v2

verus/haraka.c

@@ -26,7 +26,7 @@ Optimized Implementations for Haraka256 and Haraka512
 
 #include <stdio.h>
 #include "haraka.h"
-
+#include <stdint.h>
 u128 rc[40];
 u128 rc0[40] = {0};
 
@@ -434,7 +434,6 @@ void haraka512_keyed(unsigned char *out, const unsigned char *in, const u128 *rc
   s[3] = LOAD(in + 48);
 
   AES4(s[0], s[1], s[2], s[3], 0);
-
   MIX4(s[0], s[1], s[2], s[3]);
 
   AES4(s[0], s[1], s[2], s[3], 8);
@@ -444,17 +443,18 @@ void haraka512_keyed(unsigned char *out, const unsigned char *in, const u128 *rc
   MIX4(s[0], s[1], s[2], s[3]);
 
   AES4(s[0], s[1], s[2], s[3], 24);
-  MIX4(s[0], s[1], s[2], s[3]);
+  MIX4_LAST(s[0], s[1], s[2], s[3]);
+
+  AES4_LAST(s[0], s[1], s[2], s[3], 32);
 
-  AES4(s[0], s[1], s[2], s[3], 32);
-  MIX4LAST(s[0], s[1], s[2], s[3]);
 
  // s[0] = _mm_xor_si128(s[0], LOAD(in));
  // s[1] = _mm_xor_si128(s[1], LOAD(in + 16));
  // s[2] = _mm_xor_si128(s[2], LOAD(in + 32));
-  s[3] = _mm_xor_si128(s[3], LOAD(in + 48));
+ // s[3] = _mm_xor_si128(s[0], LOAD(in + 48));
+  ((uint32_t*)&out[0])[7] = ((uint32_t*)&s[0])[10] ^ ((uint32_t*)&in[52])[0];
 
-  TRUNCSTORE(out, s[0], s[1], s[2], s[3]);
+  //TRUNCSTORE(out, s[0],s[1], s[2], s[3]);
 }
 
 void haraka512_4x(unsigned char *out, const unsigned char *in) {

verus/haraka.h

@@ -175,15 +175,10 @@ extern u128 rc[40];
   s2 = _mm_aesenc_si128(s2, rc[rci + 6]); \
   s3 = _mm_aesenc_si128(s3, rc[rci + 7]); \
 
-#define AES4LAST(s0, s1, s2, s3, rci) \
-  s0 = _mm_aesenc_si128(s0, rc[rci]); \
-  s1 = _mm_aesenc_si128(s1, rc[rci + 1]); \
+#define AES4_LAST(s0, s1, s2, s3, rci) \
   s2 = _mm_aesenc_si128(s2, rc[rci + 2]); \
-  s3 = _mm_aesenc_si128(s3, rc[rci + 3]); \
-  s0 = _mm_aesenc_si128(s0, rc[rci + 4]); \
-  s1 = _mm_aesenc_si128(s1, rc[rci + 5]); \
-  s2 = _mm_aesenc_si128(s2, rc[rci + 6]); \
-  s3 = _mm_aesenc_si128(s3, rc[rci + 7]); \
+  s2 = _mm_aesenc_si128(s2, rc[rci + 6]); 
+
 
 #define AES4_zero(s0, s1, s2, s3, rci) \
   s0 = _mm_aesenc_si128(s0, rc0[rci]); \
@@ -220,24 +215,15 @@ extern u128 rc[40];
   s2 = _mm_unpackhi_epi32(s1, tmp); \
   s1 = _mm_unpacklo_epi32(s1, tmp);
 
-#define MIX4LAST(s0, s1, s2, s3) \
-  s0 = _mm_unpackhi_epi32(s0, s1); \
-  s2 = _mm_unpackhi_epi32(s2, s3); \
-  s3 = _mm_unpacklo_epi32(s0, s2); 
-
-#define MIX4_LASTBUT1(s0, s1, s2, s3) \
+#define MIX4_LAST(s0, s1, s2, s3) \
   tmp  = _mm_unpacklo_epi32(s0, s1); \
   s1 = _mm_unpacklo_epi32(s2, s3); \
-  s2 = _mm_unpackhi_epi32(s1, tmp); 
-
-#define AES4_LAST(s3, rci) \
-  _mm_aesenc_si128(s3, rc[rci + 2]); \
-  _mm_aesenc_si128(s3, rc[rci + 6]); \
-
-#define TRUNCSTORE(out, s0, s1, s2, s3) \
-  *(u64*)(out) = *(((u64*)&s0 + 1)); \
-  *(u64*)(out + 8) = *(((u64*)&s1 + 1)); \
-  *(u64*)(out + 16) = *(((u64*)&s2 + 0)); \
+  s2 = _mm_unpackhi_epi32(s1, tmp); 
+
+#define TRUNCSTORE(out, s0, s1, s2, s3) \
+  *(u64*)(out) = *(((u64*)&s0 + 1)); \
+  *(u64*)(out + 8) = *(((u64*)&s1 + 1)); \
+  *(u64*)(out + 16) = *(((u64*)&s2 + 0)); \
   *(u64*)(out + 24) = *(((u64*)&s3 + 0));
 
 void load_constants();

verus/verus_clhash.cpp

@@ -24,7 +24,7 @@
 
 #include <assert.h>
 #include <string.h>
-#include "immintrin.h"
+//#include <intrin.h>
 //#include "cpu_verushash.hpp"
 
 #ifdef _WIN32
@@ -76,7 +76,7 @@ static inline uint64_t precompReduction64( __m128i A) {
 }
 
 // verus intermediate hash extra
-static __m128i __verusclmulwithoutreduction64alignedrepeat(__m128i *randomsource, const __m128i buf[4], uint64_t keyMask, uint32_t *fixrand, uint32_t *fixrandex)
+static __m128i __verusclmulwithoutreduction64alignedrepeat(__m128i * __restrict randomsource, const __m128i *  __restrict buf, uint64_t keyMask, uint32_t * __restrict fixrand, uint32_t * __restrict fixrandex)
 {
     __m128i const *pbuf;
 
@@ -101,6 +101,7 @@ static __m128i __verusclmulwithoutreduction64alignedrepeat(__m128i *randomsource
     // the random buffer must have at least 32 16 byte dwords after the keymask to work with this
     // algorithm. we take the value from the last element inside the keyMask + 2, as that will never
     // be used to xor into the accumulator before it is hashed with other values first
+
 	  for (uint64_t i = 0; i < 32; i++)
     {
 
@@ -402,7 +403,7 @@ static __m128i __verusclmulwithoutreduction64alignedrepeat(__m128i *randomsource
 
 // hashes 64 bytes only by doing a carryless multiplication and reduction of the repeated 64 byte sequence 16 times, 
 // returning a 64 bit hash value
-uint64_t verusclhash(void * random, const unsigned char buf[64], uint64_t keyMask, uint32_t *fixrand, uint32_t *fixrandex) {
+uint64_t verusclhash(void * random, const unsigned char buf[64], uint64_t keyMask, uint32_t *  __restrict fixrand, uint32_t * __restrict fixrandex) {
 	const __m128i lazy = _mm_cvtsi32_si128( 0x00010000);
 	__m128i  acc = __verusclmulwithoutreduction64alignedrepeat((__m128i *)random, (const __m128i *)buf, keyMask, fixrand, fixrandex);
     acc = _mm_xor_si128(acc, lazy);
@@ -430,9 +431,9 @@ inline void haraka512_keyed_local(unsigned char *out, const unsigned char *in, c
 
   AES4(s[0], s[1], s[2], s[3], 24);
 
-  MIX4_LASTBUT1(s[0], s[1], s[2], s[3]);
+ //MIX4_LASTBUT1(s[0], s[1], s[2], s[3]);
 
-  AES4_LAST(s[2], 32);
+ // AES4_LAST(s[2], 32);
 
 //  MIX4(s[0], s[1], s[2], s[3]);
 
@@ -441,7 +442,7 @@ inline void haraka512_keyed_local(unsigned char *out, const unsigned char *in, c
 
  // s[0] = _mm_xor_si128(s[0], LOAD(in));
  // s[1] = _mm_xor_si128(s[1], LOAD(in + 16));
-  s[2] = _mm_xor_si128(s[2], LOAD(in + 46));
+ // s[2] = _mm_xor_si128(s[2], LOAD(in + 46));
  // s[3] = _mm_xor_si128(s[3], LOAD(in + 48));
 
  // TRUNCSTORE(out, s[0], s[1], s[2], s[3]);

verus/verusscan.cpp

@@ -129,7 +129,7 @@ extern "C" void VerusHashHalf(void *result2, unsigned char *data, size_t len)
 
 
 extern "C" void Verus2hash(unsigned char *hash, unsigned char *curBuf, uint32_t nonce,
-	u128 *data_key, uint8_t *gpu_init, uint32_t *fixrand, uint32_t *fixrandex, u128 *data_key_master)
+	u128 * __restrict data_key, uint8_t *gpu_init, uint32_t * __restrict fixrand, uint32_t * __restrict fixrandex, u128 * __restrict data_key_master)
 {
 	uint64_t mask = VERUS_KEY_SIZE128; //552
 	if (!gpu_init[0]) {
@@ -203,7 +203,7 @@ extern "C" int scanhash_verus(int thr_id, struct work *work, uint32_t max_nonce,
 		*hashes_done = nonce_buf + throughput;
 		Verus2hash((unsigned char *)vhash, (unsigned char *)blockhash_half, nonce_buf, data_key, &gpuinit, fixrand, fixrandex, data_key_master);
 
-		if (vhash[7] <= Htarg && fulltest(vhash, ptarget))
+		if (vhash[7] <= Htarg )
 		{
 			*((uint32_t *)full_data + 368) = nonce_buf;
 			work->valid_nonces++;
@@ -235,8 +235,8 @@ extern "C" int scanhash_verus(int thr_id, struct work *work, uint32_t max_nonce,
 	solps = (double)nonce_buf / secs;
 
 	pdata[NONCE_OFT] = endiandata[NONCE_OFT] + 1;
-	free(data_key);
-	free(data_key_master);
+	//free(data_key);
+	//free(data_key_master);
 	return work->valid_nonces;
 }