lyra2: improve cuda implementation (part 1, SM5+)

based on the new djm34 method, 2x faster than first version

cleaned and tuned for the GTX 750/960 (linux / cuda 6.5)

Algo256/cuda_groestl256.cu

@@ -176,7 +176,7 @@ void groestl256_perm_Q(uint32_t thread, uint32_t *a, char *mixtabs)
 }
 
 __global__ __launch_bounds__(256,1)
-void groestl256_gpu_hash32(uint32_t threads, uint32_t startNounce, uint64_t *outputHash, uint32_t *resNonces)
+void groestl256_gpu_hash_32(uint32_t threads, uint32_t startNounce, uint64_t *outputHash, uint32_t *resNonces)
 {
 #if USE_SHARED
 	extern __shared__ char mixtabs[];
@@ -315,7 +315,7 @@ uint32_t groestl256_cpu_hash_32(int thr_id, uint32_t threads, uint32_t startNoun
 #else
 	size_t shared_size = 0;
 #endif
-	groestl256_gpu_hash32<<<grid, block, shared_size>>>(threads, startNounce, d_outputHash, d_GNonces[thr_id]);
+	groestl256_gpu_hash_32<<<grid, block, shared_size>>>(threads, startNounce, d_outputHash, d_GNonces[thr_id]);
 
 	MyStreamSynchronize(NULL, order, thr_id);
 

README.txt

@@ -229,9 +229,10 @@ features.
 >>> RELEASE HISTORY <<<
 
   Under Dev...    v1.7
+                  Improve lyra2 (v1) cuda implementation
                   Restore whirlpool algo (and whirlcoin variant)
                   Prepare algo switch ability
-                  Add --benchmark -a all to run a benchmark for all algos
+                  Add --benchmark alone to run a benchmark for all algos
                   Add --cuda-schedule parameter
                   Add --show-diff parameter, which display shares diff,
                     and is able to detect real solved blocks on pools.

bench.cpp

@@ -98,7 +98,6 @@ bool bench_algo_switch_next(int thr_id)
 	if (algo == ALGO_DMD_GR) algo++; // same as groestl
 	if (algo == ALGO_WHIRLCOIN) algo++; // same as whirlpool
 	// and unwanted ones...
-	if (algo == ALGO_LYRA2) algo++; // weird memory leak to fix (uint2 Matrix[96][8] too big)
 	if (algo == ALGO_SCRYPT) algo++;
 	if (algo == ALGO_SCRYPT_JANE) algo++;
 

configure.sh

@@ -1,4 +1,5 @@
 # possible additional CUDA_CFLAGS
+#-gencode=arch=compute_52,code=\"sm_52,compute_52\"
 #-gencode=arch=compute_50,code=\"sm_50,compute_50\"
 #-gencode=arch=compute_35,code=\"sm_35,compute_35\"
 #-gencode=arch=compute_30,code=\"sm_30,compute_30\"