Repository navigation

[GR-79939] Improve performance of pyperformance pickle benchmarks - #1208

Merged

graalvmbot merged 39 commits into

masterfrom

msimacek/GR-79939-Improve_performance_of_pyperformance_pickle_benchmarks

Oct 13, 2026

Merged

[GR-79939] Improve performance of pyperformance pickle benchmarks#1208
graalvmbot merged 39 commits into
masterfrom
msimacek/GR-79939-Improve_performance_of_pyperformance_pickle_benchmarks

Commits on Oct 8, 2026
- Reduce temporary bytearray allocations in PPickler
  msimacek
  committed
  5edd93e
- Lazily initialize fastMemo table
  msimacek
  committed
  34df922
- Simplify pickling builtins
  msimacek
  committed
  58067cb
- Profile version and fast mode when pickling
  msimacek
  committed
  ce1f410
- Add a recursion limit to PPickler
  msimacek
  committed
  964bd5e
- Add more profiling to PPickler
  msimacek
  committed
  1011631
- Add fast path for pickling sequences of primitives
  msimacek
  committed
  f1044d3
- Add more pickler optimizations
  msimacek
  committed
  ba662f2
- Push down helper nodes into PPickler.SaveNode
  msimacek
  committed
  b60b36c
- Make pickler write helpers static
  msimacek
  committed
  afc6dd2
- Avoid graph size explosion
  msimacek
  committed
  a9f6108
- Get rid of MemoIterator
  msimacek
  committed
  ad38971
- Profile buffer bypass
  msimacek
  committed
  d867056
- Keep size estimates for buffer and memo table
  msimacek
  committed
  c31f65d
- Use more specific nodes when working with collections
  msimacek
  committed
  bc0e552
- Fix emitting OPCODE_BYTEARRAY8
  msimacek
  committed
  2d7de94
- Reintroduce BasePickleNode again
  msimacek
  committed
  2600c46
- Organize SaveNode fields
  msimacek
  committed
  c275c1d
- Move write helpers to PPickler
  msimacek
  committed
  7848db8
- Get rid of PPickler.maxOutputLen
  msimacek
  committed
  7bfa0b6
- Get rid of ByteArrayView
  msimacek
  committed
  5f78bd2
- Add unpickler fast path for decoding UTF-8
  msimacek
  committed
  eb14674
- Add profiling for seen unpickler opcodes
  msimacek
  committed
  9e2d6e5
- Pre-size dictionaries when unpickling
  msimacek
  committed
  2102b34
- Add a few branch probabilities
  msimacek
  committed
  2212a83
- Avoid copying buffer in pickle.loads
  msimacek
  committed
  16e83bf
- Profile file reads in unpickle
  msimacek
  committed
  1a5c650
- Add fast paths for pickling/unpickling large longs
  msimacek
  committed
  3047bf5
- Avoid boundary in getDottedPath
  msimacek
  committed
  538816c
- Cache unpickled names
  msimacek
  committed
  d0c6a40
- Use ImportName in unpickler
  msimacek
  committed
  2934d6d
- Push members down from BasePickleNode
  msimacek
  committed
  427c4fb
- Use the same import mechanism in pickler and unpickler
  msimacek
  committed
  a75550e
- Improve pickler graph size
  msimacek
  committed
  29988e9
- Use separate recursive SaveNodes
  msimacek
  committed
  90dc553
- Report loop counts in pickle code
  msimacek
  committed
  dff33e7
Commits on Oct 12, 2026
- Avoid out-of-bounds access on list mutation in pickler
  msimacek
  committed
  c99ab1c
Commits on Oct 13, 2026
- Downsize oversized output bytes
  msimacek
  committed
  50ce7ee
- Cap the max pickle buffer/memo size estimate at 15 MiB
  msimacek
  committed
  a1b186e

Commits

Commits on Oct 8, 2026

Commits on Oct 12, 2026

Commits on Oct 13, 2026