fix wrong cuda binary arch detection

fix wrong arch comparsion
This commit is contained in:
psychocrypt
2017-11-17 09:32:27 +01:00
parent 51c7b26e7e
commit 656c55309f
@@ -364,7 +364,7 @@ extern "C" int cuda_get_deviceinfo(nvid_ctx* ctx)
for(int i = 0; i < arch.size(); ++i)
if(minSupportedArch == 0 || (arch[i] >= 30 && arch[i] < minSupportedArch))
minSupportedArch = arch[i];
if(minSupportedArch >= 30 && gpuArch <= minSupportedArch)
if(minSupportedArch < 30 || gpuArch < minSupportedArch)
{
printf("WARNING: NVIDIA GPU %d: miner not compiled for the gpu architecture %d.\n", ctx->device_id, gpuArch);
return 5;