ARM_MAX_ARCH__!=__ARM_ARCH__ +# if __ARM_ARCH__<7 && __ARM_MAX_ARCH__>=7 && defined(__ARMEB__) +# error "can't build universal big-endian binary" +# endif +# endif + +# ifndef __ASSEMBLER__ +extern unsigned int CRYPTOGAMS_armcaps; +# endif + +# define ARMV7_NEON (1<<0) + +@ JW, MAY 2019: End defines from taken from arm_arch.h +@ Back to original Cryptogams code + +#if defined(__thumb2__) +.syntax unified +.thumb +#else +.code 32 +#endif + +.text + +.globl sha1_block_data_order +.type sha1_block_data_order,%function + +.align 5 +sha1_block_data_order: +#if __ARM_MAX_ARCH__>=7 +.Lsha1_block: + ldr r12,.LCRYPTOGAMS_armcap +# if !defined(_WIN32) + adr r3,.Lsha1_block + ldr r12,[r3,r12] @ CRYPTOGAMS_armcaps +# endif +# if defined(__APPLE__) || defined(_WIN32) + ldr r12,[r12] +# endif + tst r12,#ARMV7_NEON + bne .LNEON +#endif + stmdb sp!,{r4,r5,r6,r7,r8,r9,r10,r11,r12,lr} + add r2,r1,r2,lsl#6 @ r2 to point at the end of r1 + ldmia r0,{r3,r4,r5,r6,r7} +.Lloop: + ldr r8,.LK_00_19 + mov r14,sp + sub sp,sp,#15*4 + mov r5,r5,ror#30 + mov r6,r6,ror#30 + mov r7,r7,ror#30 @ [6] +.L_00_15: +#if __ARM_ARCH__<7 + ldrb r10,[r1,#2] + ldrb r9,[r1,#3] + ldrb r11,[r1,#1] + add r7,r8,r7,ror#2 @ E+=K_00_19 + ldrb r12,[r1],#4 + orr r9,r9,r10,lsl#8 + eor r10,r5,r6 @ F_xx_xx + orr r9,r9,r11,lsl#16 + add r7,r7,r3,ror#27 @ E+=ROR(A,27) + orr r9,r9,r12,lsl#24 +#else + ldr r9,[r1],#4 @ handles unaligned + add r7,r8,r7,ror#2 @ E+=K_00_19 + eor r10,r5,r6 @ F_xx_xx + add r7,r7,r3,ror#27 @ E+=ROR(A,27) +#ifdef __ARMEL__ + rev r9,r9 @ byte swap +#endif +#endif + and r10,r4,r10,ror#2 + add r7,r7,r9 @ E+=X[i] + eor r10,r10,r6,ror#2 @ F_00_19(B,C,D) + str r9,[r14,#-4]! + add r7,r7,r10 @ E+=F_00_19(B,C,D) +#if __ARM_ARCH__<7 + ldrb r10,[r1,#2] + ldrb r9,[r1,#3] + ldrb r11,[r1,#1] + add r6,r8,r6,ror#2 @ E+=K_00_19 + ldrb r12,[r1],#4 + orr r9,r9,r10,lsl#8 + eor r10,r4,r5 @ F_xx_xx + orr r9,r9,r11,lsl#16 + add r6,r6,r7,ror#27 @ E+=ROR(A,27) + orr r9,r9,r12,lsl#24 +#else + ldr r9,[r1],#4 @ handles unaligned + add r6,r8,r6,ror#2 @ E+=K_00_19 + eor r10,r4,r5 @ F_xx_xx + add r6,r6,r7,ror#27 @ E+=ROR(A,27) +#ifdef __ARMEL__ + rev r9,r9 @ byte swap +#endif +#endif + and r10,r3,r10,ror#2 + add r6,r6,r9 @ E+=X[i] + eor r10,r10,r5,ror#2 @ F_00_19(B,C,D) + str r9,[r14,#-4]! + add r6,r6,r10 @ E+=F_00_19(B,C,D) +#if __ARM_ARCH__<7 + ldrb r10,[r1,#2] + ldrb r9,[r1,#3] + ldrb r11,[r1,#1] + add r5,r8,r5,ror#2 @ E+=K_00_19 + ldrb r12,[r1],#4 + orr r9,r9,r10,lsl#8 + eor r10,r3,r4 @ F_xx_xx + orr r9,r9,r11,lsl#16 + add r5,r5,r6,ror#27 @ E+=ROR(A,27) + orr r9,r9,r12,lsl#24 +#else + ldr r9,[r1],#4 @ handles unaligned + add r5,r8,r5,ror#2 @ E+=K_00_19 + eor r10,r3,r4 @ F_xx_xx + add r5,r5,r6,ror#27 @ E+=ROR(A,27) +#ifdef __ARMEL__ + rev r9,r9 @ byte swap +#endif +#endif + and r10,r7,r10,ror#2 + add r5,r5,r9 @ E+=X[i] + eor r10,r10,r4,ror#2 @ F_00_19(B,C,D) + str r9,[r14,#-4]! + add r5,r5,r10 @ E+=F_00_19(B,C,D) +#if __ARM_ARCH__<7 + ldrb r10,[r1,#2] + ldrb r9,[r1,#3] + ldrb r11,[r1,#1] + add r4,r8,r4,ror#2 @ E+=K_00_19 + ldrb r12,[r1],#4 + orr r9,r9,r10,lsl#8 + eor r10,r7,r3 @ F_xx_xx + orr r9,r9,r11,lsl#16 + add r4,r4,r5,ror#27 @ E+=ROR(A,27) + orr r9,r9,r12,lsl#24 +#else + ldr r9,[r1],#4 @ handles unaligned + add r4,r8,r4,ror#2 @ E+=K_00_19 + eor r10,r7,r3 @ F_xx_xx + add r4,r4,r5,ror#27 @ E+=ROR(A,27) +#ifdef __ARMEL__ + rev r9,r9 @ byte swap +#endif +#endif + and r10,r6,r10,ror#2 + add r4,r4,r9 @ E+=X[i] + eor r10,r10,r3,ror#2 @ F_00_19(B,C,D) + str r9,[r14,#-4]! + add r4,r4,r10 @ E+=F_00_19(B,C,D) +#if __ARM_ARCH__<7 + ldrb r10,[r1,#2] + ldrb r9,[r1,#3] + ldrb r11,[r1,#1] + add r3,r8,r3,ror#2 @ E+=K_00_19 + ldrb r12,[r1],#4 + orr r9,r9,r10,lsl#8 + eor r10,r6,r7 @ F_xx_xx + orr r9,r9,r11,lsl#16 + add r3,r3,r4,ror#27 @ E+=ROR(A,27) + orr r9,r9,r12,lsl#24 +#else + ldr r9,[r1],#4 @ handles unaligned + add r3,r8,r3,ror#2 @ E+=K_00_19 + eor r10,r6,r7 @ F_xx_xx + add r3,r3,r4,ror#27 @ E+=ROR(A,27) +#ifdef __ARMEL__ + rev r9,r9 @ byte swap +#endif +#endif + and r10,r5,r10,ror#2 + add r3,r3,r9 @ E+=X[i] + eor r10,r10,r7,ror#2 @ F_00_19(B,C,D) + str r9,[r14,#-4]! + add r3,r3,r10 @ E+=F_00_19(B,C,D) +#if defined(__thumb2__) + mov r12,sp + teq r14,r12 +#else + teq r14,sp +#endif + bne .L_00_15 @ [((11+4)*5+2)*3] + sub sp,sp,#25*4 +#if __ARM_ARCH__<7 + ldrb r10,[r1,#2] + ldrb r9,[r1,#3] + ldrb r11,[r1,#1] + add r7,r8,r7,ror#2 @ E+=K_00_19 + ldrb r12,[r1],#4 + orr r9,r9,r10,lsl#8 + eor r10,r5,r6 @ F_xx_xx + orr r9,r9,r11,lsl#16 + add r7,r7,r3,ror#27 @ E+=ROR(A,27) + orr r9,r9,r12,lsl#24 +#else + ldr r9,[r1],#4 @ handles unaligned + add r7,r8,r7,ror#2 @ E+=K_00_19 + eor r10,r5,r6 @ F_xx_xx + add r7,r7,r3,ror#27 @ E+=ROR(A,27) +#ifdef __ARMEL__ + rev r9,r9 @ byte swap +#endif +#endif + and r10,r4,r10,ror#2 + add r7,r7,r9 @ E+=X[i] + eor r10,r10,r6,ror#2 @ F_00_19(B,C,D) + str r9,[r14,#-4]! + add r7,r7,r10 @ E+=F_00_19(B,C,D) + ldr r9,[r14,#15*4] + ldr r10,[r14,#13*4] + ldr r11,[r14,#7*4] + add r6,r8,r6,ror#2 @ E+=K_xx_xx + ldr r12,[r14,#2*4] + eor r9,r9,r10 + eor r11,r11,r12 @ 1 cycle stall + eor r10,r4,r5 @ F_xx_xx + mov r9,r9,ror#31 + add r6,r6,r7,ror#27 @ E+=ROR(A,27) + eor r9,r9,r11,ror#31 + str r9,[r14,#-4]! + and r10,r3,r10,ror#2 @ F_xx_xx + @ F_xx_xx + add r6,r6,r9 @ E+=X[i] + eor r10,r10,r5,ror#2 @ F_00_19(B,C,D) + add r6,r6,r10 @ E+=F_00_19(B,C,D) + ldr r9,[r14,#15*4] + ldr r10,[r14,#13*4] + ldr r11,[r14,#7*4] + add r5,r8,r5,ror#2 @ E+=K_xx_xx + ldr r12,[r14,#2*4] + eor r9,r9,r10 + eor r11,r11,r12 @ 1 cycle stall + eor r10,r3,r4 @ F_xx_xx + mov r9,r9,ror#31 + add r5,r5,r6,ror#27 @ E+=ROR(A,27) + eor r9,r9,r11,ror#31 + str r9,[r14,#-4]! + and r10,r7,r10,ror#2 @ F_xx_xx + @ F_xx_xx + add r5,r5,r9 @ E+=X[i] + eor r10,r10,r4,ror#2 @ F_00_19(B,C,D) + add r5,r5,r10 @ E+=F_00_19(B,C,D) + ldr r9,[r14,#15*4] + ldr r10,[r14,#13*4] + ldr r11,[r14,#7*4] + add r4,r8,r4,ror#2 @ E+=K_xx_xx + ldr r12,[r14,#2*4] + eor r9,r9,r10 + eor r11,r11,r12 @ 1 cycle stall + eor r10,r7,r3 @ F_xx_xx + mov r9,r9,ror#31 + add r4,r4,r5,ror#27 @ E+=ROR(A,27) + eor r9,r9,r11,ror#31 + str r9,[r14,#-4]! + and r10,r6,r10,ror#2 @ F_xx_xx + @ F_xx_xx + add r4,r4,r9 @ E+=X[i] + eor r10,r10,r3,ror#2 @ F_00_19(B,C,D) + add r4,r4,r10 @ E+=F_00_19(B,C,D) + ldr r9,[r14,#15*4] + ldr r10,[r14,#13*4] + ldr r11,[r14,#7*4] + add r3,r8,r3,ror#2 @ E+=K_xx_xx + ldr r12,[r14,#2*4] + eor r9,r9,r10 + eor r11,r11,r12 @ 1 cycle stall + eor r10,r6,r7 @ F_xx_xx + mov r9,r9,ror#31 + add r3,r3,r4,ror#27 @ E+=ROR(A,27) + eor r9,r9,r11,ror#31 + str r9,[r14,#-4]! + and r10,r5,r10,ror#2 @ F_xx_xx + @ F_xx_xx + add r3,r3,r9 @ E+=X[i] + eor r10,r10,r7,ror#2 @ F_00_19(B,C,D) + add r3,r3,r10 @ E+=F_00_19(B,C,D) + + ldr r8,.LK_20_39 @ [+15+16*4] + cmn sp,#0 @ [+3], clear carry to denote 20_39 +.L_20_39_or_60_79: + ldr r9,[r14,#15*4] + ldr r10,[r14,#13*4] + ldr r11,[r14,#7*4] + add r7,r8,r7,ror#2 @ E+=K_xx_xx + ldr r12,[r14,#2*4] + eor r9,r9,r10 + eor r11,r11,r12 @ 1 cycle stall + eor r10,r5,r6 @ F_xx_xx + mov r9,r9,ror#31 + add r7,r7,r3,ror#27 @ E+=ROR(A,27) + eor r9,r9,r11,ror#31 + str r9,[r14,#-4]! + eor r10,r4,r10,ror#2 @ F_xx_xx + @ F_xx_xx + add r7,r7,r9 @ E+=X[i] + add r7,r7,r10 @ E+=F_20_39(B,C,D) + ldr r9,[r14,#15*4] + ldr r10,[r14,#13*4] + ldr r11,[r14,#7*4] + add r6,r8,r6,ror#2 @ E+=K_xx_xx + ldr r12,[r14,#2*4] + eor r9,r9,r10 + eor r11,r11,r12 @ 1 cycle stall + eor r10,r4,r5 @ F_xx_xx + mov r9,r9,ror#31 + add r6,r6,r7,ror#27 @ E+=ROR(A,27) + eor r9,r9,r11,ror#31 + str r9,[r14,#-4]! + eor r10,r3,r10,ror#2 @ F_xx_xx + @ F_xx_xx + add r6,r6,r9 @ E+=X[i] + add r6,r6,r10 @ E+=F_20_39(B,C,D) + ldr r9,[r14,#15*4] + ldr r10,[r14,#13*4] + ldr r11,[r14,#7*4] + add r5,r8,r5,ror#2 @ E+=K_xx_xx + ldr r12,[r14,#2*4] + eor r9,r9,r10 + eor r11,r11,r12 @ 1 cycle stall + eor r10,r3,r4 @ F_xx_xx + mov r9,r9,ror#31 + add r5,r5,r6,ror#27 @ E+=ROR(A,27) + eor r9,r9,r11,ror#31 + str r9,[r14,#-4]! + eor r10,r7,r10,ror#2 @ F_xx_xx + @ F_xx_xx + add r5,r5,r9 @ E+=X[i] + add r5,r5,r10 @ E+=F_20_39(B,C,D) + ldr r9,[r14,#15*4] + ldr r10,[r14,#13*4] + ldr r11,[r14,#7*4] + add r4,r8,r4,ror#2 @ E+=K_xx_xx + ldr r12,[r14,#2*4] + eor r9,r9,r10 + eor r11,r11,r12 @ 1 cycle stall + eor r10,r7,r3 @ F_xx_xx + mov r9,r9,ror#31 + add r4,r4,r5,ror#27 @ E+=ROR(A,27) + eor r9,r9,r11,ror#31 + str r9,[r14,#-4]! + eor r10,r6,r10,ror#2 @ F_xx_xx + @ F_xx_xx + add r4,r4,r9 @ E+=X[i] + add r4,r4,r10 @ E+=F_20_39(B,C,D) + ldr r9,[r14,#15*4] + ldr r10,[r14,#13*4] + ldr r11,[r14,#7*4] + add r3,r8,r3,ror#2 @ E+=K_xx_xx + ldr r12,[r14,#2*4] + eor r9,r9,r10 + eor r11,r11,r12 @ 1 cycle stall + eor r10,r6,r7 @ F_xx_xx + mov r9,r9,ror#31 + add r3,r3,r4,ror#27 @ E+=ROR(A,27) + eor r9,r9,r11,ror#31 + str r9,[r14,#-4]! + eor r10,r5,r10,ror#2 @ F_xx_xx + @ F_xx_xx + add r3,r3,r9 @ E+=X[i] + add r3,r3,r10 @ E+=F_20_39(B,C,D) +#if defined(__thumb2__) + mov r12,sp + teq r14,r12 +#else + teq r14,sp @ preserve carry +#endif + bne .L_20_39_or_60_79 @ [+((12+3)*5+2)*4] + bcs .L_done @ [+((12+3)*5+2)*4], spare 300 bytes + + ldr r8,.LK_40_59 + sub sp,sp,#20*4 @ [+2] +.L_40_59: + ldr r9,[r14,#15*4] + ldr r10,[r14,#13*4] + ldr r11,[r14,#7*4] + add r7,r8,r7,ror#2 @ E+=K_xx_xx + ldr r12,[r14,#2*4] + eor r9,r9,r10 + eor r11,r11,r12 @ 1 cycle stall + eor r10,r5,r6 @ F_xx_xx + mov r9,r9,ror#31 + add r7,r7,r3,ror#27 @ E+=ROR(A,27) + eor r9,r9,r11,ror#31 + str r9,[r14,#-4]! + and r10,r4,r10,ror#2 @ F_xx_xx + and r11,r5,r6 @ F_xx_xx + add r7,r7,r9 @ E+=X[i] + add r7,r7,r10 @ E+=F_40_59(B,C,D) + add r7,r7,r11,ror#2 + ldr r9,[r14,#15*4] + ldr r10,[r14,#13*4] + ldr r11,[r14,#7*4] + add r6,r8,r6,ror#2 @ E+=K_xx_xx + ldr r12,[r14,#2*4] + eor r9,r9,r10 + eor r11,r11,r12 @ 1 cycle stall + eor r10,r4,r5 @ F_xx_xx + mov r9,r9,ror#31 + add r6,r6,r7,ror#27 @ E+=ROR(A,27) + eor r9,r9,r11,ror#31 + str r9,[r14,#-4]! + and r10,r3,r10,ror#2 @ F_xx_xx + and r11,r4,r5 @ F_xx_xx + add r6,r6,r9 @ E+=X[i] + add r6,r6,r10 @ E+=F_40_59(B,C,D) + add r6,r6,r11,ror#2 + ldr r9,[r14,#15*4] + ldr r10,[r14,#13*4] + ldr r11,[r14,#7*4] + add r5,r8,r5,ror#2 @ E+=K_xx_xx + ldr r12,[r14,#2*4] + eor r9,r9,r10 + eor r11,r11,r12 @ 1 cycle stall + eor r10,r3,r4 @ F_xx_xx + mov r9,r9,ror#31 + add r5,r5,r6,ror#27 @ E+=ROR(A,27) + eor r9,r9,r11,ror#31 + str r9,[r14,#-4]! + and r10,r7,r10,ror#2 @ F_xx_xx + and r11,r3,r4 @ F_xx_xx + add r5,r5,r9 @ E+=X[i] + add r5,r5,r10 @ E+=F_40_59(B,C,D) + add r5,r5,r11,ror#2 + ldr r9,[r14,#15*4] + ldr r10,[r14,#13*4] + ldr r11,[r14,#7*4] + add r4,r8,r4,ror#2 @ E+=K_xx_xx + ldr r12,[r14,#2*4] + eor r9,r9,r10 + eor r11,r11,r12 @ 1 cycle stall + eor r10,r7,r3 @ F_xx_xx + mov r9,r9,ror#31 + add r4,r4,r5,ror#27 @ E+=ROR(A,27) + eor r9,r9,r11,ror#31 + str r9,[r14,#-4]! + and r10,r6,r10,ror#2 @ F_xx_xx + and r11,r7,r3 @ F_xx_xx + add r4,r4,r9 @ E+=X[i] + add r4,r4,r10 @ E+=F_40_59(B,C,D) + add r4,r4,r11,ror#2 + ldr r9,[r14,#15*4] + ldr r10,[r14,#13*4] + ldr r11,[r14,#7*4] + add r3,r8,r3,ror#2 @ E+=K_xx_xx + ldr r12,[r14,#2*4] + eor r9,r9,r10 + eor r11,r11,r12 @ 1 cycle stall + eor r10,r6,r7 @ F_xx_xx + mov r9,r9,ror#31 + add r3,r3,r4,ror#27 @ E+=ROR(A,27) + eor r9,r9,r11,ror#31 + str r9,[r14,#-4]! + and r10,r5,r10,ror#2 @ F_xx_xx + and r11,r6,r7 @ F_xx_xx + add r3,r3,r9 @ E+=X[i] + add r3,r3,r10 @ E+=F_40_59(B,C,D) + add r3,r3,r11,ror#2 +#if defined(__thumb2__) + mov r12,sp + teq r14,r12 +#else + teq r14,sp +#endif + bne .L_40_59 @ [+((12+5)*5+2)*4] + + ldr r8,.LK_60_79 + sub sp,sp,#20*4 + cmp sp,#0 @ set carry to denote 60_79 + b .L_20_39_or_60_79 @ [+4], spare 300 bytes +.L_done: + add sp,sp,#80*4 @ "deallocate" stack frame + ldmia r0,{r8,r9,r10,r11,r12} + add r3,r8,r3 + add r4,r9,r4 + add r5,r10,r5,ror#2 + add r6,r11,r6,ror#2 + add r7,r12,r7,ror#2 + stmia r0,{r3,r4,r5,r6,r7} + teq r1,r2 + bne .Lloop @ [+18], total 1307 + +#if __ARM_ARCH__>=5 + ldmia sp!,{r4,r5,r6,r7,r8,r9,r10,r11,r12,pc} +#else + ldmia sp!,{r4,r5,r6,r7,r8,r9,r10,r11,r12,lr} + tst lr,#1 + moveq pc,lr @ be binary compatible with V4, yet +.word 0xe12fff1e @ interoperable with Thumb ISA:-) +#endif +.size sha1_block_data_order,.-sha1_block_data_order + +.align 5 +.LK_00_19:.word 0x5a827999 +.LK_20_39:.word 0x6ed9eba1 +.LK_40_59:.word 0x8f1bbcdc +.LK_60_79:.word 0xca62c1d6 +#if __ARM_MAX_ARCH__>=7 +.LCRYPTOGAMS_armcap: +# ifdef _WIN32 +.word CRYPTOGAMS_armcaps +# else +.word CRYPTOGAMS_armcaps-.Lsha1_block +# endif +#endif +.byte 83,72,65,49,32,98,108,111,99,107,32,116,114,97,110,115,102,111,114,109,32,102,111,114,32,65,82,77,118,52,47,78,69,79,78,47,65,82,77,118,56,44,32,67,82,89,80,84,79,71,65,77,83,32,98,121,32,60,97,112,112,114,111,64,111,112,101,110,115,115,108,46,111,114,103,62,0 +.align 2 +.align 5 +#if __ARM_MAX_ARCH__>=7 +.arch armv7-a +.fpu neon + +.type sha1_block_data_order_neon,%function +.align 4 +sha1_block_data_order_neon: +.LNEON: + stmdb sp!,{r4,r5,r6,r7,r8,r9,r10,r11,r12,lr} + add r2,r1,r2,lsl#6 @ r2 to point at the end of r1 + @ dmb @ errata #451034 on early Cortex A8 + @ vstmdb sp!,{d8-d15} @ ABI specification says so + mov r14,sp + sub r12,sp,#64 + adr r8,.LK_00_19 + bic r12,r12,#15 @ align for 128-bit stores + + ldmia r0,{r3,r4,r5,r6,r7} @ load context + mov sp,r12 @ alloca + + vld1.8 {q0,q1},[r1]! @ handles unaligned + veor q15,q15,q15 + vld1.8 {q2,q3},[r1]! + vld1.32 {d28[],d29[]},[r8,:32]! @ load K_00_19 + vrev32.8 q0,q0 @ yes, even on + vrev32.8 q1,q1 @ big-endian... + vrev32.8 q2,q2 + vadd.i32 q8,q0,q14 + vrev32.8 q3,q3 + vadd.i32 q9,q1,q14 + vst1.32 {q8},[r12,:128]! + vadd.i32 q10,q2,q14 + vst1.32 {q9},[r12,:128]! + vst1.32 {q10},[r12,:128]! + ldr r9,[sp] @ big RAW stall + +.Loop_neon: + vext.8 q8,q0,q1,#8 + bic r10,r6,r4 + add r7,r7,r9 + and r11,r5,r4 + vadd.i32 q13,q3,q14 + ldr r9,[sp,#4] + add r7,r7,r3,ror#27 + vext.8 q12,q3,q15,#4 + eor r11,r11,r10 + mov r4,r4,ror#2 + add r7,r7,r11 + veor q8,q8,q0 + bic r10,r5,r3 + add r6,r6,r9 + veor q12,q12,q2 + and r11,r4,r3 + ldr r9,[sp,#8] + veor q12,q12,q8 + add r6,r6,r7,ror#27 + eor r11,r11,r10 + vst1.32 {q13},[r12,:128]! + sub r12,r12,#64 + mov r3,r3,ror#2 + add r6,r6,r11 + vext.8 q13,q15,q12,#4 + bic r10,r4,r7 + add r5,r5,r9 + vadd.i32 q8,q12,q12 + and r11,r3,r7 + ldr r9,[sp,#12] + vsri.32 q8,q12,#31 + add r5,r5,r6,ror#27 + eor r11,r11,r10 + mov r7,r7,ror#2 + vshr.u32 q12,q13,#30 + add r5,r5,r11 + bic r10,r3,r6 + vshl.u32 q13,q13,#2 + add r4,r4,r9 + and r11,r7,r6 + veor q8,q8,q12 + ldr r9,[sp,#16] + add r4,r4,r5,ror#27 + veor q8,q8,q13 + eor r11,r11,r10 + mov r6,r6,ror#2 + add r4,r4,r11 + vext.8 q9,q1,q2,#8 + bic r10,r7,r5 + add r3,r3,r9 + and r11,r6,r5 + vadd.i32 q13,q8,q14 + ldr r9,[sp,#20] + vld1.32 {d28[],d29[]},[r8,:32]! + add r3,r3,r4,ror#27 + vext.8 q12,q8,q15,#4 + eor r11,r11,r10 + mov r5,r5,ror#2 + add r3,r3,r11 + veor q9,q9,q1 + bic r10,r6,r4 + add r7,r7,r9 + veor q12,q12,q3 + and r11,r5,r4 + ldr r9,[sp,#24] + veor q12,q12,q9 + add r7,r7,r3,ror#27 + eor r11,r11,r10 + vst1.32 {q13},[r12,:128]! + mov r4,r4,ror#2 + add r7,r7,r11 + vext.8 q13,q15,q12,#4 + bic r10,r5,r3 + add r6,r6,r9 + vadd.i32 q9,q12,q12 + and r11,r4,r3 + ldr r9,[sp,#28] + vsri.32 q9,q12,#31 + add r6,r6,r7,ror#27 + eor r11,r11,r10 + mov r3,r3,ror#2 + vshr.u32 q12,q13,#30 + add r6,r6,r11 + bic r10,r4,r7 + vshl.u32 q13,q13,#2 + add r5,r5,r9 + and r11,r3,r7 + veor q9,q9,q12 + ldr r9,[sp,#32] + add r5,r5,r6,ror#27 + veor q9,q9,q13 + eor r11,r11,r10 + mov r7,r7,ror#2 + add r5,r5,r11 + vext.8 q10,q2,q3,#8 + bic r10,r3,r6 + add r4,r4,r9 + and r11,r7,r6 + vadd.i32 q13,q9,q14 + ldr r9,[sp,#36] + add r4,r4,r5,ror#27 + vext.8 q12,q9,q15,#4 + eor r11,r11,r10 + mov r6,r6,ror#2 + add r4,r4,r11 + veor q10,q10,q2 + bic r10,r7,r5 + add r3,r3,r9 + veor q12,q12,q8 + and r11,r6,r5 + ldr r9,[sp,#40] + veor q12,q12,q10 + add r3,r3,r4,ror#27 + eor r11,r11,r10 + vst1.32 {q13},[r12,:128]! + mov r5,r5,ror#2 + add r3,r3,r11 + vext.8 q13,q15,q12,#4 + bic r10,r6,r4 + add r7,r7,r9 + vadd.i32 q10,q12,q12 + and r11,r5,r4 + ldr r9,[sp,#44] + vsri.32 q10,q12,#31 + add r7,r7,r3,ror#27 + eor r11,r11,r10 + mov r4,r4,ror#2 + vshr.u32 q12,q13,#30 + add r7,r7,r11 + bic r10,r5,r3 + vshl.u32 q13,q13,#2 + add r6,r6,r9 + and r11,r4,r3 + veor q10,q10,q12 + ldr r9,[sp,#48] + add r6,r6,r7,ror#27 + veor q10,q10,q13 + eor r11,r11,r10 + mov r3,r3,ror#2 + add r6,r6,r11 + vext.8 q11,q3,q8,#8 + bic r10,r4,r7 + add r5,r5,r9 + and r11,r3,r7 + vadd.i32 q13,q10,q14 + ldr r9,[sp,#52] + add r5,r5,r6,ror#27 + vext.8 q12,q10,q15,#4 + eor r11,r11,r10 + mov r7,r7,ror#2 + add r5,r5,r11 + veor q11,q11,q3 + bic r10,r3,r6 + add r4,r4,r9 + veor q12,q12,q9 + and r11,r7,r6 + ldr r9,[sp,#56] + veor q12,q12,q11 + add r4,r4,r5,ror#27 + eor r11,r11,r10 + vst1.32 {q13},[r12,:128]! + mov r6,r6,ror#2 + add r4,r4,r11 + vext.8 q13,q15,q12,#4 + bic r10,r7,r5 + add r3,r3,r9 + vadd.i32 q11,q12,q12 + and r11,r6,r5 + ldr r9,[sp,#60] + vsri.32 q11,q12,#31 + add r3,r3,r4,ror#27 + eor r11,r11,r10 + mov r5,r5,ror#2 + vshr.u32 q12,q13,#30 + add r3,r3,r11 + bic r10,r6,r4 + vshl.u32 q13,q13,#2 + add r7,r7,r9 + and r11,r5,r4 + veor q11,q11,q12 + ldr r9,[sp,#0] + add r7,r7,r3,ror#27 + veor q11,q11,q13 + eor r11,r11,r10 + mov r4,r4,ror#2 + add r7,r7,r11 + vext.8 q12,q10,q11,#8 + bic r10,r5,r3 + add r6,r6,r9 + and r11,r4,r3 + veor q0,q0,q8 + ldr r9,[sp,#4] + add r6,r6,r7,ror#27 + veor q0,q0,q1 + eor r11,r11,r10 + mov r3,r3,ror#2 + vadd.i32 q13,q11,q14 + add r6,r6,r11 + bic r10,r4,r7 + veor q12,q12,q0 + add r5,r5,r9 + and r11,r3,r7 + vshr.u32 q0,q12,#30 + ldr r9,[sp,#8] + add r5,r5,r6,ror#27 + vst1.32 {q13},[r12,:128]! + sub r12,r12,#64 + eor r11,r11,r10 + mov r7,r7,ror#2 + vsli.32 q0,q12,#2 + add r5,r5,r11 + bic r10,r3,r6 + add r4,r4,r9 + and r11,r7,r6 + ldr r9,[sp,#12] + add r4,r4,r5,ror#27 + eor r11,r11,r10 + mov r6,r6,ror#2 + add r4,r4,r11 + bic r10,r7,r5 + add r3,r3,r9 + and r11,r6,r5 + ldr r9,[sp,#16] + add r3,r3,r4,ror#27 + eor r11,r11,r10 + mov r5,r5,ror#2 + add r3,r3,r11 + vext.8 q12,q11,q0,#8 + eor r10,r4,r6 + add r7,r7,r9 + ldr r9,[sp,#20] + veor q1,q1,q9 + eor r11,r10,r5 + add r7,r7,r3,ror#27 + veor q1,q1,q2 + mov r4,r4,ror#2 + add r7,r7,r11 + vadd.i32 q13,q0,q14 + eor r10,r3,r5 + add r6,r6,r9 + veor q12,q12,q1 + ldr r9,[sp,#24] + eor r11,r10,r4 + vshr.u32 q1,q12,#30 + add r6,r6,r7,ror#27 + mov r3,r3,ror#2 + vst1.32 {q13},[r12,:128]! + add r6,r6,r11 + eor r10,r7,r4 + vsli.32 q1,q12,#2 + add r5,r5,r9 + ldr r9,[sp,#28] + eor r11,r10,r3 + add r5,r5,r6,ror#27 + mov r7,r7,ror#2 + add r5,r5,r11 + eor r10,r6,r3 + add r4,r4,r9 + ldr r9,[sp,#32] + eor r11,r10,r7 + add r4,r4,r5,ror#27 + mov r6,r6,ror#2 + add r4,r4,r11 + vext.8 q12,q0,q1,#8 + eor r10,r5,r7 + add r3,r3,r9 + ldr r9,[sp,#36] + veor q2,q2,q10 + eor r11,r10,r6 + add r3,r3,r4,ror#27 + veor q2,q2,q3 + mov r5,r5,ror#2 + add r3,r3,r11 + vadd.i32 q13,q1,q14 + eor r10,r4,r6 + vld1.32 {d28[],d29[]},[r8,:32]! + add r7,r7,r9 + veor q12,q12,q2 + ldr r9,[sp,#40] + eor r11,r10,r5 + vshr.u32 q2,q12,#30 + add r7,r7,r3,ror#27 + mov r4,r4,ror#2 + vst1.32 {q13},[r12,:128]! + add r7,r7,r11 + eor r10,r3,r5 + vsli.32 q2,q12,#2 + add r6,r6,r9 + ldr r9,[sp,#44] + eor r11,r10,r4 + add r6,r6,r7,ror#27 + mov r3,r3,ror#2 + add r6,r6,r11 + eor r10,r7,r4 + add r5,r5,r9 + ldr r9,[sp,#48] + eor r11,r10,r3 + add r5,r5,r6,ror#27 + mov r7,r7,ror#2 + add r5,r5,r11 + vext.8 q12,q1,q2,#8 + eor r10,r6,r3 + add r4,r4,r9 + ldr r9,[sp,#52] + veor q3,q3,q11 + eor r11,r10,r7 + add r4,r4,r5,ror#27 + veor q3,q3,q8 + mov r6,r6,ror#2 + add r4,r4,r11 + vadd.i32 q13,q2,q14 + eor r10,r5,r7 + add r3,r3,r9 + veor q12,q12,q3 + ldr r9,[sp,#56] + eor r11,r10,r6 + vshr.u32 q3,q12,#30 + add r3,r3,r4,ror#27 + mov r5,r5,ror#2 + vst1.32 {q13},[r12,:128]! + add r3,r3,r11 + eor r10,r4,r6 + vsli.32 q3,q12,#2 + add r7,r7,r9 + ldr r9,[sp,#60] + eor r11,r10,r5 + add r7,r7,r3,ror#27 + mov r4,r4,ror#2 + add r7,r7,r11 + eor r10,r3,r5 + add r6,r6,r9 + ldr r9,[sp,#0] + eor r11,r10,r4 + add r6,r6,r7,ror#27 + mov r3,r3,ror#2 + add r6,r6,r11 + vext.8 q12,q2,q3,#8 + eor r10,r7,r4 + add r5,r5,r9 + ldr r9,[sp,#4] + veor q8,q8,q0 + eor r11,r10,r3 + add r5,r5,r6,ror#27 + veor q8,q8,q9 + mov r7,r7,ror#2 + add r5,r5,r11 + vadd.i32 q13,q3,q14 + eor r10,r6,r3 + add r4,r4,r9 + veor q12,q12,q8 + ldr r9,[sp,#8] + eor r11,r10,r7 + vshr.u32 q8,q12,#30 + add r4,r4,r5,ror#27 + mov r6,r6,ror#2 + vst1.32 {q13},[r12,:128]! + sub r12,r12,#64 + add r4,r4,r11 + eor r10,r5,r7 + vsli.32 q8,q12,#2 + add r3,r3,r9 + ldr r9,[sp,#12] + eor r11,r10,r6 + add r3,r3,r4,ror#27 + mov r5,r5,ror#2 + add r3,r3,r11 + eor r10,r4,r6 + add r7,r7,r9 + ldr r9,[sp,#16] + eor r11,r10,r5 + add r7,r7,r3,ror#27 + mov r4,r4,ror#2 + add r7,r7,r11 + vext.8 q12,q3,q8,#8 + eor r10,r3,r5 + add r6,r6,r9 + ldr r9,[sp,#20] + veor q9,q9,q1 + eor r11,r10,r4 + add r6,r6,r7,ror#27 + veor q9,q9,q10 + mov r3,r3,ror#2 + add r6,r6,r11 + vadd.i32 q13,q8,q14 + eor r10,r7,r4 + add r5,r5,r9 + veor q12,q12,q9 + ldr r9,[sp,#24] + eor r11,r10,r3 + vshr.u32 q9,q12,#30 + add r5,r5,r6,ror#27 + mov r7,r7,ror#2 + vst1.32 {q13},[r12,:128]! + add r5,r5,r11 + eor r10,r6,r3 + vsli.32 q9,q12,#2 + add r4,r4,r9 + ldr r9,[sp,#28] + eor r11,r10,r7 + add r4,r4,r5,ror#27 + mov r6,r6,ror#2 + add r4,r4,r11 + eor r10,r5,r7 + add r3,r3,r9 + ldr r9,[sp,#32] + eor r11,r10,r6 + add r3,r3,r4,ror#27 + mov r5,r5,ror#2 + add r3,r3,r11 + vext.8 q12,q8,q9,#8 + add r7,r7,r9 + and r10,r5,r6 + ldr r9,[sp,#36] + veor q10,q10,q2 + add r7,r7,r3,ror#27 + eor r11,r5,r6 + veor q10,q10,q11 + add r7,r7,r10 + and r11,r11,r4 + vadd.i32 q13,q9,q14 + mov r4,r4,ror#2 + add r7,r7,r11 + veor q12,q12,q10 + add r6,r6,r9 + and r10,r4,r5 + vshr.u32 q10,q12,#30 + ldr r9,[sp,#40] + add r6,r6,r7,ror#27 + vst1.32 {q13},[r12,:128]! + eor r11,r4,r5 + add r6,r6,r10 + vsli.32 q10,q12,#2 + and r11,r11,r3 + mov r3,r3,ror#2 + add r6,r6,r11 + add r5,r5,r9 + and r10,r3,r4 + ldr r9,[sp,#44] + add r5,r5,r6,ror#27 + eor r11,r3,r4 + add r5,r5,r10 + and r11,r11,r7 + mov r7,r7,ror#2 + add r5,r5,r11 + add r4,r4,r9 + and r10,r7,r3 + ldr r9,[sp,#48] + add r4,r4,r5,ror#27 + eor r11,r7,r3 + add r4,r4,r10 + and r11,r11,r6 + mov r6,r6,ror#2 + add r4,r4,r11 + vext.8 q12,q9,q10,#8 + add r3,r3,r9 + and r10,r6,r7 + ldr r9,[sp,#52] + veor q11,q11,q3 + add r3,r3,r4,ror#27 + eor r11,r6,r7 + veor q11,q11,q0 + add r3,r3,r10 + and r11,r11,r5 + vadd.i32 q13,q10,q14 + mov r5,r5,ror#2 + vld1.32 {d28[],d29[]},[r8,:32]! + add r3,r3,r11 + veor q12,q12,q11 + add r7,r7,r9 + and r10,r5,r6 + vshr.u32 q11,q12,#30 + ldr r9,[sp,#56] + add r7,r7,r3,ror#27 + vst1.32 {q13},[r12,:128]! + eor r11,r5,r6 + add r7,r7,r10 + vsli.32 q11,q12,#2 + and r11,r11,r4 + mov r4,r4,ror#2 + add r7,r7,r11 + add r6,r6,r9 + and r10,r4,r5 + ldr r9,[sp,#60] + add r6,r6,r7,ror#27 + eor r11,r4,r5 + add r6,r6,r10 + and r11,r11,r3 + mov r3,r3,ror#2 + add r6,r6,r11 + add r5,r5,r9 + and r10,r3,r4 + ldr r9,[sp,#0] + add r5,r5,r6,ror#27 + eor r11,r3,r4 + add r5,r5,r10 + and r11,r11,r7 + mov r7,r7,ror#2 + add r5,r5,r11 + vext.8 q12,q10,q11,#8 + add r4,r4,r9 + and r10,r7,r3 + ldr r9,[sp,#4] + veor q0,q0,q8 + add r4,r4,r5,ror#27 + eor r11,r7,r3 + veor q0,q0,q1 + add r4,r4,r10 + and r11,r11,r6 + vadd.i32 q13,q11,q14 + mov r6,r6,ror#2 + add r4,r4,r11 + veor q12,q12,q0 + add r3,r3,r9 + and r10,r6,r7 + vshr.u32 q0,q12,#30 + ldr r9,[sp,#8] + add r3,r3,r4,ror#27 + vst1.32 {q13},[r12,:128]! + sub r12,r12,#64 + eor r11,r6,r7 + add r3,r3,r10 + vsli.32 q0,q12,#2 + and r11,r11,r5 + mov r5,r5,ror#2 + add r3,r3,r11 + add r7,r7,r9 + and r10,r5,r6 + ldr r9,[sp,#12] + add r7,r7,r3,ror#27 + eor r11,r5,r6 + add r7,r7,r10 + and r11,r11,r4 + mov r4,r4,ror#2 + add r7,r7,r11 + add r6,r6,r9 + and r10,r4,r5 + ldr r9,[sp,#16] + add r6,r6,r7,ror#27 + eor r11,r4,r5 + add r6,r6,r10 + and r11,r11,r3 + mov r3,r3,ror#2 + add r6,r6,r11 + vext.8 q12,q11,q0,#8 + add r5,r5,r9 + and r10,r3,r4 + ldr r9,[sp,#20] + veor q1,q1,q9 + add r5,r5,r6,ror#27 + eor r11,r3,r4 + veor q1,q1,q2 + add r5,r5,r10 + and r11,r11,r7 + vadd.i32 q13,q0,q14 + mov r7,r7,ror#2 + add r5,r5,r11 + veor q12,q12,q1 + add r4,r4,r9 + and r10,r7,r3 + vshr.u32 q1,q12,#30 + ldr r9,[sp,#24] + add r4,r4,r5,ror#27 + vst1.32 {q13},[r12,:128]! + eor r11,r7,r3 + add r4,r4,r10 + vsli.32 q1,q12,#2 + and r11,r11,r6 + mov r6,r6,ror#2 + add r4,r4,r11 + add r3,r3,r9 + and r10,r6,r7 + ldr r9,[sp,#28] + add r3,r3,r4,ror#27 + eor r11,r6,r7 + add r3,r3,r10 + and r11,r11,r5 + mov r5,r5,ror#2 + add r3,r3,r11 + add r7,r7,r9 + and r10,r5,r6 + ldr r9,[sp,#32] + add r7,r7,r3,ror#27 + eor r11,r5,r6 + add r7,r7,r10 + and r11,r11,r4 + mov r4,r4,ror#2 + add r7,r7,r11 + vext.8 q12,q0,q1,#8 + add r6,r6,r9 + and r10,r4,r5 + ldr r9,[sp,#36] + veor q2,q2,q10 + add r6,r6,r7,ror#27 + eor r11,r4,r5 + veor q2,q2,q3 + add r6,r6,r10 + and r11,r11,r3 + vadd.i32 q13,q1,q14 + mov r3,r3,ror#2 + add r6,r6,r11 + veor q12,q12,q2 + add r5,r5,r9 + and r10,r3,r4 + vshr.u32 q2,q12,#30 + ldr r9,[sp,#40] + add r5,r5,r6,ror#27 + vst1.32 {q13},[r12,:128]! + eor r11,r3,r4 + add r5,r5,r10 + vsli.32 q2,q12,#2 + and r11,r11,r7 + mov r7,r7,ror#2 + add r5,r5,r11 + add r4,r4,r9 + and r10,r7,r3 + ldr r9,[sp,#44] + add r4,r4,r5,ror#27 + eor r11,r7,r3 + add r4,r4,r10 + and r11,r11,r6 + mov r6,r6,ror#2 + add r4,r4,r11 + add r3,r3,r9 + and r10,r6,r7 + ldr r9,[sp,#48] + add r3,r3,r4,ror#27 + eor r11,r6,r7 + add r3,r3,r10 + and r11,r11,r5 + mov r5,r5,ror#2 + add r3,r3,r11 + vext.8 q12,q1,q2,#8 + eor r10,r4,r6 + add r7,r7,r9 + ldr r9,[sp,#52] + veor q3,q3,q11 + eor r11,r10,r5 + add r7,r7,r3,ror#27 + veor q3,q3,q8 + mov r4,r4,ror#2 + add r7,r7,r11 + vadd.i32 q13,q2,q14 + eor r10,r3,r5 + add r6,r6,r9 + veor q12,q12,q3 + ldr r9,[sp,#56] + eor r11,r10,r4 + vshr.u32 q3,q12,#30 + add r6,r6,r7,ror#27 + mov r3,r3,ror#2 + vst1.32 {q13},[r12,:128]! + add r6,r6,r11 + eor r10,r7,r4 + vsli.32 q3,q12,#2 + add r5,r5,r9 + ldr r9,[sp,#60] + eor r11,r10,r3 + add r5,r5,r6,ror#27 + mov r7,r7,ror#2 + add r5,r5,r11 + eor r10,r6,r3 + add r4,r4,r9 + ldr r9,[sp,#0] + eor r11,r10,r7 + add r4,r4,r5,ror#27 + mov r6,r6,ror#2 + add r4,r4,r11 + vadd.i32 q13,q3,q14 + eor r10,r5,r7 + add r3,r3,r9 + vst1.32 {q13},[r12,:128]! + sub r12,r12,#64 + teq r1,r2 + sub r8,r8,#16 + it eq + subeq r1,r1,#64 + vld1.8 {q0,q1},[r1]! + ldr r9,[sp,#4] + eor r11,r10,r6 + vld1.8 {q2,q3},[r1]! + add r3,r3,r4,ror#27 + mov r5,r5,ror#2 + vld1.32 {d28[],d29[]},[r8,:32]! + add r3,r3,r11 + eor r10,r4,r6 + vrev32.8 q0,q0 + add r7,r7,r9 + ldr r9,[sp,#8] + eor r11,r10,r5 + add r7,r7,r3,ror#27 + mov r4,r4,ror#2 + add r7,r7,r11 + eor r10,r3,r5 + add r6,r6,r9 + ldr r9,[sp,#12] + eor r11,r10,r4 + add r6,r6,r7,ror#27 + mov r3,r3,ror#2 + add r6,r6,r11 + eor r10,r7,r4 + add r5,r5,r9 + ldr r9,[sp,#16] + eor r11,r10,r3 + add r5,r5,r6,ror#27 + mov r7,r7,ror#2 + add r5,r5,r11 + vrev32.8 q1,q1 + eor r10,r6,r3 + add r4,r4,r9 + vadd.i32 q8,q0,q14 + ldr r9,[sp,#20] + eor r11,r10,r7 + vst1.32 {q8},[r12,:128]! + add r4,r4,r5,ror#27 + mov r6,r6,ror#2 + add r4,r4,r11 + eor r10,r5,r7 + add r3,r3,r9 + ldr r9,[sp,#24] + eor r11,r10,r6 + add r3,r3,r4,ror#27 + mov r5,r5,ror#2 + add r3,r3,r11 + eor r10,r4,r6 + add r7,r7,r9 + ldr r9,[sp,#28] + eor r11,r10,r5 + add r7,r7,r3,ror#27 + mov r4,r4,ror#2 + add r7,r7,r11 + eor r10,r3,r5 + add r6,r6,r9 + ldr r9,[sp,#32] + eor r11,r10,r4 + add r6,r6,r7,ror#27 + mov r3,r3,ror#2 + add r6,r6,r11 + vrev32.8 q2,q2 + eor r10,r7,r4 + add r5,r5,r9 + vadd.i32 q9,q1,q14 + ldr r9,[sp,#36] + eor r11,r10,r3 + vst1.32 {q9},[r12,:128]! + add r5,r5,r6,ror#27 + mov r7,r7,ror#2 + add r5,r5,r11 + eor r10,r6,r3 + add r4,r4,r9 + ldr r9,[sp,#40] + eor r11,r10,r7 + add r4,r4,r5,ror#27 + mov r6,r6,ror#2 + add r4,r4,r11 + eor r10,r5,r7 + add r3,r3,r9 + ldr r9,[sp,#44] + eor r11,r10,r6 + add r3,r3,r4,ror#27 + mov r5,r5,ror#2 + add r3,r3,r11 + eor r10,r4,r6 + add r7,r7,r9 + ldr r9,[sp,#48] + eor r11,r10,r5 + add r7,r7,r3,ror#27 + mov r4,r4,ror#2 + add r7,r7,r11 + vrev32.8 q3,q3 + eor r10,r3,r5 + add r6,r6,r9 + vadd.i32 q10,q2,q14 + ldr r9,[sp,#52] + eor r11,r10,r4 + vst1.32 {q10},[r12,:128]! + add r6,r6,r7,ror#27 + mov r3,r3,ror#2 + add r6,r6,r11 + eor r10,r7,r4 + add r5,r5,r9 + ldr r9,[sp,#56] + eor r11,r10,r3 + add r5,r5,r6,ror#27 + mov r7,r7,ror#2 + add r5,r5,r11 + eor r10,r6,r3 + add r4,r4,r9 + ldr r9,[sp,#60] + eor r11,r10,r7 + add r4,r4,r5,ror#27 + mov r6,r6,ror#2 + add r4,r4,r11 + eor r10,r5,r7 + add r3,r3,r9 + eor r11,r10,r6 + add r3,r3,r4,ror#27 + mov r5,r5,ror#2 + add r3,r3,r11 + ldmia r0,{r9,r10,r11,r12} @ accumulate context + add r3,r3,r9 + ldr r9,[r0,#16] + add r4,r4,r10 + add r5,r5,r11 + add r6,r6,r12 + it eq + moveq sp,r14 + add r7,r7,r9 + it ne + ldrne r9,[sp] + stmia r0,{r3,r4,r5,r6,r7} + itt ne + addne r12,sp,#3*16 + bne .Loop_neon + + @ vldmia sp!,{d8-d15} + ldmia sp!,{r4,r5,r6,r7,r8,r9,r10,r11,r12,pc} +.size sha1_block_data_order_neon,.-sha1_block_data_order_neon +#endif + +#if __ARM_MAX_ARCH__>=7 +.comm CRYPTOGAMS_armcaps,4,4 +#endif diff --git a/sha1_armv4.h b/sha1_armv4.h new file mode 100644 index 00000000..14e3e051 --- /dev/null +++ b/sha1_armv4.h @@ -0,0 +1,21 @@ +/* Header file for use with Cryptogam's ARMv4 SHA1. */ +/* Also see http://www.openssl.org/~appro/cryptogams/ */ +/* https://wiki.openssl.org/index.php/Cryptogams_SHA. */ + +#ifndef CRYPTOGAMS_SHA1_ARMV4_H +#define CRYPTOGAMS_SHA1_ARMV4_H + +#ifdef __cplusplus +extern "C" { +#endif + +void sha1_block_data_order(void *state, const void *data, size_t blocks); + +/* Cryptogams arm caps */ +#define ARMV7_NEON (1<<0) + +#ifdef __cplusplus +} +#endif + +#endif /* CRYPTOGAMS_SHA1_ARMV4_H */