www.morepypy.blogspot.com - morepypy

last data update: 1969/12/31, 16:00

Server
DNS
Domain
Map
Valuation
Content

Website loading time

during the test: 2.5 s

cable connection (average): 3.44 s

DSL connection (average): 4.38 s

modem (average): 54.09 s

HTTP headers

HTTP/1.0 200 OK

Content-Type: text/html; charset=UTF-8

Expires: Fri, 14 Oct 2011 20:06:22 GMT

Date: Fri, 14 Oct 2011 20:06:22 GMT

Cache-Control: private, max-age=0

Last-Modified: Fri, 14 Oct 2011 10:36:58 GMT

ETag: "72718862-ad56-4681-9005-f0fdbd815bcf"

X-Content-Type-Options: nosniff

X-XSS-Protection: 1; mode=block

Server: GSE

Information about DNS servers

morepypy.blogspot.com

CNAME

blogspot.l.google.com

3600

Received from the first DNS server

Request to the server "morepypy.blogspot.com"
Received 39 bytes from address 194.50.112.23#53 in 86 ms
Request to the server "morepypy.blogspot.com"
You used the following DNS server:
DNS Name: ns2.bouwmeestergroep.nl
DNS Server Address: 194.50.112.23#53
DNS server aliases:

Host morepypy.blogspot.com not found: 2(SERVFAIL)
Received 39 bytes from address 194.50.112.23#53 in 85 ms

Received from the second DNS server

Request to the server "morepypy.blogspot.com"
Received 39 bytes from address 194.50.112.22#53 in 83 ms
Request to the server "morepypy.blogspot.com"
You used the following DNS server:
DNS Name: ns1.bouwmeestergroep.nl
DNS Server Address: 194.50.112.22#53
DNS server aliases:

Host morepypy.blogspot.com not found: 2(SERVFAIL)
Received 39 bytes from address 194.50.112.22#53 in 81 ms

Subdomains (the first 50)

Typos (misspells)

norepypy.blogspot.com
korepypy.blogspot.com
jorepypy.blogspot.com
mirepypy.blogspot.com
mkrepypy.blogspot.com
mlrepypy.blogspot.com
mprepypy.blogspot.com
m0repypy.blogspot.com
m9repypy.blogspot.com
moeepypy.blogspot.com
modepypy.blogspot.com
mofepypy.blogspot.com
motepypy.blogspot.com
mo5epypy.blogspot.com
mo4epypy.blogspot.com
morwpypy.blogspot.com
morspypy.blogspot.com
mordpypy.blogspot.com
morrpypy.blogspot.com
mor4pypy.blogspot.com
mor3pypy.blogspot.com
moreoypy.blogspot.com

morelypy.blogspot.com
more-ypy.blogspot.com
more0ypy.blogspot.com
moreptpy.blogspot.com
morepgpy.blogspot.com
morephpy.blogspot.com
morepupy.blogspot.com
morep7py.blogspot.com
morep6py.blogspot.com
morepyoy.blogspot.com
morepyly.blogspot.com
morepy-y.blogspot.com
morepy0y.blogspot.com
morepypt.blogspot.com
morepypg.blogspot.com
morepyph.blogspot.com
morepypu.blogspot.com
morepyp7.blogspot.com
morepyp6.blogspot.com
orepypy.blogspot.com
mrepypy.blogspot.com
moepypy.blogspot.com

morpypy.blogspot.com
moreypy.blogspot.com
moreppy.blogspot.com
morepyy.blogspot.com
morepyp.blogspot.com
omrepypy.blogspot.com
mroepypy.blogspot.com
moerpypy.blogspot.com
morpeypy.blogspot.com
moreyppy.blogspot.com
moreppyy.blogspot.com
morepyyp.blogspot.com
mmorepypy.blogspot.com
moorepypy.blogspot.com
morrepypy.blogspot.com
moreepypy.blogspot.com
moreppypy.blogspot.com
morepyypy.blogspot.com
morepyppy.blogspot.com
morepypyy.blogspot.com

Location

IP: 209.85.175.132

continent: NA, country: United States (USA), city: Mountain View

Website value

rank in the traffic statistics:

There is not enough data to estimate website value.

Basic information

website build using CSS

code weight: 361.14 KB

text per all code ratio: 47 %

title: PyPy Status Blog

description:

keywords:

encoding: UTF-8

language: en

Website code analysis

one word phrases repeated minimum three times

Phrase	Quantity
the	1131
to	645
of	494
is	416
and	404
in	348
that	318
for	214
we	202
it	194
on	186
be	157
PyPy	151
are	150
as	136
by	132
with	130
this	129
not	129
you	123
can	104
or	93
The	93
at	89
have	89
which	85
Python	82
This	79
an	79
has	77
will	71
but	70
CPython	67
from	66
would	66
one	65
need	63
been	55
all	54
then	52
code	52
if	50
It	50
list	48
more	47
like	46
using	46
In	46
than	45
do	44
use	43
JIT	42
only	42
two	41
how	38
up	38
your	38
We	37
example	36
want	36
work	36
implementation	36
very	35
any	34
get	34
our	33
NumPy	33
about	33
so	32
For	32
---	32
actually	32
faster	31
other	30
possible	30
there	30
they	30
such	30
solution	30
run	30
PyPy,	30
If	30
just	30
objects	28
same	28
case	28
way	28
fast	28
now	28
some	28
post	27
every	26
C++	26
problem	26
also	26
STM	26
into	24
these	24
release	24
memory	24
because	24
when	24
people	23
threads	22
support	22
used	22
know	22
what	22
operations	22
time	22
much	22
locks	22
was	22
object	22
performance	22
make	22
Transactional	20
both	20
approach	20
cannot	20
data	20
it's	20
different	20
many	20
since	20
Jython	20
see	20
gives	20
us	20
2011	19
most	19
over	19
enough	18
each	18
being	18
lists	18
something	18
library	18
does	18
still	18
already	18
end	18
interface	18
no	18
well	18
number	18
several	18
methods	18
them	18
working	18
should	18
single	17
really	17
it.	17
There	17
instead	16
GIL	16
between	16
either	16
example,	16
transaction	16
give	16
try	16
compiled	16
global	16
first	16
out	16
here	16
Facebook	16
means	16
thread	16
lot	16
machine	16
bytecode	16
might	16
Software	16
Posted	16
things	16
ThisBlogThis!Share	16
TwitterShare	16
To	16
Email	16
Comments	16
interpreter	16
level	16
Links	16
existing	15
language	15
doing	15
point	14
store	14
start	14
needs	14
far	14
:-)	14
You	14
CPython,	14
few	14
mailing	14
makes	14
think	14
we're	14
next	14
often	14
making	14
change	14
call	14
general	14
back	14
integers	14
above	14
even	14
jitviewer	14
works	14
Memory	14
field	14
stored	14
write	14
under	14
idea	13
were	13
PyPy.	13
RPython	13
W_ListObject	12
multiple	12
optimized	12
place	12
extra	12
But	12
unwrapped	12
another	12
times	12
you're	12
comparison	12
uses	12
source	12
takes	12
processing	12
Python,	12
As	12
model	12
writing	12
their	12
lp.add_constraint(	12
Alex	12
On	12
storage	12
"%d	12
solve	12
must	12
recent	12
list,	12
similar	12
C,	12
def	12
(e.g.	12
due	12
However,	12
running	12
don't	11
without	11
long	11
branch	11
where	11
module	11
programmer	10
I'm	10
it,	10
access	10
benchmark	10
execution	10
GIL.	10
called	10
could	10
strategies	10
fps	10
able	10
One	10
time,	10
form	10
its	10
written	10
occur	10
improve	10
information	10
arbitrary	10
image	10
sprint	10
appropriate	10
take	10
why	10
operation	10
classes	10
through	10
ctypes	10
string	10
arrays	10
last	10
including	10
blog	10
actual	10
large	10
who	10
(which	10
locking	10
look	10
improvements	10
fact	10
Java	10
Since	10
return	10
new	10
program	10
did	10
reading	10
comes	10
high-level	10
optimization	10
virtual	10
types	10
version	10
kind	10
while	10
JIT,	10
issues	10
implementations	10
important	10
transformation	10
top	10
Hi	10
At	9
contains	9
my	9
automatically	9
better	9
(or	9
strings	8
complex	8
good	8
right	8
values	8
result	8
JIT.	8
1.6	8
contributed:	8
apply	8
generated	8
almost	8
bindings	8
expect	8
i);	8
currently	8
always	8
simple	8
writes	8
implements	8
mean	8
interested	8
char	8
everything	8
multi-threaded	8
reflection	8
strong	8
%d",	8
sprintf(x,	8
C-API	8
cppyy	8
By	8
wrote	8
reads	8
announce	8
Here	8
Some	8
details	8
process	8
interpreter.	8
please	8
transaction,	8
we've	8
Carl	8
results	8
Friedrich	8
and/or	8
dictionaries	8
bytecodes,	8
started	8
step	8
after	8
all.	8
minutes	8
let	8
probably	8
array	8
Global	8
tool	8
little	8
team	8
It's	8
Interpreter	8
default	8
pypy	8
i,	8
slow.	8
f(list1,	8
looking	8
int	8
bytecode,	8
code:	8
objects,	8
plan	8
directly	8
pixel	8
allows	8
loop,	8
speed	8
platform	8
dynamic	8
list2):	8
0;	8
Python.	8
fixed	8
bugs	8
quite	8
second	8
storing	8
implement	8
These	8
manual	8
add	8
video	8
getting	8
object,	8
that's	8
type	8
#include	8
lists,	8
particular,	8
those	8
careful	8
interpreter,	8
bug	8
topics	8
may	8
side	8
object.	8
GCC	8
overhead	8
going	8
adding	8
demo	8
yet	8
Now	8
needed	8
find	8
thing	7
implemented	7
Armin	7
Rigo	7
thus	7
else	6
supports	6
runs	6
extension	6
sets	6
API	6
>=	6
effect	6
guarantees	6
Jython:	6
spend	6
dropped	6
application	6
itself	6
lower	6
old	6
cpyext,	6
lodging	6
pixels	6
isn't	6
magnify.py	6
here.	6
open	6
situation	6
Also,	6
following	6
required	6
compatibility	6
optimizations	6
hard	6
solved	6
nicer	6
atomic	6
released	6
purpose	6
code.	6
That	6
room:	6
given	6
short	6
thanks	6
took	6
algorithm	6
however	6
glpk	6
interpolation	6
CPython:	6
What	6
branch.	6
progressed	6
€	6
system	6
lock	6
rather	6
paper	6
<=	6
hints	6
64	6
Please	6
1.5,	6
said	6
20	6
Windows	6
fun,	6
set	6
Lock	6
trying	6
fully	6
current	6
seen	6
cases	6
improved	6
consistency	6
main()	6
discuss	6
difference	6
wrapped	6
Jython,	6
contact	6
recently	6
append	6
language,	6
function	6
assembler	6
believe	6
seems	6
shows	6
runtime	6
items	6
wrapping	6
check	6
focus	6
stuff	6
show	6
seconds	6
switch	6
CPython's	6
Of	6
Strategy	6
pip	6
course	6
considered	6
strategy	6
We've	6
points	6
explicitly	6
installed	6
GIL,	6
Note	6
Thread	6
remove	6
Indeed,	6
amount	6
various	6
types,	6
record	6
More	6
own	6
fix	6
OS	6
arrays.	6
added	6
log	6
funding	6
mail	6
Wednesday,	6
everyone	6
"read"	6
fijal	6
found	6
perfect	6
original	6
put	6
represent	6
(STM)	6
precisely	6
list2.append(x)	6
list1.pop()	6
resulting	6
value	6
So	6
come	6
Maciej	6
doesn't	6
calls	6
down	6
Instead	6
above,	6
whole	6
From	6
benefits	6
objects.	6
order	6
built-in	6
nice	6
includes	6
users	6
context,	6
notice	6
e.g.	6
June	6
API.	6
clear	6
May	6
helps	6
rewrite	6
numpy	6
offer	6
size	6
floats	6
Cheers,	6
programs	6
internal	6
them.	6
hope	6
noted	6
install	6
special	6
reported	6
best	6
slightly	5
August	5
past	5
July	5
tries	5
is,	4
off	4
left	4
Actually,	4
With	4
Moreover,	4
state	4
moved	4
piece	4
lines	4
PyPy:	4
serialized	4
atomicity	4
pypy-dev	4
usage	4
abort.	4
involves	4
other,	4
fine	4
abort	4
well,	4
full	4
involve	4
away	4
Rigo)	4
far,	4
numbers	4
value,	4
overall	4
yet.	4
mind,	4
restart	4
clue,	4
confused	4
issue	4
precise	4
capacity	4
everywhere.	4
looks	4
fixing	4
During	4
guarantees.	4
detect	4
say	4
multi-core	4
thread-local	4
structure	4
And	4
pop()	4
shown	4
released.	4
great,	4
"write"	4
yet,	4
copying	4
"commit"	4
usage.	4
atomically	4
exactly	4
prevents	4
main	4
Bolz,	4
setting	4
Before	4
particular	4
multi-threading.	4
2.	4
(3)	4
feature	4
anything	4
languages	4
guarantees:	4
offers	4
consider	4
share	4
list2.lock)	4
here:	4
(This	4
below	4
(start	4
IRC	4
acquire	4
details,	4
list.	4
scales	4
stronger	4
(ends	4
Gaynor,	4
Because	4
impossible	4
sense	4
research.	4
common	4
abstractions	4
least	4
new,	4
Google	4
huge	4
requires	4
attempt	4
Thanks,	4
carefully	4
Comprehensive	4
problem.	4
turned	4
produced	4
27	4
First,	4
report	4
respondants	4
based	4
three	4
bit	4
previous	4
rare	4
multithreaded	4
bad	4
introducing	4
Jython's	4
soon	4
we'd	4
examples	4
optimize	4
described	4
Finally,	4
manually	4
production	4
fail	4
subtle	4
People	4
it:	4
Contention	4
provide	4
understanding	4
standard	4
solutions	4
uncommon	4
EuroPython	4
cases,	4
notably	4
Management	4
lock,	4
Genova-Pegli,	4
week	4
easily	4
maybe	4
(for	4
too.	4
explicit	4
2x	4
crowdfunding	4
GC	4
throughout	4
Jython's,	4
roughly	4
introduction	4
small	4
middle	4
(Armin	4
mutated	4
maps,	4
completely	4
Michael	4
Say	4
locks,	4
multi-threading	4
does;	4
break	4
All	4
(in	4
draw-back	4
happen.	4
beach,	4
globally	4
whole-program	4
easy,	4
all,	4
ones	4
quick	4
attention	4
protected	4
one.	4
around	4
(In	4
fact,	4
method	4
arrange	4
acquired	4
instead,	4
willing	4
rewriting	4
compiler	4
perform	4
hash	4
operator	4
hand	4
closer	4
finally	4
10*x	4
specialize	4
opcodes	4
parts	4
Report	4
List	4
smaller	4
depending	4
9.43	4
pplp	4
lives	4
lp.IntVar(),	4
year	4
explain	4
read	4
PyPy's	4
An	4
equivalent	4
inside	4
merged	4
well.	4
clearly	4
string,	4
post,	4
strings,	4
maxval	4
Bolz	4
print	4
necessary	4
coming	4
cvxopt	4
symbolically.	4
natural	4
be:	4
library.	4
task.	4
problems	4
returns	4
x[44];	4
1.5.	4
data,	4
work.	4
1.5	4
real	4
generators	4
user	4
inlined	4
c-code	4
support.	4
available	4
time.	4
Therefore,	4
representation	4
Tuesday,	4
entered	4
equations	4
Fijalkowski	4
big	4
(e.g.,	4
love	4
cross-language	4
layers	4
jit	4
stop	4
When	4
faster.	4
save	4
pure	4
via	4
creating	4
indirections	4
changed,	4
directly,	4
there's	4
(as	4
jitviewer.	4
public	4
planned.	4
Other	4
potential	4
unroll-if-alt	4
data.	4
instance,	4
help	4
development	4
i++)	4
(i	4
10000000;	4
items,	4
us,	4
storage.	4
Java.	4
ListStrategy	4
	4
problem,	4
	4
web	4
:-).	4
inline	4
describing	4
savings	4
*x	4
contain	4
instance	4
million	4
sizeof(char));	4
malloc(44	4
won't	4
types.	4
become	4
primitive	4
can't	4
(and	4
Unfortunately,	4
CPython.	4
done	4
case,	4
form,	4
content	4
too	4
rid	4
indirection	4
sprintf	4
NumPy,	4
cases.	4
datatypes.	4
C-extension	4
Antonio	4
overview	4
cpyext	4
containing	4
certain	4
it?	4
build	4
never	4
alternative	4
average	4
fails	4
back-end	4
pop	4
during	4
high	4
HEP,	4
day	4
within	4
32	4
(it	4
codespeak.net	4
include	4
(i,	4
%d"	4
made	4
improvement	4
compilation	4
i)	4
themselves	4
interfaces	4
20%	4
Linear	4
donation	4
hi-level	4
update	4
to,	4
what's	4
input	4
are:	4
missing	4
support,	4
create	4
meaning	4
loading	4
campaign	4
code,	4
archive	4
note:	4
copied,	4
details.	4
extensions.	4
jobs	4
matrices	4
bunch	4
integrated	4
prebuilt	4
formatting	4
before	4
histogram	4
analysis	4
work,	4
class	4
beta	4
values.	4
random	4
Most	4
you'll	4
Düsseldorf	4
system.	4
register	4
send	4
sprint,	4
sobel.py	4
warmup	4
Performance	4
We're	4
garbage	4
bool	4
behavior	4
specific	4
matrix	4
reuse	4
sure	4
fun.	4
transformed	4
basic	4
whether	4
output	4
Status	4
especially	4
x,	4
mplayer	4
factor	4
python	4
far).	4
allow	4
done.	4
length	4
one,	4
that,	4
progress	4
misses	4
pleased	4
status	3
Numpy	3
requested	3
Hello.	3
2,	3
features	3
describe	3
Thursday,	3
numpy-exp	3
50	3
Hakan	3
benchmarks	3
money	3
improvements,	3
September	3
longer	3
24	3
achieve	3
(i.e.	3
part	3
documented	3

two word phrases repeated minimum three times

Phrase	Quantity
of the	135
to the	74
in the	67
to be	56
it is	42
need to	42
on the	40
is not	31
and the	30
This is	30
want to	28
by the	28
from the	28
the same	26
with the	26
is an	24
the JIT	24
PyPy is	23
It is	22
implementation of	22
at the	21
you can	21
can be	21
the Python	20
if you	20
that we	20
for the	19
this post	18
will be	18
we can	18
the list	18
you need	18
that it	18
faster than	18
all the	18
for example	18
that the	18
to use	17
has been	17
number of	16
lot of	16
to Facebook	16
Email ThisBlogThis!Share	16
TwitterShare to	16
have been	16
to TwitterShare	16
ThisBlogThis!Share to	16
to get	16
Facebook Links	16
Posted by	16
to this	16
Links to	16
Comments Email	16
Software Transactional	16
how to	15
to do	14
we need	14
the implementation	14
would be	14
the GIL	14
Transactional Memory	14
to run	14
then we	14
should be	14
there are	14
as the	14
is the	14
working on	14
list of	14
and then	13
one of	13
of this	12
you have	12
case of	12
to have	12
does not	12
that you	12
is possible	12
this is	12
possible to	12
like to	12
in CPython	12
know that	12
well as	12
is to	12
has to	12
due to	12
the case	12
there is	12
we are	11
is that	11
is very	10
top of	10
Python code	10
way to	10
You can	10
the problem	10
kind of	10
This means	10
enough to	10
are not	10
solution is	10
instead of	10
to work	10
support for	10
able to	10
that is	10
code and	10
as well	10
fact that	10
to make	10
we have	10
would like	10
the JIT,	10
the fact	10
the CPython	10
do not	10
it in	10
the PyPy	10
not the	10
that they	10
cannot be	10
will see	10
in PyPy	10
In the	10
try to	10
the first	10
over the	9
using PyPy	8
interested in	8
we want	8
it would	8
you want	8
which is	8
than PyPy	8
virtual machine	8
is now	8
type of	8
Global Interpreter	8
the most	8
in which	8
that are	8
on this	8
means that	8
end up	8
change the	8
the two	8
they are	8
level of	8
only the	8
the interpreter	8
it can	8
PyPy team	8
blog post	8
the mailing	8
to solve	8
you will	8
f(list1, list2):	8
def f(list1,	8
the jitviewer	8
needs to	8
it to	8
has already	8
problem that	8
source code	8
Carl Friedrich	8
already been	8
the idea	8
the next	8
i, i);	8
There are	8
and has	8
your Python	8
or the	8
get the	8
in C,	8
has the	8
we would	8
used to	8
Here is	8
for this	8
sprintf(x, "%d	8
similar to	8
"%d %d",	8
of PyPy	8
of an	8
%d", i,	8
objects and	8
PyPy and	8
use the	7
At the	7
Armin Rigo	7
in RPython	7
but not	6
will not	6
the process	6
resulting in	6
an example	6
amount of	6
here is	6
the code	6
such as	6
the W_ListObject	6
example in	6
Some of	6
the extra	6
CPython and	6
the sprint	6
both threads	6
to represent	6
must be	6
so that	6
transformation that	6
much faster	6
locks in	6
that would	6
at all.	6
for all	6
up the	6
of our	6
fast enough	6
the actual	6
the interpreter.	6
with Python	6
switch to	6
the things	6
For example,	6
in PyPy.	6
the image	6
have to	6
time is	6
not just	6
of what	6
we will	6
lp.add_constraint( >=	6
we wrote	6
and it	6
since we	6
it will	6
is still	6
--- and	6
In this	6
is probably	6
not been	6
have not	6
to it.	6
operations on	6
minutes to	6
to write	6
execution of	6
we release	6
if it	6
Interpreter Lock	6
There is	6
solution to	6
the type	6
that in	6
list1.pop() list2.append(x)	6
the start	6
the GIL.	6
which objects	6
we do	6
look at	6
us to	6
at all	6
noted that	6
to give	6
on PyPy	6
on PyPy,	6
the Global	6
are using	6
rather than	6
in all	6
using it	6
In particular,	6
that this	6
Python source	6
on top	6
PyPy in	6
field of	6
do it	6
thanks to	6
Note that	6
run the	6
PyPy, the	6
into the	6
some of	6
PyPy, and	6
the global	6
the other	6
the end	6
an implementation	6
the whole	6
use of	6
that cannot	6
in your	6
locks that	6
the way	6
If we	6
the transaction	6
or any	6
pixel access	6
model for	6
the STM	6
that can	6
the original	6
written in	6
that of	6
in that	6
here to	6
the JIT.	6
as CPython	6
because it	6
to announce	6
and we	6
algorithm in	6
is faster	6
the storage	6
you do	6
the data	6
for PyPy	6
the best	6
for example,	6
multiple threads	6
is only	6
For the	6
the appropriate	6
CPython C-API	6
times faster	6
the execution	6
remove the	6
think that	6
for CPython	6
been compiled	6
compiled by	6
things like	6
NumPy arrays.	6
the past	5
tries to	5
on how	5
post Wednesday,	5
storage is	4
are actually	4
#include	4
Python interpreter,	4
to lock	4
the problem,	4
the list,	4
solve the	4
lists, and	4
that instead	4
and in	4
are two	4
through the	4
attempt to	4
int	4
we use	4
two threads	4
#include	4
the code:	4
#include	4
way of	4
reported by	4
to find	4
work with	4
This would	4
can see	4
this context,	4
implemented by	4
willing to	4
append or	4
code: #include	4
Now that	4
are as	4
in this	4
not as	4
fast PyPy	4
an object,	4
int 0;	4
to report	4
video processing	4
on every	4
Hi all,	4
is actually	4
can do	4
the previous	4
in getting	4
ones in	4
that gives	4
the ones	4
are in	4
main() int	4
one which	4
however the	4
0; 10000000;	4
of you	4
the equations	4
objects in	4
int main()	4
is in	4
know how	4
for (i	4
be entered	4
entered symbolically.	4
(i 0;	4
equations to	4
occur in	4
people are	4
write it	4
Contention Management	4
for Contention	4
Strategy for	4
Management in	4
in Software	4
might be:	4
be: char	4
char x[44];	4
Comprehensive Strategy	4
to know	4
but if	4
list we	4
we just	4
in several	4
like the	4
interfaces often	4
of their	4
hash maps,	4
and such	4
floats and	4
given that	4
been mutated	4
strategies for	4
which means	4
Python programmer	4
check that	4
of people	4
about how	4
is often	4
integers and	4
of thread	4
in Python	4
improvements to	4
bytecode of	4
in bytecode	4
and strings	4
as well,	4
well, instead	4
but that's	4
might get	4
operations are	4
allow the	4
rewrite the	4
*x malloc(44	4
the new	4
if you're	4
you're using	4
malloc(44 sizeof(char));	4
this was	4
>= lp.add_constraint(	4
and another	4
writes to	4
few minutes	4
then the	4
the program	4
char *x	4
methods for	4
:-) This	4
lives in	4
of PyPy,	4
hard to	4
the lines	4
the middle	4
overview of	4
not store	4
middle of	4
give the	4
is far	4
the beach,	4
like Python	4
specialize the	4
same level	4
sizeof(char)); sprintf(x,	4
believe that	4
difference is	4
if we	4
if the	4
problem has	4
mutated by	4
to change	4
with existing	4
used strategy	4
out of	4
for other	4
JIT in	4
protected by	4
20 minutes	4
and put	4
enough for	4
is able	4
global program	4
been dropped	4
mailing list,	4
mail to	4
for dynamic	4
or send	4
send us	4
point of	4
it because	4
then there	4
in thread	4
approach that	4
to our	4
in situation	4
and it's	4
This can	4
up in	4
in Genova-Pegli,	4
PyPy JIT	4
(Armin Rigo)	4
draw-back of	4
For instance,	4
this solution	4
Memory (STM)	4
types are	4
an approach	4
details of	4
threads in	4
Of course	4
the time	4
thread and	4
of reflection	4
the following	4
and thus	4
throughout the	4
reflection information	4
original NumPy	4
fps CPython:	4
the Jython	4
that list	4
list and	4
(i, i)	4
%d" (i,	4
by thread	4
"%d %d"	4
for each	4
save one	4
it gives	4
carefully adding	4
Python is	4
not possible	4
be slow.	4
we write	4
far too	4
one is	4
very carefully	4
thread 2.	4
is something	4
is needed	4
doing it	4
be in	4
the C++	4
is currently	4
would mean	4
what is	4
any operations	4
the pixels	4
of Python	4
well with	4
this kind	4
case you	4
language, as	4
or by	4
side by	4
code from	4
fixing issues	4
your system.	4
on your	4
in case	4
been solved	4
PyPy 1.5.	4
If you're	4
Düsseldorf sprint,	4
like CPython	4
you know	4
We think	4
all of	4
the garbage	4
yet, and	4
the pixel	4
existing multi-threaded	4
is, then	4
depending on	4
interface that	4
version of	4
the PyPy's	4
very recent	4
is written	4
on Software	4
example of	4
the sprintf	4
consistency is	4
Python in	4
and by	4
representation of	4
important to	4
great, but	4
it might	4
you're interested	4
STM to	4
it. The	4
on IRC	4
to try	4
to explain	4
would work	4
to remove	4
you just	4
just want	4
one or	4
prebuilt PyPy	4
objects we	4
CPython, but	4
the manual	4
and you	4
clue, but	4
no clue,	4
on such	4
PyPy are	4
The second	4
and they	4
Since we	4
Python implementation	4
as an	4
whole PyPy	4
than one	4
implementation in	4
in Python,	4
NumPy is	4
to form	4
points at	4
one possible	4
by Maciej	4
are the	4
access method	4
up to	4
not actually	4
let the	4
Maciej Fijalkowski	4
Fijalkowski at	4
post Tuesday,	4
focus of	4
strong guarantees.	4
let us	4
the last	4
We would	4
would also	4
us know	4
C-API compatibility	4
as long	4
since the	4
to, and	4
clear that	4
it should	4
state of	4
which one	4
the matrices	4
that PyPy	4
such an	4
mind, it	4
in mind,	4
campaign for	4
long as	4
by writing	4
mean that	4
Friedrich Bolz,	4
this does	4
includes the	4
needed for	4
multi-threading at	4
they did	4
or not	4
be used	4
be transformed	4
Friedrich Bolz	4
so we	4
back-end is	4
be noted	4
the current	4
than the	4
as Jython	4
very fast	4
we detect	4
the internal	4
based on	4
is slightly	4
end of	4
when the	4
factor of	4
jitviewer is	4
to apply	4
the pypy-dev	4
for these	4
you to	4
abort. This	4
are also	4
out on	4
and abort.	4
setting up	4
that takes	4
under PyPy	4
looks like	4
we only	4
only have	4
pypy-dev mailing	4
of all	4
the bytecode	4
idea of	4
on NumPy	4
is going	4
start of	4
think PyPy	4
that have	4
the bug	4
we think	4
the log	4
with PyPy	4
The jitviewer	4
more fun,	4
These are	4
in some	4
API. This	4
bunch of	4
of different	4
In case	4
to fix	4
the glpk	4
would need	4
be fast	4
looking at	4
with some	4
The PyPy	4
object that	4
to optimize	4
point to	4
the user	4
PyPy to	4
CPython or	4
The opcodes	4
the demo	4
as general	4
is model	4
but the	4
the performance	4
is clearly	4
for every	4
performance is	4
comes to	4
do one	4
run it	4
release the	4
The example	4
with ctypes	4
notice that	4
C++ code.	4
JIT has	4
it has	4
(e.g. for	4
the standard	4
CPython, the	4
when it	4
be to	4
it comes	4
have progressed	4
in pure	4
However, in	4
rid of	4
pixels in	4
be able	4
work on	4
the compiled	4
wrapped or	4
or unwrapped	4
been copied,	4
under PyPy,	4
work as	4
PyPy 1.6	4
2011 PyPy	4
place for	4
data. This	4
the above	4
is with	4
possible in	4
fix it	4
10000000; i++)	4
to rewrite	4
Alex at	4
piece of	4
problems that	4
to an	4
W_ListObject point	4
of Python,	4
This release	4
complex high-level	4
we cannot	4
something like	4
mailing lists,	4
JIT compiler	4
with locks	4
comes with	4
release of	4
interface with	4
We are	4
come with	4
make the	4
by Alex	4
pleased to	4
Python side	4
and that	4
announce the	4
far as	4
loop, the	4
find the	4
using the	4
abort and	4
and topics	4
If you	4
which has	4
then this	4
GIL. It	4
code which	4
gives the	4
bytecode we	4
data is	4
the right	4
set of	4
because the	4
acquire the	4
in many	4
many of	4
we acquire	4
see the	4
in list,	4
We believe	4
the source	4
is stored	4
in it	4
of Python.	4
It also	4
that these	4
this blog	4
global memory	4
and let	4
Alex Gaynor,	4
list strategies	4
the top	4
hope that	4
large amount	4
seconds to	4
archive contains	4
does the	4
this archive	4
in way	4
use it	4
bytecode (or	4
one bytecode	4
not yet	4
of one	4
integers is	4
(or actually	4
place to	4
up with	4
possible that	4
it still	4
at least	4
seems to	4
done by	4
of time	4
form of	4
the form	4
making it	4
acquired and	4
(ends the	4
topics of	4
this release	4
it's not	4
but you	4
two operations	4
with each	4
the difference	4
than CPython	4
to discuss	4
discuss the	4
feature that	4
that no	4
will end	4
of bytecodes,	4
called the	4
after the	4
the objects	4
which the	4
the threads	4
on our	4
time of	4
general JIT	3
Tuesday, August	3
have plan	3
PyPy has	3
Rigo at	3
and how	3
we don't	3
is one	3
on an	3
(i.e. the	3
by Armin	3
post Thursday,	3
which contains	3
show that	3
of extra	3
which are	3
PyPy. This	3
most requested	3
the existing	3
numpy in	3
of very	3
idea is	3
the module	3

three word phrases repeated minimum three times

Phrase	Quantity
to TwitterShare to	16
ThisBlogThis!Share to TwitterShare	16
Email ThisBlogThis!Share to	16
TwitterShare to Facebook	16
to Facebook Links	16
Links to this	16
Facebook Links to	16
Comments Email ThisBlogThis!Share	16
to this post	16
the implementation of	12
Software Transactional Memory	12
you need to	12
we need to	12
the case of	12
as well as	10
would like to	10
the fact that	10
it is possible	10
one of the	9
faster than PyPy	8
need to be	8
sprintf(x, "%d %d",	8
is not the	8
"%d %d", i,	8
%d", i, i);	8
we want to	8
in the case	8
def f(list1, list2):	8
has already been	8
by the JIT,	8
has to be	8
it is not	8
you want to	6
want to use	6
some of the	6
need to have	6
the type of	6
on top of	6
have not been	6
compiled by the	6
top of the	6
This means that	6
you will see	6
switch to the	6
the list of	6
the Global Interpreter	6
Global Interpreter Lock	6
that cannot be	6
times faster than	6
Some of the	6
well as the	6
of the things	6
of the list	6
this post Wednesday,	5
are using it	4
to announce the	4
(i 0; 10000000;	4
using it because	4
be entered symbolically.	4
0; 10000000; i++)	4
code: #include	4
the start of	4
will see the	4
with the JIT	4
the top of	4
much faster than	4
by the JIT.	4
This is not	4
that they are	4
let us know	4
here is an	4
the code: #include	4
int main() int	4
main() int 0;	4
Here is an	4
int main()	4
#include int	4
to be entered	4
#include #include	4
#include	4
for (i 0;	4
improvements to the	4
this post Tuesday,	4
Maciej Fijalkowski at	4
by Maciej Fijalkowski	4
possible to use	4
then there is	4
of the glpk	4
the pixel access	4
try to explain	4
Posted by Maciej	4
Python source code	4
implementation of the	4
solution to the	4
you're interested in	4
as the previous	4
the whole PyPy	4
be able to	4
to be able	4
whole PyPy team	4
this kind of	4
be noted that	4
the global memory	4
If you have	4
and abort. This	4
think PyPy is	4
we think PyPy	4
just want to	4
you just want	4
that in many	4
for example in	4
which objects we	4
due to the	4
pixel access method	4
has been solved	4
the same level	4
This is an	4
that of the	4
large amount of	4
to run the	4
the equations to	4
to get the	4
the original NumPy	4
to work with	4
amount of time	4
the ones in	4
is faster than	4
PyPy is faster	4
Contention Management in	4
Management in Software	4
in Software Transactional	4
if you need	4
"%d %d" (i,	4
%d" (i, i)	4
middle of the	4
to be fast	4
It is an	4
on Software Transactional	4
the middle of	4
we use the	4
Carl Friedrich Bolz,	4
In this context,	4
of an object,	4
It is only	4
for Contention Management	4
Strategy for Contention	4
This can be	4
to the list	4
will be slow.	4
or send us	4
details of the	4
of the CPython	4
on the beach,	4
very carefully adding	4
need to change	4
of the time	4
in bytecode of	4
well, instead of	4
if you do	4
that you can	4
to remove the	4
Comprehensive Strategy for	4
Now that we	4
up in situation	4
as well, instead	4
bytecode of thread	4
the Python programmer	4
might be: char	4
already been mutated	4
in which the	4
it is with	4
If you're interested	4
the idea of	4
fix it is	4
to fix it	4
the list we	4
you're using PyPy	4
will end up	4
way to fix	4
then we can	4
field of an	4
malloc(44 sizeof(char)); sprintf(x,	4
is model for	4
sizeof(char)); sprintf(x, "%d	4
know that in	4
*x malloc(44 sizeof(char));	4
char *x malloc(44	4
>= lp.add_constraint( >=	4
lp.add_constraint( >= lp.add_constraint(	4
the pypy-dev mailing	4
the JIT has	4
is possible that	4
the W_ListObject point	4
make the W_ListObject	4
to make the	4
is that it	4
is not possible	4
wrapped or unwrapped	4
been mutated by	4
mutated by thread	4
integers and strings	4
implementation of Python	4
we would like	4
that the implementation	4
few minutes to	4
when it comes	4
it comes to	4
the end of	4
has the problem	4
the problem that	4
Transactional Memory (STM)	4
an approach that	4
end of the	4
that we can	4
equations to be	4
the JIT in	4
the Python side	4
Posted by Alex	4
we acquire the	4
needs to be	4
fast enough to	4
is, then we	4
This is very	4
as long as	4
on NumPy arrays.	4
that it is	4
Carl Friedrich Bolz	4
for all the	4
your Python code	4
with locks that	4
is possible to	4
possible to get	4
no clue, but	4
means that it	4
problem that takes	4
to be transformed	4
than PyPy 1.5.	4
of all the	4
been compiled by	4
have been compiled	4
by Alex at	4
do it in	4
we only have	4
have been dropped	4
lot of the	4
this archive contains	4
faster than CPython	4
There are two	4
the execution of	4
it should be	4
we release the	4
this does not	4
is able to	4
one bytecode (or	4
the pixels in	4
pixels in the	4
the PyPy JIT	4
bytecode (or actually	4
that it can	4
seems to be	4
in mind, it	4
the mailing lists,	4
for the C++	4
you know that	4
on your system.	4
Armin Rigo at	3
post Tuesday, August	3
on how to	3
is one of	3
Posted by Armin	3
by Armin Rigo	3
this post Thursday,	3

B tags

Update:

590 times faster

U tags

I tags

some

it is not fun :-)

Hardware

confirm

images

file name	alternative text
pypy_small128.png
icon18_edit_allbkg.gif
overhead_before.png
with_special_impl.png
with_strategies.png
osheapsize.png
Pegli.jpg
eqsource1.png
eqsource4.png
eqsource3.png
eqsource2.png
idle.png
icon18_wrench_allbkg.png
pixel.gif
feed-icon32x32.png
PyPyStatusBlog?bg=FF9900&fg=000000&anim=0

headers

Wednesday, October 12, 2011

Tuesday, October 11, 2011

List Strategies

Microbenchmarks

Conclusion

Wednesday, September 21, 2011

Tuesday, August 30, 2011

Tuesday, August 23, 2011

The problem in complex high-level languages

Jython: fine-grained locking

CPython: coarse-grained locking

Existing usage

PyPy

Using Software Transactional Memory

Performance

The state of STM

Thursday, August 18, 2011

Friday, August 12, 2011

Tuesday, August 2, 2011

Thursday, July 7, 2011

Wednesday, June 29, 2011

Wednesday, June 8, 2011

Monday, May 23, 2011

Monday, May 16, 2011

Sunday, May 15, 2011

Wednesday, May 11, 2011

Thursday, May 5, 2011

Wednesday, May 4, 2011

Saturday, April 30, 2011

What is PyPy?

More highlights

Wednesday, April 20, 2011

Wednesday, April 6, 2011

Adding JIT

A bit about Tracing JIT Compilers

Debugging and Trace Logs

Optimizing

Final Words

Links of Interest

Donate

Blog Archive

Contributors

Subscribe Now

Subscriber Count

Google Analytics