liushuai
liushuai
@dearbird 你可以试试一帧大数据，我测得是必然阻塞，原因未知。 @yyzybb537 请问会是什么原因？感觉这个问题很严重哎。
可能就是epoll的一些问题。必现的条件是：write的数据量要比较大（如1024*1024*10） 最后会阻塞在这里： https://github.com/yyzybb537/libgo/blob/781720f7ca6cf0de2d82e5a791a9c775cbad5aa3/libgo/netio/unix/hook.cpp#L200
The code `compatible = "xlnx,axidma-chrdev";` in your device tree should be one and only one. Please check it!
1.多了个内部通道 2.采样周期比较长现在 3.采样结束有一些运算现在(DMA单次中断) 改起来麻烦 ---原始邮件--- 发件人: "cdhigh"
Hi, I reviewed the code and tested it. These are all about benchmarks. Therefore, I think that after supporting this, it is best to only display the real FPS in...
The benchmark also adds a lot of auxiliary code, as if nothing like this is needed here (at least I didn't notice it). Only showing the real one should be...
OK, I will update PR now.
@kisvegabor Done, this PR is ready to merge on CI pass. ps: some CI error, because network, can ignore it.
Yes, I think that test the time between 2 _lv_disp_refr_timer calls is all we need to do. because one frame render begin on timer `_lv_disp_refr_timer` begin, and end on disp_refr->driver->monitor_cb...
PS: If there was nothing to refresh, just record the time point. And `last_refresh_time` may be useful for other things. e.g. calculate real FPS. And, do you have a better...