Add optional CryptoNight-Lite support.

CMakeLists.txt

@@ -1,7 +1,8 @@
 cmake_minimum_required(VERSION 3.0)
 project(xmrig C)
 
-option(WITH_LIBCPUID "Use Libcpuid library" ON)
+option(WITH_LIBCPUID "Use Libcpuid" ON)
+option(WITH_AEON     "CryptoNight-Lite support" ON)
 
 set(HEADERS
     compat.h
@@ -114,8 +115,21 @@ else()
     set(SOURCES_CPUID cpu_stub.c)
 endif()
 
+if (WITH_AEON)
+    set(SOURCES_AEON
+    algo/cryptonight-lite/cryptonight_lite_av1_aesni.c
+    algo/cryptonight-lite/cryptonight_lite_av2_aesni_double.c
+    algo/cryptonight-lite/cryptonight_lite_av3_softaes.c
+    algo/cryptonight-lite/cryptonight_lite_av4_softaes_double.c
+    algo/cryptonight-lite/cryptonight_lite_aesni.h
+    algo/cryptonight-lite/cryptonight_lite_softaes.h
+    )
+else()
+    add_definitions(/DXMRIG_NO_AEON)
+endif()
+
 if (CMAKE_SIZEOF_VOID_P EQUAL 8)
-    add_executable(xmrig ${HEADERS} ${HEADERS_CRYPTO} ${SOURCES} ${SOURCES_CRYPTO} ${HEADERS_UTILS} ${SOURCES_UTILS} ${HEADERS_COMPAT} ${SOURCES_COMPAT} ${SOURCES_OS} ${SOURCES_CPUID})
+    add_executable(xmrig ${HEADERS} ${HEADERS_CRYPTO} ${SOURCES} ${SOURCES_CRYPTO} ${HEADERS_UTILS} ${SOURCES_UTILS} ${HEADERS_COMPAT} ${SOURCES_COMPAT} ${SOURCES_OS} ${SOURCES_CPUID} ${SOURCES_AEON})
     target_link_libraries(xmrig jansson curl ${CPUID_LIB} ${EXTRA_LIBS})
 else()
     add_executable(xmrig32 ${HEADERS} ${HEADERS_CRYPTO} ${SOURCES} ${SOURCES_CRYPTO} ${HEADERS_UTILS} ${SOURCES_UTILS} ${HEADERS_COMPAT} ${SOURCES_COMPAT} ${SOURCES_OS} ${SOURCES_CPUID})

algo/cryptonight-lite/cryptonight_lite_aesni.h

@@ -0,0 +1,256 @@
+/* XMRig
+ * Copyright 2010      Jeff Garzik <jgarzik@pobox.com>
+ * Copyright 2012-2014 pooler      <pooler@litecoinpool.org>
+ * 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      fireice-uk  <https://github.com/fireice-uk>
+ * Copyright 2016-2017 XMRig       <support@xmrig.com>
+ *
+ *
+ *   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 3 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, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef __CRYPTONIGHT_AESNI_H__
+#define __CRYPTONIGHT_AESNI_H__
+
+#include <x86intrin.h>
+
+
+#define aes_genkey_sub(imm8) \
+    __m128i xout1 = _mm_aeskeygenassist_si128(*xout2, (imm8)); \
+    xout1  = _mm_shuffle_epi32(xout1, 0xFF); \
+    *xout0 = sl_xor(*xout0); \
+    *xout0 = _mm_xor_si128(*xout0, xout1); \
+    xout1  = _mm_aeskeygenassist_si128(*xout0, 0x00);\
+    xout1  = _mm_shuffle_epi32(xout1, 0xAA); \
+    *xout2 = sl_xor(*xout2); \
+    *xout2 = _mm_xor_si128(*xout2, xout1); \
+
+
+// This will shift and xor tmp1 into itself as 4 32-bit vals such as
+// sl_xor(a1 a2 a3 a4) = a1 (a2^a1) (a3^a2^a1) (a4^a3^a2^a1)
+inline __m128i sl_xor(__m128i tmp1)
+{
+    __m128i tmp4;
+    tmp4 = _mm_slli_si128(tmp1, 0x04);
+    tmp1 = _mm_xor_si128(tmp1, tmp4);
+    tmp4 = _mm_slli_si128(tmp4, 0x04);
+    tmp1 = _mm_xor_si128(tmp1, tmp4);
+    tmp4 = _mm_slli_si128(tmp4, 0x04);
+    tmp1 = _mm_xor_si128(tmp1, tmp4);
+    return tmp1;
+}
+
+
+inline void aes_genkey_sub1(__m128i* xout0, __m128i* xout2)
+{
+   aes_genkey_sub(0x1)
+}
+
+
+inline void aes_genkey_sub2(__m128i* xout0, __m128i* xout2)
+{
+   aes_genkey_sub(0x2)
+}
+
+
+inline void aes_genkey_sub4(__m128i* xout0, __m128i* xout2)
+{
+   aes_genkey_sub(0x4)
+}
+
+
+inline void aes_genkey_sub8(__m128i* xout0, __m128i* xout2)
+{
+   aes_genkey_sub(0x8)
+}
+
+
+inline void aes_round(__m128i key, __m128i* x0, __m128i* x1, __m128i* x2, __m128i* x3, __m128i* x4, __m128i* x5, __m128i* x6, __m128i* x7)
+{
+    *x0 = _mm_aesenc_si128(*x0, key);
+    *x1 = _mm_aesenc_si128(*x1, key);
+    *x2 = _mm_aesenc_si128(*x2, key);
+    *x3 = _mm_aesenc_si128(*x3, key);
+    *x4 = _mm_aesenc_si128(*x4, key);
+    *x5 = _mm_aesenc_si128(*x5, key);
+    *x6 = _mm_aesenc_si128(*x6, key);
+    *x7 = _mm_aesenc_si128(*x7, key);
+}
+
+
+inline void aes_genkey(const __m128i* memory, __m128i* k0, __m128i* k1, __m128i* k2, __m128i* k3, __m128i* k4, __m128i* k5, __m128i* k6, __m128i* k7, __m128i* k8, __m128i* k9)
+{
+    __m128i xout0 = _mm_load_si128(memory);
+    __m128i xout2 = _mm_load_si128(memory + 1);
+    *k0 = xout0;
+    *k1 = xout2;
+
+     aes_genkey_sub1(&xout0, &xout2);
+    *k2 = xout0;
+    *k3 = xout2;
+
+     aes_genkey_sub2(&xout0, &xout2);
+    *k4 = xout0;
+    *k5 = xout2;
+
+     aes_genkey_sub4(&xout0, &xout2);
+    *k6 = xout0;
+    *k7 = xout2;
+
+     aes_genkey_sub8(&xout0, &xout2);
+    *k8 = xout0;
+    *k9 = xout2;
+}
+
+
+inline void cn_explode_scratchpad(const __m128i* input, __m128i* output)
+{
+    // This is more than we have registers, compiler will assign 2 keys on the stack
+    __m128i xin0, xin1, xin2, xin3, xin4, xin5, xin6, xin7;
+    __m128i k0, k1, k2, k3, k4, k5, k6, k7, k8, k9;
+
+    aes_genkey(input, &k0, &k1, &k2, &k3, &k4, &k5, &k6, &k7, &k8, &k9);
+
+    xin0 = _mm_load_si128(input + 4);
+    xin1 = _mm_load_si128(input + 5);
+    xin2 = _mm_load_si128(input + 6);
+    xin3 = _mm_load_si128(input + 7);
+    xin4 = _mm_load_si128(input + 8);
+    xin5 = _mm_load_si128(input + 9);
+    xin6 = _mm_load_si128(input + 10);
+    xin7 = _mm_load_si128(input + 11);
+
+    for (size_t i = 0; __builtin_expect(i < MEMORY_LITE / sizeof(__m128i), 1); i += 8) {
+        aes_round(k0, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k1, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k2, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k3, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k4, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k5, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k6, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k7, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k8, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k9, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+
+        _mm_store_si128(output + i + 0, xin0);
+        _mm_store_si128(output + i + 1, xin1);
+        _mm_store_si128(output + i + 2, xin2);
+        _mm_store_si128(output + i + 3, xin3);
+        _mm_store_si128(output + i + 4, xin4);
+        _mm_store_si128(output + i + 5, xin5);
+        _mm_store_si128(output + i + 6, xin6);
+        _mm_store_si128(output + i + 7, xin7);
+    }
+}
+
+
+inline void cn_implode_scratchpad(const __m128i* input, __m128i* output)
+{
+    // This is more than we have registers, compiler will assign 2 keys on the stack
+    __m128i xout0, xout1, xout2, xout3, xout4, xout5, xout6, xout7;
+    __m128i k0, k1, k2, k3, k4, k5, k6, k7, k8, k9;
+
+    aes_genkey(output + 2, &k0, &k1, &k2, &k3, &k4, &k5, &k6, &k7, &k8, &k9);
+
+    xout0 = _mm_load_si128(output + 4);
+    xout1 = _mm_load_si128(output + 5);
+    xout2 = _mm_load_si128(output + 6);
+    xout3 = _mm_load_si128(output + 7);
+    xout4 = _mm_load_si128(output + 8);
+    xout5 = _mm_load_si128(output + 9);
+    xout6 = _mm_load_si128(output + 10);
+    xout7 = _mm_load_si128(output + 11);
+
+    for (size_t i = 0; __builtin_expect(i < MEMORY_LITE / sizeof(__m128i), 1); i += 8)
+    {
+        xout0 = _mm_xor_si128(_mm_load_si128(input + i + 0), xout0);
+        xout1 = _mm_xor_si128(_mm_load_si128(input + i + 1), xout1);
+        xout2 = _mm_xor_si128(_mm_load_si128(input + i + 2), xout2);
+        xout3 = _mm_xor_si128(_mm_load_si128(input + i + 3), xout3);
+        xout4 = _mm_xor_si128(_mm_load_si128(input + i + 4), xout4);
+        xout5 = _mm_xor_si128(_mm_load_si128(input + i + 5), xout5);
+        xout6 = _mm_xor_si128(_mm_load_si128(input + i + 6), xout6);
+        xout7 = _mm_xor_si128(_mm_load_si128(input + i + 7), xout7);
+
+        aes_round(k0, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k1, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k2, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k3, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k4, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k5, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k6, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k7, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k8, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k9, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+    }
+
+    _mm_store_si128(output + 4, xout0);
+    _mm_store_si128(output + 5, xout1);
+    _mm_store_si128(output + 6, xout2);
+    _mm_store_si128(output + 7, xout3);
+    _mm_store_si128(output + 8, xout4);
+    _mm_store_si128(output + 9, xout5);
+    _mm_store_si128(output + 10, xout6);
+    _mm_store_si128(output + 11, xout7);
+}
+
+
+#if defined(__x86_64__)
+#   define EXTRACT64(X) _mm_cvtsi128_si64(X)
+
+inline uint64_t _umul128(uint64_t a, uint64_t b, uint64_t* hi)
+{
+    unsigned __int128 r = (unsigned __int128) a * (unsigned __int128) b;
+    *hi = r >> 64;
+    return (uint64_t) r;
+}
+#elif defined(__i386__)
+#   define HI32(X) \
+    _mm_srli_si128((X), 4)
+
+
+#   define EXTRACT64(X) \
+    ((uint64_t)(uint32_t)_mm_cvtsi128_si32(X) | \
+    ((uint64_t)(uint32_t)_mm_cvtsi128_si32(HI32(X)) << 32))
+
+inline uint64_t _umul128(uint64_t multiplier, uint64_t multiplicand, uint64_t *product_hi) {
+    // multiplier   = ab = a * 2^32 + b
+    // multiplicand = cd = c * 2^32 + d
+    // ab * cd = a * c * 2^64 + (a * d + b * c) * 2^32 + b * d
+    uint64_t a = multiplier >> 32;
+    uint64_t b = multiplier & 0xFFFFFFFF;
+    uint64_t c = multiplicand >> 32;
+    uint64_t d = multiplicand & 0xFFFFFFFF;
+
+    //uint64_t ac = a * c;
+    uint64_t ad = a * d;
+    //uint64_t bc = b * c;
+    uint64_t bd = b * d;
+
+    uint64_t adbc = ad + (b * c);
+    uint64_t adbc_carry = adbc < ad ? 1 : 0;
+
+    // multiplier * multiplicand = product_hi * 2^64 + product_lo
+    uint64_t product_lo = bd + (adbc << 32);
+    uint64_t product_lo_carry = product_lo < bd ? 1 : 0;
+    *product_hi = (a * c) + (adbc >> 32) + (adbc_carry << 32) + product_lo_carry;
+
+    return product_lo;
+}
+#endif
+
+
+#endif /* __CRYPTONIGHT_AESNI_H__ */

algo/cryptonight-lite/cryptonight_lite_av1_aesni.c

@@ -0,0 +1,77 @@
+/* XMRig
+ * Copyright 2010      Jeff Garzik <jgarzik@pobox.com>
+ * Copyright 2012-2014 pooler      <pooler@litecoinpool.org>
+ * 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      fireice-uk  <https://github.com/fireice-uk>
+ * Copyright 2016-2017 XMRig       <support@xmrig.com>
+ *
+ *
+ *   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 3 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, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <x86intrin.h>
+#include <string.h>
+
+#include "algo/cryptonight/cryptonight.h"
+#include "cryptonight_lite_aesni.h"
+#include "crypto/c_keccak.h"
+
+
+void cryptonight_lite_av1_aesni(const void *restrict input, size_t size, void *restrict output, struct cryptonight_ctx *restrict ctx)
+{
+    keccak((const uint8_t *) input, size, ctx->state0, 200);
+
+    cn_explode_scratchpad((__m128i*) ctx->state0, (__m128i*) ctx->memory);
+
+    const uint8_t* l0 = ctx->memory;
+    uint64_t* h0 = (uint64_t*) ctx->state0;
+
+    uint64_t al0 = h0[0] ^ h0[4];
+    uint64_t ah0 = h0[1] ^ h0[5];
+    __m128i bx0 = _mm_set_epi64x(h0[3] ^ h0[7], h0[2] ^ h0[6]);
+
+    uint64_t idx0 = h0[0] ^ h0[4];
+
+    for (size_t i = 0; __builtin_expect(i < 0x40000, 1); i++) {
+        __m128i cx;
+        cx = _mm_load_si128((__m128i *) &l0[idx0 & 0xFFFF0]);
+        cx = _mm_aesenc_si128(cx, _mm_set_epi64x(ah0, al0));
+
+        _mm_store_si128((__m128i *) &l0[idx0 & 0xFFFF0], _mm_xor_si128(bx0, cx));
+        idx0 = EXTRACT64(cx);
+        bx0 = cx;
+
+        uint64_t hi, lo, cl, ch;
+        cl = ((uint64_t*) &l0[idx0 & 0xFFFF0])[0];
+        ch = ((uint64_t*) &l0[idx0 & 0xFFFF0])[1];
+        lo = _umul128(idx0, cl, &hi);
+
+        al0 += hi;
+        ah0 += lo;
+
+        ((uint64_t*)&l0[idx0 & 0xFFFF0])[0] = al0;
+        ((uint64_t*)&l0[idx0 & 0xFFFF0])[1] = ah0;
+
+        ah0 ^= ch;
+        al0 ^= cl;
+        idx0 = al0;
+    }
+
+    cn_implode_scratchpad((__m128i*) ctx->memory, (__m128i*) ctx->state0);
+
+    keccakf(h0, 24);
+    extra_hashes[ctx->state0[0] & 3](ctx->state0, 200, output);
+}

algo/cryptonight-lite/cryptonight_lite_av2_aesni_double.c

@@ -0,0 +1,111 @@
+/* XMRig
+ * Copyright 2010      Jeff Garzik <jgarzik@pobox.com>
+ * Copyright 2012-2014 pooler      <pooler@litecoinpool.org>
+ * 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      fireice-uk  <https://github.com/fireice-uk>
+ * Copyright 2016-2017 XMRig       <support@xmrig.com>
+ *
+ *
+ *   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 3 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, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <x86intrin.h>
+#include <string.h>
+
+#include "algo/cryptonight/cryptonight.h"
+#include "cryptonight_lite_aesni.h"
+#include "crypto/c_keccak.h"
+
+
+void cryptonight_lite_av2_aesni_double(const void *restrict input, size_t size, void *restrict output, struct cryptonight_ctx *restrict ctx)
+{
+    keccak((const uint8_t *) input,        size, ctx->state0, 200);
+    keccak((const uint8_t *) input + size, size, ctx->state1, 200);
+
+    const uint8_t* l0 = ctx->memory;
+    const uint8_t* l1 = ctx->memory + MEMORY_LITE;
+    uint64_t* h0 = (uint64_t*) ctx->state0;
+    uint64_t* h1 = (uint64_t*) ctx->state1;
+
+    cn_explode_scratchpad((__m128i*) h0, (__m128i*) l0);
+    cn_explode_scratchpad((__m128i*) h1, (__m128i*) l1);
+
+    uint64_t al0 = h0[0] ^ h0[4];
+    uint64_t al1 = h1[0] ^ h1[4];
+    uint64_t ah0 = h0[1] ^ h0[5];
+    uint64_t ah1 = h1[1] ^ h1[5];
+
+    __m128i bx0 = _mm_set_epi64x(h0[3] ^ h0[7], h0[2] ^ h0[6]);
+    __m128i bx1 = _mm_set_epi64x(h1[3] ^ h1[7], h1[2] ^ h1[6]);
+
+    uint64_t idx0 = h0[0] ^ h0[4];
+    uint64_t idx1 = h1[0] ^ h1[4];
+
+    for (size_t i = 0; __builtin_expect(i < 0x40000, 1); i++) {
+        __m128i cx0 = _mm_load_si128((__m128i *) &l0[idx0 & 0xFFFF0]);
+        __m128i cx1 = _mm_load_si128((__m128i *) &l1[idx1 & 0xFFFF0]);
+
+        cx0 = _mm_aesenc_si128(cx0, _mm_set_epi64x(ah0, al0));
+        cx1 = _mm_aesenc_si128(cx1, _mm_set_epi64x(ah1, al1));
+
+        _mm_store_si128((__m128i *) &l0[idx0 & 0xFFFF0], _mm_xor_si128(bx0, cx0));
+        _mm_store_si128((__m128i *) &l1[idx1 & 0xFFFF0], _mm_xor_si128(bx1, cx1));
+
+        idx0 = EXTRACT64(cx0);
+        idx1 = EXTRACT64(cx1);
+
+        bx0 = cx0;
+        bx1 = cx1;
+
+        uint64_t hi, lo, cl, ch;
+        cl = ((uint64_t*) &l0[idx0 & 0xFFFF0])[0];
+        ch = ((uint64_t*) &l0[idx0 & 0xFFFF0])[1];
+        lo = _umul128(idx0, cl, &hi);
+
+        al0 += hi;
+        ah0 += lo;
+
+        ((uint64_t*) &l0[idx0 & 0xFFFF0])[0] = al0;
+        ((uint64_t*) &l0[idx0 & 0xFFFF0])[1] = ah0;
+
+        ah0 ^= ch;
+        al0 ^= cl;
+        idx0 = al0;
+
+        cl = ((uint64_t*) &l1[idx1 & 0xFFFF0])[0];
+        ch = ((uint64_t*) &l1[idx1 & 0xFFFF0])[1];
+        lo = _umul128(idx1, cl, &hi);
+
+        al1 += hi;
+        ah1 += lo;
+
+        ((uint64_t*) &l1[idx1 & 0xFFFF0])[0] = al1;
+        ((uint64_t*) &l1[idx1 & 0xFFFF0])[1] = ah1;
+
+        ah1 ^= ch;
+        al1 ^= cl;
+        idx1 = al1;
+    }
+
+    cn_implode_scratchpad((__m128i*) l0, (__m128i*) h0);
+    cn_implode_scratchpad((__m128i*) l1, (__m128i*) h1);
+
+    keccakf(h0, 24);
+    keccakf(h1, 24);
+
+    extra_hashes[ctx->state0[0] & 3](ctx->state0, 200, output);
+    extra_hashes[ctx->state1[0] & 3](ctx->state1, 200, (char*) output + 32);
+}

algo/cryptonight-lite/cryptonight_lite_av3_softaes.c

@@ -0,0 +1,77 @@
+/* XMRig
+ * Copyright 2010      Jeff Garzik <jgarzik@pobox.com>
+ * Copyright 2012-2014 pooler      <pooler@litecoinpool.org>
+ * 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      fireice-uk  <https://github.com/fireice-uk>
+ * Copyright 2016-2017 XMRig       <support@xmrig.com>
+ *
+ *
+ *   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 3 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, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <x86intrin.h>
+#include <string.h>
+
+#include "algo/cryptonight/cryptonight.h"
+#include "cryptonight_lite_softaes.h"
+#include "crypto/c_keccak.h"
+
+
+void cryptonight_lite_av3_softaes(const void *restrict input, size_t size, void *restrict output, struct cryptonight_ctx *restrict ctx)
+{
+    keccak((const uint8_t *) input, size, ctx->state0, 200);
+
+    cn_explode_scratchpad((__m128i*) ctx->state0, (__m128i*) ctx->memory);
+
+    const uint8_t* l0 = ctx->memory;
+    uint64_t* h0 = (uint64_t*) ctx->state0;
+
+    uint64_t al0 = h0[0] ^ h0[4];
+    uint64_t ah0 = h0[1] ^ h0[5];
+    __m128i bx0 = _mm_set_epi64x(h0[3] ^ h0[7], h0[2] ^ h0[6]);
+
+    uint64_t idx0 = h0[0] ^ h0[4];
+
+    for (size_t i = 0; __builtin_expect(i < 0x40000, 1); i++) {
+        __m128i cx;
+        cx = _mm_load_si128((__m128i *)&l0[idx0 & 0xFFFF0]);
+        cx = soft_aesenc(cx, _mm_set_epi64x(ah0, al0));
+
+        _mm_store_si128((__m128i *)&l0[idx0 & 0xFFFF0], _mm_xor_si128(bx0, cx));
+        idx0 = EXTRACT64(cx);
+        bx0 = cx;
+
+        uint64_t hi, lo, cl, ch;
+        cl = ((uint64_t*)&l0[idx0 & 0xFFFF0])[0];
+        ch = ((uint64_t*)&l0[idx0 & 0xFFFF0])[1];
+        lo = _umul128(idx0, cl, &hi);
+
+        al0 += hi;
+        ah0 += lo;
+
+        ((uint64_t*)&l0[idx0 & 0xFFFF0])[0] = al0;
+        ((uint64_t*)&l0[idx0 & 0xFFFF0])[1] = ah0;
+
+        ah0 ^= ch;
+        al0 ^= cl;
+        idx0 = al0;
+    }
+
+    cn_implode_scratchpad((__m128i*) ctx->memory, (__m128i*) ctx->state0);
+
+    keccakf(h0, 24);
+    extra_hashes[ctx->state0[0] & 3](ctx->state0, 200, output);
+}

algo/cryptonight-lite/cryptonight_lite_av4_softaes_double.c

@@ -0,0 +1,111 @@
+/* XMRig
+ * Copyright 2010      Jeff Garzik <jgarzik@pobox.com>
+ * Copyright 2012-2014 pooler      <pooler@litecoinpool.org>
+ * 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      fireice-uk  <https://github.com/fireice-uk>
+ * Copyright 2016-2017 XMRig       <support@xmrig.com>
+ *
+ *
+ *   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 3 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, see <http://www.gnu.org/licenses/>.
+ */
+
+#include <x86intrin.h>
+#include <string.h>
+
+#include "algo/cryptonight/cryptonight.h"
+#include "cryptonight_lite_softaes.h"
+#include "crypto/c_keccak.h"
+
+
+void cryptonight_lite_av4_softaes_double(const void *restrict input, size_t size, void *restrict output, struct cryptonight_ctx *restrict ctx)
+{
+    keccak((const uint8_t *) input,        size, ctx->state0, 200);
+    keccak((const uint8_t *) input + size, size, ctx->state1, 200);
+
+    const uint8_t* l0 = ctx->memory;
+    const uint8_t* l1 = ctx->memory + MEMORY_LITE;
+    uint64_t* h0 = (uint64_t*) ctx->state0;
+    uint64_t* h1 = (uint64_t*) ctx->state1;
+
+    cn_explode_scratchpad((__m128i*) h0, (__m128i*) l0);
+    cn_explode_scratchpad((__m128i*) h1, (__m128i*) l1);
+
+    uint64_t al0 = h0[0] ^ h0[4];
+    uint64_t al1 = h1[0] ^ h1[4];
+    uint64_t ah0 = h0[1] ^ h0[5];
+    uint64_t ah1 = h1[1] ^ h1[5];
+
+    __m128i bx0 = _mm_set_epi64x(h0[3] ^ h0[7], h0[2] ^ h0[6]);
+    __m128i bx1 = _mm_set_epi64x(h1[3] ^ h1[7], h1[2] ^ h1[6]);
+
+    uint64_t idx0 = h0[0] ^ h0[4];
+    uint64_t idx1 = h1[0] ^ h1[4];
+
+    for (size_t i = 0; __builtin_expect(i < 0x40000, 1); i++) {
+        __m128i cx0 = _mm_load_si128((__m128i *) &l0[idx0 & 0xFFFF0]);
+        __m128i cx1 = _mm_load_si128((__m128i *) &l1[idx1 & 0xFFFF0]);
+
+        cx0 = soft_aesenc(cx0, _mm_set_epi64x(ah0, al0));
+        cx1 = soft_aesenc(cx1, _mm_set_epi64x(ah1, al1));
+
+        _mm_store_si128((__m128i *) &l0[idx0 & 0xFFFF0], _mm_xor_si128(bx0, cx0));
+        _mm_store_si128((__m128i *) &l1[idx1 & 0xFFFF0], _mm_xor_si128(bx1, cx1));
+
+        idx0 = EXTRACT64(cx0);
+        idx1 = EXTRACT64(cx1);
+
+        bx0 = cx0;
+        bx1 = cx1;
+
+        uint64_t hi, lo, cl, ch;
+        cl = ((uint64_t*) &l0[idx0 & 0xFFFF0])[0];
+        ch = ((uint64_t*) &l0[idx0 & 0xFFFF0])[1];
+        lo = _umul128(idx0, cl, &hi);
+
+        al0 += hi;
+        ah0 += lo;
+
+        ((uint64_t*) &l0[idx0 & 0xFFFF0])[0] = al0;
+        ((uint64_t*) &l0[idx0 & 0xFFFF0])[1] = ah0;
+
+        ah0 ^= ch;
+        al0 ^= cl;
+        idx0 = al0;
+
+        cl = ((uint64_t*) &l1[idx1 & 0xFFFF0])[0];
+        ch = ((uint64_t*) &l1[idx1 & 0xFFFF0])[1];
+        lo = _umul128(idx1, cl, &hi);
+
+        al1 += hi;
+        ah1 += lo;
+
+        ((uint64_t*) &l1[idx1 & 0xFFFF0])[0] = al1;
+        ((uint64_t*) &l1[idx1 & 0xFFFF0])[1] = ah1;
+
+        ah1 ^= ch;
+        al1 ^= cl;
+        idx1 = al1;
+    }
+
+    cn_implode_scratchpad((__m128i*) l0, (__m128i*) h0);
+    cn_implode_scratchpad((__m128i*) l1, (__m128i*) h1);
+
+    keccakf(h0, 24);
+    keccakf(h1, 24);
+
+    extra_hashes[ctx->state0[0] & 3](ctx->state0, 200, output);
+    extra_hashes[ctx->state1[0] & 3](ctx->state1, 200, (char*) output + 32);
+}

algo/cryptonight-lite/cryptonight_lite_softaes.h

@@ -0,0 +1,237 @@
+/* XMRig
+ * Copyright 2010      Jeff Garzik <jgarzik@pobox.com>
+ * Copyright 2012-2014 pooler      <pooler@litecoinpool.org>
+ * 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      fireice-uk  <https://github.com/fireice-uk>
+ * Copyright 2016-2017 XMRig       <support@xmrig.com>
+ *
+ *
+ *   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 3 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, see <http://www.gnu.org/licenses/>.
+ */
+
+#ifndef __CRYPTONIGHT_SOFTAES_H__
+#define __CRYPTONIGHT_SOFTAES_H__
+
+#include <x86intrin.h>
+
+extern __m128i soft_aesenc(__m128i in, __m128i key);
+extern __m128i soft_aeskeygenassist(__m128i key, uint8_t rcon);
+
+
+// This will shift and xor tmp1 into itself as 4 32-bit vals such as
+// sl_xor(a1 a2 a3 a4) = a1 (a2^a1) (a3^a2^a1) (a4^a3^a2^a1)
+inline __m128i sl_xor(__m128i tmp1)
+{
+    __m128i tmp4;
+    tmp4 = _mm_slli_si128(tmp1, 0x04);
+    tmp1 = _mm_xor_si128(tmp1, tmp4);
+    tmp4 = _mm_slli_si128(tmp4, 0x04);
+    tmp1 = _mm_xor_si128(tmp1, tmp4);
+    tmp4 = _mm_slli_si128(tmp4, 0x04);
+    tmp1 = _mm_xor_si128(tmp1, tmp4);
+    return tmp1;
+}
+
+
+inline void aes_genkey_sub(__m128i* xout0, __m128i* xout2, uint8_t rcon)
+{
+    __m128i xout1 = soft_aeskeygenassist(*xout2, rcon);
+    xout1  = _mm_shuffle_epi32(xout1, 0xFF); // see PSHUFD, set all elems to 4th elem
+    *xout0 = sl_xor(*xout0);
+    *xout0 = _mm_xor_si128(*xout0, xout1);
+    xout1  = soft_aeskeygenassist(*xout0, 0x00);
+    xout1  = _mm_shuffle_epi32(xout1, 0xAA); // see PSHUFD, set all elems to 3rd elem
+    *xout2 = sl_xor(*xout2);
+    *xout2 = _mm_xor_si128(*xout2, xout1);
+}
+
+
+inline void aes_round(__m128i key, __m128i* x0, __m128i* x1, __m128i* x2, __m128i* x3, __m128i* x4, __m128i* x5, __m128i* x6, __m128i* x7)
+{
+    *x0 = soft_aesenc(*x0, key);
+    *x1 = soft_aesenc(*x1, key);
+    *x2 = soft_aesenc(*x2, key);
+    *x3 = soft_aesenc(*x3, key);
+    *x4 = soft_aesenc(*x4, key);
+    *x5 = soft_aesenc(*x5, key);
+    *x6 = soft_aesenc(*x6, key);
+    *x7 = soft_aesenc(*x7, key);
+}
+
+
+inline void aes_genkey(const __m128i* memory, __m128i* k0, __m128i* k1, __m128i* k2, __m128i* k3, __m128i* k4, __m128i* k5, __m128i* k6, __m128i* k7, __m128i* k8, __m128i* k9)
+{
+    __m128i xout0 = _mm_load_si128(memory);
+    __m128i xout2 = _mm_load_si128(memory + 1);
+    *k0 = xout0;
+    *k1 = xout2;
+
+    aes_genkey_sub(&xout0, &xout2, 0x1);
+    *k2 = xout0;
+    *k3 = xout2;
+
+    aes_genkey_sub(&xout0, &xout2, 0x2);
+    *k4 = xout0;
+    *k5 = xout2;
+
+    aes_genkey_sub(&xout0, &xout2, 0x4);
+    *k6 = xout0;
+    *k7 = xout2;
+
+    aes_genkey_sub(&xout0, &xout2, 0x8);
+    *k8 = xout0;
+    *k9 = xout2;
+}
+
+
+inline void cn_explode_scratchpad(const __m128i* input, __m128i* output)
+{
+    // This is more than we have registers, compiler will assign 2 keys on the stack
+    __m128i xin0, xin1, xin2, xin3, xin4, xin5, xin6, xin7;
+    __m128i k0, k1, k2, k3, k4, k5, k6, k7, k8, k9;
+
+    aes_genkey(input, &k0, &k1, &k2, &k3, &k4, &k5, &k6, &k7, &k8, &k9);
+
+    xin0 = _mm_load_si128(input + 4);
+    xin1 = _mm_load_si128(input + 5);
+    xin2 = _mm_load_si128(input + 6);
+    xin3 = _mm_load_si128(input + 7);
+    xin4 = _mm_load_si128(input + 8);
+    xin5 = _mm_load_si128(input + 9);
+    xin6 = _mm_load_si128(input + 10);
+    xin7 = _mm_load_si128(input + 11);
+
+    for (size_t i = 0; i < MEMORY_LITE / sizeof(__m128i); i += 8) {
+        aes_round(k0, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k1, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k2, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k3, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k4, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k5, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k6, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k7, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k8, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+        aes_round(k9, &xin0, &xin1, &xin2, &xin3, &xin4, &xin5, &xin6, &xin7);
+
+        _mm_store_si128(output + i + 0, xin0);
+        _mm_store_si128(output + i + 1, xin1);
+        _mm_store_si128(output + i + 2, xin2);
+        _mm_store_si128(output + i + 3, xin3);
+        _mm_store_si128(output + i + 4, xin4);
+        _mm_store_si128(output + i + 5, xin5);
+        _mm_store_si128(output + i + 6, xin6);
+        _mm_store_si128(output + i + 7, xin7);
+    }
+}
+
+
+inline void cn_implode_scratchpad(const __m128i* input, __m128i* output)
+{
+    // This is more than we have registers, compiler will assign 2 keys on the stack
+    __m128i xout0, xout1, xout2, xout3, xout4, xout5, xout6, xout7;
+    __m128i k0, k1, k2, k3, k4, k5, k6, k7, k8, k9;
+
+    aes_genkey(output + 2, &k0, &k1, &k2, &k3, &k4, &k5, &k6, &k7, &k8, &k9);
+
+    xout0 = _mm_load_si128(output + 4);
+    xout1 = _mm_load_si128(output + 5);
+    xout2 = _mm_load_si128(output + 6);
+    xout3 = _mm_load_si128(output + 7);
+    xout4 = _mm_load_si128(output + 8);
+    xout5 = _mm_load_si128(output + 9);
+    xout6 = _mm_load_si128(output + 10);
+    xout7 = _mm_load_si128(output + 11);
+
+    for (size_t i = 0; __builtin_expect(i < MEMORY_LITE / sizeof(__m128i), 1); i += 8)
+    {
+        xout0 = _mm_xor_si128(_mm_load_si128(input + i + 0), xout0);
+        xout1 = _mm_xor_si128(_mm_load_si128(input + i + 1), xout1);
+        xout2 = _mm_xor_si128(_mm_load_si128(input + i + 2), xout2);
+        xout3 = _mm_xor_si128(_mm_load_si128(input + i + 3), xout3);
+        xout4 = _mm_xor_si128(_mm_load_si128(input + i + 4), xout4);
+        xout5 = _mm_xor_si128(_mm_load_si128(input + i + 5), xout5);
+        xout6 = _mm_xor_si128(_mm_load_si128(input + i + 6), xout6);
+        xout7 = _mm_xor_si128(_mm_load_si128(input + i + 7), xout7);
+
+        aes_round(k0, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k1, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k2, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k3, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k4, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k5, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k6, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k7, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k8, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+        aes_round(k9, &xout0, &xout1, &xout2, &xout3, &xout4, &xout5, &xout6, &xout7);
+    }
+
+    _mm_store_si128(output + 4, xout0);
+    _mm_store_si128(output + 5, xout1);
+    _mm_store_si128(output + 6, xout2);
+    _mm_store_si128(output + 7, xout3);
+    _mm_store_si128(output + 8, xout4);
+    _mm_store_si128(output + 9, xout5);
+    _mm_store_si128(output + 10, xout6);
+    _mm_store_si128(output + 11, xout7);
+}
+
+
+#if defined(__x86_64__)
+#   define EXTRACT64(X) _mm_cvtsi128_si64(X)
+
+inline uint64_t _umul128(uint64_t a, uint64_t b, uint64_t* hi)
+{
+    unsigned __int128 r = (unsigned __int128) a * (unsigned __int128) b;
+    *hi = r >> 64;
+    return (uint64_t) r;
+}
+#elif defined(__i386__)
+#   define HI32(X) \
+    _mm_srli_si128((X), 4)
+
+
+#   define EXTRACT64(X) \
+    ((uint64_t)(uint32_t)_mm_cvtsi128_si32(X) | \
+    ((uint64_t)(uint32_t)_mm_cvtsi128_si32(HI32(X)) << 32))
+
+inline uint64_t _umul128(uint64_t multiplier, uint64_t multiplicand, uint64_t *product_hi) {
+    // multiplier   = ab = a * 2^32 + b
+    // multiplicand = cd = c * 2^32 + d
+    // ab * cd = a * c * 2^64 + (a * d + b * c) * 2^32 + b * d
+    uint64_t a = multiplier >> 32;
+    uint64_t b = multiplier & 0xFFFFFFFF;
+    uint64_t c = multiplicand >> 32;
+    uint64_t d = multiplicand & 0xFFFFFFFF;
+
+    //uint64_t ac = a * c;
+    uint64_t ad = a * d;
+    //uint64_t bc = b * c;
+    uint64_t bd = b * d;
+
+    uint64_t adbc = ad + (b * c);
+    uint64_t adbc_carry = adbc < ad ? 1 : 0;
+
+    // multiplier * multiplicand = product_hi * 2^64 + product_lo
+    uint64_t product_lo = bd + (adbc << 32);
+    uint64_t product_lo_carry = product_lo < bd ? 1 : 0;
+    *product_hi = (a * c) + (adbc >> 32) + (adbc_carry << 32) + product_lo_carry;
+
+    return product_lo;
+}
+#endif
+
+
+#endif /* __CRYPTONIGHT_SOFTAES_H__ */

algo/cryptonight/cryptonight.c

@@ -52,7 +52,7 @@ const static char test_input[152] = {
 };
 
 
-const static char test_output[64] = {
+const static char test_output0[64] = {
     0x1B, 0x60, 0x6A, 0x3F, 0x4A, 0x07, 0xD6, 0x48, 0x9A, 0x1B, 0xCD, 0x07, 0x69, 0x7B, 0xD1, 0x66,
     0x96, 0xB6, 0x1C, 0x8A, 0xE9, 0x82, 0xF6, 0x1A, 0x90, 0x16, 0x0F, 0x4E, 0x52, 0x82, 0x8A, 0x7F,
     0x1A, 0x3F, 0xFB, 0xEE, 0x90, 0x9B, 0x42, 0x0D, 0x91, 0xF7, 0xBE, 0x6E, 0x5F, 0xB5, 0x6D, 0xB7,
@@ -65,10 +65,28 @@ void cryptonight_av2_aesni_double(const void* input, size_t size, void* output,
 void cryptonight_av3_softaes(const void* input, size_t size, void* output, struct cryptonight_ctx* ctx);
 void cryptonight_av4_softaes_double(const void* input, size_t size, void* output, struct cryptonight_ctx* ctx);
 
+#ifndef XMRIG_NO_AEON
+const static char test_output1[64] = {
+    0x28, 0xA2, 0x2B, 0xAD, 0x3F, 0x93, 0xD1, 0x40, 0x8F, 0xCA, 0x47, 0x2E, 0xB5, 0xAD, 0x1C, 0xBE,
+    0x75, 0xF2, 0x1D, 0x05, 0x3C, 0x8C, 0xE5, 0xB3, 0xAF, 0x10, 0x5A, 0x57, 0x71, 0x3E, 0x21, 0xDD,
+    0x36, 0x95, 0xB4, 0xB5, 0x3B, 0xB0, 0x03, 0x58, 0xB0, 0xAD, 0x38, 0xDC, 0x16, 0x0F, 0xEB, 0x9E,
+    0x00, 0x4E, 0xEC, 0xE0, 0x9B, 0x83, 0xA7, 0x2E, 0xF6, 0xBA, 0x98, 0x64, 0xD3, 0x51, 0x0C, 0x88,
+};
+
+void cryptonight_lite_av1_aesni(const void* input, size_t size, void* output, struct cryptonight_ctx* ctx);
+void cryptonight_lite_av2_aesni_double(const void* input, size_t size, void* output, struct cryptonight_ctx* ctx);
+void cryptonight_lite_av3_softaes(const void* input, size_t size, void* output, struct cryptonight_ctx* ctx);
+void cryptonight_lite_av4_softaes_double(const void* input, size_t size, void* output, struct cryptonight_ctx* ctx);
+#endif
+
 void (*cryptonight_hash_ctx)(const void* input, size_t size, void* output, struct cryptonight_ctx* ctx) = NULL;
 
 
 static bool self_test() {
+    if (cryptonight_hash_ctx == NULL) {
+        return false;
+    }
+
     char output[64];
 
     struct cryptonight_ctx *ctx = (struct cryptonight_ctx*) _mm_malloc(sizeof(struct cryptonight_ctx), 16);
@@ -79,12 +97,54 @@ static bool self_test() {
     _mm_free(ctx->memory);
     _mm_free(ctx);
 
-    return memcmp(output, test_output, (opt_double_hash ? 64 : 32)) == 0;
+#   ifndef XMRIG_NO_AEON
+    if (opt_algo == ALGO_CRYPTONIGHT_LITE) {
+        return memcmp(output, test_output1, (opt_double_hash ? 64 : 32)) == 0;
+    }
+#   endif
+
+    return memcmp(output, test_output0, (opt_double_hash ? 64 : 32)) == 0;
 }
 
 
+#ifndef XMRIG_NO_AEON
+bool cryptonight_lite_init(int variant) {
+    switch (variant) {
+        case AEON_AV1_AESNI:
+            cryptonight_hash_ctx = cryptonight_lite_av1_aesni;
+            break;
+
+        case AEON_AV2_AESNI_DOUBLE:
+            opt_double_hash = true;
+            cryptonight_hash_ctx = cryptonight_lite_av2_aesni_double;
+            break;
+
+        case AEON_AV3_SOFT_AES:
+            cryptonight_hash_ctx = cryptonight_lite_av3_softaes;
+            break;
+
+        case AEON_AV4_SOFT_AES_DOUBLE:
+            opt_double_hash = true;
+            cryptonight_hash_ctx = cryptonight_lite_av4_softaes_double;
+            break;
+
+        default:
+            break;
+    }
+
+    return self_test();
+}
+#endif
+
+
 bool cryptonight_init(int variant)
 {
+#   ifndef XMRIG_NO_AEON
+    if (opt_algo == ALGO_CRYPTONIGHT_LITE) {
+        return cryptonight_lite_init(variant);
+    }
+#   endif
+
     switch (variant) {
         case XMR_AV1_AESNI:
             cryptonight_hash_ctx = cryptonight_av1_aesni;

algo/cryptonight/cryptonight.h

@@ -28,7 +28,8 @@
 #include <stdint.h>
 #include <stdbool.h>
 
-#define MEMORY 2097152 /* 2 MiB */
+#define MEMORY      2097152 /* 2 MiB */
+#define MEMORY_LITE 1048576 /* 1 MiB */
 
 struct cryptonight_ctx {
     uint8_t state0[200] __attribute__((aligned(16)));

memory.c

@@ -30,6 +30,26 @@
 static size_t offset = 0;
 
 
+#ifndef XMRIG_NO_AEON
+static void * create_persistent_ctx_lite(int thr_id) {
+    struct cryptonight_ctx *ctx = NULL;
+
+    if (!opt_double_hash) {
+        const size_t offset = MEMORY * (thr_id + 1);
+
+        ctx = (struct cryptonight_ctx *) &persistent_memory[offset + MEMORY_LITE];
+        ctx->memory = &persistent_memory[offset];
+        return ctx;
+    }
+
+    ctx = (struct cryptonight_ctx *) &persistent_memory[MEMORY - sizeof(struct cryptonight_ctx) * (thr_id + 1)];
+    ctx->memory = &persistent_memory[MEMORY * (thr_id + 1)];
+
+    return ctx;
+}
+#endif
+
+
 void * persistent_calloc(size_t num, size_t size) {
     void *mem = &persistent_memory[offset];
     offset += (num * size);
@@ -41,6 +61,12 @@ void * persistent_calloc(size_t num, size_t size) {
 
 
 void * create_persistent_ctx(int thr_id) {
+#   ifndef XMRIG_NO_AEON
+    if (opt_algo == ALGO_CRYPTONIGHT_LITE) {
+        return create_persistent_ctx_lite(thr_id);
+    }
+#   endif
+
     struct cryptonight_ctx *ctx = (struct cryptonight_ctx *) &persistent_memory[MEMORY - sizeof(struct cryptonight_ctx) * (thr_id + 1)];
 
     const int ratio = opt_double_hash ? 2 : 1;

options.c

@@ -54,6 +54,8 @@ char    *opt_userpass     = NULL;
 char    *opt_user         = NULL;
 char    *opt_pass         = NULL;
 
+enum mining_algo opt_algo = ALGO_CRYPTONIGHT;
+
 
 static char const usage[] = "\
 Usage: " APP_ID " [OPTIONS]\n\
@@ -108,16 +110,45 @@ static struct option const options[] = {
 };
 
 
-static int get_algo_variant(int variant) {
+static const char *algo_names[] = {
-   if (variant > XMR_AV0_AUTO && variant < XMR_AV_MAX) {
+    [ALGO_CRYPTONIGHT]      = "cryptonight",
-       return variant;
+#   ifndef XMRIG_NO_AEON
-   }
+    [ALGO_CRYPTONIGHT_LITE] = "cryptonight-lite"
+#   endif
+};
 
-   if (cpu_info.flags & CPU_FLAG_AES) {
-       return XMR_AV1_AESNI;
-   }
 
-   return XMR_AV3_SOFT_AES;
+#ifndef XMRIG_NO_AEON
+static int get_cryptonight_lite_variant(int variant) {
+    if (variant > AEON_AV0_AUTO && variant < AEON_AV_MAX) {
+        return variant;
+    }
+
+    if (cpu_info.flags & CPU_FLAG_AES) {
+        return AEON_AV2_AESNI_DOUBLE;
+    }
+
+    return AEON_AV4_SOFT_AES_DOUBLE;
+}
+#endif
+
+
+static int get_algo_variant(int algo, int variant) {
+#   ifndef XMRIG_NO_AEON
+    if (algo == ALGO_CRYPTONIGHT_LITE) {
+        return get_cryptonight_lite_variant(variant);
+    }
+#   endif
+
+    if (variant > XMR_AV0_AUTO && variant < XMR_AV_MAX) {
+        return variant;
+    }
+
+    if (cpu_info.flags & CPU_FLAG_AES) {
+        return XMR_AV1_AESNI;
+    }
+
+    return XMR_AV3_SOFT_AES;
 }
 
 
@@ -133,6 +164,22 @@ static void parse_arg(int key, char *arg) {
     switch (key)
     {
     case 'a':
+        for (int i = 0; i < ARRAY_SIZE(algo_names); i++) {
+            if (algo_names[i] && !strcmp(arg, algo_names[i])) {
+                opt_algo = i;
+                break;
+            }
+
+#           ifndef XMRIG_NO_AEON
+            if (i == ARRAY_SIZE(algo_names) && !strcmp(arg, "cryptonight-light")) {
+                opt_algo = i = ALGO_CRYPTONIGHT_LITE;
+            }
+#           endif
+
+            if (i == ARRAY_SIZE(algo_names)) {
+                show_usage_and_exit(1);
+            }
+        }
         break;
 
     case 'O': /* --userpass */
@@ -253,7 +300,7 @@ static void parse_arg(int key, char *arg) {
 
     case 'v': /* --av */
         v = atoi(arg);
-        if (v < 0 || v > XMR_AV_MAX) {
+        if (v < 0 || v > 1000) {
             show_usage_and_exit(1);
         }
 
@@ -404,10 +451,7 @@ void parse_cmdline(int argc, char *argv[]) {
         opt_n_threads = get_optimal_threads_count();
     }
 
-    opt_algo_variant = get_algo_variant(opt_algo_variant);
+    opt_algo_variant = get_algo_variant(opt_algo, opt_algo_variant);
-    if (!opt_algo_variant) {
-        opt_algo_variant = get_algo_variant(0);
-    }
 
     if (!cryptonight_init(opt_algo_variant)) {
         applog(LOG_ERR, "Cryptonight hash self-test failed. This might be caused by bad compiler optimizations.");
@@ -454,3 +498,8 @@ void show_version_and_exit(void) {
     #endif
     proper_exit(0);
 }
+
+
+const char* get_current_algo_name(void) {
+    return algo_names[opt_algo];
+}

options.h

@@ -32,6 +32,12 @@
 #endif
 
 
+enum mining_algo {
+    ALGO_CRYPTONIGHT,      /* CryptoNight (Monero) */
+    ALGO_CRYPTONIGHT_LITE, /* CryptoNight-Lite (AEON) */
+};
+
+
 enum xmr_algo_variant {
     XMR_AV0_AUTO,
     XMR_AV1_AESNI,
@@ -42,6 +48,18 @@ enum xmr_algo_variant {
 };
 
 
+#ifndef XMRIG_NO_AEON
+enum aeon_algo_variant {
+    AEON_AV0_AUTO,
+    AEON_AV1_AESNI,
+    AEON_AV2_AESNI_DOUBLE,
+    AEON_AV3_SOFT_AES,
+    AEON_AV4_SOFT_AES_DOUBLE,
+    AEON_AV_MAX
+};
+#endif
+
+
 extern bool opt_colors;
 extern bool opt_keepalive;
 extern bool opt_background;
@@ -59,10 +77,12 @@ extern int opt_retries;
 extern int opt_donate_level;
 extern int opt_max_cpu_usage;
 extern int64_t opt_affinity;
+extern enum mining_algo opt_algo;
 
 void parse_cmdline(int argc, char *argv[]);
 void show_usage_and_exit(int status);
 void show_version_and_exit(void);
+const char* get_current_algo_name(void);
 
 extern void proper_exit(int reason);
 

unix/memory_unix.c

@@ -38,7 +38,7 @@ int persistent_memory_flags = 0;
 
 
 const char * persistent_memory_allocate() {
-    const int ratio = opt_double_hash ? 2 : 1;
+    const int ratio = (opt_double_hash && opt_algo != ALGO_CRYPTONIGHT_LITE) ? 2 : 1;
     const int size = MEMORY * (opt_n_threads * ratio + 1);
     persistent_memory_flags |= MEMORY_HUGEPAGES_AVAILABLE;
 

utils/summary.c

@@ -72,10 +72,10 @@ static void print_cpu() {
 
 static void print_threads() {
     if (opt_colors) {
-        applog_notime(LOG_INFO, CL_LGR " * " CL_WHT "THREADS:      " CL_WHT "%d" CL_WHT ", av=%d, donate=%d%%", opt_n_threads, opt_algo_variant, opt_donate_level);
+        applog_notime(LOG_INFO, CL_LGR " * " CL_WHT "THREADS:      " CL_WHT "%d" CL_WHT ", av=%d, %s, donate=%d%%", opt_n_threads, opt_algo_variant, get_current_algo_name(), opt_donate_level);
     }
     else {
-        applog_notime(LOG_INFO, " * THREADS:      %d, av=%d, donate=%d%%", opt_n_threads, opt_algo_variant, opt_donate_level);
+        applog_notime(LOG_INFO, " * THREADS:      %d, av=%d, %s, donate=%d%%", opt_n_threads, opt_algo_variant, get_current_algo_name(), opt_donate_level);
     }
 }
 

win/memory_win.c

@@ -144,7 +144,7 @@ static BOOL TrySetLockPagesPrivilege() {
 
 
 const char * persistent_memory_allocate() {
-    const int ratio = opt_double_hash ? 2 : 1;
+    const int ratio = (opt_double_hash && opt_algo != ALGO_CRYPTONIGHT_LITE) ? 2 : 1;
     const int size  = MEMORY * (opt_n_threads * ratio + 1);
 
     if (TrySetLockPagesPrivilege()) {