[Numpy-discussion] advanced indexing bug with huge arrays?
A colleague has run into this weird behaviour with NumPy 1.6.1, EPD 7.1-2, on Linux (Fedora Core 14) 64-bit: a = numpy.array(numpy.random.randint(256,size=(500,972)),dtype='uint8') b = numpy.random.randint(500,size=(4993210,)) c = a[b] It seems c is not getting filled in full, namely: In [14]: c[100:].sum() Out[14]: 0 I haven't been able to reproduce this quite yet, I'll try to find a machine with sufficient memory tomorrow. But does anyone have any insight in the mean time? It smells like some kind of integer overflow bug. Thanks, David ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] The NumPy Mandelbrot code 16x slower than Fortran
Den 23.01.2012 10:04, skrev Dag Sverre Seljebotn: On 01/23/2012 05:35 AM, Jonathan Rocher wrote: Hi all, I was reading this while learning about Pytables in more details and the origin of its efficiency. This sounds like a problem where out of core computation using pytables would shine since the dataset doesn't fit into CPU cache: http://www.pytables.org/moin/ComputingKernel. Of course C/Cythonizing the problem would be another good way... Well, since the data certainly fits in RAM, one would use numexpr directly (which is what pytables also uses). Personally I feel this debate is asking the wrong question. It is not uncommon for NumPy code to be 16x slower than C or Fortran. But that is not really interesting. This is what I think matters: - Is the NumPy code FAST ENOUGH? If not, then go ahead and optimize. If it's fast enough, then just leave it. In this case, it seems Python takes ~13 seconds compared to ~1 second for Fortran. Sure, those extra 12 seconds could be annoying. But how much coding time should we spend to avoid them? 15 minutes? An hour? Two hours? Taking the time spent optimizing into account, then perhaps Python is 'faster' anyway? It is common to ask what is fastest for the computer. But we should really be asking what is fastest for our selves. For example: I have a computation that will take a day in Fortran or a month in Python (estimated). And I am going to run this code several times (20 or so, I think). In this case, yes, coding the bottlenecks in Fortran matters to me. But 13 seconds versus 1 second? I find that hardly interesting. Sturla ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] The NumPy Mandelbrot code 16x slower than Fortran
On Mon, Jan 23, 2012 at 12:23 PM, Sturla Molden stu...@molden.no wrote: Den 23.01.2012 10:04, skrev Dag Sverre Seljebotn: On 01/23/2012 05:35 AM, Jonathan Rocher wrote: Hi all, I was reading this while learning about Pytables in more details and the origin of its efficiency. This sounds like a problem where out of core computation using pytables would shine since the dataset doesn't fit into CPU cache: http://www.pytables.org/moin/ComputingKernel. Of course C/Cythonizing the problem would be another good way... Well, since the data certainly fits in RAM, one would use numexpr directly (which is what pytables also uses). Personally I feel this debate is asking the wrong question. It is not uncommon for NumPy code to be 16x slower than C or Fortran. But that is not really interesting. This is what I think matters: - Is the NumPy code FAST ENOUGH? If not, then go ahead and optimize. If it's fast enough, then just leave it. In this case, it seems Python takes ~13 seconds compared to ~1 second for Fortran. Sure, those extra 12 seconds could be annoying. But how much coding time should we spend to avoid them? 15 minutes? An hour? Two hours? Taking the time spent optimizing into account, then perhaps Python is 'faster' anyway? It is common to ask what is fastest for the computer. But we should really be asking what is fastest for our selves. For example: I have a computation that will take a day in Fortran or a month in Python (estimated). And I am going to run this code several times (20 or so, I think). In this case, yes, coding the bottlenecks in Fortran matters to me. But 13 seconds versus 1 second? I find that hardly interesting. Sturla I would think that interactive zooming would be quite nice (illuminating) and for that 13 secs would not be tolerable Well... it's not at the top of my priority list ... ;-) -Sebastian Haase ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] The NumPy Mandelbrot code 16x slower than Fortran
On 01/23/2012 12:23 PM, Sturla Molden wrote: Den 23.01.2012 10:04, skrev Dag Sverre Seljebotn: On 01/23/2012 05:35 AM, Jonathan Rocher wrote: Hi all, I was reading this while learning about Pytables in more details and the origin of its efficiency. This sounds like a problem where out of core computation using pytables would shine since the dataset doesn't fit into CPU cache: http://www.pytables.org/moin/ComputingKernel. Of course C/Cythonizing the problem would be another good way... Well, since the data certainly fits in RAM, one would use numexpr directly (which is what pytables also uses). Personally I feel this debate is asking the wrong question. It is not uncommon for NumPy code to be 16x slower than C or Fortran. But that is not really interesting. This is what I think matters: - Is the NumPy code FAST ENOUGH? If not, then go ahead and optimize. If it's fast enough, then just leave it. In this case, it seems Python takes ~13 seconds compared to ~1 second for Fortran. Sure, those extra 12 seconds could be annoying. But how much coding time should we spend to avoid them? 15 minutes? An hour? Two hours? Taking the time spent optimizing into account, then perhaps Python is 'faster' anyway? It is common to ask what is fastest for the computer. But we should really be asking what is fastest for our selves. For example: I have a computation that will take a day in Fortran or a month in Python (estimated). And I am going to run this code several times (20 or so, I think). In this case, yes, coding the bottlenecks in Fortran matters to me. But 13 seconds versus 1 second? I find that hardly interesting. You, me, Ondrej, and many more are happy to learn 4 languages and use them where they are most appropriate. But most scientists only want to learn and use one tool. And most scientists have both problems where performance doesn't matter, and problems where it does. So as long as examples like this exists, many people will prefer Fortran for *all* their tasks. (Of course, that's why I got involved in Cython...) Dag Sverre ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] The NumPy Mandelbrot code 16x slower than Fortran
Den 23.01.2012 13:09, skrev Sebastian Haase: I would think that interactive zooming would be quite nice (illuminating) and for that 13 secs would not be tolerable Well... it's not at the top of my priority list ... ;-) Sure, that comes under the 'fast enough' issue. But even Fortran might be too slow here? For zooming Mandelbrot I'd use PyOpenGL and a GLSL fragment shader (which would be a text string in Python): madelbrot_fragment_shader = uniform sampler1D tex; uniform vec2 center; uniform float scale; uniform int iter; void main() { vec2 z, c; c.x = 1. * (gl_TexCoord[0].x - 0.5) * scale - center.x; c.y = (gl_TexCoord[0].y - 0.5) * scale - center.y; int i; z = c; for(i=0; iiter; i++) { float x = (z.x * z.x - z.y * z.y) + c.x; float y = (z.y * z.x + z.x * z.y) + c.y; if((x * x + y * y) 4.0) break; z.x = x; z.y = y; } gl_FragColor = texture1D(tex, (i == iter ? 0.0 : float(i)) / 100.0); } The rest is just boiler-plate OpenGL... Sources: http://nuclear.mutantstargoat.com/articles/sdr_fract/ http://pyopengl.sourceforge.net/context/tutorials/shader_1.xhtml Sturla ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] The NumPy Mandelbrot code 16x slower than Fortran
On 01/23/12 13:51, Sturla Molden wrote: Den 23.01.2012 13:09, skrev Sebastian Haase: I would think that interactive zooming would be quite nice (illuminating) and for that 13 secs would not be tolerable Well... it's not at the top of my priority list ... ;-) Sure, that comes under the 'fast enough' issue. But even Fortran might be too slow here? For zooming Mandelbrot I'd use PyOpenGL and a GLSL fragment shader (which would be a text string in Python): madelbrot_fragment_shader = uniform sampler1D tex; uniform vec2 center; uniform float scale; uniform int iter; void main() { vec2 z, c; c.x = 1. * (gl_TexCoord[0].x - 0.5) * scale - center.x; c.y = (gl_TexCoord[0].y - 0.5) * scale - center.y; int i; z = c; for(i=0; iiter; i++) { float x = (z.x * z.x - z.y * z.y) + c.x; float y = (z.y * z.x + z.x * z.y) + c.y; if((x * x + y * y) 4.0) break; z.x = x; z.y = y; } gl_FragColor = texture1D(tex, (i == iter ? 0.0 : float(i)) / 100.0); } The rest is just boiler-plate OpenGL... Sources: http://nuclear.mutantstargoat.com/articles/sdr_fract/ http://pyopengl.sourceforge.net/context/tutorials/shader_1.xhtml Off-topic comment: Or use some algorithmic cleverness, see [1]. I recall Xaos had interactive, extremely fast a fluid fractal zooming more than 10 (or 15?) years ago (- on a laughable hardware by today's standards). r. [1] http://wmi.math.u-szeged.hu/xaos/doku.php ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] The NumPy Mandelbrot code 16x slower than Fortran
I'd like to add http://git.tiker.net/pyopencl.git/blob/HEAD:/examples/demo_mandelbrot.py to the discussion, since I use pyopencl (http://mathema.tician.de/software/pyopencl) with great success in my daily scientific computing. Install with pip. PyOpenCL does understand numpy arrays. You write a kernel (small c-program) directly into a python triple quoted strings and get a pythonic way to program GPU and core i5 and i7 CPUs with python Exception if something goes wrong. Whenever I hit a speed bottleneck that I cannot solve with pure numpy, I code a little part of the computation for GPU. The compilation is done just in time when you run the python code. Especially for the mandelbrot this may be a _huge_ gain in speed since its embarrassingly parallel. Samuel On 23.01.2012, at 14:02, Robert Cimrman wrote: On 01/23/12 13:51, Sturla Molden wrote: Den 23.01.2012 13:09, skrev Sebastian Haase: I would think that interactive zooming would be quite nice (illuminating) and for that 13 secs would not be tolerable Well... it's not at the top of my priority list ... ;-) Sure, that comes under the 'fast enough' issue. But even Fortran might be too slow here? For zooming Mandelbrot I'd use PyOpenGL and a GLSL fragment shader (which would be a text string in Python): madelbrot_fragment_shader = uniform sampler1D tex; uniform vec2 center; uniform float scale; uniform int iter; void main() { vec2 z, c; c.x = 1. * (gl_TexCoord[0].x - 0.5) * scale - center.x; c.y = (gl_TexCoord[0].y - 0.5) * scale - center.y; int i; z = c; for(i=0; iiter; i++) { float x = (z.x * z.x - z.y * z.y) + c.x; float y = (z.y * z.x + z.x * z.y) + c.y; if((x * x + y * y) 4.0) break; z.x = x; z.y = y; } gl_FragColor = texture1D(tex, (i == iter ? 0.0 : float(i)) / 100.0); } The rest is just boiler-plate OpenGL... Sources: http://nuclear.mutantstargoat.com/articles/sdr_fract/ http://pyopengl.sourceforge.net/context/tutorials/shader_1.xhtml Off-topic comment: Or use some algorithmic cleverness, see [1]. I recall Xaos had interactive, extremely fast a fluid fractal zooming more than 10 (or 15?) years ago (- on a laughable hardware by today's standards). r. [1] http://wmi.math.u-szeged.hu/xaos/doku.php ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Counting the Colors of RGB-Image
On Wed, Jan 18, 2012 at 1:26 AM, a...@pdauf.de wrote: Your ideas are very helpfull and the code is very fast. I'm curios -- a number of ideas were floated here -- what did you end up using? -Chris -- Christopher Barker, Ph.D. Oceanographer Emergency Response Division NOAA/NOS/ORR (206) 526-6959 voice 7600 Sand Point Way NE (206) 526-6329 fax Seattle, WA 98115 (206) 526-6317 main reception chris.bar...@noaa.gov ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] advanced indexing bug with huge arrays?
I've reproduced this (rather serious) bug myself and confirmed that it exists in master, and as far back as 1.4.1. I'd really appreciate if someone could reproduce and confirm on another machine, as so far all my testing has been on our single high-memory machine. Thanks, David On Mon, Jan 23, 2012 at 05:23:28AM -0500, David Warde-Farley wrote: A colleague has run into this weird behaviour with NumPy 1.6.1, EPD 7.1-2, on Linux (Fedora Core 14) 64-bit: a = numpy.array(numpy.random.randint(256,size=(500,972)),dtype='uint8') b = numpy.random.randint(500,size=(4993210,)) c = a[b] It seems c is not getting filled in full, namely: In [14]: c[100:].sum() Out[14]: 0 I haven't been able to reproduce this quite yet, I'll try to find a machine with sufficient memory tomorrow. But does anyone have any insight in the mean time? It smells like some kind of integer overflow bug. Thanks, David ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] advanced indexing bug with huge arrays?
Can you determine where the problem is, precisely.In other words, can you verify that c is not getting filled in correctly? You are no doubt going to get overflow in the summation as you have a uint8 parameter. But, having that overflow be exactly '0' would be surprising. Can you verify that a and b are getting created correctly? Also, 'c' should be a 2-d array, can you verify that? Can you take the sum along the -1 axis and the 0 axis separately: print a.shape print b.shape print c.shape c[100:].sum(axis=0) d = c[100:].sum(axis=-1) print d[:100] print d[-100:] On Jan 23, 2012, at 12:55 PM, David Warde-Farley wrote: I've reproduced this (rather serious) bug myself and confirmed that it exists in master, and as far back as 1.4.1. I'd really appreciate if someone could reproduce and confirm on another machine, as so far all my testing has been on our single high-memory machine. Thanks, David On Mon, Jan 23, 2012 at 05:23:28AM -0500, David Warde-Farley wrote: A colleague has run into this weird behaviour with NumPy 1.6.1, EPD 7.1-2, on Linux (Fedora Core 14) 64-bit: a = numpy.array(numpy.random.randint(256,size=(500,972)),dtype='uint8') b = numpy.random.randint(500,size=(4993210,)) c = a[b] It seems c is not getting filled in full, namely: In [14]: c[100:].sum() Out[14]: 0 I haven't been able to reproduce this quite yet, I'll try to find a machine with sufficient memory tomorrow. But does anyone have any insight in the mean time? It smells like some kind of integer overflow bug. Thanks, David ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] advanced indexing bug with huge arrays?
On Mon, Jan 23, 2012 at 7:55 PM, David Warde-Farley warde...@iro.umontreal.ca wrote: I've reproduced this (rather serious) bug myself and confirmed that it exists in master, and as far back as 1.4.1. I'd really appreciate if someone could reproduce and confirm on another machine, as so far all my testing has been on our single high-memory machine. I see the same behaviour on a Winodows machine with numpy 1.6.1. But I don't think it is an indexing problem - rather something with the random number creation. a itself is already zeros for high indexes. In [8]: b[100:110] Out[8]: array([3429029, 1251819, 4292918, 2249483, 757620, 3977130, 3455449, 2005054, 2565207, 3114930]) In [9]: a[b[100:110]] Out[9]: array([[0, 0, 0, ..., 0, 0, 0], [0, 0, 0, ..., 0, 0, 0], [0, 0, 0, ..., 0, 0, 0], ..., [0, 0, 0, ..., 0, 0, 0], [0, 0, 0, ..., 0, 0, 0], [0, 0, 0, ..., 0, 0, 0]], dtype=uint8) In [41]: a[581350:,0].sum() Out[41]: 0 Cheers Robin Thanks, David On Mon, Jan 23, 2012 at 05:23:28AM -0500, David Warde-Farley wrote: A colleague has run into this weird behaviour with NumPy 1.6.1, EPD 7.1-2, on Linux (Fedora Core 14) 64-bit: a = numpy.array(numpy.random.randint(256,size=(500,972)),dtype='uint8') b = numpy.random.randint(500,size=(4993210,)) c = a[b] It seems c is not getting filled in full, namely: In [14]: c[100:].sum() Out[14]: 0 I haven't been able to reproduce this quite yet, I'll try to find a machine with sufficient memory tomorrow. But does anyone have any insight in the mean time? It smells like some kind of integer overflow bug. Thanks, David ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Counting the Colors of RGB-Image
Am 23.01.2012 18:17, schrieb Chris Barker: On Wed, Jan 18, 2012 at 1:26 AM,a...@pdauf.de wrote: Your ideas are very helpfull and the code is very fast. I'm curios -- a number of ideas were floated here -- what did you end up using? -Chris I'am sorry but when i see the code of Torgil Svenson, I think, the game is over. I use the follow. code: t0=clock() tt = n_im2.view() tt.shape = -1,3 ifl = tt[...,0].astype(np.int)*256*256 + tt[...,1].astype(np.int)*256 + tt[...,2].astype(np.int) colors, inv = np.unique(ifl,return_inverse=True) zus = np.array([colors[-1]+1]) colplus = np.hstack((colors,zus)) ccnt = np.histogram(ifl,colplus)[0] t1=clock() print (t1-t0) t0=t1 -- Christopher Barker, Ph.D. Oceanographer Emergency Response Division NOAA/NOS/ORR(206) 526-6959 voice 7600 Sand Point Way NE (206) 526-6329 fax Seattle, WA 98115 (206) 526-6317 main reception chris.bar...@noaa.gov ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] advanced indexing bug with huge arrays?
On Mon, Jan 23, 2012 at 1:33 PM, Travis Oliphant teoliph...@gmail.comwrote: Can you determine where the problem is, precisely.In other words, can you verify that c is not getting filled in correctly? You are no doubt going to get overflow in the summation as you have a uint8 parameter. But, having that overflow be exactly '0' would be surprising. Can you verify that a and b are getting created correctly? Also, 'c' should be a 2-d array, can you verify that? Can you take the sum along the -1 axis and the 0 axis separately: print a.shape print b.shape print c.shape c[100:].sum(axis=0) d = c[100:].sum(axis=-1) print d[:100] print d[-100:] I am getting the same results as David. It looks like c just stopped filling in partway through the array. I don't think there is any overflow issue, since the result of sum() is up-promoted to uint64 when I do that. Travis, here are the outputs at my end - I cut out many zeros for brevity: In [7]: print a.shape (500, 972) In [8]: print b.shape (4993210,) In [9]: print c.shape (4993210, 972) In [10]: c[100:].sum(axis=0) Out[10]: array([0, 0, 0, , 0]) In [11]: d = c[100:].sum(axis=-1) In [12]: print d[:100] [0 0 0 ... 0 0] In [13]: print d[-100:] [0 0 0 ... 0 0 0] I looked at sparse subsamples with matplotlib - specifically, imshow(a[::1000, :]) - and the a array looks correct (random values everywhere), but c is zero past a certain row number. In fact, it looks like it becomes zero at row 575419 - I think for all rows in c beyond row 574519, the values will be zero. For lower row numbers, I think they are correctly filled (at least, by the sparse view in matplotlib). In [15]: a[b[574519], 350:360] Out[15]: array([143, 155, 11, 30, 212, 149, 110, 164, 165, 120], dtype=uint8) In [16]: c[574519, 350:360] Out[16]: array([143, 155, 11, 30, 212, 149, 0, 0, 0, 0], dtype=uint8) I'm using EPD 7.1, numpy 1.6.1, Linux installation (I don't know the kernel details) HTH, Aronne ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
[Numpy-discussion] Saving and loading a structured array from a TEXT file
Is there a way to save a structured array in a text file? My problem is not so much in the saving procedure, but rather in the 'reloading' procedure. See below In [3]: import numpy as np In [4]: r = np.ones(3,dtype=[('name', '|S5'), ('foo', 'i8'), ('bar', 'f8')]) In [5]: r.tofile('toto.txt',sep='\n') bash-4.2$ cat toto.txt ('1', 1, 1.0) ('1', 1, 1.0) ('1', 1, 1.0) In [7]: r2 = np.fromfile('toto.txt',sep='\n',dtype=r.dtype) --- ValueErrorTraceback (most recent call last) /home/cls1fs/clseng/10/ipython-input-7-b07ba265ede7 in module() 1 r2 = np.fromfile('toto.txt',sep='\n',dtype=r.dtype) ValueError: Unable to read character files of that array type -- Emmanuel ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] advanced indexing bug with huge arrays?
Hi Travis, Thanks for your reply. On Mon, Jan 23, 2012 at 01:33:42PM -0600, Travis Oliphant wrote: Can you determine where the problem is, precisely.In other words, can you verify that c is not getting filled in correctly? You are no doubt going to get overflow in the summation as you have a uint8 parameter. But, having that overflow be exactly '0' would be surprising. I've already looked at this actually. The last 440 or so rows of c are all zero, however 'a' seems to be filled in fine: import numpy a = numpy.array(numpy.random.randint(256,size=(500,972)), dtype=numpy.uint8) b = numpy.random.randint(500,size=(4993210,)) c = a[b] print c [[186 215 204 ..., 170 98 198] [ 56 98 112 ..., 32 233 1] [ 44 133 171 ..., 163 35 51] ..., [ 0 0 0 ..., 0 0 0] [ 0 0 0 ..., 0 0 0] [ 0 0 0 ..., 0 0 0]] print a [[ 30 182 56 ..., 133 162 173] [112 100 69 ..., 3 147 80] [124 70 232 ..., 114 177 11] ..., [ 22 42 31 ..., 141 196 134] [ 74 47 167 ..., 38 193 9] [162 228 190 ..., 150 18 1]] So it seems to have nothing to do with the sum, but rather the advanced indexing operation. The zeros seem to start in the middle of row 574519, in particular at element 356. This is reproducible with different random vectors of indices, it seems. So 558432824th element things go awry. I can't say it makes any sense to me why this would be the magic number. David ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] advanced indexing bug with huge arrays?
On Mon, Jan 23, 2012 at 08:38:44PM +0100, Robin wrote: On Mon, Jan 23, 2012 at 7:55 PM, David Warde-Farley warde...@iro.umontreal.ca wrote: I've reproduced this (rather serious) bug myself and confirmed that it exists in master, and as far back as 1.4.1. I'd really appreciate if someone could reproduce and confirm on another machine, as so far all my testing has been on our single high-memory machine. I see the same behaviour on a Winodows machine with numpy 1.6.1. But I don't think it is an indexing problem - rather something with the random number creation. a itself is already zeros for high indexes. In [8]: b[100:110] Out[8]: array([3429029, 1251819, 4292918, 2249483, 757620, 3977130, 3455449, 2005054, 2565207, 3114930]) In [9]: a[b[100:110]] Out[9]: array([[0, 0, 0, ..., 0, 0, 0], [0, 0, 0, ..., 0, 0, 0], [0, 0, 0, ..., 0, 0, 0], ..., [0, 0, 0, ..., 0, 0, 0], [0, 0, 0, ..., 0, 0, 0], [0, 0, 0, ..., 0, 0, 0]], dtype=uint8) In [41]: a[581350:,0].sum() Out[41]: 0 Hmm, this seems like a separate bug to mine. In mine, 'a' is indeed being filled in -- the problem arises with c alone. So, another Windows-specific bug to add to the pile, perhaps? :( David ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Saving and loading a structured array from a TEXT file
On 23 Jan 2012, at 21:15, Emmanuel Mayssat wrote: Is there a way to save a structured array in a text file? My problem is not so much in the saving procedure, but rather in the 'reloading' procedure. See below In [3]: import numpy as np In [4]: r = np.ones(3,dtype=[('name', '|S5'), ('foo', 'i8'), ('bar', 'f8')]) In [5]: r.tofile('toto.txt',sep='\n') bash-4.2$ cat toto.txt ('1', 1, 1.0) ('1', 1, 1.0) ('1', 1, 1.0) In [7]: r2 = np.fromfile('toto.txt',sep='\n',dtype=r.dtype) --- ValueErrorTraceback (most recent call last) /home/cls1fs/clseng/10/ipython-input-7-b07ba265ede7 in module() 1 r2 = np.fromfile('toto.txt',sep='\n',dtype=r.dtype) ValueError: Unable to read character files of that array type I think most of the np.fromfile functionality works for binary input; for reading text input np.loadtxt and np.genfromtxt are the (currently) recommended functions. It is bit tricky to read the format generated by tofile() in the above example, but the following should work: cnv = {0: lambda s: s.lstrip('('), -1: lambda s: s.rstrip(')')} r2 = np.loadtxt('toto.txt', delimiter=',', converters=cnv, dtype=r.dtype) Generally loadtxt works more smoothly together with savetxt, but the latter unfortunately does not offer an easy way to save structured arrays (note to self and others currently working on npyio: definitely room for improvement!). HTH, Derek ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] advanced indexing bug with huge arrays?
On 1/23/2012 12:33 PM, David Warde-Farley wrote: On Mon, Jan 23, 2012 at 08:38:44PM +0100, Robin wrote: On Mon, Jan 23, 2012 at 7:55 PM, David Warde-Farley warde...@iro.umontreal.ca wrote: I've reproduced this (rather serious) bug myself and confirmed that it exists in master, and as far back as 1.4.1. I'd really appreciate if someone could reproduce and confirm on another machine, as so far all my testing has been on our single high-memory machine. I see the same behaviour on a Winodows machine with numpy 1.6.1. But I don't think it is an indexing problem - rather something with the random number creation. a itself is already zeros for high indexes. In [8]: b[100:110] Out[8]: array([3429029, 1251819, 4292918, 2249483, 757620, 3977130, 3455449, 2005054, 2565207, 3114930]) In [9]: a[b[100:110]] Out[9]: array([[0, 0, 0, ..., 0, 0, 0], [0, 0, 0, ..., 0, 0, 0], [0, 0, 0, ..., 0, 0, 0], ..., [0, 0, 0, ..., 0, 0, 0], [0, 0, 0, ..., 0, 0, 0], [0, 0, 0, ..., 0, 0, 0]], dtype=uint8) In [41]: a[581350:,0].sum() Out[41]: 0 Hmm, this seems like a separate bug to mine. In mine, 'a' is indeed being filled in -- the problem arises with c alone. So, another Windows-specific bug to add to the pile, perhaps? :( David Maybe this explains the win-amd64 behavior: There are a couple of places in mtrand where array indices and sizes are C long instead of npy_intp, for example in the randint function: https://github.com/numpy/numpy/blob/master/numpy/random/mtrand/mtrand.pyx#L863 Christoph ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Saving and loading a structured array from a TEXT file
On 23 Jan 2012, at 22:07, Derek Homeier wrote: In [4]: r = np.ones(3,dtype=[('name', '|S5'), ('foo', 'i8'), ('bar', 'f8')]) In [5]: r.tofile('toto.txt',sep='\n') bash-4.2$ cat toto.txt ('1', 1, 1.0) ('1', 1, 1.0) ('1', 1, 1.0) cnv = {0: lambda s: s.lstrip('('), -1: lambda s: s.rstrip(')')} r2 = np.loadtxt('toto.txt', delimiter=',', converters=cnv, dtype=r.dtype) Generally loadtxt works more smoothly together with savetxt, but the latter unfortunately does not offer an easy way to save structured arrays (note to self and others currently working on npyio: definitely room for improvement!). For the record, in that example np.savetxt('toto.txt', r, fmt='%s,%d,%f') would work as well, saving you the custom converter for loadtxt - it could just become tedious to work out the format for more complex structures, so an option to construct this automatically from r.dtype could certainly be a nice enhancement. Just wondering, is there something like the inverse operator to np.format_parser, i.e. mapping each dtype to a default print format specifier? Cheers, Derek ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
[Numpy-discussion] 'Advanced' save and restore operation
After having saved data, I need to know/remember the data dtype to restore it correctly. Is there a way to save the dtype with the data? (I guess the header parameter of savedata could help, but they are only available in v2.0+ ) I would like to save several related structured array and a dictionary of parameters into a TEXT file. Is there an easy way to do that? (maybe xml file, or maybe archive zip file of other files, or . ) Any recommendation is helpful. Regards, -- Emmanuel ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
[Numpy-discussion] Working with MATLAB
Dear List, I frequently work with MATLAB and it is necessary for me many a times to adapt MATLAB codes for NumPy arrays. While for most practical purposes it works fine, I think there might be a lot of 'under the hood' things that I might be missing when I make the translations from MATLAB to Python. Are there any 'best practices' for working on this transition? Thanks ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] Working with MATLAB
Please ignore my question. I found what I needed on the scipy website. I asked the question in haste. I'm sorry. Thanks ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] 'Advanced' save and restore operation
Note sure if there's a better way, but you can do it with some custom load and save functions: with open('f.txt', 'w') as f: ... f.write(str(x.dtype) + '\n') ... numpy.savetxt(f, x) with open('f.txt') as f: ... dtype = f.readline().strip() ... y = numpy.loadtxt(f).astype(dtype) I'm not sure how that'd work with structured arrays though. For the dict of parameters you'd have to write your own load/save piece of code too if you need a clean text file. -=- Olivier 2012/1/23 Emmanuel Mayssat emays...@gmail.com After having saved data, I need to know/remember the data dtype to restore it correctly. Is there a way to save the dtype with the data? (I guess the header parameter of savedata could help, but they are only available in v2.0+ ) I would like to save several related structured array and a dictionary of parameters into a TEXT file. Is there an easy way to do that? (maybe xml file, or maybe archive zip file of other files, or . ) Any recommendation is helpful. Regards, -- Emmanuel ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] 'Advanced' save and restore operation
On 24 Jan 2012, at 01:45, Olivier Delalleau wrote: Note sure if there's a better way, but you can do it with some custom load and save functions: with open('f.txt', 'w') as f: ... f.write(str(x.dtype) + '\n') ... numpy.savetxt(f, x) with open('f.txt') as f: ... dtype = f.readline().strip() ... y = numpy.loadtxt(f).astype(dtype) I'm not sure how that'd work with structured arrays though. For the dict of parameters you'd have to write your own load/save piece of code too if you need a clean text file. -=- Olivier 2012/1/23 Emmanuel Mayssat emays...@gmail.com After having saved data, I need to know/remember the data dtype to restore it correctly. Is there a way to save the dtype with the data? (I guess the header parameter of savedata could help, but they are only available in v2.0+ ) I would like to save several related structured array and a dictionary of parameters into a TEXT file. Is there an easy way to do that? (maybe xml file, or maybe archive zip file of other files, or . ) Any recommendation is helpful. asciitable might be of some help, but to implement all of your required functionality, you'd probably still have to implement your own Reader class: http://cxc.cfa.harvard.edu/contrib/asciitable/ Cheers, Derek ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] advanced indexing bug with huge arrays?
Den 23.01.2012 22:08, skrev Christoph Gohlke: Maybe this explains the win-amd64 behavior: There are a couple of places in mtrand where array indices and sizes are C long instead of npy_intp, for example in the randint function: https://github.com/numpy/numpy/blob/master/numpy/random/mtrand/mtrand.pyx#L863 AFAIK, on AMD64 a C long is 64 bit on Linux (gcc) and 32 bit on Windows (gcc and MSVC). Sturla ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] advanced indexing bug with huge arrays?
Den 23.01.2012 22:08, skrev Christoph Gohlke: Maybe this explains the win-amd64 behavior: There are a couple of places in mtrand where array indices and sizes are C long instead of npy_intp, for example in the randint function: https://github.com/numpy/numpy/blob/master/numpy/random/mtrand/mtrand.pyx#L863 Both i and length could overflow here. It should overflow on allocation of more than 2 GB. There is also a lot of C longs in the internal state (line 55-105), as well as the other functions. Producing 2 GB of random ints twice fails: import numpy as np np.random.randint(500,size=(2*1024**3,)) array([0, 0, 0, ..., 0, 0, 0]) np.random.randint(500,size=(2*1024**3,)) Traceback (most recent call last): File pyshell#3, line 1, in module np.random.randint(500,size=(2*1024**3,)) File mtrand.pyx, line 881, in mtrand.RandomState.randint (numpy\random\mtrand\mtrand.c:6040) MemoryError Sturla ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion
Re: [Numpy-discussion] advanced indexing bug with huge arrays?
Den 24.01.2012 06:00, skrev Sturla Molden: Both i and length could overflow here. It should overflow on allocation of more than 2 GB. There is also a lot of C longs in the internal state (line 55-105), as well as the other functions. The use of C long affects all the C and Pyrex source code in mtrand module, not just mtrand.pyx. All of it is fubar on Win64. From the C standard, a C long is only quarranteed to be at least 32 bits wide. Thus a C long can only be expected to index up to 2**31 - 1, and it is not a Windows specific problem. So it seems there are hundreds of places in the mtrand module where integers can overflow on 64-bit Python. Also the crappy old Pyrex code should be updated to some more recent Cython. Sturla ___ NumPy-Discussion mailing list NumPy-Discussion@scipy.org http://mail.scipy.org/mailman/listinfo/numpy-discussion