aaru-dps/Aaru.Checksums.Native
1
0
Fork 0
mirror of https://github.com/aaru-dps/Aaru.Checksums.Native.git synced 2025-12-16 11:14:29 +00:00
Code Issues Packages Projects Releases Wiki Activity

Refactor code.

Browse Source
This commit is contained in:
Natalia Portillo
2021-10-13 03:46:47 +01:00
parent ca65d12c7e
commit 58f97a93f3
10 changed files with 103 additions and 106 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
adler32.h
View File

@@ -39,7 +39,7 @@ AARU_EXPORT void AARU_CALL adler32_slicing(uint16_t* sum1, uint16_t* sum
defined(__i386__) || defined(__THW_INTEL) || defined(_M_IX86)
AARU_EXPORT SSSE3 void AARU_CALL adler32_ssse3(uint16_t* sum1, uint16_t* sum2, const uint8_t* data, long len);
AARU_EXPORT AVX2 void AARU_CALL adler32_avx2(uint16_t* sum1, uint16_t* sum2, const uint8_t* data, long len);
AARU_EXPORT AVX2 void AARU_CALL adler32_avx2(uint16_t* sum1, uint16_t* sum2, const uint8_t* data, long len);
#endif

34
adler32_neon.c
View File

@@ -77,8 +77,8 @@ TARGET_WITH_SIMD void adler32_neon(uint16_t* sum1, uint16_t* sum2, const uint8_t
uint32x4_t v_s2 = {.n128_u32 = {0, 0, 0, s1 * n}};
uint32x4_t v_s1 = {.n128_u32 = {0, 0, 0, 0}};
#else
uint32x4_t v_s2 = (uint32x4_t){0, 0, 0, s1 * n};
uint32x4_t v_s1 = (uint32x4_t){0, 0, 0, 0};
uint32x4_t v_s2 = (uint32x4_t){0, 0, 0, s1 * n};
uint32x4_t v_s1 = (uint32x4_t){0, 0, 0, 0};
#endif
uint16x8_t v_column_sum_1 = vdupq_n_u16(0);
uint16x8_t v_column_sum_2 = vdupq_n_u16(0);
@@ -112,23 +112,23 @@ TARGET_WITH_SIMD void adler32_neon(uint16_t* sum1, uint16_t* sum2, const uint8_t
* Multiply-add bytes by [ 32, 31, 30, ... ] for s2.
*/
#ifdef _WIN32
v_s2 = vmlal_u16(v_s2, vget_low_u16(v_column_sum_1), neon_ld1m_16((uint16_t[]) {32, 31, 30, 29}));
v_s2 = vmlal_u16(v_s2, vget_low_u16(v_column_sum_1), neon_ld1m_16((uint16_t[]){32, 31, 30, 29}));
v_s2 = vmlal_u16(v_s2, vget_high_u16(v_column_sum_1), neon_ld1m_16((uint16_t[]){28, 27, 26, 25}));
v_s2 = vmlal_u16(v_s2, vget_low_u16(v_column_sum_2), neon_ld1m_16((uint16_t[]) {24, 23, 22, 21}));
v_s2 = vmlal_u16(v_s2, vget_high_u16(v_column_sum_2), neon_ld1m_16((uint16_t[]) {20, 19, 18, 17}));
v_s2 = vmlal_u16(v_s2, vget_low_u16(v_column_sum_3), neon_ld1m_16((uint16_t[]) {16, 15, 14, 13}));
v_s2 = vmlal_u16(v_s2, vget_high_u16(v_column_sum_3), neon_ld1m_16((uint16_t[]) {12, 11, 10, 9}));
v_s2 = vmlal_u16(v_s2, vget_low_u16(v_column_sum_4), neon_ld1m_16((uint16_t[]) {8, 7, 6, 5}));
v_s2 = vmlal_u16(v_s2, vget_high_u16(v_column_sum_4), neon_ld1m_16((uint16_t[]) {4, 3, 2, 1}));
v_s2 = vmlal_u16(v_s2, vget_low_u16(v_column_sum_2), neon_ld1m_16((uint16_t[]){24, 23, 22, 21}));
v_s2 = vmlal_u16(v_s2, vget_high_u16(v_column_sum_2), neon_ld1m_16((uint16_t[]){20, 19, 18, 17}));
v_s2 = vmlal_u16(v_s2, vget_low_u16(v_column_sum_3), neon_ld1m_16((uint16_t[]){16, 15, 14, 13}));
v_s2 = vmlal_u16(v_s2, vget_high_u16(v_column_sum_3), neon_ld1m_16((uint16_t[]){12, 11, 10, 9}));
v_s2 = vmlal_u16(v_s2, vget_low_u16(v_column_sum_4), neon_ld1m_16((uint16_t[]){8, 7, 6, 5}));
v_s2 = vmlal_u16(v_s2, vget_high_u16(v_column_sum_4), neon_ld1m_16((uint16_t[]){4, 3, 2, 1}));
#else
v_s2 = vmlal_u16(v_s2, vget_low_u16(v_column_sum_1), (uint16x4_t){32, 31, 30, 29});
v_s2 = vmlal_u16(v_s2, vget_high_u16(v_column_sum_1), (uint16x4_t){28, 27, 26, 25});
v_s2 = vmlal_u16(v_s2, vget_low_u16(v_column_sum_2), (uint16x4_t){24, 23, 22, 21});
v_s2 = vmlal_u16(v_s2, vget_high_u16(v_column_sum_2), (uint16x4_t){20, 19, 18, 17});
v_s2 = vmlal_u16(v_s2, vget_low_u16(v_column_sum_3), (uint16x4_t){16, 15, 14, 13});
v_s2 = vmlal_u16(v_s2, vget_high_u16(v_column_sum_3), (uint16x4_t){12, 11, 10, 9});
v_s2 = vmlal_u16(v_s2, vget_low_u16(v_column_sum_4), (uint16x4_t){8, 7, 6, 5});
v_s2 = vmlal_u16(v_s2, vget_high_u16(v_column_sum_4), (uint16x4_t){4, 3, 2, 1});
v_s2 = vmlal_u16(v_s2, vget_low_u16(v_column_sum_1), (uint16x4_t){32, 31, 30, 29});
v_s2 = vmlal_u16(v_s2, vget_high_u16(v_column_sum_1), (uint16x4_t){28, 27, 26, 25});
v_s2 = vmlal_u16(v_s2, vget_low_u16(v_column_sum_2), (uint16x4_t){24, 23, 22, 21});
v_s2 = vmlal_u16(v_s2, vget_high_u16(v_column_sum_2), (uint16x4_t){20, 19, 18, 17});
v_s2 = vmlal_u16(v_s2, vget_low_u16(v_column_sum_3), (uint16x4_t){16, 15, 14, 13});
v_s2 = vmlal_u16(v_s2, vget_high_u16(v_column_sum_3), (uint16x4_t){12, 11, 10, 9});
v_s2 = vmlal_u16(v_s2, vget_low_u16(v_column_sum_4), (uint16x4_t){8, 7, 6, 5});
v_s2 = vmlal_u16(v_s2, vget_high_u16(v_column_sum_4), (uint16x4_t){4, 3, 2, 1});
#endif
/*
* Sum epi32 ints v_s1(s2) and accumulate in s1(s2).

4
crc32.h
View File

@@ -275,9 +275,9 @@ AARU_EXPORT CLMUL uint32_t AARU_CALL crc32_clmul(uint32_t previous_crc, const ui
#if defined(__aarch64__) || defined(_M_ARM64) || defined(__arm__) || defined(_M_ARM)
#if __ARM_ARCH >= 8
AARU_EXPORT TARGET_ARMV8_WITH_CRC uint32_t AARU_CALL armv8_crc32_little(uint32_t previous_crc,
AARU_EXPORT TARGET_ARMV8_WITH_CRC uint32_t AARU_CALL armv8_crc32_little(uint32_t previous_crc,
const uint8_t* data,
uint32_t len);
uint32_t len);
#endif
AARU_EXPORT TARGET_WITH_SIMD uint32_t AARU_CALL crc32_vmull(uint32_t previous_crc, const uint8_t* data, long len);
#endif

4
crc32_vmull.c
View File

@@ -26,13 +26,13 @@
#if defined(__aarch64__) || defined(_M_ARM64) || defined(__arm__) || defined(_M_ARM)
#include <arm_neon.h>
#include <stdint.h>
#include <stddef.h>
#include <stdint.h>
#include "library.h"
#include "arm_vmull.h"
#include "crc32.h"
#include "crc32_simd.h"
#include "arm_vmull.h"
/*
* somewhat surprisingly the "naive" way of doing this, ie. with a flag and a cond. branch,

4
crc64_clmul.c
View File

@@ -114,7 +114,7 @@ AARU_EXPORT CLMUL uint64_t AARU_CALL crc64_clmul(uint64_t crc, const uint8_t* da
const __m128i P = _mm_xor_si128(A, crc0);
R = _mm_xor_si128(_mm_clmulepi64_si128(P, foldConstants1, 0x10),
_mm_xor_si128(_mm_srli_si128(P, 8), _mm_slli_si128(crc1, 8)));
_mm_xor_si128(_mm_srli_si128(P, 8), _mm_slli_si128(crc1, 8)));
}
else if(alignedLength == 2)
{
@@ -132,7 +132,7 @@ AARU_EXPORT CLMUL uint64_t AARU_CALL crc64_clmul(uint64_t crc, const uint8_t* da
const __m128i P = _mm_xor_si128(_mm_xor_si128(B, C), crc0);
R = _mm_xor_si128(_mm_clmulepi64_si128(P, foldConstants1, 0x10),
_mm_xor_si128(_mm_srli_si128(P, 8), _mm_slli_si128(crc1, 8)));
_mm_xor_si128(_mm_srli_si128(P, 8), _mm_slli_si128(crc1, 8)));
}
else
{

20
crc64_vmull.c
View File

@@ -1,16 +1,16 @@
/*
* This file is part of the Aaru Data Preservation Suite.
* Copyright (c) 2019-2021 Natalia Portillo.
*
* This file is under the public domain:
* https://github.com/rawrunprotected/crc
*/
* This file is part of the Aaru Data Preservation Suite.
* Copyright (c) 2019-2021 Natalia Portillo.
*
* This file is under the public domain:
* https://github.com/rawrunprotected/crc
*/
#if defined(__aarch64__) || defined(_M_ARM64) || defined(__arm__) || defined(_M_ARM)
#include <arm_neon.h>
#include <stdint.h>
#include <stddef.h>
#include <stdint.h>
#include "library.h"
#include "arm_vmull.h"
@@ -26,7 +26,7 @@ TARGET_WITH_SIMD FORCE_INLINE void shiftRight128(uint64x2_t in, size_t n, uint64
const uint64x2_t maskA =
vreinterpretq_u64_u32(vld1q_u32((const uint32_t*)(const uint64x2_t*)(shuffleMasks + (16 - n))));
uint64x2_t b = vreinterpretq_u64_u8(vceqq_u8(vreinterpretq_u8_u64(vreinterpretq_u64_u32(vdupq_n_u32(0))),
vreinterpretq_u8_u64(vreinterpretq_u64_u32(vdupq_n_u32(0)))));
vreinterpretq_u8_u64(vreinterpretq_u64_u32(vdupq_n_u32(0)))));
const uint64x2_t maskB = vreinterpretq_u64_u32(veorq_u32(vreinterpretq_u32_u64(maskA), vreinterpretq_u32_u64(b)));
*outLeft = mm_shuffle_epi8(in, maskB);
@@ -84,7 +84,7 @@ AARU_EXPORT TARGET_WITH_SIMD uint64_t AARU_CALL crc64_vmull(uint64_t previous_cr
const uint64x2_t P = veorq_u64(A, crc0);
R = veorq_u64(sse2neon_vmull_p64(vget_low_u64(P), vget_high_u64(foldConstants1)),
veorq_u64(mm_srli_si128(P, 8), mm_slli_si128(crc1, 8)));
veorq_u64(mm_srli_si128(P, 8), mm_slli_si128(crc1, 8)));
}
else if(alignedLength == 2)
{
@@ -102,7 +102,7 @@ AARU_EXPORT TARGET_WITH_SIMD uint64_t AARU_CALL crc64_vmull(uint64_t previous_cr
const uint64x2_t P = veorq_u64(veorq_u64(B, C), crc0);
R = veorq_u64(sse2neon_vmull_p64(vget_low_u64(P), vget_high_u64(foldConstants1)),
veorq_u64(mm_srli_si128(P, 8), mm_slli_si128(crc1, 8)));
veorq_u64(mm_srli_si128(P, 8), mm_slli_si128(crc1, 8)));
}
else
{

32
library.h
View File

@@ -1,20 +1,20 @@
/*
* This file is part of the Aaru Data Preservation Suite.
* Copyright (c) 2019-2021 Natalia Portillo.
*
* This library is free software; you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public License as
* published by the Free Software Foundation; either version 2.1 of the
* License, or (at your option) any later version.
*
* This library 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, see <http://www.gnu.org/licenses/>.
*/
* This file is part of the Aaru Data Preservation Suite.
* Copyright (c) 2019-2021 Natalia Portillo.
*
* This library is free software; you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public License as
* published by the Free Software Foundation; either version 2.1 of the
* License, or (at your option) any later version.
*
* This library 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, see <http://www.gnu.org/licenses/>.
*/
#ifndef AARU_CHECKSUMS_NATIVE_LIBRARY_H
#define AARU_CHECKSUMS_NATIVE_LIBRARY_H

51
simd.c
View File

@@ -1,20 +1,20 @@
/*
* This file is part of the Aaru Data Preservation Suite.
* Copyright (c) 2019-2021 Natalia Portillo.
*
* This library is free software; you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public License as
* published by the Free Software Foundation; either version 2.1 of the
* License, or (at your option) any later version.
*
* This library 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, see <http://www.gnu.org/licenses/>.
*/
* This file is part of the Aaru Data Preservation Suite.
* Copyright (c) 2019-2021 Natalia Portillo.
*
* This library is free software; you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public License as
* published by the Free Software Foundation; either version 2.1 of the
* License, or (at your option) any later version.
*
* This library 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, see <http://www.gnu.org/licenses/>.
*/
#include "library.h"
#include "simd.h"
@@ -112,6 +112,7 @@ int have_avx2(void)
#if defined(__aarch64__) || defined(_M_ARM64) || defined(__arm__) || defined(_M_ARM)
#if defined(_WIN32)
#include <windows.h>
#include <processthreadsapi.h>
#elif defined(__APPLE__)
#include <sys/sysctl.h>
@@ -120,29 +121,25 @@ int have_avx2(void)
#endif
#endif
#if (defined(__aarch64__) || defined(_M_ARM64) || defined(__arm__) || defined(_M_ARM)) && defined(__APPLE__)
#if(defined(__aarch64__) || defined(_M_ARM64) || defined(__arm__) || defined(_M_ARM)) && defined(__APPLE__)
int have_neon_apple()
{
int value;
int value;
size_t len = sizeof(int);
int ret=
sysctlbyname("hw.optional.neon", &value, &len, NULL, 0);
int ret = sysctlbyname("hw.optional.neon", &value, &len, NULL, 0);
if(ret != 0)
return 0;
if(ret != 0) return 0;
return value == 1;
}
int have_crc32_apple()
{
int value;
int value;
size_t len = sizeof(int);
int ret=
sysctlbyname("hw.optional.crc32", &value, &len, NULL, 0);
int ret = sysctlbyname("hw.optional.crc32", &value, &len, NULL, 0);
if(ret != 0)
return 0;
if(ret != 0) return 0;
return value == 1;
}

32
simd.h
View File

@@ -1,20 +1,20 @@
/*
* This file is part of the Aaru Data Preservation Suite.
* Copyright (c) 2019-2021 Natalia Portillo.
*
* This library is free software; you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public License as
* published by the Free Software Foundation; either version 2.1 of the
* License, or (at your option) any later version.
*
* This library 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, see <http://www.gnu.org/licenses/>.
*/
* This file is part of the Aaru Data Preservation Suite.
* Copyright (c) 2019-2021 Natalia Portillo.
*
* This library is free software; you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public License as
* published by the Free Software Foundation; either version 2.1 of the
* License, or (at your option) any later version.
*
* This library 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
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, see <http://www.gnu.org/licenses/>.
*/
#ifndef AARU_CHECKSUMS_NATIVE_SIMD_H
#define AARU_CHECKSUMS_NATIVE_SIMD_H

26
spamsum.c
View File

@@ -141,7 +141,7 @@ AARU_LOCAL inline void roll_hash(spamsum_ctx* ctx, uint8_t c)
AARU_LOCAL inline void fuzzy_try_reduce_blockhash(spamsum_ctx* ctx)
{
//assert(ctx->bh_start < ctx->bh_end);
// assert(ctx->bh_start < ctx->bh_end);
if(ctx->bh_end - ctx->bh_start < 2) /* Need at least two working hashes. */
return;
@@ -163,7 +163,7 @@ AARU_LOCAL inline void fuzzy_try_fork_blockhash(spamsum_ctx* ctx)
{
if(ctx->bh_end >= NUM_BLOCKHASHES) return;
//assert(ctx->bh_end != 0);
// assert(ctx->bh_end != 0);
uint32_t obh = ctx->bh_end - 1;
uint32_t nbh = ctx->bh_end;
@@ -184,7 +184,7 @@ AARU_EXPORT int AARU_CALL spamsum_final(spamsum_ctx* ctx, uint8_t* result)
if(!result) return -1;
/* Verify that our elimination was not overeager. */
//assert(bi == 0 || (uint64_t)SSDEEP_BS(bi) / 2 * SPAMSUM_LENGTH < ctx->total_size);
// assert(bi == 0 || (uint64_t)SSDEEP_BS(bi) / 2 * SPAMSUM_LENGTH < ctx->total_size);
/* Initial blocksize guess. */
while((uint64_t)SSDEEP_BS(bi) * SPAMSUM_LENGTH < ctx->total_size)
@@ -203,21 +203,21 @@ AARU_EXPORT int AARU_CALL spamsum_final(spamsum_ctx* ctx, uint8_t* result)
while(bi > ctx->bh_start && ctx->bh[bi].d_len < SPAMSUM_LENGTH / 2) --bi;
//assert(!(bi > 0 && ctx->bh[bi].d_len < SPAMSUM_LENGTH / 2));
// assert(!(bi > 0 && ctx->bh[bi].d_len < SPAMSUM_LENGTH / 2));
int i = snprintf((char*)result, (size_t)remain, "%lu:", (unsigned long)SSDEEP_BS(bi));
if(i <= 0) /* Maybe snprintf has set errno here? */
return -1;
//assert(i < remain);
// assert(i < remain);
remain -= i;
result += i;
i = (int)ctx->bh[bi].d_len;
//assert(i <= remain);
// assert(i <= remain);
memcpy(result, ctx->bh[bi].digest, (size_t)i);
result += i;
@@ -225,7 +225,7 @@ AARU_EXPORT int AARU_CALL spamsum_final(spamsum_ctx* ctx, uint8_t* result)
if(h != 0)
{
//assert(remain > 0);
// assert(remain > 0);
*result = b64[ctx->bh[bi].h % 64];
@@ -237,7 +237,7 @@ AARU_EXPORT int AARU_CALL spamsum_final(spamsum_ctx* ctx, uint8_t* result)
}
else if(ctx->bh[bi].digest[i] != 0)
{
//assert(remain > 0);
// assert(remain > 0);
*result = ctx->bh[bi].digest[i];
@@ -248,7 +248,7 @@ AARU_EXPORT int AARU_CALL spamsum_final(spamsum_ctx* ctx, uint8_t* result)
}
}
//assert(remain > 0);
// assert(remain > 0);
*result++ = ':';
--remain;
@@ -267,7 +267,7 @@ AARU_EXPORT int AARU_CALL spamsum_final(spamsum_ctx* ctx, uint8_t* result)
if(h != 0)
{
//assert(remain > 0);
// assert(remain > 0);
h = ctx->bh[bi].half_h;
*result = b64[h % 64];
@@ -284,7 +284,7 @@ AARU_EXPORT int AARU_CALL spamsum_final(spamsum_ctx* ctx, uint8_t* result)
if(i != 0)
{
//assert(remain > 0);
// assert(remain > 0);
*result = (uint8_t)i;
@@ -298,9 +298,9 @@ AARU_EXPORT int AARU_CALL spamsum_final(spamsum_ctx* ctx, uint8_t* result)
}
else if(h != 0)
{
//assert(ctx->bh[bi].d_len == 0);
// assert(ctx->bh[bi].d_len == 0);
//assert(remain > 0);
// assert(remain > 0);
*result++ = b64[ctx->bh[bi].h % 64];
/* No need to bother with FUZZY_FLAG_ELIMSEQ, because this