fzorb/REDACTED-rig
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

CryptoNight v8 ReverseWaltz (+asm code and all test vectors)

Browse source 
Co-Authored-By: Vitalius Parubochyi <eddragonwolf@users.noreply.github.com>
This commit is contained in:
SChernykh 2019-03-04 19:57:33 +01:00
parent 10eb579270
commit 154806ac74
18 changed files with 1353 additions and 29 deletions
Download patch file Download diff file Expand all files Collapse all files

34
src/crypto/CryptoNight_monero.h
View file

@ -83,11 +83,11 @@
sqrt_result_xmm_##part = int_sqrt_v2(cx_0 + division_result); \
} while (0)
# define VARIANT2_SHUFFLE(base_ptr, offset, _a, _b, _b1, _c) \
# define VARIANT2_SHUFFLE(base_ptr, offset, _a, _b, _b1, _c, reverse) \
do { \
const __m128i chunk1 = _mm_load_si128((__m128i *)((base_ptr) + ((offset) ^ 0x10))); \
const __m128i chunk1 = _mm_load_si128((__m128i *)((base_ptr) + ((offset) ^ (reverse ? 0x30 : 0x10)))); \
const __m128i chunk2 = _mm_load_si128((__m128i *)((base_ptr) + ((offset) ^ 0x20))); \
const __m128i chunk3 = _mm_load_si128((__m128i *)((base_ptr) + ((offset) ^ 0x30))); \
const __m128i chunk3 = _mm_load_si128((__m128i *)((base_ptr) + ((offset) ^ (reverse ? 0x10 : 0x30)))); \
_mm_store_si128((__m128i *)((base_ptr) + ((offset) ^ 0x10)), _mm_add_epi64(chunk3, _b1)); \
_mm_store_si128((__m128i *)((base_ptr) + ((offset) ^ 0x20)), _mm_add_epi64(chunk1, _b)); \
_mm_store_si128((__m128i *)((base_ptr) + ((offset) ^ 0x30)), _mm_add_epi64(chunk2, _a)); \
@ -96,15 +96,20 @@
} \
} while (0)
# define VARIANT2_SHUFFLE2(base_ptr, offset, _a, _b, _b1, hi, lo) \
# define VARIANT2_SHUFFLE2(base_ptr, offset, _a, _b, _b1, hi, lo, reverse) \
do { \
const __m128i chunk1 = _mm_xor_si128(_mm_load_si128((__m128i *)((base_ptr) + ((offset) ^ 0x10))), _mm_set_epi64x(lo, hi)); \
const __m128i chunk2 = _mm_load_si128((__m128i *)((base_ptr) + ((offset) ^ 0x20))); \
hi ^= ((uint64_t*)((base_ptr) + ((offset) ^ 0x20)))[0]; \
lo ^= ((uint64_t*)((base_ptr) + ((offset) ^ 0x20)))[1]; \
const __m128i chunk3 = _mm_load_si128((__m128i *)((base_ptr) + ((offset) ^ 0x30))); \
_mm_store_si128((__m128i *)((base_ptr) + ((offset) ^ 0x10)), _mm_add_epi64(chunk3, _b1)); \
_mm_store_si128((__m128i *)((base_ptr) + ((offset) ^ 0x20)), _mm_add_epi64(chunk1, _b)); \
if (reverse) { \
_mm_store_si128((__m128i *)((base_ptr) + ((offset) ^ 0x10)), _mm_add_epi64(chunk1, _b1)); \
_mm_store_si128((__m128i *)((base_ptr) + ((offset) ^ 0x20)), _mm_add_epi64(chunk3, _b)); \
} else { \
_mm_store_si128((__m128i *)((base_ptr) + ((offset) ^ 0x10)), _mm_add_epi64(chunk3, _b1)); \
_mm_store_si128((__m128i *)((base_ptr) + ((offset) ^ 0x20)), _mm_add_epi64(chunk1, _b)); \
} \
_mm_store_si128((__m128i *)((base_ptr) + ((offset) ^ 0x30)), _mm_add_epi64(chunk2, _a)); \
} while (0)
@ -128,11 +133,11 @@
sqrt_result_##part += ((r2 + b > sqrt_input) ? -1 : 0) + ((r2 + (1ULL << 32) < sqrt_input - s) ? 1 : 0); \
} while (0)
# define VARIANT2_SHUFFLE(base_ptr, offset, _a, _b, _b1, _c) \
# define VARIANT2_SHUFFLE(base_ptr, offset, _a, _b, _b1, _c, reverse) \
do { \
const uint64x2_t chunk1 = vld1q_u64((uint64_t*)((base_ptr) + ((offset) ^ 0x10))); \
const uint64x2_t chunk1 = vld1q_u64((uint64_t*)((base_ptr) + ((offset) ^ (reverse ? 0x30 : 0x10)))); \
const uint64x2_t chunk2 = vld1q_u64((uint64_t*)((base_ptr) + ((offset) ^ 0x20))); \
const uint64x2_t chunk3 = vld1q_u64((uint64_t*)((base_ptr) + ((offset) ^ 0x30))); \
const uint64x2_t chunk3 = vld1q_u64((uint64_t*)((base_ptr) + ((offset) ^ (reverse ? 0x10 : 0x30)))); \
vst1q_u64((uint64_t*)((base_ptr) + ((offset) ^ 0x10)), vaddq_u64(chunk3, vreinterpretq_u64_u8(_b1))); \
vst1q_u64((uint64_t*)((base_ptr) + ((offset) ^ 0x20)), vaddq_u64(chunk1, vreinterpretq_u64_u8(_b))); \
vst1q_u64((uint64_t*)((base_ptr) + ((offset) ^ 0x30)), vaddq_u64(chunk2, vreinterpretq_u64_u8(_a))); \
@ -141,15 +146,20 @@
} \
} while (0)
# define VARIANT2_SHUFFLE2(base_ptr, offset, _a, _b, _b1, hi, lo) \
# define VARIANT2_SHUFFLE2(base_ptr, offset, _a, _b, _b1, hi, lo, reverse) \
do { \
const uint64x2_t chunk1 = veorq_u64(vld1q_u64((uint64_t*)((base_ptr) + ((offset) ^ 0x10))), vcombine_u64(vcreate_u64(hi), vcreate_u64(lo))); \
const uint64x2_t chunk2 = vld1q_u64((uint64_t*)((base_ptr) + ((offset) ^ 0x20))); \
hi ^= ((uint64_t*)((base_ptr) + ((offset) ^ 0x20)))[0]; \
lo ^= ((uint64_t*)((base_ptr) + ((offset) ^ 0x20)))[1]; \
const uint64x2_t chunk3 = vld1q_u64((uint64_t*)((base_ptr) + ((offset) ^ 0x30))); \
vst1q_u64((uint64_t*)((base_ptr) + ((offset) ^ 0x10)), vaddq_u64(chunk3, vreinterpretq_u64_u8(_b1))); \
vst1q_u64((uint64_t*)((base_ptr) + ((offset) ^ 0x20)), vaddq_u64(chunk1, vreinterpretq_u64_u8(_b))); \
if (reverse) { \
vst1q_u64((uint64_t*)((base_ptr) + ((offset) ^ 0x10)), vaddq_u64(chunk1, vreinterpretq_u64_u8(_b1))); \
vst1q_u64((uint64_t*)((base_ptr) + ((offset) ^ 0x20)), vaddq_u64(chunk3, vreinterpretq_u64_u8(_b))); \
} else { \
vst1q_u64((uint64_t*)((base_ptr) + ((offset) ^ 0x10)), vaddq_u64(chunk3, vreinterpretq_u64_u8(_b1))); \
vst1q_u64((uint64_t*)((base_ptr) + ((offset) ^ 0x20)), vaddq_u64(chunk1, vreinterpretq_u64_u8(_b))); \
} \
vst1q_u64((uint64_t*)((base_ptr) + ((offset) ^ 0x30)), vaddq_u64(chunk2, vreinterpretq_u64_u8(_a))); \
} while (0)
#endif