| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9
Comments:
<0> the randomness is not in the number of nodes
<0> it is in the response time
<1> that's not possible, since the gaussian distribution is absolutely continuous, while any distribution on a finite subset of the real numbers will be discrete.
<0> Well, explain why it IS in fact, gaussian ?
<1> it's not. see above.
<0> build a ****ing machine and find out.
<2> Google for Yousef scheduling _Aegis_.
<2> Exactly the same scenario, with "random" scheduling, it will elaborate on the exclusivity of ordering.
<2> _Aegis_: Yousef scheduling GWU or something.
<3> how can i see if my code has memory leacks?
<1> valgrind ./yourcode
<3> win
<4> lose.
<3> lool
<5> I have a function that allocates a block of memory, stores something in that block then returns it. I am trying to figure out how to free the block of memory. The main function can't see the block so I can't do it in there. I can't free it in the function which allocates it because then it has nothing to return. How do I take care of this?
<3> be creative.
<5> excellent, i'll try that.
<5> :D
<3> lol
<5> I know I could copy the contents to a variable, free the block and then return the variable but it seems like there would be a better way
<6> have you thought of returning the memory address of the block
<5> How would that help?
<6> i would recommend looking up pointers
<5> If i returned the address after freeing the block then couldn't another program change the block of memory before the next function looks up whatever is in the address?
<6> well you want to return the data in your block of memory, so allocate the memory, return the address, free it in main when your done
<6> then hug a panda
<5> Oh, that makes sense
<6> hugging pandas? i agree
<5> no, the one before that
<5> heh
<6> hm
<6> what are you watering
<5> hmm
<5> if I return an address to a char * is the return type of the function char **?
<7> amazon10x no
<7> it's char 8
<7> char *
<6> i like my char 8s
<7> hehehehe
<5> ok
<4> amazon10x: Yes, if what you actually say is what you do.
<4> ie, if you have char *a; and return &a;.
<6> has everyone been drinking
<6> oops
<5> hund, yes, that is what I am doing
<7> terminus no I haven't ;)
<5> so I do make the function type = char **
<6> i didn't mean to say that i t just pooped out of my head
<5> does your head poop a lot?
<6> amazon10x: char *func(){ char *myChar = (char *)malloc(char *); mychar = "poop"; return mychar; }
<7> amazon10x if you do char *a; return &a; then yes
<4> Don't cast malloc().
<4> Uh.
<7> malloc returns void
<4> terminus: That doesn't make sense at all.
<4> malloc() returns (void *).
<6> everything makes more sense after 4 shots of vodka
<5> char ** func(){ char *ipaddy = "blahblah"; return &ipaddy}
<5> that is what i have
<7> amazon10x it's good
<5> excellent
<7> go on as usual
<7> ;-)
<5> ooh
<5> just tested
<5> and it worked
<7> well, this method is also a way, though I prefer to read then do
<7> ;-)
<4> Of course, why do you need a pointer to a pointer of an immutable value?
<6> you could also do char *func(){char *ipaddy="blahblah"; return ipaddy;}
<5> i want to return the address so I can use it in a later function and then free it
<5> because the data is allocated my malloc()
<5> so it needs to be freed
<7> amazon10x you think wisely
<4> Right.
<5> I_v0: terminus told me to do it that way, :D
<5> I asked earlier
<6> see i told you that everything makes more sense after vodka
<6> brb need more
<7> amazon10x also consider that when you exit child process memory will be freed way or another
<0> my scheduler seems to work .:)
<7> _Aegis_ what license you gaved him ?
<7> it*
<0> license ?
<7> hund freeing depends on case, we both know it
<0> all it does at the moment is pick the node with the lowest load average
<7> _Aegis_ in other words, can I have source?
<0> ah
<7> ;-)
<0> not yet. When I have it all pretty, you can get a copy
<7> it's always fascinating to see how others write something ;)
<0> it is really simple
<0> it uses a shell script to get load average
<0> then it parses the result and choses a node
<7> I see
<0> and uses system() to execute an ssh command on the proper node
<0> there is no queue at present
<7> well, then it's propably not so complicated program
<0> correct, it is very simple
<0> I'm more interested in a capability than code at the moment.
<0> I've got a simple parallel shell utility that is GPL
<7> _Aegis_ but this can be done making a cluster
<0> well, this is for a cluster
<7> then make on top of clusters another mega cluster
<7> ;-)
<0> good ides. :)
<0> maybe we can connect a lot of clusters together
<7> as usually, on linux everything can be done
<7> ;>
<0> yah
<5> I have a beowulf cluster of toasters running windows 95
<0> well this is not linux. this is freebsd, but the code is so simple it will work on linux too
<5> with a scheduler
<5> :p
<0> which is the idea
<7> amazon10x hehehe
<8> that makes no sense amazon10x, nobody has written a toaster scheduler
<0> cool beans amazon. we can all have brunch in parallele
<0> IT CAN BE DONE!
<5> Kardos: i wrote a toaster scheduler
<4> I was a toaster scheduler once.
<8> ah, well in that case... good job
<4> By hitting my sister to use the toaster first.
<7> hund hehehe
<6> at least we can all make toasts
<0> well I can only imagione the complexity of a toaster queue
<8> i can see it now, rfc2324 reborn as a toaster queueing system
<0> I wonder if we can get a toaster with an infiniband interconnect
<8> myrinet interconnect!
<0> it might make a good april 1 rfc
<0> I'm burned out on myrinet
<0> I had a cluster with myrinet problems , it was horrid
<8> infiniband faster, or just less problems?
<0> more bandwidth easier than myrinet
<0> but myrinet is way low latency
<0> you get better benchmarks with myrinet
<8> it seems to me, that it depends how much and how often you need to exchange stuff
<8> so theres no 'better' in general, its program specific
<0> well, myrinet replaced over 200 bad nics in our cluster.
<0> it was a major pain in the butt
<0> they made it good, but home boy had to change them all
<8> mmmmm
<8> haha
<0> it took a whle
<0> while
<0> but once we gotthe bad nics replaced things worked great
<0> thye had a bad batch of laser transcievers
<0> they replaced all of the myrinet cards in clusters at JPL and Goddard
<0> for free
Return to
#c
or
Go to some related
logs:
bevarian pretzel
#stocks
#asp
quotes on lazines
#beginner
#gentoo
resocialized person brainwashed
#gamedev
LEARN HOW TO USE YOUR WANG
#computers
|
|