[Numpy-discussion] Deprecating PyDataMem_RENEW ?

[David Cournapeau]

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/browser/branches/aligned_alloca

For now, I've only commited:
    - a new bunch of tests for all functions using PyDataMem_RENEW (in 
numpy/core/tests/test_renew.py)
    - If NOUSE_PYDATAMEM_RENEW is defined, PyDataMEM_RENEW is not used: 
instead, SYS_REALLOC is used in loops (this is defined to realloc), and 
when outside a loop, _fake_realloc is used, which does not use realloc, 
but use PyDataMem_NEW/FREE.

I got a small slowdown, but I am not sure it is really relevant: the 
figures change almost as much between several runs as between runs of 
different implementations. I am not convinced about the quality of my 
tests either.

cheers,

David