On Fri, Mar 24, 2017 at 6:09 AM, Vincent Torri <[email protected]> wrote: [...] >>> TL;DR: we'll never benefit from the thread pool benefits. >> >> it will if we have a lot of use of connections using ecore_con_local >> or ecore_ipc > > also, it does not consume HANDLE's,
what you mean? > contrary to what you will use with > the design you want to choose (which is btw my first implementation of > ecore_con_local, you're re-writing it...) Yes, I checked the old win32 local file when I did some grep... but it's slightly different as it uses the hEvent instead of notifying on our own with SetEvent(). > and will be faster. The > Windows port is already largely slower than the linux one. I've chosen > I/O completion port for that very reason here is the part that I think I couldn't explain, it doesn't matter the port speed if the user is not sending stuff to it. it's like having a 1Gbps network but the card is only capable of 10bmps. this is because we must go back to main thread to get more data to send. Same for read, we only read up to BUFSIZE (I'm using 16kb), then we call the user... until that is done, the kernel will idle... -- Gustavo Sverzut Barbieri -------------------------------------- Mobile: +55 (16) 99354-9890 ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ enlightenment-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/enlightenment-devel
