| Re: Accelerating a one-line program doing matrix multiplication [message #72662] |
Wed, 29 September 2010 03:48  |
Axel Martínez
Messages: 22
Registered: June 2010
|
Junior Member |
|
|
I have to admit that I did not understand this proposed use of
REPLICATE:
>> sometimes (replicate({temp:input},newsize)).(0) is faster then rebin
But it brought me a related question in mind: does IDL have a
"REPLICATE" function for vectors instead of scalar values? I am using
REBIN, but REBIN is thought for more advanced uses and probably
suboptimal for a "replicate-like" use... right?
|
|
|
|
| Re: Accelerating a one-line program doing matrix multiplication [message #72663 is a reply to message #72662] |
Wed, 29 September 2010 03:22  |
Axel Martínez
Messages: 22
Registered: June 2010
|
Junior Member |
|
|
On Sep 29, 8:57 am, chris <rog...@googlemail.com> wrote:
> On 28 Sep., 15:10, nata <bernat.puigdomen...@gmail.com> wrote:
>
>
>
>> Concatenation is a very slow action in IDL and, if you are copying
>> memory, the time of computation increases...
>> If v0, v1, v2 and v3 are each of them 3-element vectors then you will
>> not see the difference. TEMPORARY function is great when you are
>> copying large arrays. I think you can not improve your code because
>> the problem is the matrix multiplication and you can not change that.
>> Try putting timers to see what's the time to compute each instruction.
>
>> tt=SYSTIME(/SEC)
>> aux=[[v1],[v2],[v3]]
>> PRINT, SYSTIME(/SEc)-tt
>
>> tt=SYSTIME(/SEC)
>> aux=aux # vc
>> PRINT, SYSTIME(/SEC)-tt
>
>> etc.
>
>> Cheers,
>> nata
>
>> On Sep 28, 2:17 am, Axel M <axe...@gmail.com> wrote:
>
>>> On 27 Sep., 15:31, nata <bernat.puigdomen...@gmail.com> wrote:
>
>>>> You can use the TEMPORARY function if you can set the input to
>>>> undefined...
>>>> When you do [[v1],[v2],[v3]] you are duplicating data. v1, v2 and v3
>>>> are copied and you are not conserving memory.
>
>>>> You could try:
>>>> RETURN, [[TEMPORARY(v1)],[TEMPORARY(v2)],[TEMPORARY(v3)]] # vc +
>>>> REBIN(v0, SIZE(vc, /DIMENSIONS))
>
>>>> Cheers,
>>>> nata
>
>>> Thanks nata.
>
>>> v0, v1, v2 and v3 are each of them 3-element vectors. I can add that
>>> but, as I understand it, it will only save the place of 12 floating
>>> values in memory (48 bytes?).
>
>>> But I am happy that you did not see any other obvious thing. I started
>>> feeling depressed seeing that I am not being able to improve this
>>> single line of code... maybe it is ok, and the whole thing is just
>>> slow...? ahh.
>
> Yes, and you can substitute some of your calls:
>
> aux #= vc (makes no copy of aux as far as i know)
> invert -> la_invert (much much speedier)
> sometimes (replicate({temp:input},newsize)).(0) is faster then rebin
> exchange ## with matrix_multiply
>
> Cheers
>
> CR
Thanks a lot for all suggestions!
I have tried the different ideas. Unfortunately no improvements. Not
even the /SAMPLE in the rebin call helped, which I thought could make
a difference.
So I guess the matrix multiplication is just too computationally
intensive to get fast results.
What I also thought is that the method is fast enough when only a few
points are given, but very time-consuming when "vc" contains all
coordinates of the images, and they are sorted like [[0,0,0], [0,0,1],
[0, 0, 2], [0, 1, 0], ... [2, 2, 1], [2, 2, 2]]. Maybe there is some
way to use this property of vc to accelerate it. Like multiplying the
v1, v2, v3 by INDGEN and then somehow adding them smartly... I will
give it a try.
|
|
|
|
| Re: Accelerating a one-line program doing matrix multiplication [message #72665 is a reply to message #72663] |
Tue, 28 September 2010 23:57 |
rogass
Messages: 200
Registered: April 2008
|
Senior Member |
|
|
On 28 Sep., 15:10, nata <bernat.puigdomen...@gmail.com> wrote:
> Concatenation is a very slow action in IDL and, if you are copying
> memory, the time of computation increases...
> If v0, v1, v2 and v3 are each of them 3-element vectors then you will
> not see the difference. TEMPORARY function is great when you are
> copying large arrays. I think you can not improve your code because
> the problem is the matrix multiplication and you can not change that.
> Try putting timers to see what's the time to compute each instruction.
>
> tt=SYSTIME(/SEC)
> aux=[[v1],[v2],[v3]]
> PRINT, SYSTIME(/SEc)-tt
>
> tt=SYSTIME(/SEC)
> aux=aux # vc
> PRINT, SYSTIME(/SEC)-tt
>
> etc.
>
> Cheers,
> nata
>
> On Sep 28, 2:17 am, Axel M <axe...@gmail.com> wrote:
>
>
>
>> On 27 Sep., 15:31, nata <bernat.puigdomen...@gmail.com> wrote:
>
>>> You can use the TEMPORARY function if you can set the input to
>>> undefined...
>>> When you do [[v1],[v2],[v3]] you are duplicating data. v1, v2 and v3
>>> are copied and you are not conserving memory.
>
>>> You could try:
>>> RETURN, [[TEMPORARY(v1)],[TEMPORARY(v2)],[TEMPORARY(v3)]] # vc +
>>> REBIN(v0, SIZE(vc, /DIMENSIONS))
>
>>> Cheers,
>>> nata
>
>> Thanks nata.
>
>> v0, v1, v2 and v3 are each of them 3-element vectors. I can add that
>> but, as I understand it, it will only save the place of 12 floating
>> values in memory (48 bytes?).
>
>> But I am happy that you did not see any other obvious thing. I started
>> feeling depressed seeing that I am not being able to improve this
>> single line of code... maybe it is ok, and the whole thing is just
>> slow...? ahh.
Yes, and you can substitute some of your calls:
aux #= vc (makes no copy of aux as far as i know)
invert -> la_invert (much much speedier)
sometimes (replicate({temp:input},newsize)).(0) is faster then rebin
exchange ## with matrix_multiply
Cheers
CR
|
|
|
|
| Re: Accelerating a one-line program doing matrix multiplication [message #72680 is a reply to message #72665] |
Tue, 28 September 2010 08:31 |
Jeremy Bailin
Messages: 618
Registered: April 2008
|
Senior Member |
|
|
On Sep 27, 5:18 am, Axel M <axe...@gmail.com> wrote:
> Hi all,
>
> I wrote a one-line function to convert a list of points from "voxel
> coordinates" (image coordinates) to "real coordinates" (physical
> coordinates):
>
> ;input: the points "vc", the spatial origin of an image v0 and its x,
> y, and z orientation vectors (v1,v2,v3).
> FUNCTION vc2rc, v0,v1,v2,v3,vc
> RETURN, [[v1],[v2],[v3]] # vc + REBIN(v0, SIZE(vc, /DIMENSIONS))
> END
>
> For example, I give the image coordinate [8,1,0] and I want as output
> something like [34.25, 4.12, 0], indicating the location of this voxel
> in space. And the same thing but, instead of having one input point,
> having several millions.
>
> The function looks simple to me and it works great. BUT, for large
> images (e.g. 500x500x200 voxels), it is terribly slow and uses way too
> much memory... Am I doing something wrong, could I save speed
> somewhere? I guess there should be some way to accelerate it, but I am
> not able to see how...
>
> I also have the opposite function, in my opinion also too slow (though
> faster than the other)...
>
> FUNCTION rc2vc_round, v0,v1,v2,v3,rc
> RETURN, ROUND((rc - REBIN(v0, SIZE(rc, /DIMENSIONS))) ## INVERT([[v1],
> [v2],[v3]]))
> END
>
> I would be really grateful for any clue!
You can try adding /SAMPLE to the REBIN call, but I suspect that it's
not the bottleneck.
-Jeremy.
|
|
|
|
| Re: Accelerating a one-line program doing matrix multiplication [message #72681 is a reply to message #72680] |
Tue, 28 September 2010 06:10 |
natha
Messages: 482
Registered: October 2007
|
Senior Member |
|
|
Concatenation is a very slow action in IDL and, if you are copying
memory, the time of computation increases...
If v0, v1, v2 and v3 are each of them 3-element vectors then you will
not see the difference. TEMPORARY function is great when you are
copying large arrays. I think you can not improve your code because
the problem is the matrix multiplication and you can not change that.
Try putting timers to see what's the time to compute each instruction.
tt=SYSTIME(/SEC)
aux=[[v1],[v2],[v3]]
PRINT, SYSTIME(/SEc)-tt
tt=SYSTIME(/SEC)
aux=aux # vc
PRINT, SYSTIME(/SEC)-tt
etc.
Cheers,
nata
On Sep 28, 2:17 am, Axel M <axe...@gmail.com> wrote:
> On 27 Sep., 15:31, nata <bernat.puigdomen...@gmail.com> wrote:
>
>> You can use the TEMPORARY function if you can set the input to
>> undefined...
>> When you do [[v1],[v2],[v3]] you are duplicating data. v1, v2 and v3
>> are copied and you are not conserving memory.
>
>> You could try:
>> RETURN, [[TEMPORARY(v1)],[TEMPORARY(v2)],[TEMPORARY(v3)]] # vc +
>> REBIN(v0, SIZE(vc, /DIMENSIONS))
>
>> Cheers,
>> nata
>
> Thanks nata.
>
> v0, v1, v2 and v3 are each of them 3-element vectors. I can add that
> but, as I understand it, it will only save the place of 12 floating
> values in memory (48 bytes?).
>
> But I am happy that you did not see any other obvious thing. I started
> feeling depressed seeing that I am not being able to improve this
> single line of code... maybe it is ok, and the whole thing is just
> slow...? ahh.
|
|
|
|
| Re: Accelerating a one-line program doing matrix multiplication [message #72682 is a reply to message #72681] |
Mon, 27 September 2010 23:17 |
Axel Martínez
Messages: 22
Registered: June 2010
|
Junior Member |
|
|
On 27 Sep., 15:31, nata <bernat.puigdomen...@gmail.com> wrote:
> You can use the TEMPORARY function if you can set the input to
> undefined...
> When you do [[v1],[v2],[v3]] you are duplicating data. v1, v2 and v3
> are copied and you are not conserving memory.
>
> You could try:
> RETURN, [[TEMPORARY(v1)],[TEMPORARY(v2)],[TEMPORARY(v3)]] # vc +
> REBIN(v0, SIZE(vc, /DIMENSIONS))
>
> Cheers,
> nata
Thanks nata.
v0, v1, v2 and v3 are each of them 3-element vectors. I can add that
but, as I understand it, it will only save the place of 12 floating
values in memory (48 bytes?).
But I am happy that you did not see any other obvious thing. I started
feeling depressed seeing that I am not being able to improve this
single line of code... maybe it is ok, and the whole thing is just
slow...? ahh.
|
|
|
|
| Re: Accelerating a one-line program doing matrix multiplication [message #72692 is a reply to message #72682] |
Mon, 27 September 2010 06:31 |
natha
Messages: 482
Registered: October 2007
|
Senior Member |
|
|
You can use the TEMPORARY function if you can set the input to
undefined...
When you do [[v1],[v2],[v3]] you are duplicating data. v1, v2 and v3
are copied and you are not conserving memory.
You could try:
RETURN, [[TEMPORARY(v1)],[TEMPORARY(v2)],[TEMPORARY(v3)]] # vc +
REBIN(v0, SIZE(vc, /DIMENSIONS))
Cheers,
nata
|
|
|
|
| Re: Accelerating a one-line program doing matrix multiplication [message #72758 is a reply to message #72662] |
Wed, 29 September 2010 07:49 |
rogass
Messages: 200
Registered: April 2008
|
Senior Member |
|
|
On 29 Sep., 12:48, Axel M <axe...@gmail.com> wrote:
> I have to admit that I did not understand this proposed use of
> REPLICATE:
>
>>> sometimes (replicate({temp:input},newsize)).(0) is faster then rebin
>
> But it brought me a related question in mind: does IDL have a
> "REPLICATE" function for vectors instead of scalar values? I am using
> REBIN, but REBIN is thought for more advanced uses and probably
> suboptimal for a "replicate-like" use... right?
Yes, it has as I mentioned uncommented above:
sometimes (replicate({temp:input},newsize)).(0) is faster then rebin
-> this means:
IDL> a=findgen(3)
IDL> print,a
0.000000 1.00000 2.00000
IDL> print, rebin(a,3,5)
0.000000 1.00000 2.00000
0.000000 1.00000 2.00000
0.000000 1.00000 2.00000
0.000000 1.00000 2.00000
0.000000 1.00000 2.00000
IDL> print, (replicate({temp:a},5)).(0)
0.000000 1.00000 2.00000
0.000000 1.00000 2.00000
0.000000 1.00000 2.00000
0.000000 1.00000 2.00000
0.000000 1.00000 2.00000
Regards
CR
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
