A resource allocator for A/B performance tests, where A/B means comparing a reference version with a changed version. It assigns separate containers, folders, CPU sets, logs, and result files.
An environment checker for CPython and CinderX performance experiments. It determines whether an existing setup can be reused, needs basic setup, or needs a clean rebuild.
A nine-part evidence table for deeply investigating one benchmark case, from performance measurements through generated instructions, processor behavior, a proposed optimization, and an isolated A/B test.
Use when repeated CPython/CinderX validation builds are slow and the task involves cinderxlocal.setuprelease, release wheel builds, rungate.py, testcinderxrelease, or pyperformance candidate wheel builds that can safely reuse ccache without modifying the source tree.
A method for analyzing CinderX’s intermediate code and generated machine code. It covers the stages from HIR and LIR, the lower-level intermediate form, through small operations, instructions, deoptimization, and calling details.
An analysis method for benchmarks that mainly run through the interpreter, which executes code directly instead of compiling it with CinderX’s just-in-time compiler.
A check for proving that the benchmark’s own code entered the CinderX just-in-time compiler, rather than startup code, imported packages, or test-generated code.
A report format for recording CPython or CinderX optimization work, crashes, benchmark regressions, and platform differences so another person can reproduce the investigation.
Use when running or preparing NUMA/L3-aware parallel pyperformance validation for CinderX on large ARM64/Linux hosts, especially blue-server-53/Kunpeng-style machines, with adaptive 8-core or 16-core scheduling, CinderX manager/worker venv proof, pyperformance worker injection, SSH proxy setup, full-suite/subset…
A runbook for operating CinderX and CPython experiments on remote machines over SSH. It covers tmux, file copying, Docker Compose, logs, command results, time limits, and long-running tasks.
A CinderX installation and JIT smoke check. CinderX is a modified Python runtime with a just-in-time compiler, which turns some Python code into machine code while it runs.
A reference guide to compiler optimization theory, including intermediate representations, data-flow analysis, instruction scheduling, and register allocation. An intermediate representation is the compiler’s internal form of a program before machine code is produced.
A procedure for running CinderX and CPython runtime and integration tests. Runtime tests check program behavior, while integration tests check how connected parts work together.
A set of templates and guidance for writing structured software design documents, from feature proposals through architecture and detailed implementation.
A reference tool for official ARM A64 and x86-64 processor instruction specifications. These specifications describe instruction encoding, syntax, behavior, and hardware requirements.
A comparison guide for existing pyperformance result files. It checks whether baseline and candidate runs were made under comparable conditions before interpreting speed changes.