12/23 OS question: what does it mean when SGI said they were adding
asyncronous I/O support for Linux?
\- it means they are trying to boost their stock by
becoming a linux company. --psb
\_ Answered twice already. Deleted by the motd censors.
read ~mehlhaff/tmp/motd,v to escape the censorship.
\_ that SGI is trying to boost their stock is a pretty stupid
answer. Like I said, this is an OS question so unless
you give a technical answer with sound reasoning then
you might as well not answer at all. I didn't delete the
answers, by the way.
\_ You deleted all the recent motd entries fuckhead.
\_ There was a technical answer as well. It said:
\_ They're adding the POSIX.4 AIO API that programs like databases
like to use to issue a read or write command, continue working,
and then check later to see if it finished, instead of the
traditional model of the process being suspended by the OS while
it fetches the data from disk.
\_ Isn't this the same as creating a new thread, doing a read
operation, and then doing a thread join? I don't see how
you can continue execution in something like
read(fd,buf,10);
printf("%s",buf);
since the printf statement is dependent on the readoperation
finishing.
\_ well duh, if you're writing code to use asynchronous IO, you
write the code to not do anything that depends on an IO
request until after you have confirmed that the IO request
has completed.
\_ Yes, you could do it with threads, but the AIO API is more
efficient and doesn't require you to make a non-threaded
program MT-safe.
\_ [Summary: Linux takes another small step into the world of Real
OS's] |
http://csua.com/feed/