On Sat, Nov 17, 2018 at 5:46 AM Paul E. McKenney <paulmck@xxxxxxxxxxxxx> wrote:
>
> On Sat, Nov 17, 2018 at 12:01:11AM +0900, Akira Yokosawa wrote:
> > (Cc: Paul's latest address)
> > Hi Junchang,
> >
> > On 2018/11/16 16:19:57 +0800, Junchang Wang wrote:
> > > Signed-off-by: Junchang Wang <junchangwang@xxxxxxxxx>
> > > ---
> > > Hi list,
> > >
> > > There are a few typos in Section Partitioning. Please take a look.
> > >
> > > Thanks,
> > > --Junchang
> > >
> > > --
> > > SMPdesign/SMPdesign.tex | 4 ++--
> > > 1 file changed, 2 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/SMPdesign/SMPdesign.tex b/SMPdesign/SMPdesign.tex
> > > index 7b4f6eb..562c057 100644
> > > --- a/SMPdesign/SMPdesign.tex
> > > +++ b/SMPdesign/SMPdesign.tex
> > > @@ -936,7 +936,7 @@ We could simply assign each CPU one gigabyte of memory, and allow
> > > each CPU to access its own private chunk of memory, without the
> > > need for locking and its complexities and overheads.
> > > Unfortunately, this simple scheme breaks down if an algorithm happens
> > > -to have CPU~0 allocate all of the memory and CPU~1 the free it, as
> > > +to have CPU~0 allocate all of the memory and CPU~1 free it, as
> >
> > I guess Paul's intention was:
> >
> > +to have CPU~0 allocate all of the memory and CPU~1 then free it, as
> >
> > ???
>
> You are right, that was my intention, but there significant room for
> improvement. Please see below for an updated version of Junchang's
> patch. Thoughts?
>
Hi Paul and Akira,
The new patch looks good to me. Thanks a lot for the help.
Thanks,
--Junchang
> > > would happen in a simple producer-consumer workload.
> > >
> > > The other extreme, code locking, suffers from excessive lock contention
> > > @@ -1017,7 +1017,7 @@ smaller than the number of non-\co{NULL} pointers.
> > > The allocation function \co{memblock_alloc()} may be seen in
> > > Listing~\ref{lst:SMPdesign:Allocator-Cache Allocator Function}.
> > > Line~\lnref{pick} picks up the current thread's per-thread pool,
> > > -and line~8 check to see if it is empty.
> > > +and line~\lnref{chk:empty} checks to see if it is empty.
> >
> > Nice catch!
>
> Indeed, thank you both!!!
>
> > Thanks, Akira
> >
> > >
> > > If so, lines~\lnref{ack}-\lnref{rel} attempt to refill it
> > > from the global pool
>
> ------------------------------------------------------------------------
>
> commit 81c18b03d1aa2d3edc1dc28867e1f84413eb31e3
> Author: Junchang Wang <junchangwang@xxxxxxxxx>
> Date: Fri Nov 16 16:19:57 2018 +0800
>
> SMPdesign: Fix typos
>
> Signed-off-by: Junchang Wang <junchangwang@xxxxxxxxx>
> Signed-off-by: Paul E. McKenney <paulmck@xxxxxxxxxxxxx>
> [ paulmck: Rework to clarify original wording. ]
>
> diff --git a/SMPdesign/SMPdesign.tex b/SMPdesign/SMPdesign.tex
> index 7b4f6eb2ba71..fe8f445851e3 100644
> --- a/SMPdesign/SMPdesign.tex
> +++ b/SMPdesign/SMPdesign.tex
> @@ -932,12 +932,11 @@ data ownership to this problem---simply carve up memory so that
> each CPU owns its share.
> For example, suppose that a system with two CPUs has two gigabytes
> of memory (such as the one that I am typing on right now).
> -We could simply assign each CPU one gigabyte of memory, and allow
> -each CPU to access its own private chunk of memory, without the
> -need for locking and its complexities and overheads.
> -Unfortunately, this simple scheme breaks down if an algorithm happens
> -to have CPU~0 allocate all of the memory and CPU~1 the free it, as
> -would happen in a simple producer-consumer workload.
> +We could simply assign each CPU one gigabyte of memory, and allow each
> +CPU to allocate from its own gigabyte, without the need for locking and
> +its complexities and overheads.
> +Unfortunately, this scheme fails when CPU~0 only allocates memory and
> +CPU~1 only frees it, as happens in simple producer-consumer workloads.
>
> The other extreme, code locking, suffers from excessive lock contention
> and overhead~\cite{McKenney93}.
> @@ -1017,7 +1016,7 @@ smaller than the number of non-\co{NULL} pointers.
> The allocation function \co{memblock_alloc()} may be seen in
> Listing~\ref{lst:SMPdesign:Allocator-Cache Allocator Function}.
> Line~\lnref{pick} picks up the current thread's per-thread pool,
> -and line~8 check to see if it is empty.
> +and line~\lnref{chk:empty} checks to see if it is empty.
>
> If so, lines~\lnref{ack}-\lnref{rel} attempt to refill it
> from the global pool
>
