A free lunch

J

Jan Decaluwe

Guest
The Python community is about to offer us a free lunch.
A new compliant interpreter, pypy, is already 4.3x faster
than cPython, and getting faster everyday. It shows
that there is not conceptual reason why high-level
dynamic languages should be slow.

For MyHDL, an HDL implemented as a Python library, the
results are even more spectacular: my benchmarks run
8-20x faster on pypy. In a single strike, this makes
MyHDL simulation performance competitive with Verilog/VHDL.

http://myhdl.org/doku.php/performance

Jan

--
Jan Decaluwe - Resources bvba - http://www.jandecaluwe.com
Python as a HDL: http://www.myhdl.org
VHDL development, the modern way: http://www.sigasi.com
World-class digital design: http://www.easics.com
 
Jan Decaluwe <jan@jandecaluwe.com> wrote:

The Python community is about to offer us a free lunch.
A new compliant interpreter, pypy, is already 4.3x faster
than cPython, and getting faster everyday. It shows
that there is not conceptual reason why high-level
dynamic languages should be slow.
Did you even try this with your program?
http://psyco.sourceforge.net/

Sounds more than one group is speeding up Python.

--
Failure does not prove something is impossible, failure simply
indicates you are not using the right tools...
nico@nctdevpuntnl (punt=.)
--------------------------------------------------------------
 
On Aug 30, 5:27 am, Jan Decaluwe <j...@jandecaluwe.com> wrote:
For MyHDL, an HDL implemented as a Python library, the
results are even more spectacular: my benchmarks run
8-20x faster on pypy. In a single strike, this makes
MyHDL simulation performance competitive with Verilog/VHDL.

     http://myhdl.org/doku.php/performance
Impressive.
Do the pypy team have your code to use as a Version Test ?
 
On 30/08/2011 01:45, Nico Coesel wrote:
Jan Decaluwe<jan@jandecaluwe.com> wrote:

The Python community is about to offer us a free lunch.
A new compliant interpreter, pypy, is already 4.3x faster
than cPython, and getting faster everyday. It shows
that there is not conceptual reason why high-level
dynamic languages should be slow.

Did you even try this with your program?
http://psyco.sourceforge.net/

Sounds more than one group is speeding up Python.
pscyo is an interesting project, and I've used it to speed up Python
code. However, the project has been basically dead for 5 years (other
than some updated builds for newer python releases on non-Linux
platforms), since the guy behind it moved his effort over to the more
flexible and future-oriented pypy project. In particular, pscyo is
limited to 32-bit x86 cpus, while pypy is aimed at a range of targets -
including 64-bit amd64.
 
On 08/30/2011 09:19 AM, David Brown wrote:
On 30/08/2011 01:45, Nico Coesel wrote:
Jan Decaluwe<jan@jandecaluwe.com> wrote:

The Python community is about to offer us a free lunch. A new
compliant interpreter, pypy, is already 4.3x faster than cPython,
and getting faster everyday. It shows that there is not
conceptual reason why high-level dynamic languages should be
slow.

Did you even try this with your program?
http://psyco.sourceforge.net/

Sounds more than one group is speeding up Python.


pscyo is an interesting project, and I've used it to speed up Python
code. However, the project has been basically dead for 5 years (other
than some updated builds for newer python releases on non-Linux
platforms), since the guy behind it moved his effort over to the more
flexible and future-oriented pypy project. In particular, pscyo is
limited to 32-bit x86 cpus, while pypy is aimed at a range of targets
- including 64-bit amd64.
In addition, psycho doesn't accelerate generators, which are
the name of the game in MyHDL. However, through psycho I got to
track Armin Rigo as someone who can do miracles :)

--
Jan Decaluwe - Resources bvba - http://www.jandecaluwe.com
Python as a HDL: http://www.myhdl.org
VHDL development, the modern way: http://www.sigasi.com
World-class digital design: http://www.easics.com
 
On 08/30/2011 12:12 AM, Jim Granville wrote:
On Aug 30, 5:27 am, Jan Decaluwe<j...@jandecaluwe.com> wrote:
For MyHDL, an HDL implemented as a Python library, the
results are even more spectacular: my benchmarks run
8-20x faster on pypy. In a single strike, this makes
MyHDL simulation performance competitive with Verilog/VHDL.

http://myhdl.org/doku.php/performance

Impressive.
Do the pypy team have your code to use as a Version Test ?
I don't think they use it now, but they are very aware of the
project, and it's all open source, so I'll leave that to their
good judgement.

With MyHDL, I found one bug in PyPy 1.5 which has been fixed
and added to the PyPy regression test suite, which must be
very broad by now.

Jan

--
Jan Decaluwe - Resources bvba - http://www.jandecaluwe.com
Python as a HDL: http://www.myhdl.org
VHDL development, the modern way: http://www.sigasi.com
World-class digital design: http://www.easics.com
 

Welcome to EDABoard.com

Sponsor

Back
Top