Author: Maciej Fijalkowski <fij...@gmail.com> Branch: extradoc Changeset: r4110:b71362d9af8d Date: 2012-02-29 16:21 -0800 http://bitbucket.org/pypy/extradoc/changeset/b71362d9af8d/
Log: work some more on slides diff --git a/talk/pycon2012/tutorial/slides.rst b/talk/pycon2012/tutorial/slides.rst --- a/talk/pycon2012/tutorial/slides.rst +++ b/talk/pycon2012/tutorial/slides.rst @@ -10,7 +10,11 @@ |pause| -If it's not faster, you're wasting ime. +If it's not faster, you're wasting time. + +|pause| + +But if you iterate fast, you can afford wasting time Third rule of optimization? =========================== @@ -32,6 +36,8 @@ * Avoiding function calls +* Don't write Python + Forget these ============ @@ -39,6 +45,8 @@ * Which we're going to learn about now +* You cannot speak about operations in isolation (more later) + Why PyPy? ========= @@ -89,7 +97,7 @@ * **simple** python -* if you can't understand it, JIT won't either +* if I can't understand it, JIT won't either How PyPy runs your program, involved parts ========================================== @@ -111,7 +119,7 @@ * .... goes on and on -* XXX example 1 +* example Tracing JIT =========== @@ -119,11 +127,11 @@ * once the loop gets hot, it's starting tracing (1039 runs, or 1619 function calls) -* generating operations following how the interpreter would execute them +* generating operations following how the interpreter executes them * optimizing them -* compiling to assembler (x86 only for now) +* compiling to assembler (x86, ppc or arm) PyPy's specific features ======================== @@ -134,30 +142,66 @@ * Decent tools for inspecting the generated code -XXXXXXXXXXXXXXXXXXXXXXXXXXXX +Performance characteristics - runtime +===================================== +* Runtime the same or a bit slower as CPython -* Sweetspot? +* Examples of runtime: - * CPython's sweetspot: stuff written in C + * ``list.sort`` - * PyPy's sweetspot: lots of stuff written in Python + * ``long + long`` -* http://speed.pypy.org + * ``set & set`` -* How do you hit the sweetspot? + * ``unicode.join`` - * Be in this room for the next 3 hours. + * ... -Memory -====== +Performance characteristics - JIT +================================= -* PyPy memory usage is difficult to estimate. -* Very program dependent. -* Learn to predict! +* Important notion - don't consider operations in separation -Sandbox -======= +* Always working as a loop or as a function -* We're not going to talk about it here. -* Run untrusted code. +* Heuristics to what we believe is common python + +* Often much faster than CPython once warm + +Heuristics +========== + +* What to specialize on (assuming stuff is constant) + +* Data structures + +* Relative cost of operations + +Heuristic example - dicts vs objects +==================================== + +* Dicts - an unknown set of keys, potentially large + +* Objects - a relatively stable, constant set of keys + (but not enforced) + +* Performance example + +Specialized lists +================= + +* XXX Example of how much speedup you get out of not mixing + +Itertools abuse +=============== + +XXX + +Obscure stuff +============= + +* Frame access is slow + +* List comprehension vs generator expression _______________________________________________ pypy-commit mailing list pypy-commit@python.org http://mail.python.org/mailman/listinfo/pypy-commit
