fix assert in AMD backend

fix bug announced in #52

If the miner is compiled in debug mode the assert condition results in a compile time error.

xmrstak/backend/amd/amd_gpu/gpu.cpp

@@ -896,7 +896,7 @@ size_t XMRRunJob(GpuContext* ctx, cl_uint* HashOutput)
 			// round up to next multiple of w_size
 			BranchNonces[i] = ((BranchNonces[i] + w_size - 1u) / w_size) * w_size;
 			// number of global threads must be a multiple of the work group size (w_size)
-			assert(BranchNonces%w_size == 0);
+			assert(BranchNonces[i]%w_size == 0);
 			if((ret = clEnqueueNDRangeKernel(ctx->CommandQueues, ctx->Kernels[i + 3], 1, &ctx->Nonce, BranchNonces + i, &w_size, 0, NULL, NULL)) != CL_SUCCESS)
 			{
 				printer::inst()->print_msg(L1,"Error %s when calling clEnqueueNDRangeKernel for kernel %d.", err_to_str(ret), i + 3);