Merge branch 'dev' of https://github.com/aegroto/xmrig into dev

CHANGELOG.md

@@ -1,9 +1,11 @@
 # v2.6.0-beta2
+- Improved performance for `cryptonight v7` especially in double hash mode.
 - [#499](https://github.com/xmrig/xmrig/issues/499) IPv6 disabled for internal HTTP API by default, was cause issues on some systems.
 - Added short aliases for algorithm names: `cn`, `cn-lite` and `cn-heavy`.
 - Fixed regressions (v2.6.0-beta1 affected)
   - [#494](https://github.com/xmrig/xmrig/issues/494) Command line option `--donate-level` was broken.
-  - [#502](https://github.com/xmrig/xmrig/issues/502) Build without libmicrohttpd was broken.
+  - [#502](https://github.com/xmrig/xmrig/issues/502) Build without libmicrohttpd was broken.
+  - Fixed nonce calculation for `--av 4` (software AES, double hash) was cause reduction of effective hashrate and rejected shares on nicehash.
 
 # v2.6.0-beta1
  - [#476](https://github.com/xmrig/xmrig/issues/476) **Added Cryptonight-Heavy support for Sumokoin ASIC resistance fork.**

README.md

@@ -124,10 +124,10 @@ Please note performance is highly dependent on system load. The numbers above ar
 ## Release checksums
 ### SHA-256
 ```
-bd14bc3cfd9528e4a7583ab39aecc876250333e1e0faab83781584bb7f65e3eb xmrig-2.6.0-beta1-xenial-amd64.tar.gz/xmrig-2.6.0-beta1/xmrig
+232af0c5f3b1cdbc2d90b514873a764b434d5621d2790da67954b35c17e44fe3 xmrig-2.6.0-beta2-xenial-amd64.tar.gz/xmrig-2.6.0-beta2/xmrig
-32eebf71e5631029202ae5cbf6f03caad912f1722aa86a1be01a26d491801aba xmrig-2.6.0-beta1-gcc-win32.zip/xmrig.exe
+2366a06729d4de538ef511862bf11d0c7ad40fd245e7aeab3c1957307d63471a xmrig-2.6.0-beta2-gcc-win32.zip/xmrig.exe
-1cc08844ff019408e2e2c9560fee0c4e0b2dbc2a72bcc1c1da4a847a1787eca3 xmrig-2.6.0-beta1-gcc-win64.zip/xmrig.exe
+2f6538c765e001d13ca380cbc1558d51efcb97d4bccdfa40993cb872be4e9efd xmrig-2.6.0-beta2-gcc-win64.zip/xmrig.exe
-fdf99aa8e7792a34b1be0cc6c77e4e83ff9a4b21abb27989f7927dcfed82f6e2 xmrig-2.6.0-beta1-msvc-win64.zip/xmrig.exe
+3c0479acb78a3cee8fe416ee438dbff09c786acf50fbaf28a820127fcd0c6e62 xmrig-2.6.0-beta2-msvc-win64.zip/xmrig.exe
 ```
 
 ## Contacts

src/Summary.cpp

@@ -123,7 +123,7 @@ static void print_pools(xmrig::Config *config)
 
 #   ifdef APP_DEBUG
     for (size_t i = 0; i < pools.size(); ++i) {
-        Log::i()->text("%s:%d, user: %s, pass: %s, ka: %d, nicehash: %d", pools[i]->host(), pools[i]->port(), pools[i]->user(), pools[i]->password(), pools[i]->isKeepAlive(), pools[i]->isNicehash());
+        Log::i()->text("%s:%d, user: %s, pass: %s, ka: %d, nicehash: %d", pools[i]->host(), pools[i]->port(), pools[i]->user(), pools[i]->password(), pools[i]->keepAlive(), pools[i]->isNicehash());
     }
 #   endif
 }

src/crypto/CryptoNight_x86.h

@@ -389,6 +389,22 @@ static inline void cn_implode_scratchpad(const __m128i *input, __m128i *output)
 }
 
 
+static inline void cryptonight_monero_tweak(uint64_t* mem_out, __m128i tmp)
+{
+    mem_out[0] = EXTRACT64(tmp);
+
+    tmp = _mm_castps_si128(_mm_movehl_ps(_mm_castsi128_ps(tmp), _mm_castsi128_ps(tmp)));
+    uint64_t vh = EXTRACT64(tmp);
+
+    uint8_t x = vh >> 24;
+    static const uint16_t table = 0x7531;
+    const uint8_t index = (((x >> 3) & 6) | (x & 1)) << 1;
+    vh ^= ((table >> index) & 0x3) << 28;
+
+    mem_out[1] = vh;
+}
+
+
 template<xmrig::Algo ALGO, bool SOFT_AES, int VARIANT>
 inline void cryptonight_single_hash(const uint8_t *__restrict__ input, size_t size, uint8_t *__restrict__ output, cryptonight_ctx *__restrict__ ctx)
 {
@@ -403,7 +419,7 @@ inline void cryptonight_single_hash(const uint8_t *__restrict__ input, size_t si
 
     keccak(input, (int) size, ctx->state0, 200);
 
-    VARIANT1_INIT(0);
+    VARIANT1_INIT(0)
 
     cn_explode_scratchpad<ALGO, MEM, SOFT_AES>((__m128i*) ctx->state0, (__m128i*) ctx->memory);
 
@@ -416,7 +432,7 @@ inline void cryptonight_single_hash(const uint8_t *__restrict__ input, size_t si
 
     uint64_t idx0 = h0[0] ^ h0[4];
     void* mp = ((uint8_t*) l0) + ((idx0) & MASK);
-  
+    
     for (size_t i = 0; i < ITERATIONS; i++) {
         __m128i cx;
 
@@ -501,80 +517,83 @@ inline void cryptonight_double_hash(const uint8_t *__restrict__ input, size_t si
     uint64_t idx0 = h0[0] ^ h0[4];
     uint64_t idx1 = h1[0] ^ h1[4];
 
+    void* mp0 = ((uint8_t*) l0) + ((idx0) & MASK);
+    void* mp1 = ((uint8_t*) l1) + ((idx1) & MASK);
+
     for (size_t i = 0; i < ITERATIONS; i++) {
         __m128i cx0, cx1;
 
         if (SOFT_AES) {
-            cx0 = soft_aesenc((uint32_t*)&l0[idx0 & MASK], _mm_set_epi64x(ah0, al0));
+            cx0 = soft_aesenc((uint32_t*) mp0, _mm_set_epi64x(ah0, al0));
-            cx1 = soft_aesenc((uint32_t*)&l1[idx1 & MASK], _mm_set_epi64x(ah1, al1));
+            cx1 = soft_aesenc((uint32_t*) mp1, _mm_set_epi64x(ah1, al1));
-        }
+        } else {
-        else {
+            cx0 = _mm_load_si128((__m128i *) mp0);
-            cx0 = _mm_load_si128((__m128i *) &l0[idx0 & MASK]);
+            cx1 = _mm_load_si128((__m128i *)mp1);
-            cx1 = _mm_load_si128((__m128i *) &l1[idx1 & MASK]);
             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 & MASK], _mm_xor_si128(bx0, cx0));
+        if (VARIANT > 0) {
-        _mm_store_si128((__m128i *) &l1[idx1 & MASK], _mm_xor_si128(bx1, cx1));
+            cryptonight_monero_tweak((uint64_t*)mp0, _mm_xor_si128(bx0, cx0));
-        VARIANT1_1(&l0[idx0 & MASK]);
+            cryptonight_monero_tweak((uint64_t*)mp1, _mm_xor_si128(bx1, cx1));
-        VARIANT1_1(&l1[idx1 & MASK]);
+        } else {
+            _mm_store_si128((__m128i *) mp0, _mm_xor_si128(bx0, cx0));
+            _mm_store_si128((__m128i *) mp1, _mm_xor_si128(bx1, cx1));
+        }
 
-        idx0 = EXTRACT64(cx0);
+        mp0 = ((uint8_t*) l0) + ((idx0 = EXTRACT64(cx0)) & MASK);
-        idx1 = EXTRACT64(cx1);
+        mp1 = ((uint8_t*) l1) + ((idx1 = EXTRACT64(cx1)) & MASK);
 
         bx0 = cx0;
         bx1 = cx1;
 
         uint64_t hi, lo, cl, ch;
-        cl = ((uint64_t*) &l0[idx0 & MASK])[0];
+        cl = ((uint64_t*) mp0)[0];
-        ch = ((uint64_t*) &l0[idx0 & MASK])[1];
+        ch = ((uint64_t*) mp0)[1];
         lo = __umul128(idx0, cl, &hi);
 
         al0 += hi;
         ah0 += lo;
 
         VARIANT1_2(ah0, 0);
-        ((uint64_t*) &l0[idx0 & MASK])[0] = al0;
+        ((uint64_t*) mp0)[0] = al0;
-        ((uint64_t*) &l0[idx0 & MASK])[1] = ah0;
+        ((uint64_t*) mp0)[1] = ah0;
         VARIANT1_2(ah0, 0);
 
         ah0 ^= ch;
         al0 ^= cl;
-        idx0 = al0;
+        mp0 = ((uint8_t*) l0) + ((al0) & MASK);
 
         if (ALGO == xmrig::CRYPTONIGHT_HEAVY) {
-            int64_t n  = ((int64_t*)&l0[idx0 & MASK])[0];
+            int64_t n  = ((int64_t*)mp0)[0];
-            int32_t d  = ((int32_t*)&l0[idx0 & MASK])[2];
+            int32_t d  = ((int32_t*)mp0)[2];
             int64_t q = n / (d | 0x5);
 
-            ((int64_t*)&l0[idx0 & MASK])[0] = n ^ q;
+            ((int64_t*) mp0)[0] = n ^ q;
-            idx0 = d ^ q;
         }
 
-        cl = ((uint64_t*) &l1[idx1 & MASK])[0];
+        cl = ((uint64_t*) mp1)[0];
-        ch = ((uint64_t*) &l1[idx1 & MASK])[1];
+        ch = ((uint64_t*) mp1)[1];
         lo = __umul128(idx1, cl, &hi);
 
         al1 += hi;
         ah1 += lo;
 
         VARIANT1_2(ah1, 1);
-        ((uint64_t*) &l1[idx1 & MASK])[0] = al1;
+        ((uint64_t*) mp1)[0] = al1;
-        ((uint64_t*) &l1[idx1 & MASK])[1] = ah1;
+        ((uint64_t*) mp1)[1] = ah1;
         VARIANT1_2(ah1, 1);
 
         ah1 ^= ch;
         al1 ^= cl;
-        idx1 = al1;
+        mp1 = ((uint8_t*) l1) + ((al1) & MASK);
 
         if (ALGO == xmrig::CRYPTONIGHT_HEAVY) {
-            int64_t n  = ((int64_t*)&l1[idx1 & MASK])[0];
+            int64_t n  = ((int64_t*)mp1)[0];
-            int32_t d  = ((int32_t*)&l1[idx1 & MASK])[2];
+            int32_t d  = ((int32_t*)mp1)[2];
             int64_t q = n / (d | 0x5);
 
-            ((int64_t*)&l1[idx1 & MASK])[0] = n ^ q;
+            ((int64_t*)mp1)[0] = n ^ q;
-            idx1 = d ^ q;
         }
     }
 

src/version.h

@@ -27,7 +27,7 @@
 #define APP_ID        "xmrig"
 #define APP_NAME      "XMRig"
 #define APP_DESC      "XMRig CPU miner"
-#define APP_VERSION   "2.6.0-beta1"
+#define APP_VERSION   "2.6.0-beta2"
 #define APP_DOMAIN    "xmrig.com"
 #define APP_SITE      "www.xmrig.com"
 #define APP_COPYRIGHT "Copyright (C) 2016-2018 xmrig.com"
@@ -36,7 +36,7 @@
 #define APP_VER_MAJOR  2
 #define APP_VER_MINOR  6
 #define APP_VER_BUILD  0
-#define APP_VER_REV    1
+#define APP_VER_REV    2
 
 #ifdef _MSC_VER
 #   if (_MSC_VER >= 1910)

src/workers/CpuThread.cpp

@@ -146,28 +146,28 @@ xmrig::CpuThread *xmrig::CpuThread::createFromAV(size_t index, Algo algorithm, A
         softAES  = true;
         break;
 
-    case AV_DOUBLE:
-        multiway = DoubleWay;
     case AV_DOUBLE_SOFT:
         softAES  = true;
+    case AV_DOUBLE:
+        multiway = DoubleWay;
         break;
 
-    case AV_TRIPLE:
-        multiway = TripleWay;
     case AV_TRIPLE_SOFT:
         softAES  = true;
+    case AV_TRIPLE:
+        multiway = TripleWay;
         break;
 
-    case AV_QUAD:
-        multiway = QuadWay;
     case AV_QUAD_SOFT:
         softAES  = true;
+    case AV_QUAD:
+        multiway = QuadWay;
         break;
 
-    case AV_PENTA:
-        multiway = PentaWay;
     case AV_PENTA_SOFT:
         softAES  = true;
+    case AV_PENTA:
+        multiway = PentaWay;
         break;
 
     default: