shilch/fastmerkle

Compute bitcoin Merkle trees on your GPU

fastmerkle's Issues

CL_MEM_OBJECT_ALLOCATION_FAILURE on Nvidia GTX1070

I had to change merkle input argument types in merkle.cl

diff --git a/merkle.cl b/merkle.cl
index af4de86..f47b2da 100644
--- a/merkle.cl
+++ b/merkle.cl
@@ -233,7 +233,7 @@ static void sha256d_cat(global const uint* lhs, global const uint* rhs, global u

 #define EQUAL(lhs, rhs) (lhs[0] == rhs[0] && lhs[1] == rhs[1] && lhs[2] == rhs[2] && lhs[3] == rhs[3] && lhs[4] == rhs[4] && lhs[5] == rhs[5] && lhs[6] == rhs[6] && lhs[7] == rhs[7])

-kernel void merkle(global const uchar* data, const size_t leaves, const size_t round, global int* mutated){
+kernel void merkle(global const uchar* data, const ulong leaves, const ulong round, global int* mutated){
     size_t id = get_global_id(0);
     size_t size = get_global_size(0);

to be able to even start on GTX1070, because size_t is apparently restricted type from input arguments. ulong is required instead of uint because of CL_INVALID_ARG_SIZE error.

Unfortunately, this fix only partially helps (btw - still works for integrated GPU on my Intel CPU), because on 8388608 leaves I'm getting CL_MEM_OBJECT_ALLOCATION_FAILURE error. I'm guessing that some other values/data types should also be changed to ulong, but I'm not an expert in OpenCL and not sure what else should be changed.