Open dd123-a opened 4 weeks ago
里面用了个叫libfork的fork-join多线程库,但似乎问题有点多,我找时间看下。 我看你这应该是在debug模式下运行的,有崩溃的call stack吗?
I found that setting config level_count lower helps, as well as leaf_count. From having people test it out it "seems" like GPU's with Memory Size 8GB crash but i'm not sure.
auto dag_node_pool = DAGNodePool::Create(
hashdag::DefaultConfig<uint32_t>{
.level_count = 17, // change this to 8- 12.
.top_level_count = 9,
.word_bits_per_page = 14,
.page_bits_per_bucket = 2,
.bucket_bits_per_top_level = 7,
.bucket_bits_per_bottom_level = 11,
}(),
{generic_queue, sparse_queue});
auto dag_color_pool = DAGColorPool::Create(
DAGColorPool::Config{
.leaf_level = 10, // also lower this
.node_bits_per_node_page = 18,
.word_bits_per_leaf_page = 24,
.keep_history = false,
},
{generic_queue, sparse_queue});
LOG: [0x16c] /busy_pool.hpp(95:6) "worker waits" LOG: [0x820] /task.hpp(269:21) "~promise_base()" edit cost -426.013144 ms root = 2664
进程已结束,退出代码为 -1073741819 (0xC0000005)