Hi,
I have two processes running both under wine (wine-3.0 / Ubuntu 3.0-1ubuntu1) communicating with each other through a pipe ("popen"):
Process1 (stdout) -> Process2 (stdin)
Performance is important and much data is transfered through this pipe (~1GB/s) and I noticed that the "wineserver" process uses much CPU-Load when the processes communicate through the pipe. Is it possible to bypass wineserver in any manner? I cannot modificate Process1, but Process2. I tried to use a named pipe I created under Linux with "mkfifo" for the communication between the processes, but this didn't work as expected.
Regards
IPC / Pipes without wineserver
Re: IPC / Pipes without wineserver
Hello,
I have exactly the same problem. I have 2 processes running under wine. The first one produces very much amount of data (to its stdout) which has to be consumed by the 2nd one (reading from stdin). I am facing the same problem, that the wineserver process produces heavy CPU load.
I also tried to get a workaround by using a named unix pipe. The idea behind this approach was, to break down the communication to bare file access which is not handled by the wineserver. But this mechanism does not seem to work in general. Under wine the named pipe file can not be opened at all.
I got somehow a "proof of concept" that a theoretical solution exists, by using a real linux kernel fifo device file, which I symlinked into the wine filesystem. This approach could solve the problem, but is quite complex and requires additional (kernel mode!) software (e.g. https://github.com/Zildj1an/FIFO-Driver with some changes).
Is there any prossibility to get "standard" named unix pipes working? It would really be nice, if there was no need of using additional kernel modules...
Thank you and best regards
bsch
I have exactly the same problem. I have 2 processes running under wine. The first one produces very much amount of data (to its stdout) which has to be consumed by the 2nd one (reading from stdin). I am facing the same problem, that the wineserver process produces heavy CPU load.
I also tried to get a workaround by using a named unix pipe. The idea behind this approach was, to break down the communication to bare file access which is not handled by the wineserver. But this mechanism does not seem to work in general. Under wine the named pipe file can not be opened at all.
I got somehow a "proof of concept" that a theoretical solution exists, by using a real linux kernel fifo device file, which I symlinked into the wine filesystem. This approach could solve the problem, but is quite complex and requires additional (kernel mode!) software (e.g. https://github.com/Zildj1an/FIFO-Driver with some changes).
Is there any prossibility to get "standard" named unix pipes working? It would really be nice, if there was no need of using additional kernel modules...
Thank you and best regards
bsch
-
- Level 7
- Posts: 823
- Joined: Thu Aug 27, 2009 6:23 am
Re: IPC / Pipes without wineserver
Hi,
this is basically a user2user-forum, so not many people here will be able to help. You could try to ask very politely on the wine-devel list or #winehq IRC. Or file a bug report to get into contact with wine developers, though an expected lack of efficiency probably won't justify a bug report.
kind regards,
this is basically a user2user-forum, so not many people here will be able to help. You could try to ask very politely on the wine-devel list or #winehq IRC. Or file a bug report to get into contact with wine developers, though an expected lack of efficiency probably won't justify a bug report.
kind regards,