faster-cpython / ideas

faster-cpython/ideas Issues

Use UTF-8 internally for strings.
Updated a month ago11
Faster access to per-interpreter globals.
Updated a month ago3
Candidate root base cycle GC
Updated 2 months ago
String search improvements
Updated 2 months ago2
Investigation: Collect "mean instruction count" metric
Updated 2 months ago9
Change MSBuild to use the Clang backend
Updated 2 months ago9
Tail-calling interpreter
Updated 2 months ago4
Superinstructions for Copy & Patch JIT
Updated 3 months ago16
Investigation: Are we using the right statistics to show improvement in our benchmarks?
Updated 3 months ago4
Lower `BEFORE_WITH` and `BEFORE_ASYNC_WITH` to attribute lookups and calls.
Closed 3 months ago1
Better handling of accessing an object's `__dict__` attribute.
Updated 3 months ago5
Should `pyperf` detect when CPython is using a JIT?
Updated 3 months ago4
PGO-ed JIT-ted startup
Updated 3 months ago1
Make `struct _typeobject` a plain C struct and use it to build immutable, shareable `PyTypeObject`s
Updated 3 months ago16
Interesting paper comparing JIT compilers for Python
Updated 3 months ago5
Deferred reference counts.
Updated 4 months ago13
Better optimization of side exits by recording type information
Updated 4 months ago1
Tagged integers
Updated 4 months ago4
Use two call stacks instead of one.
Updated 4 months ago2
Combining deferred reference counting and tagged pointers
Updated 4 months ago
Lazily create attributes that are created often, but read rarely.
Updated 4 months ago
Speed up s.startswith()
Updated 5 months ago17
Things to do for 3.13
Updated 5 months ago
Trace stitching. Only drop into tier 1 for genuine de-opts.
Closed 5 months ago
Specialize binary ops according to reference count as well as type.
Updated 6 months ago3
String concatenation optimization of known strings
Updated 6 months ago5
Investigation: Optimizer/executor/traces are behaving strangely
Closed 6 months ago23
Tier 2 optimizer: Constant Int and Float propagation, without the refcount problems
Updated 6 months ago3
Investigation: Why do some benchmarks have significantly increased memory usage with the new incremental GC?
Updated 6 months ago1
Store both function and code object in the function version cache
Updated 6 months ago3
Can we split `_INIT_CALL_PY_EXACT_ARGS` further in Tier 2?
Updated 6 months ago1
Investigation: Group benchmarks based on their profiling "characterization"
Updated 6 months ago20
Make Linux perf profiling work well with the JIT
Closed 6 months ago4
JIT: Register passing proposal & reducing stack traffic significantly
Closed 6 months ago2
Speeding up Python-to-Python calls.
Updated 6 months ago5
Low effort, precise specialization in tier 2 using lookup tables, `replace` and `replicate`.
Updated 6 months ago5
Proposal for modified frame layout in 3.13
Updated 7 months ago6
True function inlining: design discussion
Updated 7 months ago21
Data structures and code for exits from executors.
Closed 7 months ago11
Investigation: Why are pystats changing in unexpected ways?
Closed 7 months ago11
Design discussion for https://github.com/python/cpython/pull/114059
Closed 7 months ago9
Handling unlikely events with de-optimization in tier 2.
Updated 8 months ago7
Investigation: Why is the performance of the mypy2 bench so poor on main relative to 3.12.0?
Closed 8 months ago4
Reimplement `mappingproxy` as dict-like object, not a wrapper around a dict.
Updated 8 months ago2
Consider improving Tier 2 coverage in PGO data collection
Updated 9 months ago5
Classification of tier 2 exits
Updated 9 months ago2
Is it possible to implement automatic function inlining in CPython?
Closed 10 months ago1
Tier 2 IR format is not flexible enough for LOAD_ATTR_PROPERTY
Updated 10 months ago4
Constructing Executors using stack contents
Updated 10 months ago26
Interpreter state overview
Closed a year ago6