By: Etienne (etienne_lorrain.delete@this.yahoo.fr), July 3, 2013 2:26 am
Room: Moderated Discussions
Patrick Chase (patrickjchase.delete@this.gmail.com) on July 2, 2013 10:01 am wrote:
> Etienne (etienne_lorrain.delete@this.yahoo.fr) on July 2, 2013 4:36 am wrote:
> > manage security by erasing newly allocated pages to processes,
>
> This isn't a HW function and can be done just as easily on a GPU as
> on a CPU (simply launch a 'bzero kernel' after every allocation...)
>
> > manage all the crappy hardware around (active waits because some version of that
> > chip do not allow two consecutive writes within N microseconds...), manage different
> > version of libraries (page loaded on demand, position independent code, dynamic
> > linking of files which can be in 10 different places in the filesystem)?
>
> Do you really think that these have significant cost for compute-oriented
> workloads as might be targeted to a GPU?
I was answering about replacing the CPU by a GPU (+ small CPU) on a desktop,
I just said that managing those sample tasks would not be quicker by a GPU.
For what I am seeing, none of desktop or embedded software I use is limited
by processor execution, time is lost in fetching loads of memory into the caches
where only few bytes will be used before the cache line is evicted again...
Or fetching a lot of memory into the cache so that a big area can be cleared
(without being read) and then the whole area is written back to external memory.
I am also wondering if the management of the dirty-page bit in software on ARM
is taking a lot of time.
It is easier to be bitten by those "hidden" things on a CPU, GPU software do
not have this kind of problem because it is a different "domain".
> Etienne (etienne_lorrain.delete@this.yahoo.fr) on July 2, 2013 4:36 am wrote:
> > manage security by erasing newly allocated pages to processes,
>
> This isn't a HW function and can be done just as easily on a GPU as
> on a CPU (simply launch a 'bzero kernel' after every allocation...)
>
> > manage all the crappy hardware around (active waits because some version of that
> > chip do not allow two consecutive writes within N microseconds...), manage different
> > version of libraries (page loaded on demand, position independent code, dynamic
> > linking of files which can be in 10 different places in the filesystem)?
>
> Do you really think that these have significant cost for compute-oriented
> workloads as might be targeted to a GPU?
I was answering about replacing the CPU by a GPU (+ small CPU) on a desktop,
I just said that managing those sample tasks would not be quicker by a GPU.
For what I am seeing, none of desktop or embedded software I use is limited
by processor execution, time is lost in fetching loads of memory into the caches
where only few bytes will be used before the cache line is evicted again...
Or fetching a lot of memory into the cache so that a big area can be cleared
(without being read) and then the whole area is written back to external memory.
I am also wondering if the management of the dirty-page bit in software on ARM
is taking a lot of time.
It is easier to be bitten by those "hidden" things on a CPU, GPU software do
not have this kind of problem because it is a different "domain".