prepare the 1.7 release

README.txt

@@ -1,5 +1,5 @@
 
-ccMiner release 1.6.6 (Aug 2015)   "Lyra2REv2, BMW And Vanilla"
+ccMiner release 1.7.0 (Nov 2015) "Mining Diff & basic MultiAlgo"
 ---------------------------------------------------------------
 
 ***************************************************************
@@ -228,11 +228,13 @@ features.
 
 >>> RELEASE HISTORY <<<
 
-  Under Dev...    v1.7
+  Nov. 02nd 2015  v1.7
+                  Improve old devices compatibility (x11, lyra2, qubit...)
                   Add windows support for SM 2.1 and drop SM 3.5 (x86)
-                  Improve lyra2 (v1/v2) cuda implementation
+                  Improve lyra2 (v1/v2) cuda implementations
+                  Improve most common algos on SM5+ with sp blake kernel
                   Restore whirlpool algo (and whirlcoin variant)
-                  Prepare algo switch ability
+                  Prepare algo/pool switch ability, trivial method
                   Add --benchmark alone to run a benchmark for all algos
                   Add --cuda-schedule parameter
                   Add --show-diff parameter, which display shares diff,

bench.cpp

@@ -77,9 +77,6 @@ void algo_free_all(int thr_id)
 	//free_sha256d(thr_id);
 	free_scrypt(thr_id);
 	free_scrypt_jane(thr_id);
-
-	// warn on cuda error
-	CUDA_LOG_ERROR();
 }
 
 // benchmark all algos (called once per mining thread)
@@ -118,6 +115,7 @@ bool bench_algo_switch_next(int thr_id)
 	// free current algo memory and track mem usage
 	mused = cuda_available_memory(thr_id);
 	algo_free_all(thr_id);
+	CUDA_LOG_ERROR();
 
 	// device can take some time to free
 	mfree = cuda_available_memory(thr_id);

ccminer.cpp

@@ -1636,6 +1636,8 @@ static void *miner_thread(void *userdata)
 
 			// free gpu resources
 			algo_free_all(thr_id);
+			// clear any free error (algo switch)
+			cuda_clear_lasterror();
 
 			// conditional pool switch
 			if (num_pools > 1 && conditional_pool_rotate) {

configure.ac

@@ -1,4 +1,4 @@
-AC_INIT([ccminer], [1.7-dev])
+AC_INIT([ccminer], [1.7])
 
 AC_PREREQ([2.59c])
 AC_CANONICAL_SYSTEM

cpuminer-config.h

@@ -162,7 +162,7 @@
 #define PACKAGE_NAME "ccminer"
 
 /* Define to the full name and version of this package. */
-#define PACKAGE_STRING "ccminer 1.7-dev"
+#define PACKAGE_STRING "ccminer 1.7"
 
 /* Define to the one symbol short name of this package. */
 #define PACKAGE_TARNAME "ccminer"
@@ -171,7 +171,7 @@
 #define PACKAGE_URL "http://github.com/tpruvot/ccminer"
 
 /* Define to the version of this package. */
-#define PACKAGE_VERSION "1.7-dev"
+#define PACKAGE_VERSION "1.7"
 
 /* If using the C implementation of alloca, define if you know the
    direction of stack growth for your system; otherwise it will be
@@ -185,7 +185,7 @@
 #define STDC_HEADERS 1
 
 /* Version number of package */
-#define VERSION "1.7-dev"
+#define VERSION "1.7"
 
 /* Define curl_free() as free() if our version of curl lacks curl_free. */
 /* #undef curl_free */

cuda.cpp

@@ -201,6 +201,12 @@ void cuda_log_lasterror(int thr_id, const char* func, int line)
 		gpulog(LOG_WARNING, thr_id, "%s:%d %s", func, line, cudaGetErrorString(err));
 }
 
+// Clear any cuda error in non-cuda unit (.c/.cpp)
+void cuda_clear_lasterror()
+{
+	cudaGetLastError();
+}
+
 #ifdef __cplusplus
 } /* extern "C" */
 #endif

miner.h

@@ -494,6 +494,7 @@ uint32_t cuda_default_throughput(int thr_id, uint32_t defcount);
 #define device_intensity(t,f,d) cuda_default_throughput(t,d)
 
 void cuda_log_lasterror(int thr_id, const char* func, int line);
+void cuda_clear_lasterror();
 #define CUDA_LOG_ERROR() cuda_log_lasterror(thr_id, __func__, __LINE__)
 
 #define CL_N    "\x1B[0m"

qubit/qubit.cu

@@ -23,6 +23,8 @@ extern void qubit_luffa512_cpu_hash_80(int thr_id, uint32_t threads, uint32_t st
 
 extern "C" void qubithash(void *state, const void *input)
 {
+	uint8_t _ALIGN(128) hash[64];
+
 	// luffa1-cubehash2-shavite3-simd4-echo5
 
 	sph_luffa512_context ctx_luffa;
@@ -31,8 +33,6 @@ extern "C" void qubithash(void *state, const void *input)
 	sph_simd512_context ctx_simd;
 	sph_echo512_context ctx_echo;
 
-	uint8_t hash[64];
-
 	sph_luffa512_init(&ctx_luffa);
 	sph_luffa512 (&ctx_luffa, input, 80);
 	sph_luffa512_close(&ctx_luffa, (void*) hash);
@@ -68,7 +68,7 @@ extern "C" int scanhash_qubit(int thr_id, struct work* work, uint32_t max_nonce,
 	if (init[thr_id]) throughput = min(throughput, max_nonce - first_nonce);
 
 	if (opt_benchmark)
-		((uint32_t*)ptarget)[7] = 0x0000ff;
+		ptarget[7] = 0x007f;
 
 	if (!init[thr_id])
 	{
@@ -133,11 +133,16 @@ extern "C" int scanhash_qubit(int thr_id, struct work* work, uint32_t max_nonce,
 			}
 		}
 
+		if ((uint64_t) throughput + pdata[19] >= max_nonce) {
+			pdata[19] = max_nonce;
+			break;
+		}
+
 		pdata[19] += throughput;
 
-	} while (pdata[19] < max_nonce && !work_restart[thr_id].restart);
+	} while (!work_restart[thr_id].restart);
 
-	*hashes_done = pdata[19] - first_nonce + 1;
+	*hashes_done = pdata[19] - first_nonce;
 	return 0;
 }