Repository navigation
Lower BEFORE_WITH and BEFORE_ASYNC_WITH to attribute lookups and calls. #120507
Description
With the JIT and better tier 2 optimizations, effective specialization is becoming more important that plain speed in tier 1.
BEFORE_WITH and BEFORE_ASYNC_WITH could be specialized, but they are bulky and won't optimize well in tier 2.
Instead, we should lower them to attribute lookups and calls which can then be optimized.
We should add a LOAD_SPECIAL instruction for loading dunder methods and replace BEFORE_WITH as follows:
COPY 1
LOAD_SPECIAL __enter__ + NULL|self
SWAP 3
LOAD_SPECIAL __exit__
SWAP 3
CALL 0
Likewise for BEFORE_ASYNC_WITH.
Even without any specialization of LOAD_SPECIAL, the CALL will be specialized and the JIT can eliminate the COPY and SWAPs.
Activity
markshannon commentedon Jun 17, 2024
No performance impact
https://github.com/faster-cpython/benchmarking-public/tree/main/results/bm-20240617-3.14.0a0-68df0a5
pythongh-120507: Double WASI memory
gh-120507: Double WASI memory (#120648)
GH-120507: Lower the `BEFORE_WITH` and `BEFORE_ASYNC_WITH` instructio…
pythonGH-120507: Lower the `BEFORE_WITH` and `BEFORE_ASYNC_WITH` inst…
pythongh-120507: Double WASI memory (python#120648)
pythonGH-120507: Lower the `BEFORE_WITH` and `BEFORE_ASYNC_WITH` inst…
pythongh-120507: Double WASI memory (python#120648)
pythonGH-120507: Lower the `BEFORE_WITH` and `BEFORE_ASYNC_WITH` inst…
pythongh-120507: Double WASI memory (python#120648)
pythonGH-120507: Lower the `BEFORE_WITH` and `BEFORE_ASYNC_WITH` inst…