Browse Source

Add missing real cuda arch checks

master
Tanguy Pruvot 8 years ago
parent
commit
2cdf2ddd43
  1. 2
      Algo256/blake256.cu
  2. 2
      Algo256/blake2s.cu
  3. 2
      Algo256/decred.cu
  4. 2
      Algo256/vanilla.cu
  5. 1
      lbry/lbry.cu
  6. 2
      skein.cu

2
Algo256/blake256.cu

@ -506,6 +506,8 @@ extern "C" int scanhash_blake256(int thr_id, struct work* work, uint32_t max_non @@ -506,6 +506,8 @@ extern "C" int scanhash_blake256(int thr_id, struct work* work, uint32_t max_non
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
cuda_get_arch(thr_id);
CUDA_CALL_OR_RET_X(cudaMalloc(&d_resNonce[thr_id], NBN * sizeof(uint32_t)), -1);
CUDA_CALL_OR_RET_X(cudaMallocHost(&h_resNonce[thr_id], NBN * sizeof(uint32_t)), -1);
init[thr_id] = true;

2
Algo256/blake2s.cu

@ -458,6 +458,8 @@ extern "C" int scanhash_blake2s(int thr_id, struct work *work, uint32_t max_nonc @@ -458,6 +458,8 @@ extern "C" int scanhash_blake2s(int thr_id, struct work *work, uint32_t max_nonc
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
cuda_get_arch(thr_id);
CUDA_CALL_OR_RET_X(cudaMalloc(&d_resNonce[thr_id], maxResults * sizeof(uint32_t)), -1);
CUDA_CALL_OR_RET_X(cudaMallocHost(&h_resNonce[thr_id], maxResults * sizeof(uint32_t)), -1);
init[thr_id] = true;

2
Algo256/decred.cu

@ -374,6 +374,8 @@ extern "C" int scanhash_decred(int thr_id, struct work* work, uint32_t max_nonce @@ -374,6 +374,8 @@ extern "C" int scanhash_decred(int thr_id, struct work* work, uint32_t max_nonce
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
cuda_get_arch(thr_id);
CUDA_CALL_OR_RET_X(cudaMalloc(&d_resNonce[thr_id], MAX_RESULTS*sizeof(uint32_t)), -1);
CUDA_CALL_OR_RET_X(cudaMallocHost(&h_resNonce[thr_id], MAX_RESULTS*sizeof(uint32_t)), -1);
init[thr_id] = true;

2
Algo256/vanilla.cu

@ -394,6 +394,8 @@ extern "C" int scanhash_vanilla(int thr_id, struct work* work, uint32_t max_nonc @@ -394,6 +394,8 @@ extern "C" int scanhash_vanilla(int thr_id, struct work* work, uint32_t max_nonc
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
cuda_get_arch(thr_id);
CUDA_CALL_OR_RET_X(cudaMalloc(&d_resNonce[thr_id], NBN * sizeof(uint32_t)), -1);
CUDA_CALL_OR_RET_X(cudaMallocHost(&h_resNonce[thr_id], NBN * sizeof(uint32_t)), -1);
cudaStreamCreate(&streams[thr_id]);

1
lbry/lbry.cu

@ -119,6 +119,7 @@ extern "C" int scanhash_lbry(int thr_id, struct work *work, uint32_t max_nonce, @@ -119,6 +119,7 @@ extern "C" int scanhash_lbry(int thr_id, struct work *work, uint32_t max_nonce,
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
cuda_get_arch(thr_id);
if (CUDART_VERSION == 6050) {
applog(LOG_ERR, "This lbry kernel is not compatible with CUDA 6.5!");

2
skein.cu

@ -374,6 +374,8 @@ extern "C" int scanhash_skeincoin(int thr_id, struct work* work, uint32_t max_no @@ -374,6 +374,8 @@ extern "C" int scanhash_skeincoin(int thr_id, struct work* work, uint32_t max_no
}
gpulog(LOG_INFO, thr_id, "Intensity set to %g, %u cuda threads", throughput2intensity(throughput), throughput);
cuda_get_arch(thr_id);
if (sm5) {
skeincoin_init(thr_id);
} else {

Loading…
Cancel
Save