[HN Gopher] HPy: A better C API for Python?
___________________________________________________________________
HPy: A better C API for Python?
Author : mkesper
Score : 127 points
Date : 2021-04-02 07:57 UTC (15 hours ago)
(HTM) web link (lwn.net)
(TXT) w3m dump (lwn.net)
| pansa2 wrote:
| HPy was recently discussed here:
| https://news.ycombinator.com/item?id=26625398
| JetAlone wrote:
| I remember reading an article or two about Python being quite
| energy inefficient as languages go, reaching very low on a chart
| like. I don't know if and/or how much Hpy will improve this.
| Definitely have to parrot the article and say the situation a
| matter of "wait, keep watching and see", or better yet if you can
| manage it, "contribute, keep contributing, and look ahead".
| ketralnis wrote:
| HPy is (sort of) a replacement for CPython's normal API for C
| Extensions. It has nothing to do with "efficiency" or the
| performance of python itself.
| virtue3 wrote:
| No one has ever claimed that python is energy conservative.
|
| There is a general understanding that you are trading machine
| efficiency for developer efficiency.
|
| There were attempts at making it more efficient (see
| ironpython, which is dead now).
|
| If you drop a serious bottleneck of code into C you are going
| to see massive improvements provided you don't need to marshal
| between python/c a lot (aka, if you can dump a ton of data, do
| a ton of work in C, then dump it back to python, you'll be
| sitting pretty).
| nullserver wrote:
| Had to help out with some python the other day. About 15
| years since I used python. A couple of hours to figure out
| how to get it running locally. Then I fixed the various
| issues. Mostly horrible performance problems.
|
| All told half a day of effort.
|
| Felt amazing being able to be productive so quickly on
| ancient abandoned code. The badly written algorithm was just
| obvious in its wrongness.
| dragonwriter wrote:
| > There were attempts at making it more efficient (see
| ironpython, which is dead now).
|
| IronPython seems to be on life support (it had a bugfix
| release for .NET 5 late last year), and no longer an MS
| project, but its not quite dead.
|
| But pointing to IronPython alone is weird, as that was more
| about interface woth the rest of .NET than machine
| efficiency, and there are a number of major, live projects
| with efficiency as one of their central goals: PyPy, mypyc,
| Cython, for instance.
___________________________________________________________________
(page generated 2021-04-02 23:01 UTC)