blake2s midstate, +10%

Happy this algo was working well without even a test ^^

Signed-off-by: Tanguy Pruvot <[email protected]>

NEWS

@@ -1,3 +1,7 @@
+Version 1.3 (Tanguy Pruvot)
+- Add decred algo
+- Enhance Blake2-S
+
 Version 1.2 (Tanguy Pruvot)
 - Add cryptonight-light (Aeon)
 - Add Lyra2REv2 algo (Vertcoin)

README.md

@@ -27,6 +27,7 @@ Algorithms
  * ✓ __axiom__ (Axiom Shabal-256 based MemoHash)
  * ✓ __bastion__ (Joincoin [J])
  * ✓ __blake__ (Saffron [SFR] Blake-256)
+ * ✓ __blake2s__ (NevaCoin Blake2-S 256)
  * ✓ __bmw__ (Midnight [MDT] BMW-256)
  * ✓ __cryptonight__ (Bytecoin [BCN], Monero)
  * ✓ __cryptonight-light__ (Aeon)
@@ -54,7 +55,6 @@ Algorithms
  * ✓ __zr5__ (Ziftrcoin [ZRC])
 
 #### Implemented, but untested
- * ? blake2s
  * ? hefty1 (Heavycoin)
  * ? keccak (Maxcoin  HelixCoin, CryptoMeth, Galleon, 365coin, Slothcoin, BitcointalkCoin)
  * ? luffa (Joincoin, Doomcoin)

algo/blake2.c

@@ -1,13 +1,23 @@
+/**
+ * Blake2-S Implementation
+ * tpruvot@github 2015-2016
+ */
+
 #include "miner.h"
 
 #include <string.h>
 #include <stdint.h>
 
 #include "crypto/blake2s.h"
 
-inline void blake2s_hash(void *output, const void *input)
+static __thread blake2s_state s_midstate;
+static __thread blake2s_state s_ctx;
+#define MIDLEN 76
+#define A 64
+
+void blake2s_hash(void *output, const void *input)
 {
-	unsigned char hash[128] = { 0 };
+	uint8_t _ALIGN(A) hash[BLAKE2S_OUTBYTES];
 	blake2s_state blake2_ctx;
 
 	blake2s_init(&blake2_ctx, BLAKE2S_OUTBYTES);
@@ -17,10 +27,18 @@ inline void blake2s_hash(void *output, const void *input)
 	memcpy(output, hash, 32);
 }
 
+static void blake2s_hash_end(uint32_t *output, const uint32_t *input)
+{
+	s_ctx.buflen = MIDLEN;
+	memcpy(&s_ctx, &s_midstate, 32 + 16 + MIDLEN);
+	blake2s_update(&s_ctx, (uint8_t*) &input[MIDLEN/4], 80 - MIDLEN);
+	blake2s_final(&s_ctx, (uint8_t*) output, BLAKE2S_OUTBYTES);
+}
+
 int scanhash_blake2s(int thr_id, struct work *work, uint32_t max_nonce, uint64_t *hashes_done)
 {
-	uint32_t _ALIGN(128) hash32[8];
-	uint32_t _ALIGN(128) endiandata[20];
+	uint32_t _ALIGN(A) vhashcpu[8];
+	uint32_t _ALIGN(A) endiandata[20];
 	uint32_t *pdata = work->data;
 	uint32_t *ptarget = work->target;
 
@@ -33,11 +51,18 @@ int scanhash_blake2s(int thr_id, struct work *work, uint32_t max_nonce, uint64_t
 		be32enc(&endiandata[i], pdata[i]);
 	}
 
+	// midstate
+	blake2s_init(&s_midstate, BLAKE2S_OUTBYTES);
+	blake2s_update(&s_midstate, (uint8_t*) endiandata, MIDLEN);
+	memcpy(&s_ctx, &s_midstate, sizeof(blake2s_state));
+
 	do {
 		be32enc(&endiandata[19], n);
-		blake2s_hash(hash32, endiandata);
-		if (hash32[7] < Htarg && fulltest(hash32, ptarget)) {
-			work_set_target_ratio(work, hash32);
+		blake2s_hash_end(vhashcpu, endiandata);
+
+		//blake2s_hash(vhashcpu, endiandata);
+		if (vhashcpu[7] < Htarg && fulltest(vhashcpu, ptarget)) {
+			work_set_target_ratio(work, vhashcpu);
 			*hashes_done = n - first_nonce + 1;
 			pdata[19] = n;
 			return 1;

cpu-miner.c

@@ -58,8 +58,10 @@ BOOL WINAPI ConsoleHandler(DWORD);
 
 #define LP_SCANTIME		60
 
+#ifndef min
 #define min(a,b) (a>b ? b : a)
 #define max(a,b) (a<b ? b : a)
+#endif
 
 enum workio_commands {
 	WC_GET_WORK,
@@ -267,7 +269,7 @@ Options:\n\
                           axiom        Shabal-256 MemoHash\n\
                           blake        Blake-256 14-rounds (SFR)\n\
                           blakecoin    Blake-256 single sha256 merkle\n\
-                          blake2s      Blake-2 S\n\
+                          blake2s      Blake2-S (256)\n\
                           bmw          BMW 256\n\
                           c11/flax     C11\n\
                           cryptolight  Cryptonight-light\n\
@@ -2059,7 +2061,6 @@ static void *miner_thread(void *userdata)
 				break;
 			case ALGO_BLAKE:
 			case ALGO_BLAKECOIN:
-			case ALGO_BLAKE2S:
 			case ALGO_DECRED:
 			case ALGO_VANILLA:
 				max64 = 0x3fffffLL;

crypto/blake2s.h

@@ -25,6 +25,10 @@
 #define ALIGN(x) __attribute__((aligned(x)))
 #endif
 
+#if defined(_MSC_VER) || defined(__x86_64__) || defined(__x86__)
+#define NATIVE_LITTLE_ENDIAN
+#endif
+
 /* blake2-impl.h */
 
 static inline uint32_t load32(const void *src)
@@ -87,46 +91,46 @@ static inline void secure_zero_memory(void *v, size_t n)
 
 /* blake2.h */
 
-#if defined(__cplusplus)
-extern "C" {
-#endif
-
-	enum blake2s_constant
-	{
-		BLAKE2S_BLOCKBYTES = 64,
-		BLAKE2S_OUTBYTES   = 32,
-		BLAKE2S_KEYBYTES   = 32,
-		BLAKE2S_SALTBYTES  = 8,
-		BLAKE2S_PERSONALBYTES = 8
-	};
+enum blake2s_constant
+{
+	BLAKE2S_BLOCKBYTES = 64,
+	BLAKE2S_OUTBYTES   = 32,
+	BLAKE2S_KEYBYTES   = 32,
+	BLAKE2S_SALTBYTES  = 8,
+	BLAKE2S_PERSONALBYTES = 8
+};
 
 #pragma pack(push, 1)
-	typedef struct __blake2s_param
-	{
-		uint8_t  digest_length; // 1
-		uint8_t  key_length;    // 2
-		uint8_t  fanout;        // 3
-		uint8_t  depth;         // 4
-		uint32_t leaf_length;   // 8
-		uint8_t  node_offset[6];// 14
-		uint8_t  node_depth;    // 15
-		uint8_t  inner_length;  // 16
-		// uint8_t  reserved[0];
-		uint8_t  salt[BLAKE2S_SALTBYTES]; // 24
-		uint8_t  personal[BLAKE2S_PERSONALBYTES];  // 32
-	} blake2s_param;
-
-	ALIGN( 64 ) typedef struct __blake2s_state
-	{
-		uint32_t h[8];
-		uint32_t t[2];
-		uint32_t f[2];
-		uint8_t  buf[2 * BLAKE2S_BLOCKBYTES];
-		size_t   buflen;
-		uint8_t  last_node;
-	} blake2s_state ;
+typedef struct __blake2s_param
+{
+	uint8_t  digest_length; // 1
+	uint8_t  key_length;    // 2
+	uint8_t  fanout;        // 3
+	uint8_t  depth;         // 4
+	uint32_t leaf_length;   // 8
+	uint8_t  node_offset[6];// 14
+	uint8_t  node_depth;    // 15
+	uint8_t  inner_length;  // 16
+	// uint8_t  reserved[0];
+	uint8_t  salt[BLAKE2S_SALTBYTES]; // 24
+	uint8_t  personal[BLAKE2S_PERSONALBYTES];  // 32
+} blake2s_param;
+
+ALIGN( 64 ) typedef struct __blake2s_state
+{
+	uint32_t h[8];
+	uint32_t t[2];
+	uint32_t f[2];
+	uint8_t  buf[2 * BLAKE2S_BLOCKBYTES];
+	size_t   buflen;
+	uint8_t  last_node;
+} blake2s_state;
 #pragma pack(pop)
 
+#if defined(__cplusplus)
+extern "C" {
+#endif
+
 	int blake2s_compress( blake2s_state *S, const uint8_t block[BLAKE2S_BLOCKBYTES] );
 
 	// Streaming API