rpm/python-pythran
SHA256
1
0
Fork 0
forked from pool/python-pythran
Code Pull Requests Activity

- update to 0.13.1:

Browse Source 
* Bump xsimd dependency to 11.0.0, with a backported patch.
  * Improve performance of functions revieving scalar arguments.
    This changes the internal function call API.
  * Improve performance of fix-stride slicing, using a new slice
    representation.
  * Improve numpy.copyto performance, and detect copyto pattern usage.
  * Force internal linkage of generated functions, which gives more
    optimization room to the C++ compiler.
  * Provide entry points pythran.import_pythrancode and
    pythran.import_pythranfile, as a poor man JIT option.
  * Optimize numpy.argmax(cst * val) into numpy.argmax(val) when cst is
    positive.
  * Avoid copies upon numpy.array_split
  * Get rid of unused functions C++ warnings
  * Avoid generating a loop footer when the loop index is not used
    outside of the loop.

OBS-URL: https://build.opensuse.org/package/show/devel:languages:python:numeric/python-pythran?expand=0&rev=18
This commit is contained in:
Dirk Mueller
2023-05-15 15:24:30 +00:00
committed by Git OBS Bridge
parent 3b913b8990
commit 2212963a8a
4 changed files with 25 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
python-pythran.changes
View File

@@ -1,3 +1,24 @@
-------------------------------------------------------------------
Mon May 15 15:19:41 UTC 2023 - Dirk Müller <dmueller@suse.com>
- update to 0.13.1:
* Bump xsimd dependency to 11.0.0, with a backported patch.
* Improve performance of functions revieving scalar arguments.
This changes the internal function call API.
* Improve performance of fix-stride slicing, using a new slice
representation.
* Improve numpy.copyto performance, and detect copyto pattern usage.
* Force internal linkage of generated functions, which gives more
optimization room to the C++ compiler.
* Provide entry points pythran.import_pythrancode and
pythran.import_pythranfile, as a poor man JIT option.
* Optimize numpy.argmax(cst * val) into numpy.argmax(val) when cst is
positive.
* Avoid copies upon numpy.array_split
* Get rid of unused functions C++ warnings
* Avoid generating a loop footer when the loop index is not used
outside of the loop.
-------------------------------------------------------------------
Wed Jan 25 10:22:53 UTC 2023 - Dirk Müller <dmueller@suse.com>