Spin buffers


#1

Hmm. Just being reading the Dr. Dobbs.

I certainly have hit the bottleneck talked about in this article…
http://www.ddj.com/dept/architect/199902669
when using the ‘thread.rb’ Queue class.

I hacked around it by creating a Queue of Arrays, but that probably
creates more garbage than I should.

This “Spin Buffer” trick looks Good…
http://www.ddj.com/dept/architect/199902669?pgno=2

Anyone have a ruby Spin Buffer implementation lying around?

John C. Phone : (64)(3) 358 6639
Tait Electronics Fax : (64)(3) 359 4632
PO Box 1645 Christchurch Email : removed_email_address@domain.invalid
New Zealand


#2

On 16.07.2007 04:20, John C. wrote:

Hmm. Just being reading the Dr. Dobbs.

I certainly have hit the bottleneck talked about in this article…
http://www.ddj.com/dept/architect/199902669
when using the ‘thread.rb’ Queue class.

What exactly makes you believe that? Can you show some code and / or
figures that back this theory?

I hacked around it by creating a Queue of Arrays, but that probably
creates more garbage than I should.

Difficult to say without seeing the code. Generally I’d say that Queue
is pretty efficient already since it’s implemented in C AFAIR.

This “Spin Buffer” trick looks Good…
http://www.ddj.com/dept/architect/199902669?pgno=2

Anyone have a ruby Spin Buffer implementation lying around?

I’d check with the RAA.

Kind regards

robert


#3

On Jul 15, 2007, at 8:20 PM, John C. wrote:

Anyone have a ruby Spin Buffer implementation lying around?

no, but i’d be inclined to try writing one on top of guy’s mmap ext -
the interface is that of string so the example should map (no pun
intended) rather well.

cheers.

a @ http://drawohara.com/


#4

On Mon, 2007-07-16 at 14:30 +0900, Robert K. wrote:

What exactly makes you believe that? Can you show some code and / or
figures that back this theory?

Seconded. Measure first, then optimize.

I hacked around it by creating a Queue of Arrays, but that probably
creates more garbage than I should.

That probably performs slower than Queue.

Difficult to say without seeing the code. Generally I’d say that Queue
is pretty efficient already since it’s implemented in C AFAIR.

Depends on the version of Ruby. It certainly is if you’re using
fastthread. If he’s not using fastthread, I’d recommend giving it a
try.

This “Spin Buffer” trick looks Good…
http://www.ddj.com/dept/architect/199902669?pgno=2

Anyone have a ruby Spin Buffer implementation lying around?

“Spin Buffers” are snake oil; they get their “speed” at the expense of
correctness – the implementation given in the DDJ article is not
threadsafe. If it were rewritten to be threadsafe, it would actually be
slower than some simpler alternatives.

-mental


#5

On Wed, 18 Jul 2007, MenTaLguY wrote:

And yes… I did, as always, measure before I optimized and my Queue
of Array trick did speed things measurably.

Depends on the version of Ruby. It certainly is if you’re using
fastthread. If he’s not using fastthread, I’d recommend giving it a
try.

I’m not using fastthread since it caused code that had been working
fine for several years to curl up and die. So I disabled it.

I’ll admit I never got to the bottom of why it did… (and no, it
wasn’t the code that used my queue of arrays trick)

This “Spin Buffer” trick looks Good…
http://www.ddj.com/dept/architect/199902669?pgno=2

Anyone have a ruby Spin Buffer implementation lying around?

“Spin Buffers” are snake oil; they get their “speed” at the expense of
correctness – the implementation given in the DDJ article is not
threadsafe.

Well that was a bit of a problem with the article… it didn’t
actually include the code so I can’t say one way or the other on
that. Any pointers as to what the incorrect bit is?

John C. Phone : (64)(3) 358 6639
Tait Electronics Fax : (64)(3) 359 4632
PO Box 1645 Christchurch Email : removed_email_address@domain.invalid
New Zealand


#6

On Mon, 2007-07-16 at 14:46 +0900, ara.t.howard wrote:

no, but i’d be inclined to try writing one on top of guy’s mmap ext -
the interface is that of string so the example should map (no pun
intended) rather well.

What does mmap have to do with communication between Ruby threads?

(and again, I’d like to emphasize that “spin buffers” are snake oil)

-mental


#7

On Wed, 2007-07-18 at 10:32 +0900, John C. wrote:

And yes… I did, as always, measure before I optimized and my Queue
of Array trick did speed things measurably.

Hmm, did you just mean sending batches of objects as arrays, rather than
sending individual objects? That would make sense and it’s a reasonable
optimization, provided you’re careful about not touching the array on
the sending side once it’s been added to the queue. I’d just finished
reading the DDJ article and was imagining something more elaborate at
first…

Depends on the version of Ruby. It certainly is if you’re using
fastthread. If he’s not using fastthread, I’d recommend giving it a
try.

I’m not using fastthread since it caused code that had been working
fine for several years to curl up and die. So I disabled it.

It might be worth looking into why – if it doesn’t work with
fastthread, it’s unlikely to work with future versions of Ruby
(including 1.8.6 or later) or alternate Ruby implementations like JRuby.

Absent fastthread bugs (which do crop up occasionally, though rarely at
this point), there are two main sources of problems:

  1. Code that tries to manipulate the internal implementation of
    thread.rb objects (Mutex, ConditionVariable, Queue, etc); not always a
    bug, but it isn’t portable or future-proof.

  2. Code which has existing concurrency bugs which show up more clearly
    when the scheduling behavior changes.

“Spin Buffers” are snake oil; they get their “speed” at the expense of
correctness – the implementation given in the DDJ article is not
threadsafe.

Well that was a bit of a problem with the article… it didn’t
actually include the code so I can’t say one way or the other on
that. Any pointers as to what the incorrect bit is?

The code is included in the source archive on the DDJ ftp site as
spin.txt (and spin.zip, which has his test harness):

ftp://66.77.27.238/sourcecode/ddj/2007/0707.zip

It’s not just a one-line bug, but a systemic problem: the author assumes
that the only reason he needs synchronization is to prevent two threads
from modifying the same data at the same time (hence the elaborate dance
with the buffers); in reality, however, it’s also important to protect
the code from compiler (and CPU) optimizations which will alter its
behavior in undesirable ways if it is used in a multi-threaded context
(e.g. it becomes possible for the reader to see the writer’s addition of
the object but see the object in an uninitalized state!). He’s
basically trying to get a performance “free lunch” by not using
synchronization.

Concurrency bugs are notoriously hard to find through testing, and it
doesn’t help that the author only ever tested on hardware which is
extremely forgiving about concurrency (a single hyperthreaded x86 CPU).
It also didn’t help that all his test did was count how many objects
were read from the queue.

-mental


#8

On Wed, 2007-07-18 at 11:29 +0900, MenTaLguY wrote:

in reality, however, it’s also important to protect
the code from compiler (and CPU) optimizations which will alter its
behavior in undesirable ways if it is used in a multi-threaded context
(e.g. it becomes possible for the reader to see the writer’s addition of
the object but see the object in an uninitalized state!).

It is worth noting that these are not issues if you’re only writing for
Ruby 1.8, which has neither an optimizing compiler nor native threads
(where instruction ordering and cache effects become a consideration).

So, does that mean a Spin Buffer implementation would theoretically be
okay for Ruby 1.8? Maybe as far as those issues go. But note that the
author himself identifies a number of issues with spin buffers when the
readers and writers get out of sync, including poor performance and the
reader getting “stuck” on a non-empty buffer if it hits a buffer
boundary when no writers are writing. Those are actually inherent to
the data structure’s design.

-mental


#9

On Wed, 2007-07-18 at 10:32 +0900, John C. wrote:

I’m not using fastthread since it caused code that had been working
fine for several years to curl up and die. So I disabled it.

At any rate, I think your best option is something like fastthread’s
Queue. It shouldn’t be too hard to change the code to work with it, and
it’s possible that you’ve uncovered a bug in that “working” code which
really needs to be fixed.

-mental