Yangshen Deng

Results 10 issues of Yangshen Deng

brpc 的bthread中提供一个`bthread_start_urgent`接口,可以启动一个协程后立即将它swap in当前的线程上执行。这种接口的应用场景是新的协程要读写当前线程上的数据时,可以让他拥有更好的cache locality。 libgo我阅读了一下发现并没有提供类似的接口,请问是否有这样的接口以及是否可以考虑增加一个呢?我非常乐意参与到接口的开发上。

注释中将send_chain后的delay计算和定时器重设解释为“发送速度过快”具有误导性,其实应该是根据发送的字节数设置定时器进行补偿。

![image](https://user-images.githubusercontent.com/47519522/69218893-a3e20300-0bac-11ea-82cf-72c2dbaa4269.png) ![image](https://user-images.githubusercontent.com/47519522/69219021-e86d9e80-0bac-11ea-8446-bd818e9dd578.png) SE should be placed after Depthwise and before Pointwise. However, you set it after Pointwise. Is there any reason or just a mistake?

Our team is currently working on GPU DB. We have read your blog and pretty much appreciate your work. Now we are trying to integrate CURA into other DB such...

I'm an MPhil student doing research on GPU database. I want to extend heavydb's execution engine to introduce heterogeneous parallelism and new decision model. Here's what I want to achieve....

I'm working importing SSB into heavydb. My source file `date.tbl` looks like. ``` 19920101,"January 1, 1992","Thursday","January",1992,199201,"Jan1992",5,1,1,1,1,"Winter","0","1","1","1" 19920102,"January 2, 1992","Friday","January",1992,199201,"Jan1992",6,2,2,1,1,"Winter","0","1","0","1" 19920103,"January 3, 1992","Saturday","January",1992,199201,"Jan1992",7,3,3,1,1,"Winter","1","1","0","0" 19920104,"January 4, 1992","Sunday","January",1992,199201,"Jan1992",1,4,4,1,1,"Winter","0","1","0","0" 19920105,"January 5, 1992","Monday","January",1992,199201,"Jan1992",2,5,5,1,1,"Winter","0","1","0","1" 19920106,"January 6,...

1. Fix cmake to compile comm.cpp. 2. Add support of CMake install. Headers are installed to `include/libco`

I can't find BlockScan and BlockSuffle API which are mentioned in Table 1 in your paper. ![image](https://user-images.githubusercontent.com/47519522/134734572-fce38205-0cf2-4c57-9731-99c8ecc1ab18.png)

### Motivation. Currently models like `llava-hf/llava-next-video*` recognize image and video inputs with different tokens, and do different computations. Therefore vLLM should provide new APIs and inference support for video input....

RFC