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

Added reference implementation for cn/gpu (cn-gpu).

Browse source
This commit is contained in:
XMRig 2019-02-03 14:44:23 +07:00
parent 7605c4241f
commit bd4bc9ba4d
18 changed files with 614 additions and 31 deletions
Download patch file Download diff file Expand all files Collapse all files

66
src/crypto/CryptoNight_x86.h
View file

@ -4,7 +4,7 @@
* Copyright 2014 Lucas Jones <https://github.com/lucasjones>
* Copyright 2014-2016 Wolf9466 <https://github.com/OhGodAPet>
* Copyright 2016 Jay D Dee <jayddee246@gmail.com>
* Copyright 2017-2018 XMR-Stak <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
* Copyright 2017-2019 XMR-Stak <https://github.com/fireice-uk>, <https://github.com/psychocrypt>
* Copyright 2018 Lee Clagett <https://github.com/vtnerd>
* Copyright 2018-2019 SChernykh <https://github.com/SChernykh>
* Copyright 2016-2019 XMRig <https://github.com/xmrig>, <support@xmrig.com>
@ -35,6 +35,7 @@
#endif
#include "common/cpu/Cpu.h"
#include "common/crypto/keccak.h"
#include "crypto/CryptoNight.h"
#include "crypto/CryptoNight_constants.h"
@ -289,6 +290,34 @@ static inline void cn_explode_scratchpad(const __m128i *input, __m128i *output)
}
#ifndef XMRIG_NO_CN_GPU
template<xmrig::Algo ALGO, size_t MEM>
void cn_explode_scratchpad_gpu(const uint8_t *input, uint8_t *output)
{
constexpr size_t hash_size = 200; // 25x8 bytes
alignas(16) uint64_t hash[25];
for (uint64_t i = 0; i < MEM / 512; i++)
{
memcpy(hash, input, hash_size);
hash[0] ^= i;
xmrig::keccakf(hash, 24);
memcpy(output, hash, 160);
output += 160;
xmrig::keccakf(hash, 24);
memcpy(output, hash, 176);
output += 176;
xmrig::keccakf(hash, 24);
memcpy(output, hash, 176);
output += 176;
}
}
#endif
template<xmrig::Algo ALGO, size_t MEM, bool SOFT_AES>
static inline void cn_implode_scratchpad(const __m128i *input, __m128i *output)
{
@ -566,6 +595,41 @@ inline void cryptonight_single_hash(const uint8_t *__restrict__ input, size_t si
}
#ifndef XMRIG_NO_CN_GPU
template<size_t ITER, uint32_t MASK>
void cn_gpu_inner_avx(const uint8_t *spad, uint8_t *lpad);
template<size_t ITER, uint32_t MASK>
void cn_gpu_inner_ssse3(const uint8_t *spad, uint8_t *lpad);
template<xmrig::Algo ALGO, bool SOFT_AES, xmrig::Variant VARIANT>
inline void cryptonight_single_hash_gpu(const uint8_t *__restrict__ input, size_t size, uint8_t *__restrict__ output, cryptonight_ctx **__restrict__ ctx)
{
constexpr size_t MASK = xmrig::CRYPTONIGHT_GPU_MASK;
constexpr size_t ITERATIONS = xmrig::cn_select_iter<ALGO, VARIANT>();
constexpr size_t MEM = xmrig::cn_select_memory<ALGO>();
static_assert(MASK > 0 && ITERATIONS > 0 && MEM > 0, "unsupported algorithm/variant");
xmrig::keccak(input, size, ctx[0]->state);
cn_explode_scratchpad_gpu<ALGO, MEM>(ctx[0]->state, ctx[0]->memory);
if (xmrig::Cpu::info()->hasAVX2()) {
cn_gpu_inner_avx<ITERATIONS, MASK>(ctx[0]->state, ctx[0]->memory);
} else {
cn_gpu_inner_ssse3<ITERATIONS, MASK>(ctx[0]->state, ctx[0]->memory);
}
cn_implode_scratchpad<xmrig::CRYPTONIGHT_HEAVY, MEM, SOFT_AES>((__m128i*) ctx[0]->memory, (__m128i*) ctx[0]->state);
xmrig::keccakf((uint64_t*) ctx[0]->state, 24);
memcpy(output, ctx[0]->state, 32);
}
#endif
#ifndef XMRIG_NO_ASM
extern "C" void cnv2_mainloop_ivybridge_asm(cryptonight_ctx *ctx);
extern "C" void cnv2_mainloop_ryzen_asm(cryptonight_ctx *ctx);