Anne Archibald wrote:
>
> How much does this matter? I mean, what if you simply left all the
> reallocs as-is? The arrays that resulted from reallocs would not
> typically be aligned, but we cannot in any case expect all arrays to
> be aligned.
The problem would be the interaction between the alig
2008/5/5 David Cournapeau <[EMAIL PROTECTED]>:
> Basically, what I have in mind is, in a first step (for numpy 1.2):
> - define functions to allocate on a given alignement
> - make PyMemData_NEW 16 byte aligned by default (to be compatible
> with SSE and co).
>
> The problem was, and st
Robert Kern wrote:
>
> Since there are only 6 places where PyMemData_RENEW is used, all 6
> uses should be benchmarked. I would prefer a more targeted benchmark
> so we know exactly what we are measuring.
>
Ok, I started a new branch for aligned allocator:
http://projects.scipy.org/scipy/numpy/br
On Mon, May 5, 2008 at 1:30 PM, David Cournapeau <[EMAIL PROTECTED]> wrote:
> On Tue, May 6, 2008 at 2:11 AM, Robert Kern <[EMAIL PROTECTED]> wrote:
> >
> > I am in favor of at least trying this out. We will have to have a set
> > of benchmarks to make sure we haven't hurt the current uses of
2008/5/5 Robert Kern <[EMAIL PROTECTED]>:
> On Mon, May 5, 2008 at 7:44 AM, David Cournapeau
> <[EMAIL PROTECTED]> wrote:
>
> > In numpy, we can always replace realloc by malloc/free, because we know
> > the size of the old block: would deprecating PyMemData_RENEW and
> > replacing them by Py
On Tue, May 6, 2008 at 2:11 AM, Robert Kern <[EMAIL PROTECTED]> wrote:
>
> I am in favor of at least trying this out. We will have to have a set
> of benchmarks to make sure we haven't hurt the current uses of
> PyMemData_RENEW which Tim points out.
What would be a good stress test for PyArray_
On Tue, May 6, 2008 at 1:59 AM, Timothy Hochberg <[EMAIL PROTECTED]> wrote:
> I don't think you would want to do this in the core of PyArray_FromIter;
> presumably realloc can sometimes reuse the existing pointer and save on
> allocating a new chunk of memory. Since there are lots of allocations i
On Mon, May 5, 2008 at 7:44 AM, David Cournapeau
<[EMAIL PROTECTED]> wrote:
> In numpy, we can always replace realloc by malloc/free, because we know
> the size of the old block: would deprecating PyMemData_RENEW and
> replacing them by PyMemeData_NEW/PyMemData_FREE be possible, such as to
> ma
On Mon, May 5, 2008 at 5:44 AM, David Cournapeau <
[EMAIL PROTECTED]> wrote:
> Hi,
>
>While working again on the fftpack module, to clean things up and
> speed some backends (in particular fftw3, which is really sub-optimal
> right now), I remembered how much unaligned data pointer in numpy ar
Hi,
While working again on the fftpack module, to clean things up and
speed some backends (in particular fftw3, which is really sub-optimal
right now), I remembered how much unaligned data pointer in numpy arrays
hurt performances. So I would like to relaunch the discussion on aligned
allo
10 matches
Mail list logo