Jens Axel Søgaard writes:
> See my post in this thread 3 days ago. It is attached.
It isn't, but I found it in the thread, thanks! I am mostly interested
in using it as an example for the Racket FFI, given that I know LAPACK
rather well.
> > Transposition can be avoided in many practically rel
On May 14, 2014, at 10:40 AM, Jens Axel Søgaard wrote:
> 2014-05-13 23:24 GMT+02:00 Matthias Felleisen :
>> On May 13, 2014, at 4:19 PM, Neil Toronto wrote:
>>> We need a predicate like
>>>
>>> (: flonum-matrix? (All (A) (-> (Matrix A) Boolean : (Matrix Flonum
>>
>>
>> I think in our wor
2014-05-13 23:24 GMT+02:00 Matthias Felleisen :
> On May 13, 2014, at 4:19 PM, Neil Toronto wrote:
>> We need a predicate like
>>
>> (: flonum-matrix? (All (A) (-> (Matrix A) Boolean : (Matrix Flonum
>
>
> I think in our world of types we could even have
>
> (: flonum-matrix? (All (A) (-> (M
2014-05-14 15:31 GMT+02:00 Konrad Hinsen :
> Neil Toronto writes:
>
> > One thing we should really do is get your LAPACK FFI into the math
> > library and have `flmatrix-solve` use that, but fail over to Racket code
> > systems that don't have LAPACK. If I remember right, it would have to
> > t
Neil Toronto writes:
> One thing we should really do is get your LAPACK FFI into the math
> library and have `flmatrix-solve` use that, but fail over to Racket code
> systems that don't have LAPACK. If I remember right, it would have to
> transpose the data because LAPACK is column-major.
On May 13, 2014, at 4:19 PM, Neil Toronto wrote:
> We need a predicate like
>
> (: flonum-matrix? (All (A) (-> (Matrix A) Boolean : (Matrix Flonum
I think in our world of types we could even have
(: flonum-matrix? (All (A) (-> (Matrix A) Boolean : (TriangularMatrix A
and such and
We need a predicate like
(: flonum-matrix? (All (A) (-> (Matrix A) Boolean : (Matrix Flonum
Then `matrix-solve` could dispatch to `flmatrix-solve` and still be
well-typed. We could/should do something similar for every operation for
which checking flonum-ness is cheap compared to computi
That's great!
The question is now how to automate this sort of thing.
/Jens Axel
2014-05-13 1:39 GMT+02:00 Neil Toronto :
> When I change it to operate on (Vectorof FlVector) instead of (Vectorof
> (Vectorof Flonum)), I get this:
>
> cpu time: 996 real time: 995 gc time: 22
> 1.9
When I change it to operate on (Vectorof FlVector) instead of (Vectorof
(Vectorof Flonum)), I get this:
cpu time: 996 real time: 995 gc time: 22
1.9335
cpu time: 15387 real time: 15384 gc time: 13006
1.9335
cpu time: 1057 real time: 1056 gc time: 85
1.9335
cpu
Hi Eric,
You were absolute right. The version below cuts the time in half.
It is mostly cut and paste from existing functions and removing
non-Flonum cases.
/Jens Axel
#lang typed/racket
(require math/matrix
math/array
math/private/matrix/utils
math/private/vector/vect
Where is the time spent in the algorithm? I assume that most of it is
in the matrix manipulation work not the orchestration of finding a
pivot and reducing based on that. I.e. `elim-rows!` is the expensive
part. Given that you only specialized the outer part of the loop, I
wouldn't expect huge perf
I tried restricting the matrix-solve and matrix-gauss-elim to (Matrix Flonum).
I can't observe a change in the timings.
#lang typed/racket
(require math/matrix
math/array
math/private/matrix/utils
math/private/vector/vector-mutate
math/private/unsafe
(o
The garbage collection time is probably from cleaning up boxed flonums,
and possibly intermediate vectors. If so, a separate implementation of
Gaussian elimination for the FlArray type would cut the GC time to
nearly zero.
Neil ⊥
On 05/11/2014 01:36 PM, Jens Axel Søgaard wrote:
Or ... you co
Or ... you could take a look at
https://github.com/plt/racket/blob/master/pkgs/math-pkgs/math-lib/math/private/matrix/matrix-gauss-elim.rkt
at see if something can be improved.
/Jens Axel
2014-05-11 21:30 GMT+02:00 Jens Axel Søgaard :
> Hi Eduardo,
>
> The math/matrix library uses the arrays f
Hi Eduardo,
The math/matrix library uses the arrays from math/array to represent matrices.
If you want to try the same representation as Bigloo, you could try Will Farr's
matrix library:
http://planet.racket-lang.org/package-source/wmfarr/simple-matrix.plt/1/1/planet-docs/simple-matrix/index.htm
What is bothering me is the time Racket is spending in garbage collection.
~/wrk/scm/rkt/matrix$ racket matrix.rkt
0.99967226
cpu time: 61416 real time: 61214 gc time: 32164
If I am reading the output correctly, Racket is spending 32 seconds out of
61 seconds in garbage collection.
I am
2014-05-11 6:09 GMT+02:00 Eduardo Costa :
> The documentation says that one should expect typed/racket to be faster than
> racket. I tested the math/matrix library and it seems to be almost as slow
> in typed/racket as in racket.
What was (is?) slow was a call in an untyped module A to a function
Where are you seeing the slowness in that code? For me the majority of
the time is spent in matrix-solve, and since that is another module it
doesn't matter what you write your module in.
In untyped racket, (set the first line to #lang
typed/racket/no-check), it is much slower. 36 seconds to creat
The documentation says that one should expect typed/racket to be faster
than racket. I tested the math/matrix library and it seems to be almost as
slow in typed/racket as in racket. Here is the program in typed racket:
#lang typed/racket ; file: matrix.rkt
(require math/matrix)
(: mx Index)
(defi
On Wed, Feb 13, 2013 at 6:51 PM, Danny Yoo wrote:
> On Wed, Feb 13, 2013 at 10:49 AM, Pierpaolo Bernardi
> wrote:
>> On Wed, Feb 13, 2013 at 6:14 PM, Danny Yoo wrote:
>>
>>> Other than that, this should work ok. Can you confirm?
>>
>> Yes, it works. Thanks!
>>
>> What's the difference between _
On Wed, Feb 13, 2013 at 10:49 AM, Pierpaolo Bernardi
wrote:
> On Wed, Feb 13, 2013 at 6:14 PM, Danny Yoo wrote:
>
>> Other than that, this should work ok. Can you confirm?
>
> Yes, it works. Thanks!
>
> What's the difference between _byte and _int8 ?
Wow. There should have not been any differ
On Wed, Feb 13, 2013 at 6:14 PM, Danny Yoo wrote:
> Other than that, this should work ok. Can you confirm?
Yes, it works. Thanks!
What's the difference between _byte and _int8 ?
Racket Users list:
http://lists.racket-lang.org/users
On Wed, Feb 13, 2013 at 9:41 AM, Pierpaolo Bernardi wrote:
> Another FFI question:
>
> I want to interface a C function which has the following signature:
>
> void iauA2tf(int ndp, double angle, char *sign, int ihmsf[4])
Example written:
https://github.com/dyoo/ffi-tutorial/tree/master/ffi/
Another FFI question:
I want to interface a C function which has the following signature:
void iauA2tf(int ndp, double angle, char *sign, int ihmsf[4])
The "char *sign" parameter is a single char, not a string, output only, and
is giving me problems.
The C code uses this parameter simply:
*
On 01/22/2013 12:33 PM, Tobias Hammer wrote:
When you pass it directly to c-code, maybe you can completely bypass the
array and convert
matrix -> flarray -> flvector -> cpointer
the last 3 conversions should be blazing fast but i guess that won't
matter much for 16 elements.
Neil, can you neverth
When you pass it directly to c-code, maybe you can completely bypass the
array and convert
matrix -> flarray -> flvector -> cpointer
the last 3 conversions should be blazing fast but i guess that won't
matter much for 16 elements.
Neil, can you nevertheless tell me more about the expected perf
On 01/21/2013 02:48 AM, Pierpaolo Bernardi wrote:
Hello,
what's the most convenient way to convert between math/matrices and
FFI/arrays?
To reduce the scope of the question, at the moment I'm only interested
in conversions from and to the following FFI types:
(_array _double 3 3)
(_array _
Hello,
what's the most convenient way to convert between math/matrices and
FFI/arrays?
To reduce the scope of the question, at the moment I'm only interested in
conversions from and to the following FFI types:
(_array _double 3 3)
(_array _double 3)
(_array _double 2 3)
Thanks!
P.
__
28 matches
Mail list logo
