Re: [Numpy-discussion] Creating a sine wave with exponential decay

[Francesc Alted]

(disclosure: I have been a long time maintainer of numexpr)

An important thing to be noted is that, besides avoiding creating big
temporaries, numexpr has support for multi-threading out of the box and
Intel VML for optimal evaluation times on Intel CPUs.  For choosing your
best bet, there is no replacement for experimentation:

https://gist.github.com/FrancescAlted/203be8a44d02566f31dae11a22c179f3

I have no time now to check for memory consumption, but you can expect
numexpr and Numba consuming barely the same amount of memory.  Performance
wise things are quite different, but this is probably due to my
inexperience with Numba (in particular, paralellism does not seem to work
for this example in Numba 0.45, but I am not sure why).

Cheers!


  Probably numba has this too, but my attempts to use parallelism failed

Missatge de Sebastian Berg  del dia dc., 24 de
jul. 2019 a les 1:32:

> On Tue, 2019-07-23 at 13:38 -0500, Stanley Seibert wrote:
> > (Full disclosure: I work on Numba...)
> >
> > Just to note, the NumPy implementation will allocate (and free) more
> > than 2 arrays to compute that expression.  It has to allocate the
> > result array for each operation as Python executes.  That expression
> > is equivalent to:
> >
>
> That is mostly true, although – as Hameer mentioned – on many platforms
> (gcc compiler is needed I think) a bit of magic happens.
>
> If an array is temporary, the operation is replaced with an in-place
> operation for most python operators calls. For example:
> `-abs(arr1 * arr2 / arr3 - arr4)`
> should only create a single new array and keep reusing it in many
> cases[0]. You would achieve similar things with `arr1 *= arr2`
> manually.
>
> Another thing is that numpy will cache some arrays, so that the
> allocation cost itself may be avoided in many cases.
>
> NumPy does no "loop fusing", i.e. each operation is finished before the
> next is started. In many cases, with simple math loop fusing can give a
> very good speedup (which is wher Numba or numexpr come in). Larger
> speedups are likely if you have large arrays and very simple math
> (addition). [1]
>
> As Stanley noted, you probably should not worry too much about it. You
> have `exp`/`sin` in there, which are slow by nature. You can try, but
> it is likely that you simply cannot gain much speed there.
>
> Best,
>
> Sebastian
>
>
> [0] It requires that the shapes all match and that the result arrays
> are obviously temporary.
> [1] For small arrays overheads may be avoided using tools such as
> numba, which can help a lot as well. If you want to use multiple
> threads for a specific function that may also be worth a look.
>
> > s1 = newfactor * x
> > s2 = np.exp(s1)
> > s3 = np.sin(x)
> > y = s3 * s2
> >
> > However, memory allocation is still pretty fast compared to special
> > math functions (exp and sin), which dominate that calculation.  I
> > find this expression takes around 20 milliseconds for a million
> > elements on my older laptop, so that might be negligible in your
> > program execution time unless you need to recreate this decaying
> > exponential thousands of times.  Tools like Numba or numexpr will be
> > useful to fuse loops so you only do one allocation, but they aren't
> > necessary unless this becomes the bottleneck in your code.
> >
> > If you are getting started with NumPy, I would suggest not worrying
> > about these issues too much, and focus on making good use of arrays,
> > NumPy array functions, and array expressions in your code.  If you
> > have to write for loops (if there is no good way to do the operation
> > with existing NumPy functions), I would reach for something like
> > Numba, and if you want to speed up complex array expressions, both
> > Numba and Numexpr will do a good job.
> >
> >
> > On Tue, Jul 23, 2019 at 10:38 AM Hameer Abbasi <
> > einstein.edi...@gmail.com> wrote:
> > > Hi Ram,
> > >
> > > No, NumPy doesn’t have a way. And it newer versions, it probably
> > > won’t create two arrays if all the dtypes match, it’ll do some
> > > magic to re use the existing ones, although it will use multiple
> > > loops instead of just one.
> > >
> > > You might want to look into NumExpr or Numba if you want an
> > > efficient implementation.
> > >
> > > Get Outlook for iOS
> > >
> > > From: NumPy-Discussion <
> > > numpy-discussion-bounces+einstein.edison=gmail@python.org> on
> > > behalf of Ram Rachum 
> > > Sent: Tuesday, July 23, 2019 7:29 pm
> > > To: numpy-discussion@python.org
> > > Subject: [Numpy-discussion] Creating a sine wave with exponential
> > > decay
> > >
> > > Hi everyone! Total Numpy newbie here.
> > >
> > > I'd like to create an array with a million numbers, that has a sine
> > > wave with exponential decay on the amplitude.
> > >
> > > In other words, I want the value of each cell n to be sin(n)
> > >  * 2 ** (-n * factor).
> > >
> > > What would be the most efficient way to do that?
> > >
> > > Someone suggested I do something like this:
> > >
> > > y = np.sin(x) 

[Numpy-discussion] The FREE Demystifying Data Science Conference is happening again! July 30-31!!!

[Michael Elnajami]

Metis  is hosting our 3rd annual *FREE *Demystifying
Data Science Conference
 July 30-31 from
10am - 5pm ET!

There will be 16 interactive data science talks and 6 workshops led by
industry-leading speakers. Experience interactivity via real-time chat with
world-wide audience, polling, and social sharing. Plus, get access to
recordings of every presentation and workshop.

I know many of you NumPy users also are data scientists, so please check
this out! This is a great opportunity to learn and connect with like minded
people!

Day 1 - July 30 will be for aspiring data scientists.

Day 2 - July 31 will be for business leaders and practitioners

Registration page: https://www.thisismetis.com/demystifying-data-science
___
NumPy-Discussion mailing list
NumPy-Discussion@python.org
https://mail.python.org/mailman/listinfo/numpy-discussion

Re: [Numpy-discussion] Season of Docs applications

[Ralf Gommers]

Hi all,

Here is a short update on Season of Docs. We are very happy and impressed
with the number and quality of tech writers who applied to NumPy and SciPy.
We decided we would really like to accept more people than the number of
slots Google is likely to give us. We have found some more funding and
mentoring bandwidth to be able to do so. After Google announces which
candidates have been accepted, we will reach out to the next people on our
shortlist and offer them the same mentoring and funding for their proposed
projects as the accepted GSoD writers.

I'm really looking forward to an interesting and productive fall working
with several of you. Stay tuned!

Cheers,
Ralf



On Wed, Jul 17, 2019 at 5:12 PM Ralf Gommers  wrote:

> Hi GSoD applicants,
>
> To start with, I want to thank all of you who applied! There's a lot of
> interest in helping improve NumPy's documentation and web presence, which
> is awesome to see. During the application phase, we have interacted with
> some of you already, and I will reach out to others over the next few days.
> I'm just writing this email to say thank you and to confirm that yes we did
> receive your application. Due to the volume of applications, it will be
> difficult to reach out personally to all of you (I'll do my best though!),
> hence this email.
>
> Really looking forward to what will happen next!
>
> Cheers,
> Ralf
>
>
___
NumPy-Discussion mailing list
NumPy-Discussion@python.org
https://mail.python.org/mailman/listinfo/numpy-discussion

Re: [Numpy-discussion] Creating a sine wave with exponential decay

[Juan Nunez-Iglesias]

Hi Francesc,

Those numbers are really eye-popping! But the formatting of the code as a 
string still bugs me a lot. Asking this as a totally naive user, do you know 
whether PEP523  (adding a frame 
evaluation API) would allow numexpr to have a more Pythonic syntax? eg. 

with numexpr:
y = np.sin(x) * np.exp(newfactor * x)

?

Juan.


> On 24 Jul 2019, at 6:39 pm, Francesc Alted  wrote:
> 
> (disclosure: I have been a long time maintainer of numexpr)
> 
> An important thing to be noted is that, besides avoiding creating big 
> temporaries, numexpr has support for multi-threading out of the box and Intel 
> VML for optimal evaluation times on Intel CPUs.  For choosing your best bet, 
> there is no replacement for experimentation:
> 
> https://gist.github.com/FrancescAlted/203be8a44d02566f31dae11a22c179f3 
> 
> 
> I have no time now to check for memory consumption, but you can expect 
> numexpr and Numba consuming barely the same amount of memory.  Performance 
> wise things are quite different, but this is probably due to my inexperience 
> with Numba (in particular, paralellism does not seem to work for this example 
> in Numba 0.45, but I am not sure why).
> 
> Cheers!
> 
> 
>   Probably numba has this too, but my attempts to use parallelism failed
> 
> Missatge de Sebastian Berg  > del dia dc., 24 de jul. 2019 a les 1:32:
> On Tue, 2019-07-23 at 13:38 -0500, Stanley Seibert wrote:
> > (Full disclosure: I work on Numba...) 
> > 
> > Just to note, the NumPy implementation will allocate (and free) more
> > than 2 arrays to compute that expression.  It has to allocate the
> > result array for each operation as Python executes.  That expression
> > is equivalent to:
> > 
> 
> That is mostly true, although – as Hameer mentioned – on many platforms
> (gcc compiler is needed I think) a bit of magic happens.
> 
> If an array is temporary, the operation is replaced with an in-place
> operation for most python operators calls. For example:
> `-abs(arr1 * arr2 / arr3 - arr4)`
> should only create a single new array and keep reusing it in many
> cases[0]. You would achieve similar things with `arr1 *= arr2`
> manually.
> 
> Another thing is that numpy will cache some arrays, so that the
> allocation cost itself may be avoided in many cases.
> 
> NumPy does no "loop fusing", i.e. each operation is finished before the
> next is started. In many cases, with simple math loop fusing can give a
> very good speedup (which is wher Numba or numexpr come in). Larger
> speedups are likely if you have large arrays and very simple math
> (addition). [1]
> 
> As Stanley noted, you probably should not worry too much about it. You
> have `exp`/`sin` in there, which are slow by nature. You can try, but
> it is likely that you simply cannot gain much speed there.
> 
> Best,
> 
> Sebastian
> 
> 
> [0] It requires that the shapes all match and that the result arrays
> are obviously temporary.
> [1] For small arrays overheads may be avoided using tools such as
> numba, which can help a lot as well. If you want to use multiple
> threads for a specific function that may also be worth a look.
> 
> > s1 = newfactor * x
> > s2 = np.exp(s1)
> > s3 = np.sin(x)
> > y = s3 * s2
> > 
> > However, memory allocation is still pretty fast compared to special
> > math functions (exp and sin), which dominate that calculation.  I
> > find this expression takes around 20 milliseconds for a million
> > elements on my older laptop, so that might be negligible in your
> > program execution time unless you need to recreate this decaying
> > exponential thousands of times.  Tools like Numba or numexpr will be
> > useful to fuse loops so you only do one allocation, but they aren't
> > necessary unless this becomes the bottleneck in your code.
> > 
> > If you are getting started with NumPy, I would suggest not worrying
> > about these issues too much, and focus on making good use of arrays,
> > NumPy array functions, and array expressions in your code.  If you
> > have to write for loops (if there is no good way to do the operation
> > with existing NumPy functions), I would reach for something like
> > Numba, and if you want to speed up complex array expressions, both
> > Numba and Numexpr will do a good job.
> > 
> > 
> > On Tue, Jul 23, 2019 at 10:38 AM Hameer Abbasi <
> > einstein.edi...@gmail.com > wrote:
> > > Hi Ram,
> > > 
> > > No, NumPy doesn’t have a way. And it newer versions, it probably
> > > won’t create two arrays if all the dtypes match, it’ll do some
> > > magic to re use the existing ones, although it will use multiple
> > > loops instead of just one.
> > > 
> > > You might want to look into NumExpr or Numba if you want an
> > > efficient implementation.
> > > 
> > > Get Outlook for iOS
> > >  
> > > From: NumPy-Discussion <
> > > numpy-discussion-bounces+ei